0
登录后你可以
  • 下载海量资料
  • 学习在线课程
  • 观看技术视频
  • 写文章/发帖/加入社区
会员中心
创作中心
发布
  • 发文章

  • 发资料

  • 发帖

  • 提问

  • 发视频

创作活动
A3P1000L-1FG144

A3P1000L-1FG144

  • 厂商:

    MICROSEMI(美高森美)

  • 封装:

    FBGA

  • 描述:

    IC FPGA 97 I/O 144FBGA

  • 详情介绍
  • 数据手册
  • 价格&库存
A3P1000L-1FG144 数据手册
ProASIC3L FPGA Fabric User’s Guide ProASIC3L FPGA Fabric User’s Guide Table of Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Revision History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Related Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1 FPGA Array Architecture in Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Device Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 FPGA Array Architecture Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Device Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2 Flash*Freeze Technology and Low Power Modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Flash*Freeze Technology and Low Power Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flash Families Support the Flash*Freeze Feature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Low Power Modes Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Static (Idle) Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flash*Freeze Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sleep and Shutdown Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Flash*Freeze Design Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 22 23 23 24 32 34 42 42 42 3 Global Resources in Low Power Flash Devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Resource Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VersaNet Global Network Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chip and Quadrant Global I/Os . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Spine Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Using Clock Aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Design Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 47 48 49 51 57 60 62 74 74 75 4 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs . . . . . . . . . . . . . . 77 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overview of Clock Conditioning Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CCC Support in Microsemi’s Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Buffers with No Programmable Delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Buffer with Programmable Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Buffers with PLL Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global Input Selections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Revision 4 77 77 79 80 80 83 87 2 ProASIC3L FPGA Fabric User’s Guide Device-Specific Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 PLL Core Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 Functional Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Software Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Detailed Usage Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Recommended Board-Level Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 5 FlashROM in Microsemi’s Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Architecture of User Nonvolatile FlashROM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 FlashROM Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 FlashROM Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 FlashROM Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 Programming and Accessing FlashROM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 FlashROM Design Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 Custom Serialization Using FlashROM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 6 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . 147 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Device Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 SRAM/FIFO Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 SRAM and FIFO Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Memory Blocks and Macros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Initializing the RAM/FIFO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 Software Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 7 I/O Structures in IGLOO and ProASIC3 Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Low Power Flash Device I/O Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 Advanced I/Os—IGLOO, ProASIC3L, and ProASIC3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 I/O Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 I/O Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 I/O Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 Simultaneously Switching Outputs (SSOs) and Printed Circuit Board Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 I/O Software Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 User I/O Naming Convention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 Board-Level Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 8 I/O Structures in IGLOOe and ProASIC3E Devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Revision 4 3 Table of Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 Low Power Flash Device I/O Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 Pro I/Os—IGLOOe, ProASIC3EL, and ProASIC3E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 I/O Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 I/O Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 I/O Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Simultaneously Switching Outputs (SSOs) and Printed Circuit Board Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 I/O Software Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 User I/O Naming Convention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Board-Level Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 9 I/O Software Control in Low Power Flash Devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251 Flash FPGAs I/O Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Software-Controlled I/O Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implementing I/Os in Microsemi Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Assigning Technologies and VREF to I/O Banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 253 254 264 269 269 270 10 DDR for Microsemi’s Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 Double Data Rate (DDR) Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 DDR Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 I/O Cell Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 Input Support for DDR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Output Support for DDR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Instantiating DDR Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 Design Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 11 Programming Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Summary of Programming Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Programming Support in Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 General Flash Programming Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 Important Programming Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298 12 Security in Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301 Security in Programmable Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Security Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Security Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Security Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Security in Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 R e vi s i o n 4 301 302 303 304 308 ProASIC3L FPGA Fabric User’s Guide FlashROM Security Use Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Generating Programming Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 313 324 324 324 325 325 13 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X . . . 327 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 ISP Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 ISP Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328 Programming Voltage (VPUMP) and VJTAG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Nonvolatile Memory (NVM) Programming Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 IEEE 1532 (JTAG) Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330 Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330 Security in ARM-Enabled Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 FlashROM and Programming Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333 Programming Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334 ISP Programming Header Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 Board-Level Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 14 Core Voltage Switching Circuit for IGLOO and ProASIC3L In-System Programming . . . . . . . . . . . . 341 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 Microsemi’s Flash Families Support Voltage Switching Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Circuit Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Circuit Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 DirectC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347 15 Microprocessor Programming of Microsemi’s Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . 349 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 Microprocessor Programming Support in Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 Programming Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Implementation Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Hardware Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354 Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356 16 Boundary Scan in Low Power Flash Devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 Boundary Scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TAP Controller State Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Microsemi’s Flash Devices Support the JTAG Feature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary Scan Support in Low Power Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Boundary Scan Opcodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Revision 4 357 357 358 359 359 5 Table of Contents Boundary Scan Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Board-Level Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Advanced Boundary Scan Register Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359 360 361 362 17 UJTAG Applications in Microsemi’s Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 UJTAG Support in Flash-Based Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364 UJTAG Macro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 UJTAG Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366 Typical UJTAG Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372 18 Power-Up/-Down Behavior of Low Power Flash Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Flash Devices Support Power-Up Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374 Power-Up/-Down Sequence and Transient Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375 I/O Behavior at Power-Up/-Down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377 Cold-Sparing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382 Hot-Swapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383 Related Documents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 List of Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 A Summary of Changes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385 History of Revision to Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385 B Product Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 Customer Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Customer Technical Support Center . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Technical Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Website . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contacting the Customer Technical Support Center . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ITAR Technical Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 387 387 387 387 388 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389 6 R e vi s i o n 4 Introduction Contents This user’s guide contains information to help designers understand and use Microsemi's ProASIC®3L devices. Each chapter addresses a specific topic. Most of these chapters apply to other Microsemi device families as well. When a feature or description applies only to a specific device family, this is made clear in the text. Revision History The revision history for each chapter is listed at the end of the chapter. Most of these chapters were formerly included in device handbooks. Some were originally application notes or information included in device datasheets. A "Summary of Changes" table at the end of this user’s guide lists the chapters that were changed in each revision of the document, with links to the "List of Changes" sections for those chapters. Related Information Refer to the ProASIC3L Flash Family FPGAs datasheet for detailed specifications, timing, and package and pin information. The website page for ProASIC3L devices is /www.microsemi.com/soc/products/pa3l/default.aspx. Revision 4 7 1 – FPGA Array Architecture in Low Power Flash Devices Device Architecture Advanced Flash Switch Unlike SRAM FPGAs, the low power flash devices use a live-at-power-up ISP flash switch as their programming element. Flash cells are distributed throughout the device to provide nonvolatile, reconfigurable programming to connect signal lines to the appropriate VersaTile inputs and outputs. In the flash switch, two transistors share the floating gate, which stores the programming information (Figure 1-1). One is the sensing transistor, which is only used for writing and verification of the floating gate voltage. The other is the switching transistor. The latter is used to connect or separate routing nets, or to configure VersaTile logic. It is also used to erase the floating gate. Dedicated high-performance lines are connected as required using the flash switch for fast, low-skew, global signal distribution throughout the device core. Maximum core utilization is possible for virtually any design. The use of the flash switch technology also removes the possibility of firm errors, which are increasingly common in SRAM-based FPGAs. Floating Gate Sensing Switch In Switching Word Switch Out Figure 1-1 • Flash-Based Switch Revision 4 9 FPGA Array Architecture in Low Power Flash Devices FPGA Array Architecture Support The flash FPGAs listed in Table 1-1 support the architecture features described in this document. Table 1-1 • Flash-Based FPGAs Family* Series IGLOO® ® ProASIC 3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 1-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 1-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 10 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Device Overview Low power flash devices consist of multiple distinct programmable architectural features (Figure 1-5 on page 13 through Figure 1-7 on page 14): • FPGA fabric/core (VersaTiles) • Routing and clock resources (VersaNets) • FlashROM • Dedicated SRAM and/or FIFO – 30 k gate and smaller device densities do not support SRAM or FIFO. – Automotive devices do not support FIFO operation. • I/O structures • Flash*Freeze technology and low power modes Bank 1* I/Os Bank 1 Bank 0 VersaTile User Nonvolatile FlashROM Flash*Freeze† Technology Charge Pumps CCC-GL Bank 1 Notes: * Bank 0 for the 30 k devices † Flash*Freeze mode is supported on IGLOO devices. Figure 1-2 • IGLOO and ProASIC3 nano Device Architecture Overview with Two I/O Banks (applies to 10 k and 30 k device densities, excluding IGLOO PLUS devices) Revision 4 11 FPGA Array Architecture in Low Power Flash Devices Bank 0 Bank 0 Bank 1 CCC RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block I/Os ISP AES Decryption User Nonvolatile FlashRom Flash*Freeze† Technology Charge Pumps Bank 0 Bank 1 VersaTile Bank 1 Note: † Flash*Freeze mode is supported on IGLOO devices. Figure 1-3 • IGLOO Device Architecture Overview with Two I/O Banks with RAM and PLL (60 k and 125 k gate densities) Bank 1 I/Os Bank 1 Bank 0 VersaTile User Nonvolatile FlashROM Flash*Freeze† Technology Charge Pumps CCC-GL Bank 1 Note: † Flash*Freeze mode is supported on IGLOO devices. Figure 1-4 • IGLOO Device Architecture Overview with Three I/O Banks (AGLN015, AGLN020, A3PN015, and A3PN020) 12 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Bank 0 RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 1 Bank 3 CCC I/Os Bank 1 Bank 3 VersaTile ISP AES Decryption* User Nonvolatile FlashRom Flash*Freeze Technology † RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps Bank 2 Note: Flash*Freeze technology only applies to IGLOO and ProASIC3L families. Figure 1-5 • IGLOO, IGLOO nano, ProASIC3 nano, and ProASIC3/L Device Architecture Overview with Four I/O Banks (AGL600 device is shown) Bank 0 Bank 1 Bank 3 CCC* RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block I/Os ISP AES Decryption* User Nonvolatile FlashRom Flash*Freeze Technology Charge Pumps Bank 1 Bank 3 VersaTile Bank 2 Note: * AGLP030 does not contain a PLL or support AES security. Figure 1-6 • IGLOO PLUS Device Architecture Overview with Four I/O Banks Revision 4 13 FPGA Array Architecture in Low Power Flash Devices Bank 0 Bank 1 CCC Bank 2 Bank 7 RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Pro I/Os Bank 3 Bank 6 VersaTile ISP AES Decryption User Nonvolatile FlashRom Bank 5 Flash*Freeze† Technology Charge Pumps RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 4 Note: Flash*Freeze technology only applies to IGLOOe devices. Figure 1-7 • IGLOOe and ProASIC3E Device Architecture Overview (AGLE600 device is shown) I/O State of Newly Shipped Devices Devices are shipped from the factory with a test design in the device. The power-on switch for VCC is OFF by default in this test design, so I/Os are tristated by default. Tristated means the I/O is not actively driven and floats. The exact value cannot be guaranteed when it is floating. Even in simulation software, a tristate value is marked as unknown. Due to process variations and shifts, tristated I/Os may float toward High or Low, depending on the particular device and leakage level. If there is concern regarding the exact state of unused I/Os, weak pull-up/pull-down should be added to the floating I/Os so their state is controlled and stabilized. 14 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Core Architecture VersaTile The proprietary IGLOO and ProASIC3 device architectures provide granularity comparable to gate arrays. The device core consists of a sea-of-VersaTiles architecture. As illustrated in Figure 1-8, there are four inputs in a logic VersaTile cell, and each VersaTile can be configured using the appropriate flash switch connections: • Any 3-input logic function • Latch with clear or set • D-flip-flop with clear or set • Enable D-flip-flop with clear or set (on a 4th input) VersaTiles can flexibly map the logic and sequential gates of a design. The inputs of the VersaTile can be inverted (allowing bubble pushing), and the output of the tile can connect to high-speed, very-long-line routing resources. VersaTiles and larger functions can be connected with any of the four levels of routing hierarchy. When the VersaTile is used as an enable D-flip-flop, SET/CLR is supported by a fourth input. The SET/CLR signal can only be routed to this fourth input over the VersaNet (global) network. However, if, in the user’s design, the SET/CLR signal is not routed over the VersaNet network, a compile warning message will be given, and the intended logic function will be implemented by two VersaTiles instead of one. The output of the VersaTile is F2 when the connection is to the ultra-fast local lines, or YL when the connection is to the efficient long-line or very-long-line resources. 0 1 Data X3 Y Pin 1 0 1 0 1 F2 YL CLK X2 0 1 CLR/ Enable X1 CLR XC* Legend: Via (hard connection) Switch (flash connection) Ground * This input can only be connected to the global clock distribution network. Figure 1-8 • Low Power Flash Device Core VersaTile Revision 4 15 FPGA Array Architecture in Low Power Flash Devices Array Coordinates During many place-and-route operations in the Microsemi Designer software tool, it is possible to set constraints that require array coordinates. Table 1-2 provides array coordinates of core cells and memory blocks for IGLOO and ProASIC3 devices. Table 1-3 provides the information for IGLOO PLUS devices. Table 1-4 on page 17 provides the information for IGLOO nano and ProASIC3 nano devices. The array coordinates are measured from the lower left (0, 0). They can be used in region constraints for specific logic groups/blocks, designated by a wildcard, and can contain core cells, memories, and I/Os. I/O and cell coordinates are used for placement constraints. Two coordinate systems are needed because there is not a one-to-one correspondence between I/O cells and core cells. In addition, the I/O coordinate system changes depending on the die/package combination. It is not listed in Table 1-2. The Designer ChipPlanner tool provides the array coordinates of all I/O locations. I/O and cell coordinates are used for placement constraints. However, I/O placement is easier by package pin assignment. Figure 1-9 on page 17 illustrates the array coordinates of a 600 k gate device. For more information on how to use array coordinates for region/placement constraints, see the Designer User's Guide or online help (available in the software) for software tools. Table 1-2 • IGLOO and ProASIC3 Array Coordinates VersaTiles Device Min. Memory Rows Max. Entire Die Bottom Top Min. Max. IGLOO ProASIC3/ ProASIC3L x y x y (x, y) (x, y) (x, y) (x, y) AGL015 A3P015 3 2 34 13 None None (0, 0) (37, 15) AGL030 A3P030 3 3 66 13 None None (0, 0) (69, 15) AGL060 A3P060 3 2 66 25 None (3, 26) (0, 0) (69, 29) AGL125 A3P125 3 2 130 25 None (3, 26) (0, 0) (133, 29) AGL250 A3P250/L 3 2 130 49 None (3, 50) (0, 0) (133, 53) AGL400 A3P400 3 2 194 49 None (3, 50) (0, 0) (197, 53) AGL600 A3P600/L 3 4 194 75 (3, 2) (3, 76) (0, 0) (197, 79) AGL1000 A3P1000/L 3 4 258 99 (3, 2) (3, 100) (0, 0) (261, 103) AGLE600 A3PE600/L, RT3PE600L 3 4 194 75 (3, 2) (3, 76) (0, 0) (197, 79) A3PE1500 3 4 322 123 (3, 2) (3, 124) (0, 0) (325, 127) A3PE3000/L, RT3PE3000L 3 6 450 173 (3, 2) or (3, 4) (3, 174) or (3, 176) (0, 0) (453, 179) AGLE3000 Table 1-3 • IGLOO PLUS Array Coordinates VersaTiles Device 16 Min. Memory Rows Max. Entire Die Bottom Top Min. Max. IGLOO PLUS x y x y (x, y) (x, y) (x, y) (x, y) AGLP030 2 3 67 13 None None (0, 0) (69, 15) AGLP060 2 2 67 25 None (3, 26) (0, 0) (69, 29) AGLP125 2 2 131 25 None (3, 26) (0, 0) (133, 29) R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Table 1-4 • IGLOO nano and ProASIC3 nano Array Coordinates VersaTiles Device Memory Rows Entire Die Min. Max. Bottom Top Min. Max. IGLOO nano ProASIC3 nano (x, y) (x, y) (x, y) (x, y) (x, y) (x, y) AGLN010 A3P010 (0, 2) (32, 5) None None (0, 0) (34, 5) AGLN015 A3PN015 (0, 2) (32, 9) None None (0, 0) (34, 9) AGLN020 A3PN020 (0, 2) 32, 13) None None (0, 0) (34, 13) AGLN060 A3PN060 (3, 2) (66, 25) None (3, 26) (0, 0) (69, 29) AGLN125 A3PN125 (3, 2) (130, 25) None (3, 26) (0, 0) (133, 29) AGLN250 A3PN250 (3, 2) (130, 49) None (3, 50) (0, 0) (133, 49) Top Row (7, 79) to (189, 79) Bottom Row (5, 78) to (192, 78) (0, 79) I/O Tile (197, 79) Memory (3, 77) Blocks (3, 76) (194, 77) Memory (194, 76) Blocks VersaTile (Core) (3, 75) (194, 75) VersaTile (Core) (194, 4) VersaTile (Core) VersaTile (Core) (3, 4) (194, 3) Memory (194, 2) Blocks Memory (3, 3) Blocks (3, 2) (197, 1) (0, 0) I/O Tile UJTAG FlashROM Top Row (5, 1) to (168, 1) Bottom Row (7, 0) to (165, 0) (197, 0) Top Row (169, 1) to (192, 1) Note: The vertical I/O tile coordinates are not shown. West-side coordinates are {(0, 2) to (2, 2)} to {(0, 77) to (2, 77)}; east-side coordinates are {(195, 2) to (197, 2)} to {(195, 77) to (197, 77)}. Figure 1-9 • Array Coordinates for AGL600, AGLE600, A3P600, and A3PE600 Revision 4 17 FPGA Array Architecture in Low Power Flash Devices Routing Architecture The routing structure of low power flash devices is designed to provide high performance through a flexible four-level hierarchy of routing resources: ultra-fast local resources; efficient long-line resources; high-speed, very-long-line resources; and the high-performance VersaNet networks. The ultra-fast local resources are dedicated lines that allow the output of each VersaTile to connect directly to every input of the eight surrounding VersaTiles (Figure 1-10). The exception to this is that the SET/CLR input of a VersaTile configured as a D-flip-flop is driven only by the VersaTile global network. The efficient long-line resources provide routing for longer distances and higher-fanout connections. These resources vary in length (spanning one, two, or four VersaTiles), run both vertically and horizontally, and cover the entire device (Figure 1-11 on page 19). Each VersaTile can drive signals onto the efficient long-line resources, which can access every input of every VersaTile. Routing software automatically inserts active buffers to limit loading effects. The high-speed, very-long-line resources, which span the entire device with minimal delay, are used to route very long or high-fanout nets: length ±12 VersaTiles in the vertical direction and length ±16 in the horizontal direction from a given core VersaTile (Figure 1-12 on page 19). Very long lines in low power flash devices have been enhanced over those in previous ProASIC families. This provides a significant performance boost for long-reach signals. The high-performance VersaNet global networks are low-skew, high-fanout nets that are accessible from external pins or internal logic. These nets are typically used to distribute clocks, resets, and other highfanout nets requiring minimum skew. The VersaNet networks are implemented as clock trees, and signals can be introduced at any junction. These can be employed hierarchically, with signals accessing every input of every VersaTile. For more details on VersaNets, refer to the "Global Resources in Low Power Flash Devices" section on page 47. Long Lines L Inputs L L L Output L L L L Ultra-Fast Local Lines (connects a VersaTile to the adjacent VersaTile, I/O buffer, or memory block) L Note: Input to the core cell for the D-flip-flop set and reset is only available via the VersaNet global network connection. Figure 1-10 • Ultra-Fast Local Lines Connected to the Eight Nearest Neighbors 18 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Spans 4 VersaTiles Spans 2 VersaTiles Spans 1 VersaTile VersaTile L L L L L L L L L L L L L L L L L L L L L L L L L L L L L L Spans 1 VersaTile Spans 2 VersaTiles Spans 4 VersaTiles Figure 1-11 • Efficient Long-Line Resources High-Speed, Very-Long-Line Resources Pad Ring SRAM I/O Ring Pad Ring I/O Ring 16×12 Block of VersaTiles Pad Ring Figure 1-12 • Very-Long-Line Resources Revision 4 19 FPGA Array Architecture in Low Power Flash Devices Related Documents User’s Guides Designer User's Guide http://www.microsemi.com/soc/documents/designer_ug.pdf List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page August 2012 The "I/O State of Newly Shipped Devices" section is new (SAR 39542). 14 July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 1-1 • Flash-Based FPGAs. 10 Figure 1-2 • IGLOO and ProASIC3 nano Device Architecture Overview with Two I/O Banks (applies to 10 k and 30 k device densities, excluding IGLOO PLUS devices) through Figure 1-5 • IGLOO, IGLOO nano, ProASIC3 nano, and ProASIC3/L Device Architecture Overview with Four I/O Banks (AGL600 device is shown) are new. 11, 12 v1.3 (October 2008) v1.2 (June 2008) v1.1 (March 2008) 20 Table 1-4 • IGLOO nano and ProASIC3 nano Array Coordinates is new. 17 The title of this document was changed from "Core Architecture of IGLOO and ProASIC3 Devices" to "FPGA Array Architecture in Low Power Flash Devices." 9 The "FPGA Array Architecture Support" section was revised to include new families and make the information more concise. 10 Table 1-2 • IGLOO and ProASIC3 Array Coordinates was updated to include Military ProASIC3/EL and RT ProASIC3 devices. 16 The following changes were made to the family descriptions in Table 1-1 • FlashBased FPGAs: 10 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. Table 1-1 • Flash-Based FPGAs and the accompanying text was updated to include the IGLOO PLUS family. The "IGLOO Terminology" section and "Device Overview" section are new. 10 The "Device Overview" section was updated to note that 15 k devices do not support SRAM or FIFO. 11 Figure 1-6 • IGLOO PLUS Device Architecture Overview with Four I/O Banks is new. 13 Table 1-2 • IGLOO and ProASIC3 Array Coordinates was updated to add A3P015 and AGL015. 16 Table 1-3 • IGLOO PLUS Array Coordinates is new. 16 R e visio n 4 2 – Flash*Freeze Technology and Low Power Modes Flash*Freeze Technology and Low Power Modes Microsemi IGLOO,® IGLOO nano, IGLOO PLUS, ProASIC®3L, and Radiation-Tolerant (RT) ProASIC3 FPGAs with Flash*Freeze technology are designed to meet the most demanding power and area challenges of today’s portable electronics products with a reprogrammable, small-footprint, full-featured flash FPGA. These devices offer lower power consumption in static and dynamic modes, utilizing the unique Flash*Freeze technology, than any other FPGA or CPLD. IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 devices offer various power-saving modes that enable every system to utilize modes that achieve the lowest total system power. Low Power Active capability (static idle) allows for ultra-low power consumption while the device is operational in the system by maintaining SRAM, registers, I/Os, and logic functions. Flash*Freeze technology provides an ultra-low power static mode (Flash*Freeze mode) that retains all SRAM and register information with rapid recovery to Active (operating) mode. IGLOO nano and IGLOO PLUS devices have an additional feature when operating in Flash*Freeze mode, allowing them to retain I/O states as well as SRAM and register states. This mechanism enables the user to quickly (within 1 µs) enter and exit Flash*Freeze mode by activating the Flash*Freeze (FF) pin while all power supplies are kept in their original states. In addition, I/Os and clocks connected to the FPGA can still be toggled without impact on device power consumption. While in Flash*Freeze mode, the device retains all core register states and SRAM information. This mode can be configured so that no power is consumed by the I/O banks, clocks, JTAG pins, or PLLs; and the IGLOO and IGLOO PLUS devices consume as little as 5 µW, while IGLOO nano devices consume as little as 2 µW. Microsemi offers a state management IP core to aid users in gating clocks and managing data before entering Flash*Freeze mode. This document will guide users in selecting the best low power mode for their applications, and introduces Microsemi's Flash*Freeze management IP core. Revision 4 21 Flash*Freeze Technology and Low Power Modes Flash Families Support the Flash*Freeze Feature The low power flash FPGAs listed in Table 2-1 support the Flash*Freeze feature and the functions described in this document. Table 2-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 2-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 2-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 22 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Low Power Modes Overview Table 2-2 summarizes the low power modes that achieve power consumption reduction when the FPGA or system is idle. Table 2-2 • Power Modes Summary Mode VCCI VCC Core Clocks ULSICC Macro To Enter Mode Active On On On On N/A Initiate clock None – Idle On On On Off N/A Stop clock Initiate clock External Flash*Freeze type 1 On On On On* N/A Assert FF pin Deassert FF pin External Flash*Freeze type 2 On On On On* Used to enter Flash*Freeze mode Assert FF pin and assert LSICC Deassert FF pin External Sleep On Off Off Off N/A Shut down VCC Shutdown Off Off Off Off N/A Shut down VCC and VCCI supplies Static To Resume Operation Trigger Turn on External VCC supply Turn on VCC and VCCI supplies External * External clocks can be left toggling while the device is in Flash*Freeze mode. Clocks generated by the embedded PLL will be turned off automatically. Static (Idle) Mode In Static (Idle) mode, none of the clock inputs is switching, and static power is the only power consumed by the device. This mode can be achieved by switching off the incoming clocks to the FPGA, thus benefitting from reduced power consumption. In addition, I/Os draw only minimal leakage current. In this mode, embedded SRAM, I/Os, and registers retain their values so the device can enter and exit this mode just by switching the clocks on or off. If the device-embedded PLL is used as the clock source, Static (Idle) mode can easily be entered by pulling the PLL POWERDOWN pin LOW (active Low), which will turn off the PLL. Revision 4 23 Flash*Freeze Technology and Low Power Modes Flash*Freeze Mode IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 FPGAs offer an ultra-low static power mode to reduce power consumption while preserving the state of the registers, SRAM contents, and I/O states (IGLOO nano and IGLOO PLUS only) without switching off any power supplies, inputs, or input clocks. Flash*Freeze technology enables the user to switch to Flash*Freeze mode within 1 µs, thus simplifying low power design implementation. The Flash*Freeze (FF) pin (active Low) is a dedicated pin used to enter or exit Flash*Freeze mode directly; or the pin can be routed internally to the FPGA core and state management IP to allow the user's application to decide if and when it is safe to transition to this mode. If the FF pin is not used, it can be used as a regular I/O. The FF pin has a built-in glitch filter and optional Schmitt trigger (not available for all devices) to prevent entering or exiting Flash*Freeze mode accidentally. There are two ways to use Flash*Freeze mode. In Flash*Freeze type 1, entering and exiting the mode is exclusively controlled by the assertion and deassertion of the FF pin. This enables an external processor or human interface device to directly control Flash*Freeze mode; however, valid data must be preserved using standard procedures (refer to the "Flash*Freeze Mode Device Behavior" section on page 30). In Flash*Freeze mode type 2, entering and exiting the mode is controlled by both the FF pin AND userdefined logic. Flash*Freeze management IP may be used in type 2 mode for clock and data management while entering and exiting Flash*Freeze mode. Flash*Freeze Type 1: Control by Dedicated Flash*Freeze Pin Flash*Freeze type 1 is intended for systems where either the device will be reset upon exiting Flash*Freeze mode, or data and clock are managed externally. The device enters Flash*Freeze mode 1 µs after the dedicated FF pin is asserted (active Low), and returns to normal operation when the FF pin is deasserted (High) (Figure 2-1 on page 25). In this mode, FF pin assertion or deassertion is the only condition that determines entering or exiting Flash*Freeze mode. In Libero® System-on-Chip (SoC) software v8.2 and before, this mode is implemented by enabling Flash*Freeze mode (default setting) in the Compile options of the Microsemi Designer software. To simplify usage of Flash*Freeze mode, beginning with Libero software v8.3, an INBUF_FF I/O macro was introduced. An INBUF_FF I/O buffer must be used to identify the Flash*Freeze input. Microsemi recommends switching to the new implementation. In Libero software v8.3 and later, the user must manually instantiate the INBUF_FF macro in the top level of the design to implement Flash*Freeze Type 1, as shown in Figure 2-1 on page 25. 24 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Figure 2-1 shows the concept of FF pin control in Flash*Freeze mode type 1. IGLOO, IGLOO PLUS, IGLOO nano, ProASIC3L, or RT ProASIC3 Device INBUF_FF Flash*Freeze Mode Control To FPGA Core or Floating Flash*Freeze (FF) Pin Flash*Freeze Signal Enables Entering Flash*Freeze Mode 1 AND Flash*Freeze Mode Flash*Freeze Technology User Design Figure 2-1 • Flash*Freeze Mode Type 1 – Controlled by the Flash*Freeze Pin Figure 2-2 shows the timing diagram for entering and exiting Flash*Freeze mode type 1. Normal Operation Normal Operation Flash*Freeze Mode Flash*Freeze Pin t = 1 μs t = 1 μs Figure 2-2 • Flash*Freeze Mode Type 1 – Timing Diagram Revision 4 25 Flash*Freeze Technology and Low Power Modes Flash*Freeze Type 2: Control by Dedicated Flash*Freeze Pin and Internal Logic The device can be made to enter Flash*Freeze mode by activating the FF pin together with Microsemi's Flash*Freeze management IP core (refer to the "Flash*Freeze Management IP" section on page 36 for more information) or user-defined control logic (Figure 2-3 on page 27) within the FPGA core. This method enables the design to perform important activities before allowing the device to enter Flash*Freeze mode, such as transitioning into a safe state, completing the processing of a critical event. Designers are encouraged to take advantage of Microsemi's Flash*Freeze Management IP to handle clean entry and exit of Flash*Freeze mode (described later in this document). The device will only enter Flash*Freeze mode when the Flash*Freeze pin is asserted (active Low) and the User Low Static ICC (ULSICC) macro input signal, called the LSICC signal, is asserted (High). One condition is not sufficient to enter Flash*Freeze mode type 2; both the FF pin and LSICC signal must be asserted. When Flash*Freeze type 2 is implemented in the design, the ULSICC macro needs to be instantiated by the user. There are no functional differences in the device whether the ULSICC macro is instantiated or not, and whether the LSICC signal is asserted or deasserted. The LSICC signal is used only to control entering Flash*Freeze mode. Figure 2-4 on page 27 shows the timing diagram for entering and exiting Flash*Freeze mode type 2. After exiting Flash*Freeze mode type 2 by deasserting the Flash*Freeze pin, the LSICC signal must be deasserted by the user design. This will prevent entering Flash*Freeze mode by asserting the Flash*Freeze pin only. Refer to Table 2-3 for Flash*Freeze (FF) pin and LSICC signal assertion and deassertion values. Table 2-3 • Flash*Freeze Mode Type 1 and Type 2 – Signal Assertion and Deassertion Values Signal Assertion Value Deassertion Value Flash*Freeze (FF) pin Low High LSICC signal High Low Notes: 1. The Flash*Freeze (FF) pin is an active-Low signal, and LSICC is an active-High signal. 2. The LSICC signal is used only in Flash*Freeze mode type 2. 26 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide IGLOO, IGLOO PLUS, IGLOO nano, ProASIC3L, or RT ProASIC3 Device Flash*Freeze (FF) Pin Connect to Top-Level Port Auto-Connected to IP INBUF_FF Flash*Freeze Signal ULSICC Macro Enables Entering Flash*Freeze Mode Flash*Freeze Management IP AND Flash*Freeze Mode Flash*Freeze Technology User Design Figure 2-3 • Flash*Freeze Mode Type 2 – Controlled by Flash*Freeze Pin and Internal Logic (LSICC signal) Normal Operation Flash*Freeze Mode Normal Operation Flash*Freeze Pin LSICC Signal t = 1 μs t = 1 μs Figure 2-4 • Flash*Freeze Mode Type 2 – Timing Diagram Revision 4 27 Flash*Freeze Technology and Low Power Modes Table 2-4 summarizes the Flash*Freeze mode implementations. Table 2-4 • Flash*Freeze Mode Usage Flash*Freeze Mode Type 1 2 Description Flash*Freeze Instantiate Pin State ULSICC Macro Flash*Freeze mode is Deasserted controlled only by the Asserted FF pin. Flash*Freeze mode is "Don’t care" controlled by the FF Deasserted pin and LSICC signal. Asserted LSICC Signal Operating Mode No N/A Normal operation No N/A Flash*Freeze mode Yes Deasserted Normal operation Yes "Don’t care" Normal operation Yes Asserted Flash*Freeze mode Note: Refer to Table 2-3 on page 26 for Flash*Freeze pin and LSICC signal assertion and deassertion values. IGLOO, ProASIC3L, and RT ProASIC3 I/O State in Flash*Freeze Mode In IGLOO and ProASIC3L devices, when the device enters Flash*Freeze mode, I/Os become tristated. If the weak pull-up or pull-down feature is used, the I/Os will maintain the configured weak pull-up or pulldown status. This feature enables the design to set the I/O state to a certain level that is determined by the pull-up/-down configuration. Table 2-5 shows the I/O pad state based on the configuration and buffer type. Note that configuring weak pull-up or pull-down for the FF pin is not allowed. The FF pin can be configured as a Schmitt trigger input in IGLOOe, IGLOO nano, IGLOO PLUS, and ProASIC3EL devices. Table 2-5 • IGLOO, ProASIC3L, and RT ProASIC3 Flash*Freeze Mode (type 1 and type 2)—I/O Pad State Buffer Type I/O Pad Weak Pull-Up/-Down I/O Pad State in Flash*Freeze Mode Input/Global Enabled Weak pull-up/pull-down* Output Bidirectional / Tristate Buffer Disabled Tristate* Enabled Weak pull-up/pull-down Disabled Tristate Enabled Weak pull-up/pull-down* E=0 (input/tristate) Disabled Tristate* E = 1 (output) Enabled Weak pull-up/pull-down Disabled Tristate * Internal core logic driven by this input/global buffer will be tied High as long as the device is in Flash*Freeze mode. 28 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide IGLOO nano and IGLOO PLUS I/O State in Flash*Freeze Mode In IGLOO nano and IGLOO PLUS devices, users have multiple options in how to configure I/Os during Flash*Freeze mode: 1. Hold the previous state 2. Set I/O pad to weak pull-up or pull-down 3. Tristate I/O pads The I/O configuration must be configured by the user in the I/O Attribute Editor or in a PDC constraint file, and can be done on a pin-by-pin basis. The output hold feature will hold the output in the last registered state, using the I/O pad weak pull-up or pull-down resistor when the FF pin is asserted. When inputs are configured with the hold feature enabled, the FPGA core side of the input will hold the last valid state of the input pad before the device entered Flash*Freeze mode. The input pad can be driven to any value, configured as tristate, or configured with the weak pull-up or pull-down I/O pad feature during Flash*Freeze mode without affecting the hold state. If the weak pull-up or pull-down feature is used without the output hold feature, the input and output pads will maintain the configured weak pull-up or pull-down status during Flash*Freeze mode and normal operation. If a fixed weak pull-up or pull-down is defined on an output buffer or as bidirectional in output mode, and a hold state is also defined for the same pin, the pin will be configured in hold state mode during Flash*Freeze mode. During normal operation, the pin will be configured with the predefined weak pull-up or pull-down. Any I/Os that do not use the hold state or I/O pad weak pull-up or pull-down features will be tristated during Flash*Freeze mode and the FPGA core will be driven High by inputs. Inputs that are tristated during Flash*Freeze mode may be left floating without any reliability concern or impact to power consumption. Table 2-6 shows the I/O pad state based on the configuration and buffer type. Note that configuring weak pull-up or pull-down for the FF pin is not allowed. Table 2-6 • IGLOO nano and IGLOO PLUS Flash*Freeze Mode (type 1 and type 2)—I/O Pad State Buffer Type Input Output Bidirectional / Tristate E = 0 Buffer (input/tristate) E = 1 (output) Hold State I/O Pad Weak Pull-Up/-Down I/O Pad State in Flash*Freeze Mode Enabled Enabled Weak pull-up/pull-down 1 Disabled Enabled Weak pull-up/pull-down 2 Enabled Disabled Tristate 1 Disabled Disabled Tristate 2 Enabled "Don't care" Weak pull to hold state Disabled Enabled Weak pull-up/pull-down Disabled Disabled Tristate Enabled Enabled Weak pull-up/pull-down 1 Disabled Enabled Weak pull-up/pull-down 2 Enabled Disabled Tristate 1 Disabled Disabled Tristate 2 Enabled "Don't care" Weak pull to hold state 3 Disabled Enabled Weak pull-up/pull-down Disabled Disabled Tristate Notes: 1. Internal core logic driven by this input buffer will be set to the value this I/O had when entering Flash*Freeze mode. 2. Internal core logic driven by this input buffer will be tied High as long as the device is in Flash*Freeze mode. 3. For bidirectional buffers: Internal core logic driven by the input portion of the bidirectional buffer will be set to the hold state. Revision 4 29 Flash*Freeze Technology and Low Power Modes Flash*Freeze Mode Device Behavior Entering Flash*Freeze Mode • IGLOO, IGLOO nano, IGLOO PLUS, ProASCI3L, and RT ProASIC3 devices are designed and optimized to enter Flash*Freeze mode only when power supplies are stable. If the device is being powered up while the FF pin is asserted (Flash*Freeze mode type 1), or while both FF pin and LSICC signal are asserted (Flash*Freeze mode type 2), the device is expected to enter Flash*Freeze mode within 5 µs after the I/Os and FPGA core have reached their activation levels. • If the device is already powered up when the FF pin is asserted, the device will enter Flash*Freeze mode within 1 µs (type 1). In Flash*Freeze mode type 2 operation, entering Flash*Freeze mode is completed within 1 µs after both FF pin and LSICC signal are asserted. Exiting Flash*Freeze mode is completed within 1 µs after deasserting the FF pin only. PLLs • If an embedded PLL is used, entering Flash*Freeze mode will automatically power down the PLL. • The PLL output clocks will stop toggling within 1 µs after the assertion of the FF pin in type 1, or after both FF pin and LSICC signal are asserted in type 2. At the same time, I/Os will transition into the state specified in Table 2-6 on page 29. The user design must ensure it is safe to enter Flash*Freeze mode. I/Os and Globals • While entering Flash*Freeze mode, inputs, globals, and PLLs will enter their Flash*Freeze state asynchronously to each other. As a result, clock and data glitches and narrow pulses may be generated while entering Flash*Freeze mode, as shown in Figure 2-5. Flash*Freeze Pin External Clock Internal Clock Enters Flash*Freeze Mode Exits Flash*Freeze Mode Figure 2-5 • Narrow Clock Pulses During Flash*Freeze Entrance and Exit 30 • I/O banks are not all deactivated simultaneously when entering Flash*Freeze mode. This can cause clocks and inputs to become disabled at different times, resulting in unexpected data being captured. • Upon entering Flash*Freeze mode, all inputs and globals become tied High internally (except when an input hold state is used on IGLOO nano or IGLOO PLUS devices). If any of these signals are driven Low or tied Low externally, they will experience a Low to High transition internally when entering Flash*Freeze mode. • Upon entering type 2 Flash*Freeze mode, ensure the LSICC signal (active High) does not deassert. This can prevent the device from entering Flash*Freeze mode. • Asynchronous input to output paths may experience output glitches. For example, on a direct into-out path, if the current state is '0' and the input bank turns off first, the input and then the output will transition to '1' before the output enters its Flash*Freeze state. This can be prevented by using latches in asynchronous in-to-out paths. • The above situations can cause glitches or invalid data to be clocked into and preserved in the device. Refer to the "Flash*Freeze Design Guide" section on page 34 for solutions. R e visio n 4 ProASIC3L FPGA Fabric User’s Guide During Flash*Freeze Mode • PLLs are turned off during Flash*Freeze mode. • I/O pads are configured according to Table 2-5 on page 28 and Table 2-6 on page 29. • Inputs and input clocks to the FPGA can toggle without any impact on static power consumption, assuming weak pull-up or pull-down is not selected. • If weak pull-up or pull-down is selected and the input is driven to the opposite direction, power dissipation will occur. • Any toggling signals will be charging and discharging the package pin capacitance. • IGLOO and ProASIC3L outputs will be tristated unless the I/O is configured with weak pull-up or pull-down. The output of the I/O to the FPGA core is logic High regardless of whether the I/O pin is configured with a weak pull-up or pull-down. Refer to Table 2-5 on page 28 for more information. • IGLOO nano and IGLOO PLUS output behavior will be based on the configuration defined by the user. Refer to Table 2-6 on page 29 for a description of output behavior during Flash*Freeze mode. • The JTAG circuit is active; however, JTAG operations, such as JTAG commands, JTAG bypass, programming, and authentication, cannot be executed. The device must exit Flash*Freeze mode before JTAG commands can be sent. TCK should be static to avoid extra power consumption from the JTAG state machine. • The FF pin must be externally asserted for the device to stay in Flash*Freeze mode. • The FF pin is still active; i.e., the pin is used to exit Flash*Freeze mode when deasserted. Exiting Flash*Freeze Mode I/Os and Globals • While exiting Flash*Freeze mode, inputs and globals will exit their Flash*Freeze state asynchronously to each other. As a result, clock and data glitches and narrow pulses may be generated while exiting Flash*Freeze mode, unless clock gating schemes are used. • I/O banks are not all activated simultaneously when exiting Flash*Freeze mode. This can cause clocks and inputs to become enabled at different times, resulting in unexpected data being captured. • Upon exiting Flash*Freeze mode, inputs and globals will no longer be tied High internally (does not apply to input hold state on IGLOO nano and IGLOO PLUS). If any of these signals are driven Low or tied Low externally, they will experience a High-to-Low transition internally when exiting Flash*Freeze mode. • Applies only to IGLOO nano and IGLOO PLUS: Output hold state is asynchronously controlled by the signal driving the output buffer (output signal). This ensures a clean, glitch-free transition from hold state to output drive. However, any glitches on the output signal during exit from Flash*Freeze mode may result in glitches on the output pad. • The above situations can cause glitches or invalid data to be clocked into and preserved in the device. Refer to the "Flash*Freeze Design Guide" on page 34 for solutions. PLLs • If the embedded PLL is used, the design must allow maximum acquisition time (per device datasheet) for the PLL to acquire the lock signal. Flash*Freeze Pin Locations Refer to the Pin Descriptions and Packaging chapter of specific device datasheets for information regarding Flash*Freeze pin location on the available packages. The Flash*Freeze pin location is independent of the device, allowing migration to larger or smaller devices while maintaining the same pin location on the board. Revision 4 31 Flash*Freeze Technology and Low Power Modes Sleep and Shutdown Modes Sleep Mode IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 FPGAs support Sleep mode when device functionality is not required. In Sleep mode, VCC (core voltage), VJTAG (JTAG DC voltage), and VPUMP (programming voltage) are grounded, resulting in the FPGA core being turned off to reduce power consumption. While the device is in Sleep mode, the rest of the system can still be operating and driving the input buffers of the device. The driven inputs do not pull up the internal power planes, and the current draw is limited to minimal leakage current. Table 2-7 shows the power supply status in Sleep mode. Table 2-7 • Sleep Mode—Power Supply Requirement for IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 Devices Power Supplies Power Supply State VCC Powered off VCCI = VMV Powered on VJTAG Powered off VPUMP Powered off Refer to the "Power-Up/-Down Behavior" section on page 33 for more information about I/O states during Sleep mode and the timing diagram for entering and exiting Sleep mode. Shutdown Mode Shutdown mode is supported for all IGLOO nano and IGLOO PLUS devices as well the following IGLOO/e devices: AGL015, AGL030, AGLE600, AGLE3000, and A3PE3000L. Shutdown mode can be used by turning off all power supplies when the device function is not needed. Cold-sparing and hotinsertion features enable these devices to be powered down without turning off the entire system. When power returns, the live-at-power-up feature enables operation of the device after reaching the voltage activation point. 32 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Using Sleep and Shutdown Modes in the System Depending on the power supply and the components used in an application, there are many ways to power on or off the power supplies connected to the device. For example, Figure 2-6 shows how a microprocessor can be used to control a power FET. Microsemi recommends that power FETs with low resistance be used to perform the switching action. Power Supply P-Channel Power FET Microprocessor VCC, VJTAG, and VPUMP Pins Power-On/Off Control Signal IGLOO, IGLOO PLUS, IGLOO nano, or ProASIC3L Devices Figure 2-6 • Controlling Power-On/-Off State Using Microprocessor and Power FET Figure 2-7 shows how a microprocessor can be used with a voltage regulator’s shutdown pin to turn on or off the power supplies connected to the device. Microprocessor Shutdown Control Signal for VCC Shutdown Control Signal for VCCI VCCI Power Pin Power Supply Voltage Regulator IGLOO, IGLOO PLUS, IGLOO nano, or ProASIC3L Device VCC , VJTAG, and VPUMP Pins Figure 2-7 • Controlling Power-On/-Off State Using Microprocessor and Voltage Regulator Power-Up/-Down Behavior By design, all IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 I/Os are in tristate mode before device power-up. The I/Os remain tristated until the last voltage supply (VCC or VCCI ) is powered to its activation level. After the last supply reaches its functional level, the outputs exit the tristate mode and drive the logic at the input of the output buffer. The behavior of user I/Os is independent of the VCC and VCCI sequence or the state of other voltage supplies of the FPGA (VPUMP and VJTAG). During power-down, device I/Os become tristated once the first power supply (VCC or VCCI) drops below its deactivation voltage level. The I/O behavior during power-down is also independent of voltage supply sequencing. Figure 2-8 on page 34 shows a timing diagram when the VCC power supply crosses the activation and deactivation trip points in a typical application when the VCC power supply ramp-rate is 100 µs (ramping from 0 V to 1.5 V in this example). This is the timing diagram for the FPGA entering and exiting Sleep mode, as this function is dependent on powering VCC down or up. Depending on the ramp-rate of the Revision 4 33 Flash*Freeze Technology and Low Power Modes power supply and board-level configurations, the user can easily calculate how long it will take for the core to become inactive or active. For more information, refer to the "Power-Up/-Down Behavior of Low Power Flash Devices" section on page 373. VCC Deactivation Trip Point Vd = 0.75 ± 0.25 V VCC = 1.5 V Activation Trip Point Va = 0.85 ± 0.25 V t = 50 μs Sleep Mode t = 56.6 μs Figure 2-8 • Entering and Exiting Sleep Mode, Typical Timing Diagram Context Save and Restore in Sleep or Shutdown Mode In Sleep mode or Shutdown mode, the contents of the SRAM, state of the I/Os, and state of the registers are lost when the device is powered off, if no other measure is taken. A low-cost external serial EEPROM can be used to save and restore the contents of the device when entering and exiting Sleep mode or Shutdown mode. In the Embedded SRAM Initialization Using External Serial EEPROM application note, detailed information and a reference design are provided for initializing the embedded SRAM using an external serial EEPROM. The user can easily customize the reference design to save and restore the FPGA state when entering and exiting Sleep mode or Shutdown mode. The microcontroller will need to manage this activity; hence, before powering down VCC , the data will be read from the FPGA and stored externally. In a similar way, after the FPGA is powered up, the microcontroller will allow the FPGA to load the data from external memory and restore its original state. Flash*Freeze Design Guide This section describes how designers can create reliable designs that use ultra-low power Flash*Freeze modes optimally. The section below provides guidance on how to select the best Flash*Freeze mode for any application. The "Design Solutions" section on page 35 gives specific recommendations on how to design and configure clocks, set/reset signals, and I/Os. This section also gives an overview of the design flow and provides details concerning Microsemi's Flash*Freeze Management IP, which enables clean clock gating and housekeeping. The "Additional Power Conservation Techniques" section on page 41 describes board-level considerations for entering and exiting Flash*Freeze mode. Selecting the Right Flash*Freeze Mode Both Flash*Freeze modes will bring an FPGA into an ultra-low power static mode that retains register and SRAM content and sets I/Os to a predetermined configuration. There are two primary differences that distinguish type 2 mode from type 1, and they must be considered when creating a design using Flash*Freeze technology. First, with type 2 mode, the device has an opportunity to wait for a second signal to enable activation of Flash*Freeze mode. This allows processes to complete prior to deactivating the device, and can be useful to control task completion, data preservation, accidental Flash*Freeze activation, system shutdown, or any other housekeeping function. The second signal may be derived from an external or into-out internal source. The second difference between type 1 and type 2 modes is that a design for type 2 mode has an opportunity to cleanly manage clocks and data activity before entering and exiting Flash*Freeze mode. This is particularly important when data preservation is needed, as it ensures valid data is stored prior to entering, and upon exiting, Flash*Freeze mode. Type 1 Flash*Freeze mode is ideally suited for applications with the following design criteria: 34 • Entering Flash*Freeze mode is not dependent on any signal other than the external FF pin. • Internal housekeeping is not required prior to entering Flash*Freeze. R e visio n 4 ProASIC3L FPGA Fabric User’s Guide • The device is reset upon exiting Flash*Freeze mode or internal state saving is not required. • State saving is required, but data and clock management is performed external to the FPGA. In other words, incoming data is externally guaranteed and held valid prior to entering Flash*Freeze mode. Type 2 Flash*Freeze mode is ideally suited for applications with the following design criteria: • Entering Flash*Freeze mode is dependent on an internal or external signal in addition to the external FF pin. • State saving is required and incoming data is not externally guaranteed valid. • The designer wants to use his/her own Flash*Freeze management IP for clock and data management. • The designer wants to use his/her own Flash*Freeze management logic for clock and data management. • Internal housekeeping is required prior to entering Flash*Freeze mode. Housekeeping activities may include loading data to SRAM, system shutdown, completion of current task, or ensuring valid Flash*Freeze pin assertion. There is no downside to type 2 mode, and Microsemi's Flash*Freeze management IP offers a very low tile count clock and data management solution. Microsemi's recommendation for most designs is to use type 2 Flash*Freeze mode with Flash*Freeze management IP. Design Solutions Clocks • Microsemi recommends using a completely synchronous design in Type 2 mode with Flash*Freeze management IP cleanly gating all internal and external clocks. This will prevent narrow pulses upon entrance and exit from Flash*Freeze mode (Figure 2-5 on page 30). • Upon entering Flash*Freeze mode, external clocks become tied off High, internal to the clock pin (unless hold state is used on IGLOO nano or IGLOO PLUS), and PLLs are turned off. Any clock that is externally Low will realize a Low to High transition internal to the device while entering Flash*Freeze. If clocks will float during Flash*Freeze mode, Microsemi recommends using the weak pull-up feature. If clocks will continue to drive the device during Flash*Freeze mode, the clock gating (filter) available in Flash*Freeze management IP can help to filter unwanted narrow clock pulses upon Flash*Freeze mode entry and exit. • Clocks may continue to drive FPGA pins while the device is in Flash*Freeze mode, with virtually no power consumption. The weak pull-up/-down configuration will result in unnecessary power consumption if used in this scenario. • Floating clocks can cause totem pole currents on the input I/O circuitry when the device is in active mode. If clocks are externally gated prior to entering Flash*Freeze mode, Microsemi recommends gating them to a known value (preferably '1', to avoid a possible narrow pulse upon Flash*Freeze mode exit), and not leaving them floating. However, during Flash*Freeze mode, all inputs and clocks are internally tied off to prevent totem pole currents, so they can be left floating. • Upon exiting Flash*Freeze mode, the design must allow maximum acquisition time for the PLL to acquire the lock signal, and for a PLL clock to become active. If a PLL output clock is used as the primary clock for Flash*Freeze management IP, it is important to note that the clock gating circuit will only release other clocks after the primary PLL output clock becomes available. Revision 4 35 Flash*Freeze Technology and Low Power Modes Set/Reset Since all I/Os and globals are tied High in Flash*Freeze mode (unless hold state is used on IGLOO nano or IGLOO PLUS), Microsemi recommends using active low set/reset at the top-level port. If needed, the signal can be inverted internally. • If the intention is to always set/reset in Flash*Freeze mode, a self set/reset circuit may be implemented to accomplish this, as shown in Figure 2-9. Configure an active High set/reset input pin so it uses the internal pull-up during Flash*Freeze mode, and drives Low during active mode. When the device exits Flash*Freeze mode, the input will transition from High to Low, releasing the set/reset. Note that this circuit may release set/reset before all outputs become active, since outputs are enabled up to 200 ns after inputs when exiting Flash*Freeze mode. Input Pull-Up '0' Set/Reset Figure 2-9 • Flash*Freeze Self-Reset Circuit I/Os • Floating inputs can cause totem pole currents on the input I/O circuitry when the device is in active mode. If inputs will be released (undriven) during Flash*Freeze mode, Microsemi recommends that they are only released after the device enters Flash*Freeze mode. • As mentioned earlier, asynchronous input to output paths are subject to possible glitching when entering Flash*Freeze mode. For example, on a direct in-to-out path, if the current state is '0' and the input bank deactivates first, the input and then the output will transition to '1' before the output enters its Flash*Freeze state. This can be prevented by using latches along with Flash*Freeze management IP to gate asynchronous in-to-out paths prior to entering Flash*Freeze mode. JTAG • The JTAG state machine is powered but not active during Flash*Freeze mode. • TCK should be held in a static state to prevent dynamic power consumption of the JTAG circuit during Flash*Freeze. • Specific JTAG pin tie-off recommendations suitable for Flash*Freeze mode can be found in the "Pin Descriptions and Packaging" chapter of the device datasheet. ULSICC • The User Low Static ICC (ULSICC) macro acts as an access point to the hard Flash*Freeze technology block in the device. The ULSICC macro represents a hard, fixed location block in the device. When the LSICC input of the ULSICC macro is driven Low, the Flash*Freeze pin is blocked, and when LSICC is driven High, the Flash*Freeze pin is enabled. • If the user decides to build his/her own Flash*Freeze type 2 clock and data management logic, note that the LSICC signal on the ULSICC macro is ANDed internally with the Flash*Freeze signal. In order to reliably enter Flash*Freeze, the LSICC signal must remain asserted High while entering and during Flash*Freeze mode. Flash*Freeze Management IP One of the key benefits of Microsemi's Flash*Freeze mode is the ability to preserve the state of all internal registers, SRAM content, and I/Os (IGLOO nano and IGLOO PLUS only). This feature enables seamless continuation of data processing before and after Flash*Freeze, without the need to reload or reinitialize the FPGA system. Microsemi's Flash*Freeze management IP, available for type 2 implementation, offers a robust RTL block that ensures clean clock gating of all system clocks before entering and upon exiting Flash*Freeze mode. This IP also gives users the option to perform housekeeping prior to entering Flash*Freeze mode. This section will provide an overview of the 36 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Flash*Freeze management IP. Additional information on this IP core can be found in the Libero online help. The Flash*Freeze management IP is comprised of three blocks: the Flash*Freeze finite state machine (FSM), the clock gating (filter) block, and the ULSICC macro, as shown in Figure 2-10. IGLOO, IGLOO PLUS, IGLOO nano, ProASIC3L, or RT ProASIC3 Device Flash*Freeze Management IP Clock Gating (filter ) CLKINT From Array To Array INBUF CLKINT Flash*Freeze Pin Connect to Top-Level Port Flash*Freeze FSM Housekeeping (optional) INBUF_FF ULSICC Macro Flash*Freeze Technology User Design LEGEND Net Logical Connection Hardwired Connection Figure 2-10 • Flash*Freeze Management IP Block Diagram Flash*Freeze Management FSM The Flash*Freeze FSM block is a simple, robust, fully encoded 3-bit state machine that ensures clean entrance to and exit from Flash*Freeze mode by controlling activities of the clock gating, ULSICC, and optional housekeeping blocks. The state diagram for the FSM is shown in Figure 2-11 on page 38. In normal operation, the state machine waits for Flash*Freeze pin assertion, and upon detection of a request, it waits for a short period of time to ensure the assertion persists; then it asserts WAIT_HOUSEKEEPING (active High) synchronous to the user’s designated system clock. This flag can be used by user logic to perform any needed shutdown processes prior to entering Flash*Freeze mode, such as storing data into SRAM, notifying other system components of the request, or timing/validating the Flash*Freeze request. The FSM also asserts Flash_Freeze_Enabled whenever the device enters Flash*Freeze mode. This occurs after all housekeeping and clock gating functions have completed. The Flash_Freeze_Enabled signal remains asserted, even during Flash*Freeze mode, until the Flash*Freeze pin is deasserted. Use the Flash_Freeze_Enabled signal to drive any logic in the design that needs to be in a particular state during Flash*Freeze mode. The DONE_HOUSEKEEPING (active High) signal should be asserted to notify the FSM when all the housekeeping tasks are completed. If the user chooses not to use housekeeping, the Flash*Freeze management IP core generator in Libero SoC will connect WAIT_HOUSEKEEPING to DONE_HOUSEKEEPING. Revision 4 37 Flash*Freeze Technology and Low Power Modes System Reset Deassert ULSICC Flash*Freeze Asserted Wait on Flash*Freeze Flash*Freeze Deasserted Flash*Freeze Deasserted Persistent Flash*Freeze House-Keeping Request to User Logic Safe to Turn Off Clock From User Logic House Keeping Request to User Logic Fail Safe Dummy State 1 Flash*Freeze Deasserted Flash*Freeze Flash*Freeze Asserted Deasserted Stop Clock to User Clock Gating Circuit Flash*Freeze Deasserted Trigger ULSICC Gated Clock to User Design Figure 2-11 • FSM State Diagram 38 Turn On User Clock Trigger ULSICC Gate User Clocks User Clock Fail Safe Dummy State 2 Flash*Freeze Deasserted Flash*Freeze Deasserted Flash*Freeze Asserted Un-Gate User Clocks R e visio n 4 ULSICC ProASIC3L FPGA Fabric User’s Guide Clock Gating Block Once DONE_HOUSEKEEPING is detected, the FSM will initiate the clock gating circuit by asserting ASSERT_GATE (active Low). ASSERT_GATE is named control_user_clock_net in the IP block. Upon assertion of the ASSERT_GATE signal, the clock will be gated in less than two cycles. The clock gating circuit is comprised of a flip-flop, latch, AND gate, and CLKINT, as shown in Figure 2-12. The clock gating block can support gating of up to 17 clocks. F*F FSM ASSERT_GATE Q D Q D Flip-Flop Latch CLK G AND System Clock CLKINT Figure 2-12 • Clock Gating Circuit After initiating the clock gating circuit, the FSM will assert and hold the LSICC signal (active High), feeding the ULSICC macro. This will initiate the 1 µs entrance into Flash*Freeze mode. Upon deassertion of the Flash*Freeze pin, the FSM will set ASSERT_GATE High. Once the I/O banks become active, the clock will enter the device and register the ASSERT_GATE signal, cleanly releasing the clock gate. Design Flow1 Microsemi has developed a convenient and intuitive design flow for configuring and integrating Flash*Freeze technology into an FPGA design. Flash*Freeze type 1 is implemented by instantiating the INBUF_FF macro in the top level of a design. Flash*Freeze type 2 with management IP can be generated by the Libero core generator or SmartGen and instantiated as a single block in the user's design. This single block will include an INBUF_FF macro and the optional Flash*Freeze management IP, which includes the ULSICC macro. If designers do not wish to use this core generator, the INBUF_FF macro and the optional ULSICC macro may be instantiated in the design, and custom Flash*Freeze management IP can be developed by the user. The remainder of this section will cover configuration details of the INBUF_FF macro, the ULSICC macro, and the Flash*Freeze management IP. Additional information on the tools discussed within this section may be found in the Libero online help. INBUF_FF The INBUF_FF macro is a special-purpose input buffer macro that is interpreted downstream in the design flow by Microsemi's Designer software. When this macro is used, the top-level port will be forced to the dedicated FF pin in the FPGA, and Flash*Freeze mode will be available for use in the device. The following are the design rules for INBUF_FF: 1. • If INBUF_FF is not used in the design, the device will not be configured to support Flash*Freeze mode. • When the INBUF_FF macro is used, the FF pin will establish a hardwired connection to the Flash*Freeze technology circuit in the device, as shown in Figure 2-1 on page 25, Figure 2-3 on page 27, and Figure 2-10 on page 37, and described in the "Flash*Freeze Type 1: Control by Dedicated Flash*Freeze Pin" section on page 24. This section applies to Libero / Designer software v8.3 and later. Microsemi recommends that designs created in earlier versions of the software be modified to accommodate this flow by instantiating the INBUF_FF macro or the Flash*Freeze management IP. Refer to the Libero / Designer software v8.3 release notes and the Libero online help for more information on migrating designs from older software versions. Revision 4 39 Flash*Freeze Technology and Low Power Modes • The INBUF_FF must be driven by a top-level input port of the design. • The INBUF_FF AND the ULSICC macro must be used to enable type 2 Flash*Freeze mode. • For type 2 Flash*Freeze mode, the INBUF_FF MUST drive some logic in the design. • For type 1 Flash*Freeze mode, the INBUF_FF may drive some logic in the design, but it may also be left floating. • Only one INBUF_FF may be instantiated in a device. • The FF pin threshold voltages are defined by VCCI and the supported single-ended I/O standard in the corresponding I/O bank. • The FF pin Schmitt trigger option may be configured in the I/O attribute editor in Microsemi's Designer software. The Schmitt trigger option is only available for IGLOOe, IGLOO nano, IGLOO PLUS, ProASIC3EL, and RT ProASIC3 devices. • A 2 ns glitch filter resides in the Flash*Freeze Technology block to filter unwanted glitches on the FF pin. ULSICC The User Low Static ICC (ULSICC) macro allows the FPGA core to access the Flash*Freeze Technology block so that entering and exiting Flash*Freeze mode can be controlled by the user's design. The ULSICC macro enables a hard block with an available LSICC input port, as shown in Figure 2-3 on page 27 and Figure 2-10 on page 37. Design rules for the ULSICC macro are as follows: • The ULSICC macro by itself cannot enable Flash*Freeze mode. The INBUF_FF AND the ULSICC macro must both be used to enable type 2 Flash*Freeze mode. • The ULSICC controls entering the Flash*Freeze mode by asserting the LSICC input (logic '1') of the ULSICC macro. The FF pin must also be asserted (logic '0') to enter Flash*Freeze mode. • When the LSICC signal is '0', the device cannot enter Flash*Freeze mode; and if already in Flash*Freeze mode, it will exit. • When the ULSICC macro is not instantiated in the user's design, the LSICC port will be tied High. Flash*Freeze Management IP The Flash*Freeze management IP can be configured with the Libero (or SmartGen) core generator in a simple, intuitive interface. With the core configuration tool, users can select the number of clocks to be gated, and select whether or not to implement housekeeping. All port names on the Flash*Freeze management IP block can be renamed by the user. • The clock gating (filter) blocks include CLKINT buffers for each gated clock output (version 8.3). • When housekeeping is NOT used, the WAIT_HOUSEKEEPING signal will be automatically fed back into DONE_HOUSEKEEPING inside the core, and the ports will not be available at the IP core interface. • The INBUF_FF macro is automatically instantiated within the IP core. • The INBUF_FF port (default name is "Flash_Freeze_N") must be connected to a top-level input port of the design. • The ULSICC macro is automatically instantiated within the IP core, and the LSICC signal is driven by the FSM. • Timing analysis can be performed on the clock domain of the source clock (i.e., input to the clock gating filters). For example, if CLKin becomes CLKin_gated, the timing can be performed on the CLKin domain in SmartTime. • The gated clocks can be added to the clock list if the user wishes to analyze these clocks specifically. The user can locate the gated clocks by looking for instance names such as those below: Top/ff1/ff_1_wrapper_inst/user_ff_1_wrapper/Primary_Filter_Instance/ Latch_For_Clock_Gating:Q Top/ff1/ff_1_wrapper_inst/user_ff_1_wrapper/genblk1.genblk2.secondary_filter[0]. seconday_filter_instance/Latch_For_Clock_Gating:Q Top/ff1/ff_1_wrapper_inst/user_ff_1_wrapper/genblk1.genblk2.secondary_filter[1]. seconday_filter_instance/Latch_For_Clock_Gating:Q 40 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide • There will be added skew and clock insertion delay due to the clock gating circuit. The user should analyze external setup/hold times carefully. The user should also ensure the additional skew across the clock gating filter circuit is accounted for in any paths where the launch register is driven from the filter input clock and captured by a register driven by the gated clock filter output clock. Power Analysis SmartPower identifies static and dynamic power consumption problems quickly within a design. It provides a hierarchical view, allowing users to drill down and estimate the power consumption of individual components or events. SmartPower analyzes power consumption for nets, gates, I/Os, memories, clocks, cores, clock domains, power supply rails, peak power during a clock cycle, and switching transitions. SmartPower generates detailed hierarchical reports of the dynamic power consumption of a design for easy inspection. These reports include design-level power summary, average switching activity, and ambient and junction temperature readings. Enter the target clock and data frequencies for a design, and let SmartPower perform a detailed and accurate power analysis. SmartPower supports importing files in the VCD (Value-Change Dump) format as specified in the IEEE 1364 standard. It also supports the Synopsys® Switching Activity Interchange Format (SAIF) standard. Support for these formats lets designers generate switching activity information in a variety of simulators and then import this information directly into SmartPower. For portable or battery-operated applications, a power profile feature enables you to measure power and battery life, based on a sequence of operational modes of the design. In most portable and batteryoperated applications, the system is seldom fully "on" 100 percent of the time. "On" is a combination of fully active, standby, sleep, or other functional modes. SmartPower allows users to create a power profile for a design by specifying operational modes and the percent of time the device will run in each of the modes. Power is calculated for each of the modes, and total power is calculated based on the weighted average of all modes. SmartPower also provides an estimated battery life based on the power profile. The current capacity for a given battery is entered and used to estimate the life of the battery. The result is an accurate and realistic indication of battery life. More information on SmartPower can be found on the Microsemi SoC Products Group website: http://www.microsemi.com/soc/products/software/libero/smartpower.aspx. Additional Power Conservation Techniques IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 FPGAs provide many ways to inherently conserve power; however, there are also several design techniques that can be used to reduce power on the board. • • • • • • • • • Microsemi recommends that the designer use the minimum number of I/O banks possible and tie any unused power supplies (such as VCCPLL, VCCI, VMV, and VPUMP) to ground. Leave unused I/O ports floating. Unused I/Os are configured by the software as follows: – Output buffer is disabled (with tristate value of high impedance) – Input buffer is disabled (with tristate value of high impedance) Use the lowest available voltage I/O standard, the lowest drive strength, and the slowest slew rate to reduce I/O switching contribution to power consumption. Advanced and pro I/O banks may consume slightly higher static current than standard and standard plus banks—avoid using advanced and pro banks whenever practical. – The small static power benefit obtained by avoiding advanced or pro I/O banks is usually negligible compared to the benefit of using a low power I/O standard. Deselect RAM blocks that are not being used. Only enable read and write ports on RAM blocks when they are needed. Gating clocks LOW offers improved static power of RAM blocks. Drive the FF port of RAM blocks with the Flash_Freeze_Enabled signal from the Flash*Freeze management IP. Drive inputs to the full voltage level so that all transistors are turned on or off completely. Revision 4 41 Flash*Freeze Technology and Low Power Modes • • Avoid using pull-ups and pull-downs on I/Os because these resistors draw some current. Avoid driving resistive loads or bipolar transistors, since these draw a continuous current, thereby adding to the static current. When partitioning the design across multiple devices, minimize I/O usage among the devices. Conclusion Microsemi IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, and RT ProASIC3 family architectures are designed to achieve ultra-low power consumption based on enhanced nonvolatile and live-at-power-up flash-based technology. Power consumption can be reduced further by using Flash*Freeze, Static (Idle), Sleep, and Shutdown power modes. All these features result in a low power, cost-effective, single-chip solution designed specifically for power-sensitive and battery-operated electronics applications. Related Documents Application Notes Embedded SRAM Initialization Using External Serial EEPROM http://www.microsemi.com/soc/documents/EmbeddedSRAMInit_AN.pdf List of Changes The following table lists critical changes that were made in each version of the chapter. Date Changes Page July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v2.3 (November 2009) The "Sleep Mode" section was revised to state the VJTAG and VPUMP, as well as VCC, are grounded during Sleep mode (SAR 22517). 32 Figure 2-6 • Controlling Power-On/-Off State Using Microprocessor and Power FET and Figure 2-7 • Controlling Power-On/-Off State Using Microprocessor and Voltage Regulator were revised to show that VJTAG and VPUMP are powered off during Sleep mode. 33 IGLOO nano devices were added as a supported family. N/A The "Prototyping for IGLOO and ProASIC3L Devices Using ProASIC3" section was removed, as these devices are now in production. N/A The "Additional Power Conservation Techniques" section was revised to add RT ProASIC3 devices. 41 The "Flash*Freeze Management FSM" section was updated with the following information: The FSM also asserts Flash_Freeze_Enabled whenever the device enters Flash*Freeze mode. This occurs after all housekeeping and clock gating functions have completed. 37 v2.2 (December 2008) v2.0 (October 2008) 42 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Date v2.1 (October 2008) Changes Page The title changed from "Flash*Freeze Technology and Low Power Modes in IGLOO, IGLOO PLUS, and ProASIC3L Devices" to Actel’s Flash*Freeze Technology and Low Power Modes." N/A The "Flash Families Support the Flash*Freeze Feature" section was updated. 22 Significant changes were made to this document to support Libero IDE v8.4 and later functionality. RT ProASIC3 device support information is new. In addition to the other major changes, the following tables and figures were updated or are new: Figure 2-3 • Flash*Freeze Mode Type 2 – Controlled by Flash*Freeze Pin and Internal Logic (LSICC signal) – updated 27 Figure 2-5 • Narrow Clock Pulses During Flash*Freeze Entrance and Exit – new Figure 2-10 • Flash*Freeze Management IP Block Diagram – new 30 Figure 2-11 • FSM State Diagram – new 37 Table 2-6 • IGLOO nano and IGLOO PLUS Flash*Freeze Mode (type 1 and type 2)—I/O Pad State – updated 38 29 Please review the entire document carefully. v1.3 (June 2008) The family description for ProASIC3L in Table 2-1 • Flash-Based FPGAs was updated to include 1.5 V. 22 v1.2 (March 2008) The part number for this document was changed from 51700094-003-1 to 51700094-004-2. N/A The title of the document was changed to "Flash*Freeze Technology and Low Power Modes in IGLOO, IGLOO PLUS, and ProASIC3L Devices." N/A The "Flash*Freeze Technology and Low Power Modes" section was updated to remove the parenthetical phrase, "from 25 µW," in the second paragraph. The following sentence was added to the third paragraph: "IGLOO PLUS has an additional feature when operating in Flash*Freeze mode, allowing it to retain I/O states as well as SRAM and register states." 21 The "Power Conservation Techniques" section was updated to add VJTAG to the parenthetical list of power supplies that should be tied to the ground plane if unused. Additional information was added regarding how the software configures unused I/Os. 2-1 Table 2-1 • Flash-Based FPGAs and the accompanying text was updated to include the IGLOO PLUS family. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 22 The "Flash*Freeze Mode" section was revised to include that I/O states are preserved in Flash*Freeze mode for IGLOO PLUS devices. The last sentence in the second paragraph was changed to, "If the FF pin is not used, it can be used as a regular I/O." The following sentence was added for Flash*Freeze mode type 2: "Exiting the mode is controlled by either the FF pin OR the user-defined LSICC signal." 24 The "Flash*Freeze Type 1: Control by Dedicated Flash*Freeze Pin" section was revised to change instructions for implementing this mode, including instructions for implementation with Libero IDE v8.3. 24 Figure 2-1 • Flash*Freeze Mode Type 1 – Controlled by the Flash*Freeze Pin was updated. 25 The "Flash*Freeze Type 2: Control by Dedicated Flash*Freeze Pin and Internal Logic" section was renamed from "Type 2 Software Implementation." 26 The "Type 2 Software Implementation for Libero IDE v8.3" section is new. 2-6 Revision 4 43 Flash*Freeze Technology and Low Power Modes Date Changes Page Figure 2-3 • Flash*Freeze Mode Type 2 – Controlled by Flash*Freeze Pin and Internal Logic (LSICC signal) was updated. 27 Figure 2-4 • Flash*Freeze Mode Type 2 – Timing Diagram was revised to show deasserting LSICC after the device has exited Flash*Freeze mode. 27 The "IGLOO nano and IGLOO PLUS I/O State in Flash*Freeze Mode" section was added to include information for IGLOO PLUS devices. Table 2-6 • IGLOO nano and IGLOO PLUS Flash*Freeze Mode (type 1 and type 2)—I/O Pad State is new. 28, 29 The "During Flash*Freeze Mode" section was revised to include a new bullet pertaining to output behavior for IGLOO PLUS. The bullet on JTAG operation was revised to provide more detail. 31 Figure 2-6 • Controlling Power-On/-Off State Using Microprocessor and Power FET and Figure 2-7 • Controlling Power-On/-Off State Using Microprocessor and Voltage Regulator were updated to include IGLOO PLUS. 33, 33 The first sentence of the "Shutdown Mode" section was updated to list the devices for which it is supported. 32 The first paragraph of the "Power-Up/-Down Behavior" section was revised. The second sentence was changed to, "The I/Os remain tristated until the last voltage supply (VCC or VCCI ) is powered to its activation level." The word "activation" replaced the word "functional." The sentence, "During power-down, device I/Os become tristated once the first power supply (VCC or VCCI) drops below its deactivation voltage level" was revised. The word "deactivation" replaced the word "brownout." 33 The "Prototyping for IGLOO and ProASIC3L Devices Using ProASIC3" section was revised to state that prototyping in ProASIC3 does not apply for the IGLOO PLUS family. 2-21 Table 2-8 • Prototyping/Migration Solutions, Table 2-9 • Device Migration—IGLOO Supported Packages in ProASIC3 Devices, and Table 2-10 • Device Migration— ProASIC3L Supported Packages in ProASIC3 Devices were updated with a table note stating that device migration is not supported for IGLOO PLUS devices. 2-21, 2-23 The text following Table 2-10 • Device Migration—ProASIC3L Supported Packages in ProASIC3 Devices was moved to a new section: the "Flash*Freeze Design Guide" section. 34 v1.1 (February 2008) Table 2-1 • Flash-Based FPGAs was updated to remove the ProASIC3, ProASIC3E, and Automotive ProASIC3 families, which were incorrectly included. 22 v1.0 (January 2008) Detailed descriptions of low power modes are described in the advanced datasheets. This application note was updated to describe how to use the features in an IGLOO/e application. N/A Figure 2-1 • Flash*Freeze Mode Type 1 – Controlled by the Flash*Freeze Pin was updated. 25 Figure 2-2 • Flash*Freeze Mode Type 1 – Timing Diagram is new. 25 Steps 4 and 5 are new in the "Flash*Freeze Type 2: Control by Dedicated Flash*Freeze Pin and Internal Logic" section. 26 v1.2 (continued) 44 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Date 51900147-2/5.07 Changes Page In the following sentence, located in the "Flash*Freeze Mode" section, the bold text was changed from active high to active Low. 24 The Flash*Freeze pin (active low) is a dedicated pin used to enter or exit Flash*Freeze mode directly, or alternatively the pin can be routed internally to the FPGA core to allow the user's logic to decide if it is safe to transition to this mode. Figure 2-2 • Flash*Freeze Mode Type 1 – Timing Diagram was updated. Information about ULSICC was added to the "Prototyping for IGLOO and ProASIC3L Devices Using ProASIC3" section. 51900147-1/3.07 In the "Flash*Freeze Mode" section, "active high" was changed to "active low." The "Prototyping for IGLOO and ProASIC3L Devices Using ProASIC3" section was updated with information concerning the Flash*Freeze pin. Revision 4 25 2-21 24 2-21 45 3 – Global Resources in Low Power Flash Devices Introduction IGLOO, Fusion, and ProASIC3 FPGA devices offer a powerful, low-delay VersaNet global network scheme and have extensive support for multiple clock domains. In addition to the Clock Conditioning Circuits (CCCs) and phase-locked loops (PLLs), there is a comprehensive global clock distribution network called a VersaNet global network. Each logical element (VersaTile) input and output port has access to these global networks. The VersaNet global networks can be used to distribute low-skew clock signals or high-fanout nets. In addition, these highly segmented VersaNet global networks contain spines (the vertical branches of the global network tree) and ribs that can reach all the VersaTiles inside their region. This allows users the flexibility to create low-skew local clock networks using spines. This document describes VersaNet global networks and discusses how to assign signals to these global networks and spines in a design flow. Details concerning low power flash device PLLs are described in the "Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs" section on page 77. This chapter describes the low power flash devices’ global architecture and uses of these global networks in designs. Global Architecture Low power flash devices offer powerful and flexible control of circuit timing through the use of global circuitry. Each chip has up to six CCCs, some with PLLs. • In IGLOOe, ProASIC3EL, and ProASIC3E devices, all CCCs have PLLs—hence, 6 PLLs per device (except the PQ208 package, which has only 2 PLLs). • In IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3, and ProASIC3L devices, the west CCC contains a PLL core (except in 10 k through 30 k devices). • In Fusion devices, the west CCC also contains a PLL core. In the two larger devices (AFS600 and AFS1500), the west and east CCCs each contain a PLL. Refer to Table 4-6 on page 100 for details. Each PLL includes delay lines, a phase shifter (0°, 90°, 180°, 270°), and clock multipliers/dividers. Each CCC has all the circuitry needed for the selection and interconnection of inputs to the VersaNet global network. The east and west CCCs each have access to three chip global lines on each side of the chip (six chip global lines total). The CCCs at the four corners each have access to three quadrant global lines in each quadrant of the chip (except in 10 k through 30 k gate devices). The nano 10 k, 15 k, and 20 k devices support four VersaNet global resources, and 30 k devices support six global resources. The 10 k through 30 k devices have simplified CCCs called CCC-GLs. The flexible use of the VersaNet global network allows the designer to address several design requirements. User applications that are clock-resource-intensive can easily route external or gated internal clocks using VersaNet global routing networks. Designers can also drastically reduce delay penalties and minimize resource usage by mapping critical, high-fanout nets to the VersaNet global network. Note: Microsemi recommends that you choose the appropriate global pin and use the appropriate global resource so you can realize these benefits. The following sections give an overview of the VersaNet global network, the structure of the global network, access point for the global networks, and the clock aggregation feature that enables a design to have very low clock skew using spines. Revision 4 47 Global Resources in Low Power Flash Devices Global Resource Support in Flash-Based Devices The flash FPGAs listed in Table 3-1 support the global resources and the functions described in this document. Table 3-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities IGLOO nano The industry’s lowest-power, smallest-size solution ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO products as listed in Table 3-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 3-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 48 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide VersaNet Global Network Distribution One of the architectural benefits of low power flash architecture is the set of powerful, low-delay VersaNet global networks that can access the VersaTiles, SRAM, and I/O tiles of the device. Each device offers a chip global network with six global lines (except for nano 10 k, 15 k, and 20 k gate devices) that are distributed from the center of the FPGA array. In addition, each device (except the 10 k through 30 k gate device) has four quadrant global networks, each consisting of three quadrant global net resources. These quadrant global networks can only drive a signal inside their own quadrant. Each VersaTile has access to nine global line resources—three quadrant and six chip-wide (main) global networks—and a total of 18 globals are available on the device (3 × 4 regional from each quadrant and 6 global). Figure 3-1 shows an overview of the VersaNet global network and device architecture for devices 60 k and above. Figure 3-2 and Figure 3-3 on page 50 show simplified VersaNet global networks. The VersaNet global networks are segmented and consist of spines, global ribs, and global multiplexers (MUXes), as shown in Figure 3-1. The global networks are driven from the global rib at the center of the die or quadrant global networks at the north or south side of the die. The global network uses the MUX trees to access the spine, and the spine uses the clock ribs to access the VersaTile. Access is available to the chip or quadrant global networks and the spines through the global MUXes. Access to the spine using the global MUXes is explained in the "Spine Architecture" section on page 57. These VersaNet global networks offer fast, low-skew routing resources for high-fanout nets, including clock signals. In addition, these highly segmented global networks offer users the flexibility to create lowskew local clock networks using spines for up to 252 internal/external clocks or other high-fanout nets in low power flash devices. Optimal usage of these low-skew networks can result in significant improvement in design performance. Quadrant Global Pads T1 T2 High-Performance Global Network T3 I/ORing Pad Ring Pad Ring Top Spine Chip (main) Global Pads Chip (main) Global Pads Spine Ribs I/O Ring Bottom Spine Scope of Spine (shaded area plus local RAMs and I/Os) Spine-Selection MUX Embedded RAM Blocks Pad Ring B1 B2 B3 Logic Tiles Note: Not applicable to 10 k through 30 k gate devices Figure 3-1 • Overview of VersaNet Global Network and Device Architecture Revision 4 49 Global Resources in Low Power Flash Devices 2 2 2 2 Chip (main) Global Network 2 2 Global Drivers Global Drivers Figure 3-2 • Simplified VersaNet Global Network (30 k gates and below) North Quadrant Global Network Quadrant Global Spine CCC 3 3 Chip Global Spine 3 Chip (main) Global 6 Network 6 3 6 6 3 CCC 6 6 3 CCC 3 CCC 3 6 3 6 3 South Quadrant Global Network Figure 3-3 • Simplified VersaNet Global Network (60 k gates and above) 50 R e visio n 4 CCC CCC ProASIC3L FPGA Fabric User’s Guide Chip and Quadrant Global I/Os The following sections give an overview of naming conventions and other related I/O information. Naming of Global I/Os In low power flash devices, the global I/Os have access to certain clock conditioning circuitry and have direct access to the global network. Additionally, the global I/Os can be used as regular I/Os, since they have identical capabilities to those of regular I/Os. Due to the comprehensive and flexible nature of the I/Os in low power flash devices, a naming scheme is used to show the details of the I/O. The global I/O uses the generic name Gmn/IOuxwByVz. Note that Gmn refers to a global input pin and IOuxwByVz refers to a regular I/O Pin, as these I/Os can be used as either global or regular I/Os. Refer to the I/O Structures chapter of the user’s guide for the device that you are using for more information on this naming convention. Figure 3-4 represents the global input pins connection. It shows all 54 global pins available to access the 18 global networks in ProASIC3E families. Quadrant Global Location A + + + GAAO/IOuxwByVz GAA1/IOuxwByVz GAA2/IOuxwByVz GABO/IOuxwByVz GAB1/IOuxwByVz GAB2/IOuxwByVz GACO/IOuxwByVz GAC1/IOuxwByVz GAC2/IOuxwByVz + + + + + Bankx Bankx + + 3 Quadrant Global Spine Bankx 3 Chip Global Location C 6 6 6 6 GCAO/IOuxwByVz GCA1/IOuxwByVz GCA2/IOuxwByVz GCBO/IOuxwByVz GCB1/IOuxwByVz GCB2/IOuxwByVz GCCO/IOuxwByVz GCC1/IOuxwByVz GCC2/IOuxwByVz + 6 3 + 3 + + 3 Bankx + 6 + + Chip Global Spine + 3 3 3 3 Bankx Bankx + 6 + 6 + Chip Global Location F GFAO/IOuxwByVz GFA1/IOuxwByVz GFA2/IOuxwByVz GFBO/IOuxwByVz GFB1/IOuxwByVz GFB2/IOuxwByVz GFCO/IOuxwByVz GFC1/IOuxwByVz GFC2/IOuxwByVz 3 + 3 3 3 3 + Bankx + Bankx + + + + + + CCC w it h PLL + CCC w it h or w it hout PLL + + Quadrant Global Location E GBAO/IOuxwByVz GBA1/IOuxwByVz GBA2/IOuxwByVz GBBO/IOuxwByVz GBB1/IOuxwByVz GBB2/IOuxwByVz GBCO/IOuxwByVz GBC1/IOuxwByVz GBC2/IOuxwByVz + + GEAO/IOuxwByVz GEAC/IOuxwByVz GEA2/IOuxwByVz GEBO/IOuxwByVz GEB1/IOuxwByVz GEB2/IOuxwByVz GECO/IOuxwByVz GEC1/IOuxwByVz GEC2/IOuxwByVz Quadrant Global Location B + CCC w it hout PLL Quadrant Global Location D GDAO/IOuxwByVz GDA1/IOuxwByVz GDA2/IOuxwByVz GDBO/IOuxwByVz GDB1/IOuxwByVz GDB2/IOuxwByVz GDCO/IOuxwByVz GDC1/IOuxwByVz GDC2/IOuxwByVz Figure 3-4 • Global Connections Details Revision 4 51 Global Resources in Low Power Flash Devices Figure 3-5 shows more detailed global input connections. It shows the global input pins connection to the northwest quadrant global networks. Each global buffer, as well as the PLL reference clock, can be driven from one of the following: • 3 dedicated single-ended I/Os using a hardwired connection • 2 dedicated differential I/Os using a hardwired connection (not supported for IGLOO nano or ProASIC3 nano devices) • The FPGA core Each shaded box represents an INBUF or INBUF_LVDS/LVPECL macro, as appropriate. To Core Sample Pin Names GAA0/IO0NDB0V0 1 GAA1/IO00PDB0V0 1 + Source for CCC (CLKA or CLKB or CLKC) GAA2/IO13PDB7V1 1 Routed Clock 2 (from FPGA core) + GAA[0:2]: GA represents global in the northwest corner of the device. A[0:2]: designates specific A clock source. Note: Differential inputs are not supported for IGLOO nano or ProASIC3 nano devices. Figure 3-5 • Global I/O Overview 52 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Figure 3-6 shows all nine global inputs for the location A connected to the top left quadrant global network via CCC. GAAO/ IOuxwByVz GAA1/ IOuxwByVz MUX CLKA MUX CLKB Quadrant Global for CLKA GAA2/ IOuxwByVz GABO/ IOuxwByVz GAB1/ IOuxwByVz CCC Quadrant Global for CLKB GAB2/ IOuxwByVz GACO/ IOuxwByVz GAC1/ IOuxwByVz MUX CLKC Quadrant Global for CLKC GAC2/ IOuxwByVz Figure 3-6 • Global Inputs Since each bank can have a different I/O standard, the user should be careful to choose the correct global I/O for the design. There are 54 global pins available to access 18 global networks. For the singleended and voltage-referenced I/O standards, you can use any of these three available I/Os to access the global network. For differential I/O standards such as LVDS and LVPECL, the I/O macro needs to be placed on (A0, A1), (B0, B1), (C0, C1), or a similar location. The unassigned global I/Os can be used as regular I/Os. Note that pin names starting with GF and GC are associated with the chip global networks, and GA, GB, GD, and GE are used for quadrant global networks. Table 3-2 on page 54 and Table 3-3 on page 55 show the general chip and quadrant global pin names. Revision 4 53 Global Resources in Low Power Flash Devices Table 3-2 • Chip Global Pin Name I/O Type Single-Ended Beginning of I/O Name Notes GFAO/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GFA1/IOuxwByVz GFA2/IOuxwByVz GFBO/IOuxwByVz GFB1/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GFB2/IOuxwByVz GFC0/IOuxwByVz GFC1/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GFC2/IOuxwByVz GCAO/IOuxwByVz GCA1/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GCA2/IOuxwByVz GCBO/IOuxwByVz GCB1/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GCB2/IOuxwByVz GCC0/IOuxwByVz GCC1/IOuxwByVz Only one of the I/Os can be directly connected to a chip global at a time. GCC2/IOuxwByVz Differential I/O Pairs GFAO/IOuxwByVz The output of the different pair will drive the chip global. GFA1/IOuxwByVz GFBO/IOuxwByVz The output of the different pair will drive the chip global. GFB1/IOuxwByVz GFCO/IOuxwByVz The output of the different pair will drive the chip global. GFC1/IOuxwByVz GCAO/IOuxwByVz The output of the different pair will drive the chip global. GCA1/IOuxwByVz GCBO/IOuxwByVz The output of the different pair will drive the chip global. GCB1/IOuxwByVz GCCO/IOuxwByVz The output of the different pair will drive the chip global. GCC1/IOuxwByVz Note: Only one of the I/Os can be directly connected to a quadrant at a time. 54 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Table 3-3 • Quadrant Global Pin Name I/O Type Single-Ended Beginning of I/O Name Notes GAAO/IOuxwByVz Only one of the I/Os can be directly connected to a quadrant global at a time GAA1/IOuxwByVz GAA2/IOuxwByVz GABO/IOuxwByVz GAB1/IOuxwByVz Only one of the I/Os can be directly connected to a quadrant global at a time. GAB2/IOuxwByVz GAC0/IOuxwByVz GAC1/IOuxwByVz Only one of the I/Os can be directly connected to a quadrant global at a time. GAC2/IOuxwByVz GBAO/IOuxwByVz GBA1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GBA2/IOuxwByVz GBBO/IOuxwByVz GBB1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GBB2/IOuxwByVz GBC0/IOuxwByVz GBC1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GBC2/IOuxwByVz GDAO/IOuxwByVz GDA1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GDA2/IOuxwByVz GDBO/IOuxwByVz GDB1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GDB2/IOuxwByVz GDC0/IOuxwByVz GDC1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GDC2/IOuxwByVz GEAO/IOuxwByVz GEA1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GEA2/IOuxwByVz GEBO/IOuxwByVz GEB1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GEB2/IOuxwByVz GEC0/IOuxwByVz GEC1/IOuxwByVz Only one of the I/Os can be directly connected to a global at a time. GEC2/IOuxwByVz Note: Only one of the I/Os can be directly connected to a quadrant at a time. Revision 4 55 Global Resources in Low Power Flash Devices Table 3-3 • Quadrant Global Pin Name (continued) Differential I/O Pairs GAAO/IOuxwByVz The output of the different pair will drive the global. GAA1/IOuxwByVz GABO/IOuxwByVz The output of the different pair will drive the global. GAB1/IOuxwByVz GACO/IOuxwByVz The output of the different pair will drive the global. GAC1/IOuxwByVz GBAO/IOuxwByVz The output of the different pair will drive the global. GBA1/IOuxwByVz GBBO/IOuxwByVz The output of the different pair will drive the global. GBB1/IOuxwByVz GBCO/IOuxwByVz The output of the different pair will drive the global. GBC1/IOuxwByVz GDAO/IOuxwByVz The output of the different pair will drive the global. GDA1/IOuxwByVz GDBO/IOuxwByVz The output of the different pair will drive the global. GDB1/IOuxwByVz GDCO/IOuxwByVz The output of the different pair will drive the global. GDC1/IOuxwByVz GEAO/IOuxwByVz The output of the different pair will drive the global. GEA1/IOuxwByVz GEBO/IOuxwByVz The output of the different pair will drive the global. GEB1/IOuxwByVz GECO/IOuxwByVz The output of the different pair will drive the global. GEC1/IOuxwByVz Note: Only one of the I/Os can be directly connected to a quadrant at a time. Unused Global I/O Configuration The unused clock inputs behave similarly to the unused Pro I/Os. The Microsemi Designer software automatically configures the unused global pins as inputs with pull-up resistors if they are not used as regular I/O. I/O Banks and Global I/O Standards In low power flash devices, any I/O or internal logic can be used to drive the global network. However, only the global macro placed at the global pins will use the hardwired connection between the I/O and global network. Global signal (signal driving a global macro) assignment to I/O banks is no different from regular I/O assignment to I/O banks with the exception that you are limited to the pin placement location available. Only global signals compatible with both the VCCI and VREF standards can be assigned to the same bank. 56 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Spine Architecture The low power flash device architecture allows the VersaNet global networks to be segmented. Each of these networks contains spines (the vertical branches of the global network tree) and ribs that can reach all the VersaTiles inside its region. The nine spines available in a vertical column reside in global networks with two separate regions of scope: the quadrant global network, which has three spines, and the chip (main) global network, which has six spines. Note that the number of quadrant globals and globals/spines per tree varies depending on the specific device. Refer to Table 3-4 for the clocking resources available for each device. The spines are the vertical branches of the global network tree, shown in Figure 3-3 on page 50. Each spine in a vertical column of a chip (main) global network is further divided into two spine segments of equal lengths: one in the top and one in the bottom half of the die (except in 10 k through 30 k gate devices). Top and bottom spine segments radiating from the center of a device have the same height. However, just as in the ProASICPLUS® family, signals assigned only to the top and bottom spine cannot access the middle two rows of the die. The spines for quadrant clock networks do not cross the middle of the die and cannot access the middle two rows of the architecture. Each spine and its associated ribs cover a certain area of the device (the "scope" of the spine; see Figure 3-3 on page 50). Each spine is accessed by the dedicated global network MUX tree architecture, which defines how a particular spine is driven—either by the signal on the global network from a CCC, for example, or by another net defined by the user. Details of the chip (main) global network spine-selection MUX are presented in Figure 3-8 on page 60. The spine drivers for each spine are located in the middle of the die. Quadrant spines can be driven from user I/Os or an internal signal from the north and south sides of the die. The ability to drive spines in the quadrant global networks can have a significant effect on system performance for high-fanout inputs to a design. Access to the top quadrant spine regions is from the top of the die, and access to the bottom quadrant spine regions is from the bottom of the die. The A3PE3000 device has 28 clock trees and each tree has nine spines; this flexible global network architecture enables users to map up to 252 different internal/external clocks in an A3PE3000 device. Table 3-4 • Globals/Spines/Rows for IGLOO and ProASIC3 Devices ProASIC3/ ProASIC3L Devices IGLOO Devices Rows Globals/ Total in Quadrant Spines Spines VersaTiles in Each Total Each Globals Clock per per Chip Trees Tree Device Tree VersaTiles Spine Globals (4×3) A3PN010 AGLN010 4 0 1 0 0 260 260 4 A3PN015 AGLN015 4 0 1 0 0 384 384 6 A3PN020 AGLN020 4 0 1 0 0 520 520 6 A3PN060 AGLN060 6 12 4 9 36 384 1,536 12 A3PN125 AGLN125 6 12 8 9 72 384 3,072 12 A3PN250 AGLN250 6 12 8 9 72 768 6,144 24 A3P015 AGL015 6 0 1 9 9 384 384 12 A3P030 AGL030 6 0 2 9 18 384 768 12 A3P060 AGL060 6 12 4 9 36 384 1,536 12 A3P125 AGL125 6 12 8 9 72 384 3,072 12 A3P250/L AGL250 6 12 8 9 72 768 6,144 24 A3P400 AGL400 6 12 12 9 108 768 9,216 24 A3P600/L AGL600 6 12 12 9 108 1,152 13,824 36 A3P1000/L AGL1000 6 12 16 9 144 1,536 24,576 48 A3PE600/L AGLE600 6 12 12 9 108 1,120 13,440 35 A3PE1500 6 12 20 9 180 1,888 37,760 59 A3PE3000/L AGLE3000 6 12 28 9 252 2,656 74,368 83 Revision 4 57 Global Resources in Low Power Flash Devices Table 3-5 • Globals/Spines/Rows for IGLOO PLUS Devices IGLOO PLUS Devices Quadrant Chip Globals Clock Globals (4×3) Trees Rows Globals/ Total in Spines Spines VersaTiles Total Each per Tree per Device in Each Tree VersaTiles Spine AGLP030 6 0 2 9 18 384* 792 12 AGLP060 6 12 4 9 36 384* 1,584 12 AGLP125 6 12 8 9 72 384* 3,120 12 Note: *Clock trees that are located at far left and far right will support more VersaTiles. Table 3-6 • Globals/Spines/Rows for Fusion Devices 58 Clock Trees Globals/ Spines per Tree Total Spines per Device VersaTiles in Each Tree Total VersaTiles Rows in Each Spine Fusion Device Chip Globals Quadrant Globals (4×3) AFS090 6 12 6 9 54 384 2,304 12 AFS250 6 12 8 9 72 768 6,144 24 AFS600 6 12 12 9 108 1,152 13,824 36 AFS1500 6 12 20 9 180 1,920 38,400 60 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Spine Access The physical location of each spine is identified by the letter T (top) or B (bottom) and an accompanying number (Tn or Bn). The number n indicates the horizontal location of the spine; 1 refers to the first spine on the left side of the die. Since there are six chip spines in each spine tree, there are up to six spines available for each combination of T (or B) and n (for example, six T1 spines). Similarly, there are three quadrant spines available for each combination of T (or B) and n (for example, four T1 spines), as shown in Figure 3-7. Tn Tn+1 Tn+2 Tn+3 Tn+4 Tn Tn+1 Tn+2 Tn+3 Tn+4 C Global Network B A A B Global Network C Figure 3-7 • Chip Global Aggregation A spine is also called a local clock network, and is accessed by the dedicated global MUX architecture. These MUXes define how a particular spine is driven. Refer to Figure 3-8 on page 60 for the global MUX architecture. The MUXes for each chip global spine are located in the middle of the die. Access to the top and bottom chip global spine is available from the middle of the die. There is no control dependency between the top and bottom spines. If a top spine, T1, of a chip global network is assigned to a net, B1 is not wasted and can be used by the global clock network. The signal assigned only to the top or bottom spine cannot access the middle two rows of the architecture. However, if a spine is using the top and bottom at the same time (T1 and B1, for instance), the previous restriction is lifted. The MUXes for each quadrant global spine are located in the north and south sides of the die. Access to the top and bottom quadrant global spines is available from the north and south sides of the die. Since the MUXes for quadrant spines are located in the north and south sides of the die, you should not try to drive T1 and B1 quadrant spines from the same signal. Revision 4 59 Global Resources in Low Power Flash Devices Using Clock Aggregation Clock aggregation allows for multi-spine clock domains to be assigned using hardwired connections, without adding any extra skew. A MUX tree, shown in Figure 3-8, provides the necessary flexibility to allow long lines, local resources, or I/Os to access domains of one, two, or four global spines. Signal access to the clock aggregation system is achieved through long-line resources in the central rib in the center of the die, and also through local resources in the north and south ribs, allowing I/Os to feed directly into the clock system. As Figure 3-9 indicates, this access system is contiguous. There is no break in the middle of the chip for the north and south I/O VersaNet access. This is different from the quadrant clocks located in these ribs, which only reach the middle of the rib. Internal/External Signals Internal/External Signals Tree Node MUX Tree Node MUX Internal/External Signal Tree Node MUX Global Rib Internal/External Signal Global Driver MUX Spine Figure 3-8 • Spine Selection MUX of Global Tree Global Spine Global Rib Global Driver and MUX Tree Node MUX I/O Access Internal Signal Access Global Signal Access Figure 3-9 • Clock Aggregation Tree Architecture 60 R e visio n 4 I/O Tiles ProASIC3L FPGA Fabric User’s Guide Clock Aggregation Architecture This clock aggregation feature allows a balanced clock tree, which improves clock skew. The physical regions for clock aggregation are defined from left to right and shift by one spine. For chip global networks, there are three types of clock aggregation available, as shown in Figure 3-10: • Long lines that can drive up to four adjacent spines (A) • Long lines that can drive up to two adjacent spines (B) • Long lines that can drive one spine (C) There are three types of clock aggregation available for the quadrant spines, as shown in Figure 3-10: • I/Os or local resources that can drive up to four adjacent spines • I/Os or local resources that can drive up to two adjacent spines • I/Os or local resources that can drive one spine As an example, A3PE600 and AFS600 devices have twelve spine locations: T1, T2, T3, T4, T5, T6, B1, B2, B3, B4, B5, and B6. Table 3-7 shows the clock aggregation you can have in A3PE600 and AFS600. A B C Tn Tn + 1 Tn + 2 Tn + 3 Tn + 4 Figure 3-10 • Four Spines Aggregation Table 3-7 • Spine Aggregation in A3PE600 or AFS600 Clock Aggregation Spine 1 spine T1, T2, T3, T4, T5, T6, B1, B2, B3, B4, B5, B6 2 spines T1:T2, T2:T3, T3:T4, T4:T5, T5:T6, B1:B2, B2:B3, B3:B4, B4:B5, B5:B6 4 spines B1:B4, B2:B5, B3:B6, T1:T4, T2:T5, T3:T6 The clock aggregation for the quadrant spines can cross over from the left to right quadrant, but not from top to bottom. The quadrant spine assignment T1:T4 is legal, but the quadrant spine assignment T1:B1 is not legal. Note that this clock aggregation is hardwired. You can always assign signals to spine T1 and B2 by instantiating a buffer, but this may add skew in the signal. Revision 4 61 Global Resources in Low Power Flash Devices Design Recommendations The following sections provide design flow recommendations for using a global network in a design. • "Global Macros and I/O Standards" • "Global Macro and Placement Selections" on page 64 • "Using Global Macros in Synplicity" on page 66 • "Global Promotion and Demotion Using PDC" on page 67 • "Spine Assignment" on page 68 • "Designer Flow for Global Assignment" on page 69 • "Simple Design Example" on page 71 • "Global Management in PLL Design" on page 73 • "Using Spines of Occupied Global Networks" on page 74 Global Macros and I/O Standards The larger low power flash devices have six chip global networks and four quadrant global networks. However, the same clock macros are used for assigning signals to chip globals and quadrant globals. Depending on the clock macro placement or assignment in the Physical Design Constraint (PDC) file or MultiView Navigator (MVN), the signal will use the chip global network or quadrant network. Table 3-8 lists the clock macros available for low power flash devices. Refer to the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide for details. Table 3-8 • Clock Macros Macro Name CLKBUF Description Symbol Input macro for Clock Network CLKBUF Y PAD CLKBUF_x Input macro for Clock Network with specific I/O standard CLKBUF_X CLKBUF_LVDS/LVPECL LVDS or LVPECL input macro PADP for Clock Network (not supported for IGLOO nano or CLKBUF_LVDS ProASIC3 nano devices) PADN CLKINT Y PAD PADP Y CLKBUF_LVPECL Y PADN Macro for internal clock interface A Y CLKINT CLKBIBUF Bidirectional macro with input dedicated to routed Clock Network D Y E PAD CLKBIBUF Use these available macros to assign a signal to the global network. In addition to these global macros, PLL and CLKDLY macros can also drive the global networks. Use I/O–standard–specific clock macros (CLKBUF_x) to instantiate a specific I/O standard for the global signals. Table 3-9 on page 63 shows the list of these I/O–standard–specific macros. Note that if you use these I/O–standard–specific clock macros, you cannot change the I/O standard later in the design stage. If you use the regular CLKBUF macro, you can use MVN or the PDC file in Designer to change the I/O standard. The default I/O 62 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide standard for CLKBUF is LVTTL in the current Microsemi Libero® System-on-Chip (SoC) and Designer software. Table 3-9 • I/O Standards within CLKBUF Name Description CLKBUF_LVCMOS5 LVCMOS clock buffer with 5.0 V CMOS voltage level CLKBUF_LVCMOS33 LVCMOS clock buffer with 3.3 V CMOS voltage level CLKBUF_LVCMOS25 LVCMOS clock buffer with 2.5 V CMOS voltage level1 CLKBUF_LVCMOS18 LVCMOS clock buffer with 1.8 V CMOS voltage level CLKBUF_LVCMOS15 LVCMOS clock buffer with 1.5 V CMOS voltage level CLKBUF_LVCMOS12 LVCMOS clock buffer with 1.2 V CMOS voltage level CLKBUF_PCI PCI clock buffer CLKBUF_PCIX PCIX clock buffer CLKBUF_GTL25 GTL clock buffer with 2.5 V CMOS voltage level1 CLKBUF_GTL33 GTL clock buffer with 3.3 V CMOS voltage level1 CLKBUF_GTLP25 GTL+ clock buffer with 2.5 V CMOS voltage level1 CLKBUF_GTLP33 GTL+ clock buffer with 3.3 V CMOS voltage level1 CLKBUF_ HSTL _I HSTL Class I clock buffer1 CLKBUF_ HSTL _II HSTL Class II clock buffer1 CLKBUF_SSTL2_I SSTL2 Class I clock buffer1 CLKBUF_SSTL2_II SSTL2 Class II clock buffer1 CLKBUF_SSTL3_I SSTL3 Class I clock buffer1 CLKBUF_SSTL3_II SSTL3 Class II clock buffer1 Notes: 1. Supported in only the IGLOOe, ProASIC3E, AFS600, and AFS1500 devices 2. By default, the CLKBUF macro uses the 3.3 V LVTTL I/O technology. The current synthesis tool libraries only infer the CLKBUF or CLKINT macros in the netlist. All other global macros must be instantiated manually into your HDL code. The following is an example of CLKBUF_LVCMOS25 global macro instantiations that you can copy and paste into your code: VHDL component clkbuf_lvcmos25 port (pad : in std_logic; y : out std_logic); end component begin -- concurrent statements u2 : clkbuf_lvcmos25 port map (pad => end ext_clk, y => int_clk); Verilog module design (______); input _____; output ______; clkbuf_lvcmos25 u2 (.y(int_clk), .pad(ext_clk); endmodule Revision 4 63 Global Resources in Low Power Flash Devices Global Macro and Placement Selections Low power flash devices provide the flexibility of choosing one of the three global input pad locations available to connect to a global / quadrant global network. For 60K gate devices and above, if the single-ended I/O standard is chosen, there is flexibility to choose one of the global input pads (the first, second, and fourth input). Once chosen, the other I/O locations are used as regular I/Os. If the differential I/O standard is chosen, the first and second inputs are considered as paired, and the third input is paired with a regular I/O. The user then has the choice of selecting one of the two sets to be used as the global input source. There is also the option to allow an internal clock signal to feed the global network. A multiplexer tree selects the appropriate global input for routing to the desired location. Note that the global I/O pads do not need to feed the global network; they can also be used as regular I/O pads. Hardwired I/O Clock Source Hardwired I/O refers to global input pins that are hardwired to the multiplexer tree, which directly accesses the global network. These global input pins have designated pin locations and are indicated with the I/O naming convention Gmn (m refers to any one of the positions where the global buffers is available, and n refers to any one of the three global input MUXes and the pin number of the associated global location, m). Choosing this option provides the benefit of directly connecting to the global buffers, which provides less delay. See Figure 3-11 for an example illustration of the connections, shown in red. If a CLKBUF macro is initiated, the clock input can be placed at one of nine dedicated global input pin locations: GmA0, GmA1, GmA2, GmB0, GmB1, GmB2, GmC0, GmC1, or GmC2. Note that the placement of the global will determine whether you are using chip global or quadrant global. For example, if the CLKBIF is placed in one of the GF pin locations, it will use the chip global network; if the CLKBIF is placed in one of the GA pin locations, it will use quadrant global network. This is shown in Figure 3-12 on page 65 and Figure 3-13 on page 65. To Core GFA0 GFA1 + To global network GFA2 + From FPGA core Figure 3-11 • CLKBUF Macro 64 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Bankx Bankx Location B Bankx Bankx Location C Chip Global Region Bankx Location F Bankx CLKBUF placed at one of the GF pin locations Location A Bankx Location E Bankx Location D Figure 3-12 • Chip Global Region CLKBUF placed at one of the GA pin locations Bankx Bankx Location B Bankx Bankx Location A Quadrant Global Region Location E Bankx Location C Bankx Location F Bankx Bankx Location D Figure 3-13 • Quadrant Global Region Revision 4 65 Global Resources in Low Power Flash Devices External I/O or Local signal as Clock Source External I/O refers to regular I/O pins are labeled with the I/O convention IOuxwByVz. You can allow the external I/O or internal signal to access the global. To allow the external I/O or internal signal to access the global network, you need to instantiate the CLKINT macro. Refer to Figure 3-4 on page 51 for an example illustration of the connections. Instead of using CLKINT, you can also use PDC to promote signals from external I/O or internal signal to the global network. However, it may cause layout issues because of synthesis logic replication. Refer to the "Global Promotion and Demotion Using PDC" section on page 67 for details. CLKINT INBUF To Core GFA0 GFA1 + To global network GFA2 + From FPGA core INBUF Figure 3-14 • CLKINT Macro Using Global Macros in Synplicity The Synplify® synthesis tool automatically inserts global buffers for nets with high fanout during synthesis. By default, Synplicity® puts six global macros (CLKBUF or CLKINT) in the netlist, including any global instantiation or PLL macro. Synplify always honors your global macro instantiation. If you have a PLL (only primary output is used) in the design, Synplify adds five more global buffers in the netlist. Synplify uses the following global counting rule to add global macros in the netlist: 1. CLKBUF: 1 global buffer 2. CLKINT: 1 global buffer 3. CLKDLY: 1 global buffer 4. PLL: 1 to 3 global buffers 66 – GLA, GLB, GLC, YB, and YC are counted as 1 buffer. – GLB or YB is used or both are counted as 1 buffer. – GLC or YC is used or both are counted as 1 buffer. R e visio n 4 ProASIC3L FPGA Fabric User’s Guide CLKA EXTFB POWERDOWN OADIVRST GLA LOCK GLB YB GLC YC Note: OAVDIVRST exists only in the Fusion PLL. Figure 3-15 • PLLs in Low Power Flash Devices You can use the syn_global_buffers attribute in Synplify to specify a maximum number of global macros to be inserted in the netlist. This can also be used to restrict the number of global buffers inserted. In the Synplicity 8.1 version or newer, a new attribute, syn_global_minfanout, has been added for low power flash devices. This enables you to promote only the high-fanout signal to global. However, be aware that you can only have six signals assigned to chip global networks, and the rest of the global signals should be assigned to quadrant global networks. So, if the netlist has 18 global macros, the remaining 12 global macros should have fanout that allows the instances driven by these globals to be placed inside a quadrant. Global Promotion and Demotion Using PDC The HDL source file or schematic is the preferred place for defining which signals should be assigned to a clock network using clock macro instantiation. This method is preferred because it is guaranteed to be honored by the synthesis tools and Designer software and stop any replication on this net by the synthesis tool. Note that a signal with fanout may have logic replication if it is not promoted to global during synthesis. In that case, the user cannot promote that signal to global using PDC. See Synplicity Help for details on using this attribute. To help you with global management, Designer allows you to promote a signal to a global network or demote a global macro to a regular macro from the user netlist using the compile options and/or PDC commands. The following are the PDC constraints you can use to promote a signal to a global network: 1. PDC syntax to promote a regular net to a chip global clock: assign_global_clock –net netname The following will happen during promotion of a regular signal to a global network: – If the net is external, the net will be driven by a CLKINT inserted automatically by Compile. – The I/O macro will not be changed to CLKBUF macros. – If the net is an internal net, the net will be driven by a CLKINT inserted automatically by Compile. 2. PDC syntax to promote a net to a quadrant clock: assign_local_clock –net netname –type quadrant UR|UL|LR|LL This follows the same rule as the chip global clock network. The following PDC command demotes the clock nets to regular nets. unassign_global_clock -net netname Revision 4 67 Global Resources in Low Power Flash Devices The following will happen during demotion of a global signal to regular nets: • CLKBUF_x becomes INBUF_x; CLKINT is removed from the netlist. • The essential global macro, such as the output of the Clock Conditioning Circuit, cannot be demoted. • No automatic buffering will happen. Since no automatic buffering happens when a signal is demoted, this net may have a high delay due to large fanout. This may have a negative effect on the quality of the results. Microsemi recommends that the automatic global demotion only be used on small-fanout nets. Use clock networks for high-fanout nets to improve timing and routability. Spine Assignment The low power flash device architecture allows the global networks to be segmented and used as clock spines. These spines, also called local clock networks, enable the use of PDC or MVN to assign a signal to a spine. PDC syntax to promote a net to a spine/local clock: assign_local_clock –net netname –type [quadrant|chip] Tn|Bn|Tn:Bm If the net is driven by a clock macro, Designer automatically demotes the clock net to a regular net before it is assigned to a spine. Nets driven by a PLL or CLKDLY macro cannot be assigned to a local clock. When assigning a signal to a spine or quadrant global network using PDC (pre-compile), the Designer software will legalize the shared instances. The number of shared instances to be legalized can be controlled by compile options. If these networks are created in MVN (only quadrant globals can be created), no legalization is done (as it is post-compile). Designer does not do legalization between nonclock nets. As an example, consider two nets, net_clk and net_reset, driving the same flip-flop. The following PDC constraints are used: assign_local_clock –net net_clk –type chip T3 assign_local_clock –net net_reset –type chip T1:T2 During Compile, Designer adds a buffer in the reset net and places it in the T1 or T2 region, and places the flip-flop in the T3 spine region (Figure 3-16). After Compile Before Compile D D net_clk CLK CLK net_clk CLR CLR net_reset net_reset T1 T2 Added buffer T3 assign_local_clock -net net_clk -type chip T3 assign_local_clock -net net_reset -type chip T1:T2 Figure 3-16 • Adding a Buffer for Shared Instances 68 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide You can control the maximum number of shared instances allowed for the legalization to take place using the Compile Option dialog box shown in Figure 3-17. Refer to Libero SoC / Designer online help for details on the Compile Option dialog box. A large number of shared instances most likely indicates a floorplanning problem that you should address. Figure 3-17 • Shared Instances in the Compile Option Dialog Box Designer Flow for Global Assignment To achieve the desired result, pay special attention to global management during synthesis and placeand-route. The current Synplify tool does not insert more than six global buffers in the netlist by default. Thus, the default flow will not assign any signal to the quadrant global network. However, you can use attributes in Synplify and increase the default global macro assignment in the netlist. Designer v6.2 supports automatic quadrant global assignment, which was not available in Designer v6.1. Layout will make the choice to assign the correct signals to global. However, you can also utilize PDC and perform manual global assignment to overwrite any automatic assignment. The following step-by-step suggestions guide you in the layout of your design and help you improve timing in Designer: 1. Run Compile and check the Compile report. The Compile report has global information in the "Device Utilization" section that describes the number of chip and quadrant signals in the design. A "Net Report" section describes chip global nets, quadrant global nets, local clock nets, a list of nets listed by fanout, and net candidates for local clock assignment. Review this information. Note that YB or YC are counted as global only when they are used in isolation; if you use YB only and not GLB, this net is not shown in the global/quadrant nets report. Instead, it appears in the Global Utilization report. 2. If some signals have a very high fanout and are candidates for global promotion, promote those signals to global using the compile options or PDC commands. Figure 3-18 on page 70 shows the Globals Management section of the compile options. Select Promote regular nets whose fanout is greater than and enter a reasonable value for fanouts. Revision 4 69 Global Resources in Low Power Flash Devices Figure 3-18 • Globals Management GUI in Designer 3. Occasionally, the synthesis tool assigns a global macro to clock nets, even though the fanout is significantly less than other asynchronous signals. Select Demote global nets whose fanout is less than and enter a reasonable value for fanouts. This frees up some global networks from the signals that have very low fanouts. This can also be done using PDC. 4. Use a local clock network for the signals that do not need to go to the whole chip but should have low skew. This local clock network assignment can only be done using PDC. 5. Assign the I/O buffer using MVN if you have fixed I/O assignment. As shown in Figure 3-10 on page 61, there are three sets of global pins that have a hardwired connection to each global network. Do not try to put multiple CLKBUF macros in these three sets of global pins. For example, do not assign two CLKBUFs to GAA0x and GAA2x pins. 6. You must click Commit at the end of MVN assignment. This runs the pre-layout checker and checks the validity of global assignment. 7. Always run Compile with the Keep existing physical constraints option on. This uses the quadrant clock network assignment in the MVN assignment and checks if you have the desired signals on the global networks. 8. Run Layout and check the timing. 70 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Simple Design Example Consider a design consisting of six building blocks (shift registers) and targeted for an A3PE600-PQ208 (Figure 3-16 on page 68). The example design consists of two PLLs (PLL1 has GLA only; PLL2 has both GLA and GLB), a global reset (ACLR), an enable (EN_ALL), and three external clock domains (QCLK1, QCLK2, and QCLK3) driving the different blocks of the design. Note that the PQ208 package only has two PLLs (which access the chip global network). Because of fanout, the global reset and enable signals need to be assigned to the chip global resources. There is only one free chip global for the remaining global (QCLK1, QCLK2, QCLK3). Place two of these signals on the quadrant global resource. The design example demonstrates manually assignment of QCLK1 and QCLK2 to the quadrant global using the PDC command. PDOWN PLLZ_CLKA PLL2 POWER-DOWN LOCK GLA CLKA GLB \$116 reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock REG_PLLCLK2GLA reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock DATA_QCLK1 QCLK1 REG_PLLCLK2GLA_OUT REG_QCLK1_OUT REG_QCLK1 DATA_PLLCQCLK2 EN_ALL DATA_QCLK2 ACLR QCLK2 reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock REG_QCLK2 REG_QCLK2_OUT reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock REG_PLLCLK2GLB_OUT REG_PLLCLK2GLB reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock DATA_QCLK3 QCLK3 REG_QCLK3_OUT REG_QCLK3 DATA_PLLCLK1 PLL1_CLKA PLL1 POWER-DOWN LOCK GLA CLKA \$115 reg256_behave Shhl_In Shhl_In Shhl_out Adr Clock REG_PLLCLK1_OUT REG_PLLCLK1 Figure 3-19 • Block Diagram of the Global Management Example Design Revision 4 71 Global Resources in Low Power Flash Devices Step 1 Run Synthesis with default options. The Synplicity log shows the following device utilization: Cell usage: cell count area count*area DFN1E1C1 1536 2.0 3072.0 BUFF 278 1.0 278.0 INBUF 10 0.0 0.0 VCC 9 0.0 0.0 GND 9 0.0 0.0 OUTBUF 6 0.0 0.0 CLKBUF 3 0.0 0.0 PLL 2 0.0 0.0 TOTAL 3350.0 1853 Step 2 Run Compile with the Promote regular nets whose fanout is greater than option selected in Designer; you will see the following in the Compile report: Device utilization report: ========================== CORE Used: 1536 Total: 13824 (11.11%) IO (W/ clocks) Used: 19 Total: 147 (12.93%) Differential IO Used: 0 Total: 65 (0.00%) GLOBAL Used: 8 Total: 18 (44.44%) PLL Used: 2 Total: 2 (100.00%) RAM/FIFO Used: 0 Total: 24 (0.00%) FlashROM Used: 0 Total: 1 (0.00%) …………………… The following nets have been assigned to a global resource: Fanout Type Name -------------------------1536 INT_NET Net : EN_ALL_c Driver: EN_ALL_pad_CLKINT Source: AUTO PROMOTED 1536 SET/RESET_NET Net : ACLR_c Driver: ACLR_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : QCLK1_c Driver: QCLK1_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : QCLK2_c Driver: QCLK2_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : QCLK3_c Driver: QCLK3_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : $1N14 Driver: $1I5/Core Source: ESSENTIAL 256 CLK_NET Net : $1N12 Driver: $1I6/Core Source: ESSENTIAL 256 CLK_NET Net : $1N10 Driver: $1I6/Core Source: ESSENTIAL Designer will promote five more signals to global due to high fanout. There are eight signals assigned to global networks. 72 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide During Layout, Designer will assign two of the signals to quadrant global locations. Step 3 (optional) You can also assign the QCLK1_c and QCLK2_c nets to quadrant regions using the following PDC commands: assign_local_clock –net QCLK1_c assign_local_clock –net QCLK2_c –type quadrant UL –type quadrant LL Step 4 Import this PDC with the netlist and run Compile again. You will see the following in the Compile report: The following nets have been assigned to a global resource: Fanout Type Name -------------------------1536 INT_NET Net : EN_ALL_c Driver: EN_ALL_pad_CLKINT Source: AUTO PROMOTED 1536 SET/RESET_NET Net : ACLR_c Driver: ACLR_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : QCLK3_c Driver: QCLK3_pad_CLKINT Source: AUTO PROMOTED 256 CLK_NET Net : $1N14 Driver: $1I5/Core Source: ESSENTIAL 256 CLK_NET Net : $1N12 Driver: $1I6/Core Source: ESSENTIAL 256 CLK_NET Net : $1N10 Driver: $1I6/Core Source: ESSENTIAL The following nets have been assigned to a quadrant clock resource using PDC: Fanout Type Name -------------------------256 CLK_NET Net : QCLK1_c Driver: QCLK1_pad_CLKINT Region: quadrant_UL 256 CLK_NET Net : QCLK2_c Driver: QCLK2_pad_CLKINT Region: quadrant_LL Step 5 Run Layout. Global Management in PLL Design This section describes the legal global network connections to PLLs in the low power flash devices. For detailed information on using PLLs, refer to "Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs" section on page 77. Microsemi recommends that you use the dedicated global pins to directly drive the reference clock input of the associated PLL for reduced propagation delays and clock distortion. However, low power flash devices offer the flexibility to connect other signals to reference clock inputs. Each PLL is associated with three global networks (Figure 3-5 on page 52). There are some limitations, such as when trying to use the global and PLL at the same time: • If you use a PLL with only primary output, you can still use the remaining two free global networks. • If you use three globals associated with a PLL location, you cannot use the PLL on that location. • If the YB or YC output is used standalone, it will occupy one global, even though this signal does not go to the global network. Revision 4 73 Global Resources in Low Power Flash Devices Using Spines of Occupied Global Networks When a signal is assigned to a global network, the flash switches are programmed to set the MUX select lines (explained in the "Clock Aggregation Architecture" section on page 61) to drive the spines of that network with the global net. However, if the global net is restricted from reaching into the scope of a spine, the MUX drivers of that spine are available for other high-fanout or critical signals (Figure 3-20). For example, if you want to limit the CLK1_c signal to the left half of the chip and want to use the right side of the same global network for CLK2_c, you can add the following PDC commands: define_region -name region1 -type inclusive 0 0 34 29 assign_net_macros region1 CLK1_c assign_local_clock –net CLK2_c –type chip B2 Figure 3-20 • Design Example Using Spines of Occupied Global Networks Conclusion IGLOO, Fusion, and ProASIC3 devices contain 18 global networks: 6 chip global networks and 12 quadrant global networks. These global networks can be segmented into local low-skew networks called spines. The spines provide low-skew networks for the high-fanout signals of a design. These allow you up to 252 different internal/external clocks in an A3PE3000 device. This document describes the architecture for the global network, plus guidelines and methodologies in assigning signals to globals and spines. Related Documents User’s Guides IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide http://www.microsemi.com/soc/documents/pa3_libguide_ug.pdf 74 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide List of Changes The following table lists critical changes that were made in each revision of the chapter. Date July 2010 v1.4 (December 2008) v1.3 (October 2008) v1.2 (June 2008) Changes Page This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A Notes were added where appropriate to point out that IGLOO nano and ProASIC3 nano devices do not support differential inputs (SAR 21449). N/A The "Global Architecture" section and "VersaNet Global Network Distribution" section were revised for clarity (SARs 20646, 24779). 47, 49 The "I/O Banks and Global I/Os" section was moved earlier in the document, renamed to "Chip and Quadrant Global I/Os", and revised for clarity. Figure 3-4 • Global Connections Details, Figure 3-6 • Global Inputs, Table 3-2 • Chip Global Pin Name, and Table 3-3 • Quadrant Global Pin Name are new (SARs 20646, 24779). 51 The "Clock Aggregation Architecture" section was revised (SARs 20646, 24779). 57 Figure 3-7 • Chip Global Aggregation was revised (SARs 20646, 24779). 59 The "Global Macro and Placement Selections" section is new (SARs 20646, 24779). 64 The "Global Architecture" section was updated to include 10 k devices, and to include information about VersaNet global support for IGLOO nano devices. 47 The Table 3-1 • Flash-Based FPGAs was updated to include IGLOO nano and ProASIC3 nano devices. 48 The "VersaNet Global Network Distribution" section was updated to include 10 k devices and to note an exception in global lines for nano devices. 49 Figure 3-2 • Simplified VersaNet Global Network (30 k gates and below) is new. 50 The "Spine Architecture" section was updated to clarify support for 10 k and nano devices. 57 Table 3-4 • Globals/Spines/Rows for IGLOO and ProASIC3 Devices was updated to include IGLOO nano and ProASIC3 nano devices. 57 The figure in the CLKBUF_LVDS/LVPECL row of Table 3-8 • Clock Macros was updated to change CLKBIBUF to CLKBUF. 62 A third bullet was added to the beginning of the "Global Architecture" section: In Fusion devices, the west CCC also contains a PLL core. In the two larger devices (AFS600 and AFS1500), the west and east CCCs each contain a PLL. 47 The "Global Resource Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 48 Table 3-4 • Globals/Spines/Rows for IGLOO and ProASIC3 Devices was updated to include A3PE600/L in the device column. 57 Table note 1 was revised in Table 3-9 • I/O Standards within CLKBUF to include AFS600 and AFS1500. 63 The following changes were made to the family descriptions in Table 3-1 • FlashBased FPGAs: 48 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. Revision 4 75 Global Resources in Low Power Flash Devices Date Changes Page v1.1 (March 2008) The "Global Architecture" section was updated to include the IGLOO PLUS family. The bullet was revised to include that the west CCC does not contain a PLL core in 15 k and 30 k devices. Instances of "A3P030 and AGL030 devices" were replaced with "15 k and 30 k gate devices." 47 v1.1 (continued) Table 3-1 • Flash-Based FPGAs and the accompanying text was updated to include the IGLOO PLUS family. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 48 The "VersaNet Global Network Distribution" section, "Spine Architecture" section, the note in Figure 3-1 • Overview of VersaNet Global Network and Device Architecture, and the note in Figure 3-3 • Simplified VersaNet Global Network (60 k gates and above) were updated to include mention of 15 k gate devices. 49, 50 Table 3-4 • Globals/Spines/Rows for IGLOO and ProASIC3 Devices was updated to add the A3P015 device, and to revise the values for clock trees, globals/spines per tree, and globals/spines per device for the A3P030 and AGL030 devices. 57 Table 3-5 • Globals/Spines/Rows for IGLOO PLUS Devices is new. 58 CLKBUF_LVCMOS12 was added to Table 3-9 • I/O Standards within CLKBUF. 63 The "User’s Guides" section was updated to include the three different I/O Structures chapters for ProASIC3 and IGLOO device families. 74 Figure 3-3 • Simplified VersaNet Global Network (60 k gates and above) was updated. 50 The "Naming of Global I/Os" section was updated. 51 The "Using Global Macros in Synplicity" section was updated. 66 The "Global Promotion and Demotion Using PDC" section was updated. 67 The "Designer Flow for Global Assignment" section was updated. 69 The "Simple Design Example" section was updated. 71 Table 3-4 • Globals/Spines/Rows for IGLOO and ProASIC3 Devices was updated. 57 v1.0 (January 2008) 51900087-0/1.05 (January 2005) 76 R e visio n 4 4 – Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Introduction This document outlines the following device information: Clock Conditioning Circuit (CCC) features, PLL core specifications, functional descriptions, software configuration information, detailed usage information, recommended board-level considerations, and other considerations concerning clock conditioning circuits and global networks in low power flash devices or mixed signal FPGAs. Overview of Clock Conditioning Circuitry In Fusion, IGLOO, and ProASIC3 devices, the CCCs are used to implement frequency division, frequency multiplication, phase shifting, and delay operations. The CCCs are available in six chip locations—each of the four chip corners and the middle of the east and west chip sides. For devicespecific variations, refer to the "Device-Specific Layout" section on page 94. The CCC is composed of the following: • PLL core • 3 phase selectors • 6 programmable delays and 1 fixed delay that advances/delays phase • 5 programmable frequency dividers that provide frequency multiplication/division (not shown in Figure 4-6 on page 87 because they are automatically configured based on the user's required frequencies) • 1 dynamic shift register that provides CCC dynamic reconfiguration capability Figure 4-1 provides a simplified block diagram of the physical implementation of the building blocks in each of the CCCs. 3 Global I/Os 3 Global I/Os 3 Global I/Os To Core Multiplexer Tree CLKA To Global Network A From Core To Core CLKB From Core To Core CCC Function Block (with or without PLL) CLKC To Global Network B To Global Network C From Core Multiple Signals Single Signals Figure 4-1 • Overview of the CCCs Offered in Fusion, IGLOO, and ProASIC3 Revision 4 77 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Each CCC can implement up to three independent global buffers (with or without programmable delay) or a PLL function (programmable frequency division/multiplication, phase shift, and delays) with up to three global outputs. Unused global outputs of a PLL can be used to implement independent global buffers, up to a maximum of three global outputs for a given CCC. CCC Programming The CCC block is fully configurable, either via flash configuration bits set in the programming bitstream or through an asynchronous interface. This asynchronous dedicated shift register interface is dynamically accessible from inside the low power flash devices to permit parameter changes, such as PLL divide ratios and delays, during device operation. To increase the versatility and flexibility of the clock conditioning system, the CCC configuration is determined either by the user during the design process, with configuration data being stored in flash memory as part of the device programming procedure, or by writing data into a dedicated shift register during normal device operation. This latter mode allows the user to dynamically reconfigure the CCC without the need for core programming. The shift register is accessed through a simple serial interface. Refer to the "UJTAG Applications in Microsemi’s Low Power Flash Devices" section on page 363 or the application note Using Global Resources in Actel Fusion Devices. Global Resources Low power flash and mixed signal devices provide three global routing networks (GLA, GLB, and GLC) for each of the CCC locations. There are potentially many I/O locations; each global I/O location can be chosen from only one of three possibilities. This is controlled by the multiplexer tree circuitry in each global network. Once the I/O location is selected, the user has the option to utilize the CCCs before the signals are connected to the global networks. The CCC in each location (up to six) has the same structure, so generating the CCC macros is always done with an identical software GUI. The CCCs in the corner locations drive the quadrant global networks, and the CCCs in the middle of the east and west chip sides drive the chip global networks. The quadrant global networks span only a quarter of the device, while the chip global networks span the entire device. For more details on global resources offered in low power flash devices, refer to the "Global Resources in Low Power Flash Devices" section on page 47. A global buffer can be placed in any of the three global locations (CLKA-GLA, CLKB-GLB, or CLKC-GLC) of a given CCC. A PLL macro uses the CLKA CCC input to drive its reference clock. It uses the GLA and, optionally, the GLB and GLC global outputs to drive the global networks. A PLL macro can also drive the YB and YC regular core outputs. The GLB (or GLC) global output cannot be reused if the YB (or YC) output is used. Refer to the "PLL Macro Signal Descriptions" section on page 84 for more information. Each global buffer, as well as the PLL reference clock, can be driven from one of the following: 78 • 3 dedicated single-ended I/Os using a hardwired connection • 2 dedicated differential I/Os using a hardwired connection (not supported for IGLOO nano or ProASIC3 nano devices) • The FPGA core R e visio n 4 ProASIC3L FPGA Fabric User’s Guide CCC Support in Microsemi’s Flash Devices The flash FPGAs listed in Table 4-1 support the CCC feature and the functions described in this document. Table 4-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities IGLOO nano The industry’s lowest-power, smallest-size solution ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 4-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 4-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. Revision 4 79 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Global Buffers with No Programmable Delays Access to the global / quadrant global networks can be configured directly from the global I/O buffer, bypassing the CCC functional block (as indicated by the dotted lines in Figure 4-1 on page 77). Internal signals driven by the FPGA core can use the global / quadrant global networks by connecting via the routed clock input of the multiplexer tree. There are many specific CLKBUF macros supporting the wide variety of single-ended I/O inputs (CLKBUF) and differential I/O standards (CLKBUF_LVDS/LVPECL) in the low power flash families. They are used when connecting global I/Os directly to the global/quadrant networks. Note: IGLOO nano and ProASIC nano devices do not support differential inputs. When an internal signal needs to be connected to the global/quadrant network, the CLKINT macro is used to connect the signal to the routed clock input of the network's MUX tree. To utilize direct connection from global I/Os or from internal signals to the global/quadrant networks, CLKBUF, CLKBUF_LVPECL/LVDS, and CLKINT macros are used (Figure 4-2). • The CLKBUF and CLKBUF_LVPECL/LVDS1 macros are composite macros that include an I/O macro driving a global buffer, which uses a hardwired connection. • The CLKBUF, CLKBUF_LVPECL/LVDS1 and CLKINT macros are pass-through clock sources and do not use the PLL or provide any programmable delay functionality. • The CLKINT macro provides a global buffer function driven internally by the FPGA core. The available CLKBUF macros are described in the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide. Clock Source CLKBUF_LVDS/LVPECL Macro PADN Y PADP CLKBIBUF Macro D Y CLKINT Macro A Clock Conditioning Y E None Output GLA, GLB, or GLC PAD CLKBIBUF CLKBUF Macro PAD Y Note: IGLOO nano and ProASIC nano devices do not support differential inputs. Figure 4-2 • CCC Options: Global Buffers with No Programmable Delay Global Buffer with Programmable Delay Clocks requiring clock adjustments can utilize the programmable delay cores before connecting to the global / quadrant global networks. A maximum of 18 CCC global buffers can be instantiated in a device— three per CCC and up to six CCCs per device. Each CCC functional block contains a programmable delay element for each of the global networks (up to three), and users can utilize these features by using the corresponding macro (Figure 4-3 on page 81). 1. 80 B-LVDS and M-LVDS are supported with the LVDS macro. R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Clock Source Clock Conditioning Output Input LVDS/LVPECL Macro PADN GLA CLK Y GL PADP or GLB or DLYGL[4:0] INBUF* Macro Y PAD GLC Notes: 1. For INBUF* driving a PLL macro or CLKDLY macro, the I/O will be hard-routed to the CCC; i.e., will be placed by software to a dedicated Global I/O. 2. IGLOO nano and ProASIC3 nano devices do not support differential inputs. Figure 4-3 • CCC Options: Global Buffers with Programmable Delay The CLKDLY macro is a pass-through clock source that does not use the PLL, but provides the ability to delay the clock input using a programmable delay. The CLKDLY macro takes the selected clock input and adds a user-defined delay element. This macro generates an output clock phase shift from the input clock. The CLKDLY macro can be driven by an INBUF* macro to create a composite macro, where the I/O macro drives the global buffer (with programmable delay) using a hardwired connection. In this case, the software will automatically place the dedicated global I/O in the appropriate locations. Many specific INBUF macros support the wide variety of single-ended and differential I/O standards supported by the low power flash family. The available INBUF macros are described in the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide. The CLKDLY macro can be driven directly from the FPGA core. The CLKDLY macro can also be driven from an I/O that is routed through the FPGA regular routing fabric. In this case, users must instantiate a special macro, PLLINT, to differentiate the clock input driven by the hardwired I/O connection. The visual CLKDLY configuration in the SmartGen area of the Microsemi Libero System-on-Chip (SoC) and Designer tools allows the user to select the desired amount of delay and configures the delay elements appropriately. SmartGen also allows the user to select the input clock source. SmartGen will automatically instantiate the special macro, PLLINT, when needed. CLKDLY Macro Signal Descriptions The CLKDLY macro supports one input and one output. Each signal is described in Table 4-2. Table 4-2 • Input and Output Description of the CLKDLY Macro Signal Name I/O CLK Reference Clock Input GL Global Output Description Reference clock input Output Primary output clock to respective global/quadrant clock networks Revision 4 81 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs CLKDLY Macro Usage When a CLKDLY macro is used in a CCC location, the programmable delay element is used to allow the clock delays to go to the global network. In addition, the user can bypass the PLL in a CCC location integrated with a PLL, but use the programmable delay that is associated with the global network by instantiating the CLKDLY macro. The same is true when using programmable delay elements in a CCC location with no PLLs (the user needs to instantiate the CLKDLY macro). There is no difference between the programmable delay elements used for the PLL and the CLKDLY macro. The CCC will be configured to use the programmable delay elements in accordance with the macro instantiated by the user. As an example, if the PLL is not used in a particular CCC location, the designer is free to specify up to three CLKDLY macros in the CCC, each of which can have its own input frequency and delay adjustment options. If the PLL core is used, assuming output to only one global clock network, the other two global clock networks are free to be used by either connecting directly from the global inputs or connecting from one or two CLKDLY macros for programmable delay. The programmable delay elements are shown in the block diagram of the PLL block shown in Figure 4-6 on page 87. Note that any CCC locations with no PLL present contain only the programmable delay blocks going to the global networks (labeled "Programmable Delay Type 2"). Refer to the "Clock Delay Adjustment" section on page 102 for a description of the programmable delay types used for the PLL. Also refer to Table 4-14 on page 110 for Programmable Delay Type 1 step delay values, and Table 4-15 on page 110 for Programmable Delay Type 2 step delay values. CCC locations with a PLL present can be configured to utilize only the programmable delay blocks (Programmable Delay Type 2) going to the global networks A, B, and C. Global network A can be configured to use only the programmable delay element (bypassing the PLL) if the PLL is not used in the design. Figure 4-6 on page 87 shows a block diagram of the PLL, where the programmable delay elements are used for the global networks (Programmable Delay Type 2). 82 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Global Buffers with PLL Function Clocks requiring frequency synthesis or clock adjustments can utilize the PLL core before connecting to the global / quadrant global networks. A maximum of 18 CCC global buffers can be instantiated in a device—three per CCC and up to six CCCs per device. Each PLL core can generate up to three global/quadrant clocks, while a clock delay element provides one. The PLL functionality of the clock conditioning block is supported by the PLL macro. Clock Source Clock Conditioning Input LVDS/LVPECL Macro Output PLL Macro GLA PADN Y PADP INBUF* Macro PAD Y CLKA EXTFB POWERDOWN OADIVRST1 GLA LOCK GLB YB GLC YC OADIVHALF1 2 OADIV[4:0] OAMUX[2:0] 2 DLYGLA[4:0] 2 OBDIV[4:0] 2 OBMUX[2:0] 2 2 DLYYB[4:0] 2 DLYGLB[4:0] OCDIV[4:0] 2 2 OCMUX[2:0] 2 DLYYC[4:0] DLYGLC[4:0]2 2 FINDIV[6:0] FBDIV[6:0] 2 FBDLY[4:0]2 FBSEL[1:0]2 XDLYSEL2 VCOSEL[2:0]2 or GLA and (GLB or YB) or GLA and (GLC or YC) or GLA and (GLB or YB) and (GLC or YC) Notes: 1. For Fusion only. 2. Refer to the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide for more information. 3. For INBUF* driving a PLL macro or CLKDLY macro, the I/O will be hard-routed to the CCC; i.e., will be placed by software to a dedicated Global I/O. 4. IGLOO nano and ProASIC3 nano devices do not support differential inputs. Figure 4-4 • CCC Options: Global Buffers with PLL The PLL macro provides five derived clocks (three independent) from a single reference clock. The PLL macro also provides power-down input and lock output signals. The additional inputs shown on the macro are configuration settings, which are configured through the use of SmartGen. For manual setting of these bits refer to the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide for details. Figure 4-6 on page 87 illustrates the various clock output options and delay elements. Revision 4 83 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs PLL Macro Signal Descriptions The PLL macro supports two inputs and up to six outputs. Table 4-3 gives a description of each signal. Table 4-3 • Input and Output Signals of the PLL Block Signal Name I/O Description CLKA Reference Clock Input Reference clock input for PLL core; input clock for primary output clock, GLA OADIVRST Reset Signal for the Output Divider A Input For Fusion only. OADIVRST can be used when you bypass the PLL core (i.e., OAMUX = 001). The purpose of the OADIVRST signals is to reset the output of the final clock divider to synchronize it with the input to that divider when the PLL is bypassed. The signal is active on a low to high transition. The signal must be low for at least one divider input. If PLL core is used, this signal is "don't care" and the internal circuitry will generate the reset signal for the synchronization purpose. OADIVHALF Output A Division by Half Input For Fusion only. Active high. Division by half feature. This feature can only be used when users bypass the PLL core (i.e., OAMUX = 001) and the RC Oscillator (RCOSC) drives the CLKA input. This can be used to divide the 100 MHz RC oscillator by a factor of 1.5, 2.5, 3.5, 4.5 ... 14.5). Refer to Table 4-18 on page 111 for more information. EXTFB External Feedback Input Allows an external signal to be compared to a reference clock in the PLL core's phase detector. Input Active low input that selects power-down mode and disables the PLL. With the POWERDOWN signal asserted, the PLL core sends 0 V signals on all of the outputs. POWERDOWN Power Down GLA Primary Output Output Primary output clock to respective global/quadrant clock networks GLB Secondary 1 Output Output Secondary 1 output clock to respective global/quadrant clock networks YB Core 1 Output Output Core 1 output clock to local routing network GLC Secondary 2 Output Output Secondary 2 output clock to respective global/quadrant clock networks YC Core 2 Output Output Core 2 output clock to local routing network LOCK PLL Lock Indicator Output Active high signal indicating that steady-state lock has been achieved between CLKA and the PLL feedback signal Input Clock The inputs to the input reference clock (CLKA) of the PLL can come from global input pins, regular I/O pins, or internally from the core. For Fusion families, the input reference clock can also be from the embedded RC oscillator or crystal oscillator. Global Output Clocks GLA (Primary), GLB (Secondary 1), and GLC (Secondary 2) are the outputs of Global Multiplexer 1, Global Multiplexer 2, and Global Multiplexer 3, respectively. These signals (GLx) can be used to drive the high-speed global and quadrant networks of the low power flash devices. A global multiplexer block consists of the input routing for selecting the input signal for the GLx clock and the output multiplexer, as well as delay elements associated with that clock. Core Output Clocks YB and YC are known as Core Outputs and can be used to drive internal logic without using global network resources. This is especially helpful when global network resources must be conserved and utilized for other timing-critical paths. 84 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide YB and YC are identical to GLB and GLC, respectively, with the exception of a higher selectable final output delay. The SmartGen PLL Wizard will configure these outputs according to user specifications and can enable these signals with or without the enabling of Global Output Clocks. The above signals can be enabled in the following output groupings in both internal and external feedback configurations of the static PLL: • One output – GLA only • Two outputs – GLA + (GLB and/or YB) • Three outputs – GLA + (GLB and/or YB) + (GLC and/or YC) PLL Macro Block Diagram As illustrated, the PLL supports three distinct output frequencies from a given input clock. Two of these (GLB and GLC) can be routed to the B and C global network access, respectively, and/or routed to the device core (YB and YC). There are five delay elements to support phase control on all five outputs (GLA, GLB, GLC, YB, and YC). There are delay elements in the feedback loop that can be used to advance the clock relative to the reference clock. The PLL macro reference clock can be driven in the following ways: 1. By an INBUF* macro to create a composite macro, where the I/O macro drives the global buffer (with programmable delay) using a hardwired connection. In this case, the I/O must be placed in one of the dedicated global I/O locations. 2. Directly from the FPGA core. 3. From an I/O that is routed through the FPGA regular routing fabric. In this case, users must instantiate a special macro, PLLINT, to differentiate from the hardwired I/O connection described earlier. During power-up, the PLL outputs will toggle around the maximum frequency of the voltage-controlled oscillator (VCO) gear selected. Toggle frequencies can range from 40 MHz to 250 MHz. This will continue as long as the clock input (CLKA) is constant (HIGH or LOW). This can be prevented by LOW assertion of the POWERDOWN signal. The visual PLL configuration in SmartGen, a component of the Libero SoC and Designer tools, will derive the necessary internal divider ratios based on the input frequency and desired output frequencies selected by the user. Revision 4 85 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Implementing EXTFB in ProASIC3/E Devices When the external feedback (EXTFB) signal of the PLL in the ProASIC3/E devices is implemented, the phase detector of the PLL core receives the reference clock (CLKA) and EXTFB as inputs. EXTFB must be sourced as an INBUF macro and located at the global/chip clock location associated with the target PLL by Designer software. EXTFB cannot be sourced from the FPGA fabric. The following example shows CLKA and EXTFB signals assigned to two global I/Os in the same global area of ProASIC3E device. To Core GxA0 The reference clock, CLKA, can be assigned on GxA0 or GxA1. GxA1 – + Source for CCC (CLKA or CLKB or CLKC) GxA2 Routed Clok (from FPGA core) – + To Core GxB0 External Feedback (EXTFB) signal is assigned on GxB1 by Designer automatically. GxB1 – + Source for CCC (CLKA or CLKB or CLKC) GxB2 Routed Clok (from FPGA core) – + x represents global location; can be A, B, C, D, E, or F Figure 4-5 • CLKA and EXTFB Assigned to Global I/Os 86 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide SmartGen also allows the user to select the various delays and phase shift values necessary to adjust the phases between the reference clock (CLKA) and the derived clocks (GLA, GLB, GLC, YB, and YC). SmartGen allows the user to select the input clock source. SmartGen automatically instantiates the special macro, PLLINT, when needed. CLKA PLL Core Programmable Delay Four-Phase Output Phase Select Programmable Delay Type 2 GLA Programmable Delay Type 2 GLB Programmable Delay Type 1 YB Programmable Delay Type 2 GLC Programmable Delay Type 1 YC Programmable Delay Type 1 EXTFB Phase Select Phase Select Note: Clock divider and clock multiplier blocks are not shown in this figure or in SmartGen. They are automatically configured based on the user's required frequencies. Figure 4-6 • CCC with PLL Block Global Input Selections Low power flash devices provide the flexibility of choosing one of the three global input pad locations available to connect to a CCC functional block or to a global / quadrant global network. Figure 4-7 on page 88 and Figure 4-8 on page 88 show the detailed architecture of each global input structure for 30 k gate devices and below, as well as 60 k gate devices and above, respectively. For 60 k gate devices and above (Figure 4-7 on page 88), if the single-ended I/O standard is chosen, there is flexibility to choose one of the global input pads (the first, second, and fourth input). Once chosen, the other I/O locations are used as regular I/Os. If the differential I/O standard is chosen (not applicable for IGLOO nano and ProASIC3 nano devices), the first and second inputs are considered as paired, and the third input is paired with a regular I/O. The user then has the choice of selecting one of the two sets to be used as the clock input source to the CCC functional block. There is also the option to allow an internal clock signal to feed the global network or the CCC functional block. A multiplexer tree selects the appropriate global input for routing to the desired location. Note that the global I/O pads do not need to feed the global network; they can also be used as regular I/O pads. Revision 4 87 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs To Core Dedicated I/O Pad Sample Pin Names Drives the global network directly (GLA or GLC) GEC0/IO37RSB1 Routed Clock (from FPGA core) Figure 4-7 • Clock Input Sources (30 k gates devices and below) Each shaded box represents an INBUF or INBUF_LVDS/LVPECL macro, as appropriate. To Core Sample Pin Names GAA0/IO0NDB0V0 1 GAA1/IO00PDB0V0 1 + Source for CCC (CLKA or CLKB or CLKC) GAA2/IO13PDB7V1 1 + Routed Clock 2 (from FPGA core) GAA[0:2]: GA represents global in the northwest corner of the device. A[0:2]: designates specific A clock source. Notes: 1. Represents the global input pins. Globals have direct access to the clock conditioning block and are not routed via the FPGA fabric. Refer to the "User I/O Naming Conventions in I/O Structures" chapter of the appropriate device user’s guide. 2. Instantiate the routed clock source input as follows: a) Connect the output of a logic element to the clock input of a PLL, CLKDLY, or CLKINT macro. b) Do not place a clock source I/O (INBUF or INBUF_LVPECL/LVDS/B-LVDS/M-LVDS/DDR) in a relevant global pin location. 3. IGLOO nano and ProASIC3 nano devices do not support differential inputs. Figure 4-8 • Clock Input Sources Including CLKBUF, CLKBUF_LVDS/LVPECL, and CLKINT (60 k gates devices and above) 88 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Each global buffer, as well as the PLL reference clock, can be driven from one of the following: • 3 dedicated single-ended I/Os using a hardwired connection • 2 dedicated differential I/Os using a hardwired connection (not applicable for IGLOO nano and ProASIC3 nano devices) • The FPGA core Since the architecture of the devices varies as size increases, the following list details I/O types supported for globals: IGLOO and ProASIC3 • LVDS-based clock sources are available only on 250 k gate devices and above (IGLOO nano and ProASIC3 nano devices do not support differential inputs). • 60 k and 125 k gate devices support single-ended clock sources only. • 15 k and 30 k gate devices support these inputs for CCC only and do not contain a PLL. • nano devices: – 10 k, 15 k, and 20 k devices do not contain PLLs in the CCCs, and support only CLKBUF and CLKINT. – 60 k, 125 k, and 250 k devices support one PLL in the middle left CCC position. In the absence of the PLL, this CCC can be used by CLKBUF, CLKINT, and CLKDLY macros. The corner CCCs support CLKBUF, CLKINT, and CLKDLY. Fusion • AFS600 and AFS1500: All single-ended, differential, and voltage-referenced I/O standards (Pro I/O). • AFS090 and AFS250: All single-ended and differential I/O standards. Clock Sources for PLL and CLKDLY Macros The input reference clock (CLKA for a PLL macro, CLK for a CLKDLY macro) can be accessed from different sources via the associated clock multiplexer tree. Each CCC has the option of choosing the source of the input clock from one of the following: • Hardwired I/O • External I/O • Core Logic • RC Oscillator (Fusion only) • Crystal Oscillator (Fusion only) The SmartGen macro builder tool allows users to easily create the PLL and CLKDLY macros with the desired settings. Microsemi strongly recommends using SmartGen to generate the CCC macros. Hardwired I/O Clock Source Hardwired I/O refers to global input pins that are hardwired to the multiplexer tree, which directly accesses the CCC global buffers. These global input pins have designated pin locations and are indicated with the I/O naming convention Gmn (m refers to any one of the positions where the PLL core is available, and n refers to any one of the three global input MUXes and the pin number of the associated global location, m). Choosing this option provides the benefit of directly connecting to the CCC reference clock input, which provides less delay. See Figure 4-9 on page 90 for an example illustration of the connections, shown in red. If a CLKDLY macro is initiated to utilize the programmable delay element of the CCC, the clock input can be placed at one of nine dedicated global input pin locations. In other words, if Hardwired I/O is chosen as the input source, the user can decide to place the input pin in one of the GmA0, GmA1, GmA2, GmB0, GmB1, GmB2, GmC0, GmC1, or GmC2 locations of the low power flash devices. When a PLL macro is used to utilize the PLL core in a CCC location, the clock input of the PLL can only be connected to one of three GmA* global pin locations: GmA0, GmA1, or GmA2. Revision 4 89 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs To Core Gmn0 Gmn1 Multiplexer Tree _ + PLL or CLKDLY Macro To Global (or local) Routing Network CLKA IOuxwByVz Gmn2 _ + Routed Clock (from FPGA core) Gmn* = Global Input Pin IOuxwByVz = Regular I/O Pin PLLINT Note: Fusion CCCs have additional source selections (RCOSC, XTAL). Figure 4-9 • Illustration of Hardwired I/O (global input pins) Usage for IGLOO and ProASIC3 devices 60 k Gates and Larger Dedicated I/O Pad To Core Sample Pin Names GEC0/IO37RSB1 Directly Drives Global Network (GLA or GLC) Routed Clock (from the FPGA core) Figure 4-10 • Illustration of Hardwired I/O (global input pins) Usage for IGLOO and ProASIC3 devices 30 k Gates and Smaller 90 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide External I/O Clock Source External I/O refers to regular I/O pins. The clock source is instantiated with one of the various INBUF options and accesses the CCCs via internal routing. The user has the option of assigning this input to any of the I/Os labeled with the I/O convention IOuxwByVz. Refer to the "User I/O Naming Conventions in I/O Structures" chapter of the appropriate device user’s guide, and for Fusion, refer to the Fusion Family of Mixed Signal FPGAs datasheet for more information. Figure 4-11 gives a brief explanation of external I/O usage. Choosing this option provides the freedom of selecting any user I/O location but introduces additional delay because the signal connects to the routed clock input through internal routing before connecting to the CCC reference clock input. For the External I/O option, the routed signal would be instantiated with a PLLINT macro before connecting to the CCC reference clock input. This instantiation is conveniently done automatically by SmartGen when this option is selected. Microsemi recommends using the SmartGen tool to generate the CCC macro. The instantiation of the PLLINT macro results in the use of the routed clock input of the I/O to connect to the PLL clock input. If not using SmartGen, manually instantiate a PLLINT macro before the PLL reference clock to indicate that the regular I/O driving the PLL reference clock should be used (see Figure 4-11 for an example illustration of the connections, shown in red). In the above two options, the clock source must be instantiated with one of the various INBUF macros. The reference clock pins of the CCC functional block core macros must be driven by regular input macros (INBUFs), not clock input macros. To Core Gmn* Gmn* Multiplexer Tree _ + PLL or CLKDLY Macro CLKA To Global (or Local) Routing Network IOuxwByVz* Gmn* _ + PLLINT Routed Clock (from FPGA Core) Gmn* = Global Input Pin IOuxwByVz = Regular I/O Pin IOuxwByVz* Figure 4-11 • Illustration of External I/O Usage For Fusion devices, the input reference clock can also be from the embedded RC oscillator and crystal oscillator. In this case, the CCC configuration is the same as the hardwired I/O clock source, and users are required to instantiate the RC oscillator or crystal oscillator macro and connect its output to the input reference clock of the CCC block. Revision 4 91 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Core Logic Clock Source Core logic refers to internal routed nets. Internal routed signals access the CCC via the FPGA Core Fabric. Similar to the External I/O option, whenever the clock source comes internally from the core itself, the routed signal is instantiated with a PLLINT macro before connecting to the CCC clock input (see Figure 4-12 for an example illustration of the connections, shown in red). To Core Gmn* Gmn* Multiplexer Tree _ + PLL or CLKDLY Macro CLKA To Global (or Local) Routing Network IOuxwByVz* Gmn* _ + Routed Clock (from FPGA Core) PLLINT Gmn* = Global Input Pin IOuxwByVz = Regular I/O Pin From Internal Signals Figure 4-12 • Illustration of Core Logic Usage For Fusion devices, the input reference clock can also be from the embedded RC oscillator and crystal oscillator. In this case, the CCC configuration is the same as the hardwired I/O clock source, and users are required to instantiate the RC oscillator or crystal oscillator macro and connect its output to the input reference clock of the CCC block. 92 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Available I/O Standards Table 4-4 • Available I/O Standards within CLKBUF and CLKBUF_LVDS/LVPECL Macros CLKBUF_LVCMOS5 CLKBUF_LVCMOS33 1 CLKBUF_LVCMOS25 2 CLKBUF_LVCMOS18 CLKBUF_LVCMOS15 CLKBUF_PCI CLKBUF_PCIX 3 CLKBUF_GTL25 2,3 CLKBUF_GTL33 2,3 CLKBUF_GTLP25 2,3 CLKBUF_GTLP33 2,3 CLKBUF_HSTL_I 2,3 CLKBUF_HSTL_II 2,3 CLKBUF_SSTL3_I 2,3 CLKBUF_SSTL3_II 2,3 CLKBUF_SSTL2_I 2,3 CLKBUF_SSTL2_II 2,3 CLKBUF_LVDS 4,5 CLKBUF_LVPECL5 Notes: 1. By default, the CLKBUF macro uses 3.3 V LVTTL I/O technology. For more details, refer to the IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide. 2. I/O standards only supported in ProASIC3E and IGLOOe families. 3. I/O standards only supported in the following Fusion devices: AFS600 and AFS1500. 4. B-LVDS and M-LVDS standards are supported by CLKBUF_LVDS. 5. Not supported for IGLOO nano and ProASIC3 nano devices. Global Synthesis Constraints The Synplify® synthesis tool, by default, allows six clocks in a design for Fusion, IGLOO, and ProASIC3. When more than six clocks are needed in the design, a user synthesis constraint attribute, syn_global_buffers, can be used to control the maximum number of clocks (up to 18) that can be inferred by the synthesis engine. High-fanout nets will be inferred with clock buffers and/or internal clock buffers. If the design consists of CCC global buffers, they are included in the count of clocks in the design. The subsections below discuss the clock input source (global buffers with no programmable delays) and the clock conditioning functional block (global buffers with programmable delays and/or PLL function) in detail. Revision 4 93 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Device-Specific Layout Two kinds of CCCs are offered in low power flash devices: CCCs with integrated PLLs, and CCCs without integrated PLLs (simplified CCCs). Table 4-5 lists the number of CCCs in various devices. Table 4-5 • Number of CCCs by Device Size and Package Device ProASIC3 IGLOO Package CCCs with Integrated PLLs A3PN010 AGLN010 All 0 2 A3PN015 AGLN015 All 0 2 A3PN020 AGLN020 All 0 2 AGLN060 CS81 0 6 AGLN060 All other packages 1 5 AGLN125 CS81 0 6 A3PN125 AGLN125 All other packages 1 5 AGLN250 CS81 0 6 A3PN250 AGLN250 All other packages 1 5 AGL015 All 0 2 A3PN060 A3P015 A3P030 CCCs without Integrated PLLs (simplified CCC) AGL030/AGLP030 All 0 2 AGL060/AGLP060 CS121/CS201 0 6 A3P060 AGL060/AGLP060 All other packages 1 5 A3P125 AGL125/AGLP125 All 1 5 A3P250/L AGL250 All 1 5 A3P400 AGL400 All 1 5 A3P600/L AGL600 All 1 5 A3P1000/L AGL1000 All 1 5 A3PE600 AGLE600 PQ208 2 4 All other packages 6 0 A3PE600/L A3PE1500 PQ208 2 4 A3PE1500 All other packages 6 0 PQ208 2 4 All other packages 6 0 All 1 5 A3PE3000/L A3PE3000/L AGLE3000 Fusion Devices AFS090 AFS250, M1AFS250 All 1 5 AFS600, M7AFS600, M1AFS600 All 2 4 AFS1500, M1AFS1500 All 2 4 Note: nano 10 k, 15 k, and 20 k offer 6 global MUXes instead of CCCs. 94 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide This section outlines the following device information: CCC features, PLL core specifications, functional descriptions, software configuration information, detailed usage information, recommended board-level considerations, and other considerations concerning global networks in low power flash devices. Clock Conditioning Circuits with Integrated PLLs Each of the CCCs with integrated PLLs includes the following: • 1 PLL core, which consists of a phase detector, a low-pass filter, and a four-phase voltagecontrolled oscillator • 3 global multiplexer blocks that steer signals from the global pads and the PLL core onto the global networks • 6 programmable delays and 1 fixed delay for time advance/delay adjustments • 5 programmable frequency divider blocks to provide frequency synthesis (automatically configured by the SmartGen macro builder tool) Clock Conditioning Circuits without Integrated PLLs There are two types of simplified CCCs without integrated PLLs in low power flash devices. 1. The simplified CCC with programmable delays, which is composed of the following: – 3 global multiplexer blocks that steer signals from the global pads and the programmable delay elements onto the global networks – 3 programmable delay elements to provide time delay adjustments 2. The simplified CCC (referred to as CCC-GL) without programmable delay elements, which is composed of the following: – A global multiplexer block that steer signals from the global pads onto the global networks Revision 4 95 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs CCC Locations CCCs located in the middle of the east and west sides of the device access the three VersaNet global networks on each side (six total networks), while the four CCCs located in the four corners access three quadrant global networks (twelve total networks). See Figure 4-13. Northwest Quadrant Global Networks Quadrant Global Spine CCC Location A 3 3 3 3 Chip-Wide (main) Global Networks 6 6 CCC Location B 3 6 6 CCC Location C CCC Location F Global Spine 3 CCC Location E 6 6 3 6 3 6 3 3 Southeast Quadrant Global Networks CCC Location D Figure 4-13 • Global Network Architecture for 60 k Gate Devices and Above The following explains the locations of the CCCs in IGLOO and ProASIC3 devices: In Figure 4-15 on page 98 through Figure 4-16 on page 98, CCCs with integrated PLLs are indicated in red, and simplified CCCs are indicated in yellow. There is a letter associated with each location of the CCC, in clockwise order. The upper left corner CCC is named "A," the upper right is named "B," and so on. These names finish up at the middle left with letter "F." 96 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide IGLOO and ProASIC3 CCC Locations In all IGLOO and ProASIC3 devices (except 10 k through 30 k gate devices, which do not contain PLLs), six CCCs are located in the same positions as the IGLOOe and ProASIC3E CCCs. Only one of the CCCs has an integrated PLL and is located in the middle of the west (middle left) side of the device. The other five CCCs are simplified CCCs and are located in the four corners and the middle of the east side of the device (Figure 4-14). A B Bank 0 RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 1 Bank 3 CCC I/Os C F Bank 1 Bank 3 VersaTile ISP AES Decryption User Nonvolatile FlashROM (FROM) RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps D E Bank 2 = CCC with integrated PLL = Simplified CCC with programmable delay elements (no PLL) Figure 4-14 • CCC Locations in IGLOO and ProASIC3 Family Devices (except 10 k through 30 k gate devices) Note: The number and architecture of the banks are different for some devices. 10 k through 30 k gate devices do not support PLL features. In these devices, there are two CCC-GLs at the lower corners (one at the lower right, and one at the lower left). These CCC-GLs do not have programmable delays. Revision 4 97 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs IGLOOe and ProASIC3E CCC Locations IGLOOe and ProASIC3E devices have six CCCs—one in each of the four corners and one each in the middle of the east and west sides of the device (Figure 4-15). All six CCCs are integrated with PLLs, except in PQFP-208 package devices. PQFP-208 package devices also have six CCCs, of which two include PLLs and four are simplified CCCs. The CCCs with PLLs are implemented in the middle of the east and west sides of the device (middle right and middle left). The simplified CCCs without PLLs are located in the four corners of the device (Figure 4-16). CCC A B RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Pro I/Os C F VersaTile ISP AES Decryption User Nonvolatile FlashRom Flash*Freeze Technology RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps E D = CCC with integrated PLL Figure 4-15 • CCC Locations in IGLOOe and ProASIC3E Family Devices (except PQFP-208 package) Bank 0 A B RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 1 Bank 3 CCC I/Os C F Bank 1 Bank 3 VersaTile ISP AES Decryption* User Nonvolatile FlashRom Flash*Freeze Technology RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps D E Bank 2 = CCC with integrated PLL = Simplified CCC with programmable delay elements (no PLL) Figure 4-16 • CCC Locations in ProASIC3E Family Devices (PQFP-208 package) 98 R e visio n 4 ProASIC3L FPGA Fabric User’s Guide Fusion CCC Locations Fusion devices have six CCCs: one in each of the four corners and one each in the middle of the east and west sides of the device (Figure 4-17 and Figure 4-18). The device can have one integrated PLL in the middle of the west side of the device or two integrated PLLs in the middle of the east and west sides of the device (middle right and middle left). A B Bank 0 RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 1 Bank 3 CCC I/Os C F Bank 1 Bank 3 VersaTile ISP AES Decryption User Nonvolatile FlashROM (FROM) RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps D E Bank 2 = CCC with integrated PLL = Simplified CCC with programmable delay elements (no PLL) Figure 4-17 • CCC Locations in Fusion Family Devices (AFS090, AFS250, M1AFS250) Bank 0 A B RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 1 Bank 3 CCC I/Os C F Bank 1 Bank 3 VersaTile ISP AES Decryption* User Nonvolatile FlashRom Flash*Freeze Technology RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Charge Pumps D E Bank 2 = CCC with integrated PLL = Simplified CCC with programmable delay elements (no PLL) Figure 4-18 • CCC Locations in Fusion Family Devices (except AFS090, AFS250, M1AFS250) Revision 4 99 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs PLL Core Specifications PLL core specifications can be found in the DC and Switching Characteristics chapter of the appropriate family datasheet. Loop Bandwidth Common design practice for systems with a low-noise input clock is to have PLLs with small loop bandwidths to reduce the effects of noise sources at the output. Table 4-6 shows the PLL loop bandwidth, providing a measure of the PLL's ability to track the input clock and jitter. Table 4-6 • –3 dB Frequency of the PLL Typical Maximum Minimum (Ta = +125°C, VCCA = 1.4 V) (Ta = +25°C, VCCA = 1.5 V) (Ta = –55°C, VCCA = 1.6 V) 15 kHz 25 kHz 45 kHz –3 dB Frequency PLL Core Operating Principles This section briefly describes the basic principles of PLL operation. The PLL core is composed of a phase detector (PD), a low-pass filter (LPF), and a four-phase voltage-controlled oscillator (VCO). Figure 4-19 illustrates a basic single-phase PLL core with a divider and delay in the feedback path. Frequency Reference Input FIN Phase Detector Voltage Controlled Oscillator Low-Pass Filter Divide by M Counter Frequency Output M × FIN Delay Figure 4-19 • Simplified PLL Core with Feedback Divider and Delay The PLL is an electronic servo loop that phase-aligns the PD feedback signal with the reference input. To achieve this, the PLL dynamically adjusts the VCO output signal according to the average phase difference between the input and feedback signals. The first element is the PD, which produces a voltage proportional to the phase difference between its inputs. A simple example of a digital phase detector is an Exclusive-OR gate. The second element, the LPF, extracts the average voltage from the phase detector and applies it to the VCO. This applied voltage alters the resonant frequency of the VCO, thus adjusting its output frequency. Consider Figure 4-19 with the feedback path bypassing the divider and delay elements. If the LPF steadily applies a voltage to the VCO such that the output frequency is identical to the input frequency, this steady-state condition is known as lock. Note that the input and output phases are also identical. The PLL core sets a LOCK output signal HIGH to indicate this condition. Should the input frequency increase slightly, the PD detects the frequency/phase difference between its reference and feedback input signals. Since the PD output is proportional to the phase difference, the change causes the output from the LPF to increase. This voltage change increases the resonant frequency of the VCO and increases the feedback frequency as a result. The PLL dynamically adjusts in this manner until the PD senses two phase-identical signals and steady-state lock is achieved. The opposite (decreasing PD output signal) occurs when the input frequency decreases. Now suppose the feedback divider is inserted in the feedback path. As the division factor M (shown in Figure 4-20 on page 101) is increased, the average phase difference increases. The average phase 100 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide difference will cause the VCO to increase its frequency until the output signal is phase-identical to the input after undergoing division. In other words, lock in both frequency and phase is achieved when the output frequency is M times the input. Thus, clock division in the feedback path results in multiplication at the output. A similar argument can be made when the delay element is inserted into the feedback path. To achieve steady-state lock, the VCO output signal will be delayed by the input period less the feedback delay. For periodic signals, this is equivalent to time-advancing the output clock by the feedback delay. Another key parameter of a PLL system is the acquisition time. Acquisition time is the amount of time it takes for the PLL to achieve lock (i.e., phase-align the feedback signal with the input reference clock). For example, suppose there is no voltage applied to the VCO, allowing it to operate at its free-running frequency. Should an input reference clock suddenly appear, a lock would be established within the maximum acquisition time. Functional Description This section provides detailed descriptions of PLL block functionality: clock dividers and multipliers, clock delay adjustment, phase adjustment, and dynamic PLL configuration. Clock Dividers and Multipliers The PLL block contains five programmable dividers. Figure 4-20 shows a simplified PLL block. n CLKA PLL Core m Fixed Delay System Delay 90° 180° 270° 0° u D1 Feedback Delay Output Delay D2 Output Delay D2 GLB Secondary 1 v D1 Output Delay Output Delay D2 D1 = Programmable Delay Type 1 D2 = Programmable Delay Type 2 GLA Primary w YB GLC Secondary 2 D1 Output Delay YC Figure 4-20 • PLL Block Diagram Revision 4 101 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Dividers n and m (the input divider and feedback divider, respectively) provide integer frequency division factors from 1 to 128. The output dividers u, v, and w provide integer division factors from 1 to 32. Frequency scaling of the reference clock CLKA is performed according to the following formulas: fGLA = fCLKA × m / (n × u) – GLA Primary PLL Output Clock EQ 4-1 fGLB = fYB = fCLKA × m / (n × v) – GLB Secondary 1 PLL Output Clock(s) EQ 4-2 fGLC = fYC = fCLKA × m / (n × w) – GLC Secondary 2 PLL Output Clock(s) EQ 4-3 SmartGen provides a user-friendly method of generating the configured PLL netlist, which includes automatically setting the division factors to achieve the closest possible match to the requested frequencies. Since the five output clocks share the n and m dividers, the achievable output frequencies are interdependent and related according to the following formula: fGLA = fGLB × (v / u) = fGLC × (w / u) EQ 4-4 Clock Delay Adjustment There are a total of seven configurable delay elements implemented in the PLL architecture. Two of the delays are located in the feedback path, entitled System Delay and Feedback Delay. System Delay provides a fixed delay of 2 ns (typical), and Feedback Delay provides selectable delay values from 0.6 ns to 5.56 ns in 160 ps increments (typical). For PLLs, delays in the feedback path will effectively advance the output signal from the PLL core with respect to the reference clock. Thus, the System and Feedback delays generate negative delay on the output clock. Additionally, each of these delays can be independently bypassed if necessary. The remaining five delays perform traditional time delay and are located at each of the outputs of the PLL. Besides the fixed global driver delay of 0.755 ns for each of the global networks, the global multiplexer outputs (GLA, GLB, and GLC) each feature an additional selectable delay value, as given in Table 4-7. Table 4-7 • Delay Values in Libero SoC Software per Device Family Device Typical Starting Values Increments Ending Value ProASIC3 200 ps 0 to 735 ps 200 ps 6.735 ns IGLOO/ProASIC3L 1.5 V 360 ps 0 to 1.610 ns 360 ps 12.410 ns IGLOO/ProASIC3L 1.2 V 580 ps 0 to 2.880 ns 580 ps 20.280 ns The additional YB and YC signals have access to a selectable delay from 0.6 ns to 5.56 ns in 160 ps increments (typical). This is the same delay value as the CLKDLY macro. It is similar to CLKDLY, which bypasses the PLL core just to take advantage of the phase adjustment option with the delay value. The following parameters must be taken into consideration to achieve minimum delay at the outputs (GLA, GLB, GLC, YB, and YC) relative to the reference clock: routing delays from the PLL core to CCC outputs, core outputs and global network output delays, and the feedback path delay. The feedback path delay acts as a time advance of the input clock and will offset any delays introduced beyond the PLL core output. The routing delays are determined from back-annotated simulation and are configurationdependent. 102 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Phase Adjustment The four phases available (0, 90, 180, 270) are phases with respect to VCO (PLL output). The VCO is divided to achieve the user's CCC required output frequency (GLA, YB/GLB, YC/GLC). The division happens after the selection of the VCO phase. The effective phase shift is actually the VCO phase shift divided by the output divider. This is why the visual CCC shows both the actual achievable phase and more importantly the actual delay that is equivalent to the phase shift that can be achieved. Dynamic PLL Configuration The CCCs can be configured both statically and dynamically. In addition to the ports available in the Static CCC, the Dynamic CCC has the dynamic shift register signals that enable dynamic reconfiguration of the CCC. With the Dynamic CCC, the ports CLKB and CLKC are also exposed. All three clocks (CLKA, CLKB, and CLKC) can be configured independently. The CCC block is fully configurable. The following two sources can act as the CCC configuration bits. Flash Configuration Bits The flash configuration bits are the configuration bits associated with programmed flash switches. These bits are used when the CCC is in static configuration mode. Once the device is programmed, these bits cannot be modified. They provide the default operating state of the CCC. Dynamic Shift Register Outputs This source does not require core reprogramming and allows core-driven dynamic CCC reconfiguration. When the dynamic register drives the configuration bits, the user-defined core circuit takes full control over SDIN, SDOUT, SCLK, SSHIFT, and SUPDATE. The configuration bits can consequently be dynamically changed through shift and update operations in the serial register interface. Access to the logic core is accomplished via the dynamic bits in the specific tiles assigned to the PLLs. Figure 4-21 illustrates a simplified block diagram of the MUX architecture in the CCCs. SDIN SDOUT SCLK SSHIFT SUPDATE * Dynamic Shift Register RESET_ENABLE Flash Programming Configuration Bits * MODE Configuration Bits Note: *For Fusion, bit is also needed. Figure 4-21 • The CCC Configuration MUX Architecture The selection between the flash configuration bits and the bits from the configuration register is made using the MODE signal shown in Figure 4-21. If the MODE signal is logic HIGH, the dynamic shift register configuration bits are selected. There are 81 control bits to configure the different functions of the CCC. Revision 4 103 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Each group of control bits is assigned a specific location in the configuration shift register. For a list of the 81 configuration bits (C[80:0]) in the CCC and a description of each, refer to "PLL Configuration Bits Description" on page 106. The configuration register can be serially loaded with the new configuration data and programmed into the CCC using the following ports: • SDIN: The configuration bits are serially loaded into a shift register through this port. The LSB of the configuration data bits should be loaded first. • SDOUT: The shift register contents can be shifted out (LSB first) through this port using the shift operation. • SCLK: This port should be driven by the shift clock. • SSHIFT: The active-high shift enable signal should drive this port. The configuration data will be shifted into the shift register if this signal is HIGH. Once SSHIFT goes LOW, the data shifting will be halted. • SUPDATE: The SUPDATE signal is used to configure the CCC with the new configuration bits when shifting is complete. To access the configuration ports of the shift register (SDIN, SDOUT, SSHIFT, etc.), the user should instantiate the CCC macro in his design with appropriate ports. Microsemi recommends that users choose SmartGen to generate the CCC macros with the required ports for dynamic reconfiguration. Users must familiarize themselves with the architecture of the CCC core and its input, output, and configuration ports to implement the desired delay and output frequency in the CCC structure. Figure 4-22 shows a model of the CCC with configurable blocks and switches. 104 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide CLKA /n C 90° 180° 270° 0° PLL Core /m (7) (6) (5) (4) M U X A D /u C (2) Internal C C (1) (0) GLA C (2) D (1) D M U X B C C C D /v YB C C D GLB CLKB C Internal M U X C D /w YC C C C D GLC CLKC C Internal Figure 4-22 • CCC Block Control Bits – Graphical Representation of Assignments Revision 4 105 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Loading the Configuration Register The most important part of CCC dynamic configuration is to load the shift register properly with the configuration bits. There are different ways to access and load the configuration shift register: • JTAG interface • Logic core • Specific I/O tiles JTAG Interface The JTAG interface requires no additional I/O pins. The JTAG TAP controller is used to control the loading of the CCC configuration shift register. Low power flash devices provide a user interface macro between the JTAG pins and the device core logic. This macro is called UJTAG. A user should instantiate the UJTAG macro in his design to access the configuration register ports via the JTAG pins. For more information on CCC dynamic reconfiguration using UJTAG, refer to the "UJTAG Applications in Microsemi’s Low Power Flash Devices" section on page 363. Logic Core If the logic core is employed, the user must design a module to provide the configuration data and control the shifting and updating of the CCC configuration shift register. In effect, this is a user-designed TAP controller, which requires additional chip resources. Specific I/O Tiles If specific I/O tiles are used for configuration, the user must provide the external equivalent of a TAP controller. This does not require additional core resources but does use pins. Shifting the Configuration Data To enter a new configuration, all 81 bits must shift in via SDIN. After all bits are shifted, SSHIFT must go LOW and SUPDATE HIGH to enable the new configuration. For simulation purposes, bits and are "don't care." The SUPDATE signal must be LOW during any clock cycle where SSHIFT is active. After SUPDATE is asserted, it must go back to the LOW state until a new update is required. PLL Configuration Bits Description Table 4-8 • Configuration Bit Descriptions for the CCC Blocks Config. Bits Signal Name 1 GLMUXCFG [1:0] NGMUX configuration Description The configuration bits specify the input clocks to the NGMUX (refer to Table 4-17 on page 110).2 86 OCDIVHALF1 Division by half When the PLL is bypassed, the 100 MHz RC oscillator can be divided by the divider factor in Table 4-18 on page 111. 85 OBDIVHALF1 Division by half When the PLL is bypassed, the 100 MHz RC oscillator can be divided by a 0.5 factor (refer to Table 4-18 on page 111). 84 OADIVHALF1 Division by half When the PLL is bypassed, the 100 MHz RC oscillator can be divided by certain 0.5 factor (refer to Table 4-16 on page 110). Notes: 1. The configuration bits are only for the Fusion dynamic CCC. 2. This value depends on the input clock source, so Layout must complete before these bits can be set. After completing Layout in Designer, generate the "CCC_Configuration" report by choosing Tools > Report > CCC_Configuration. The report contains the appropriate settings for these bits. 106 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 4-8 • Configuration Bit Descriptions for the CCC Blocks (continued) Config. Bits Signal Name Description 83 RXCSEL1 CLKC input selection Select the CLKC input clock source between RC oscillator and crystal oscillator (refer to Table 4-16 on page 110).2 82 RXBSEL1 CLKB input selection Select the CLKB input clock source between RC oscillator and crystal oscillator (refer to Table 4-16 on page 110).2 81 RXASEL1 CLKA input selection Select the CLKA input clock source between RC oscillator and crystal oscillator (refer to Table 4-16 on page 110).2 80 RESETEN Reset Enable Enables (active high) the synchronization of PLL output dividers after dynamic reconfiguration (SUPDATE). The Reset Enable signal is READ-ONLY. 79 DYNCSEL Clock Input C Dynamic Select Configures clock input C to be sent to GLC for dynamic control.2 78 DYNBSEL Clock Input B Dynamic Select Configures clock input B to be sent to GLB for dynamic control.2 77 DYNASEL Clock Input A Dynamic Select Configures clock input A for dynamic PLL configuration.2 VCO Gear Control Three-bit VCO Gear Control for four frequency ranges (refer to Table 4-19 on page 111 and Table 4-20 on page 111). VCOSEL[2:0] 73 STATCSEL MUX Select on Input C MUX selection for clock input C2 72 STATBSEL MUX Select on Input B MUX selection for clock input B2 71 STATASEL MUX Select on Input A MUX selection for clock input A2 DLYC[4:0] YC Output Delay Sets the output delay value for YC. DLYB[4:0] YB Output Delay Sets the output delay value for YB. DLYGLC[4:0] GLC Output Delay Sets the output delay value for GLC. DLYGLB[4:0] GLB Output Delay Sets the output delay value for GLB. DLYGLA[4:0] Primary Output Delay Primary GLA output delay XDLYSEL System Delay Select When selected, inserts System Delay in the feedback path in Figure 4-20 on page 101. Sets the feedback delay value for the feedback element in Figure 4-20 on page 101. 45 FBDLY[4:0] Feedback Delay FBSEL[1:0] Primary Feedback Delay Controls the feedback MUX: no delay, include Select programmable delay element, or use external feedback. OCMUX[2:0] Secondary 2 Output Select Selects from the VCO’s four phase outputs for GLC/YC. OBMUX[2:0] Secondary 1 Output Select Selects from the VCO’s four phase outputs for GLB/YB. Notes: 1. The configuration bits are only for the Fusion dynamic CCC. 2. This value depends on the input clock source, so Layout must complete before these bits can be set. After completing Layout in Designer, generate the "CCC_Configuration" report by choosing Tools > Report > CCC_Configuration. The report contains the appropriate settings for these bits. Revision 4 107 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Table 4-8 • Configuration Bit Descriptions for the CCC Blocks (continued) Config. Bits Signal OAMUX[2:0] GLA Output Select Selects from the VCO’s four phase outputs for GLA. OCDIV[4:0] Secondary 2 Output Divider Sets the divider value for the GLC/YC outputs. Also known as divider w in Figure 4-20 on page 101. The divider value will be OCDIV[4:0] + 1. OBDIV[4:0] Secondary 1 Output Divider Sets the divider value for the GLB/YB outputs. Also known as divider v in Figure 4-20 on page 101. The divider value will be OBDIV[4:0] + 1. OADIV[4:0] Primary Output Divider Sets the divider value for the GLA output. Also known as divider u in Figure 4-20 on page 101. The divider value will be OADIV[4:0] + 1. FBDIV[6:0] Feedback Divider Sets the divider value for the PLL core feedback. Also known as divider m in Figure 4-20 on page 101. The divider value will be FBDIV[6:0] + 1. FINDIV[6:0] Input Divider Input Clock Divider (/n). Sets the divider value for the input delay on CLKA. The divider value will be FINDIV[6:0] + 1. Name Description Notes: 1. The configuration bits are only for the Fusion dynamic CCC. 2. This value depends on the input clock source, so Layout must complete before these bits can be set. After completing Layout in Designer, generate the "CCC_Configuration" report by choosing Tools > Report > CCC_Configuration. The report contains the appropriate settings for these bits. 108 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 4-9 to Table 4-15 on page 110 provide descriptions of the configuration data for the configuration bits. Table 4-9 • Input Clock Divider, FINDIV[6:0] (/n) FINDIV State Divisor New Frequency Factor 0 1 1.00000 1 2 0.50000 … … … 127 128 0.0078125 Divisor New Frequency Factor 0 1 1 1 2 2 Table 4-10 • Feedback Clock Divider, FBDIV[6:0] (/m) FBDIV State … … … 127 128 128 Divisor New Frequency Factor 0 1 1.00000 1 2 0.50000 Table 4-11 • Output Frequency Dividers A Output Divider, OADIV (/u); B Output Divider, OBDIV (/v); C Output Divider, OCDIV (/w) OADIV; OBDIV; CDIV State … … … 31 32 0.03125 Table 4-12 • MUXA, MUXB, MUXC OAMUX; OBMUX; OCMUX State MUX Input Selected 0 None. Six-input MUX and PLL are bypassed. Clock passes only through global MUX and goes directly into HC ribs. 1 Not available 2 PLL feedback delay line output 3 Not used 4 PLL VCO 0° phase shift 5 PLL VCO 270° phase shift 6 PLL VCO 180° phase shift 7 PLL VCO 90° phase shift Revision 4 109 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Table 4-13 • 2-Bit Feedback MUX FBSEL State MUX Input Selected 0 Ground. Used for power-down mode in power-down logic block. 1 PLL VCO 0° phase shift 2 PLL delayed VCO 0° phase shift 3 N/A Table 4-14 • Programmable Delay Selection for Feedback Delay and Secondary Core Output Delays FBDLY; DLYYB; DLYYC State Delay Value 0 Typical delay = 600 ps 1 Typical delay = 760 ps 2 Typical delay = 920 ps … … 31 Typical delay = 5.56 ns Table 4-15 • Programmable Delay Selection for Global Clock Output Delays DLYGLA; DLYGLB; DLYGLC State Delay Value 0 Typical delay = 225 ps 1 Typical delay = 760 ps 2 Typical delay = 920 ps … … 31 Typical delay = 5.56 ns Table 4-16 • Fusion Dynamic CCC Clock Source Selection DYNASEL Source of CLKA 1 RXASEL 0 RC Oscillator 1 1 Crystal Oscillator DYNBSEL Source of CLKB RXBSEL 1 0 RC Oscillator 1 1 Crystal Oscillator DYNCSEL Source of CLKC 1 0 RC Oscillator 1 1 Crystal Oscillator RXBSEL Table 4-17 • Fusion Dynamic CCC NGMUX Configuration GLMUXCFG NGMUX Select Signal Supported Input Clocks to NGMUX 0 GLA 1 GLC 01 0 GLA 1 GLINT 10 0 GLC 1 GLINT 00 110 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 4-18 • Fusion Dynamic CCC Division by Half Configuration OADIVHALF / OBDIVHALF / OCDIVHALF OADIV / OBDIV / OCDIV (in decimal) Divider Factor 2 1.5 4 2.5 6 3.5 28.6 8 4.5 22.2 10 5.5 18.2 12 6.5 15.4 14 7.5 13.3 16 8.5 11.8 18 9.5 10.5 20 10.5 9.5 22 11.5 8.7 24 12.5 8.0 26 13.5 7.4 28 14.5 6.9 0–31 1–32 1 0 Input Clock Frequency Output Clock Frequency (MHz) 100 MHz RC Oscillator 66.7 Other Clock Sources 40.0 Depends on other divider settings Table 4-19 • Configuration Bit / VCOSEL Selection for All Families VCOSEL[2:1] 00 Min. (MHz) Voltage 01 10 11 Max. (MHz) Min. (MHz) Max. (MHz) Min. (MHz) Max. (MHz) Min. (MHz) Max. (MHz) IGLOO and IGLOO PLUS 1.2 V ± 5% 24 35 30 70 60 140 135 160 1.5 V ± 5% 24 43.75 30 87.5 60 175 135 250 ProASIC3L, RT ProASIC3, and Military ProASIC3/L 1.2 V ± 5% 24 35 30 70 60 140 135 250 1.5 V ± 5% 24 43.75 30 70 60 175 135 350 24 43.75 33.75 87.5 67.5 175 135 350 ProASIC3 and Fusion 1.5 V ± 5% Table 4-20 • Configuration Bit / VCOSEL Selection for All Families VCOSEL[0] Description 0 Fast PLL lock acquisition time with high tracking jitter. Refer to the corresponding datasheet for specific value and definition. 1 Slow PLL lock acquisition time with low tracking jitter. Refer to the corresponding datasheet for specific value and definition. Revision 4 111 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Software Configuration SmartGen automatically generates the desired CCC functional block by configuring the control bits, and allows the user to select two CCC modes: Static PLL and Delayed Clock (CLKDLY). Static PLL Configuration The newly implemented Visual PLL Configuration Wizard feature provides the user a quick and easy way to configure the PLL with the desired settings (Figure 4-23). The user can invoke SmartGen to set the parameters and generate the netlist file with the appropriate flash configuration bits set for the CCCs. As mentioned in "PLL Macro Block Diagram" on page 85, the input reference clock CLKA can be configured to be driven by Hardwired I/O, External I/O, or Core Logic. The user enters the desired settings for all the parameters (output frequency, output selection, output phase adjustment, clock delay, feedback delay, and system delay). Notice that the actual values (divider values, output frequency, delay values, and phase) are shown to aid the user in reaching the desired design frequency in real time. These values are typical-case data. Best- and worst-case data can be observed through static timing analysis in SmartTime within Designer. For dynamic configuration, the CCC parameters are defined using either the external JTAG port or an internally defined serial interface via the built-in dynamic shift register. This feature provides the ability to compensate for changes in the external environment. VCO Clock Frequency Programmable Output Delay Elements Input Selection Fixed System Delay Output Selection Feedback Selection (Feedback MUX) Figure 4-23 • Visual PLL Configuration Wizard 112 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Feedback Configuration The PLL provides both internal and external feedback delays. Depending on the configuration, various combinations of feedback delays can be achieved. Internal Feedback Configuration This configuration essentially sets the feedback multiplexer to route the VCO output of the PLL core as the input to the feedback of the PLL. The feedback signal can be processed with the fixed system and the adjustable feedback delay, as shown in Figure 4-24. The dividers are automatically configured by SmartGen based on the user input. Indicated below is the System Delay pull-down menu. The System Delay can be bypassed by setting it to 0. When set, it adds a 2 ns delay to the feedback path (which results in delay advancement of the output clock by 2 ns). Figure 4-24 • Internal Feedback with Selectable System Delay Figure 4-25 shows the controllable Feedback Delay. If set properly in conjunction with the fixed System Delay, the total output delay can be advanced significantly. Figure 4-25 • Internal Feedback with Selectable Feedback Delay Revision 4 113 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs External Feedback Configuration For certain applications, such as those requiring generation of PCB clocks that must be matched with existing board delays, it is useful to implement an external feedback, EXTFB. The Phase Detector of the PLL core will receive CLKA and EXTFB as inputs. EXTFB may be processed by the fixed System Delay element as well as the M divider element. The EXTFB option is currently not supported. After setting all the required parameters, users can generate one or more PLL configurations with HDL or EDIF descriptions by clicking the Generate button. SmartGen gives the option of saving session results and messages in a log file: **************** Macro Parameters **************** Name Family Output Format Type Input Freq(MHz) CLKA Source Feedback Delay Value Index Feedback Mux Select XDLY Mux Select Primary Freq(MHz) Primary PhaseShift Primary Delay Value Index Primary Mux Select Secondary1 Freq(MHz) Use GLB Use YB GLB Delay Value Index YB Delay Value Index Secondary1 PhaseShift Secondary1 Mux Select Secondary2 Freq(MHz) Use GLC Use YC GLC Delay Value Index YC Delay Value Index Secondary2 PhaseShift Secondary2 Mux Select : : : : : : : : : : : : : : : : : : : : : : : : : : : test_pll ProASIC3E VHDL Static PLL 10.000 Hardwired I/O 1 2 No 33.000 0 1 4 66.000 YES YES 1 1 0 4 101.000 YES NO 1 1 0 4 … … … Primary Clock frequency 33.333 Primary Clock Phase Shift 0.000 Primary Clock Output Delay from CLKA 0.180 Secondary1 Secondary1 Secondary1 Secondary1 Clock Clock Clock Clock frequency 66.667 Phase Shift 0.000 Global Output Delay from CLKA 0.180 Core Output Delay from CLKA 0.625 Secondary2 Clock frequency 100.000 Secondary2 Clock Phase Shift 0.000 Secondary2 Clock Global Output Delay from CLKA 0.180 Below is an example Verilog HDL description of a legal PLL core configuration generated by SmartGen: module test_pll(POWERDOWN,CLKA,LOCK,GLA); input POWERDOWN, CLKA; output LOCK, GLA; 114 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide wire VCC, GND; VCC VCC_1_net(.Y(VCC)); GND GND_1_net(.Y(GND)); PLL Core(.CLKA(CLKA), .EXTFB(GND), .POWERDOWN(POWERDOWN), .GLA(GLA), .LOCK(LOCK), .GLB(), .YB(), .GLC(), .YC(), .OADIV0(GND), .OADIV1(GND), .OADIV2(GND), .OADIV3(GND), .OADIV4(GND), .OAMUX0(GND), .OAMUX1(GND), .OAMUX2(VCC), .DLYGLA0(GND), .DLYGLA1(GND), .DLYGLA2(GND), .DLYGLA3(GND) , .DLYGLA4(GND), .OBDIV0(GND), .OBDIV1(GND), .OBDIV2(GND), .OBDIV3(GND), .OBDIV4(GND), .OBMUX0(GND), .OBMUX1(GND), .OBMUX2(GND), .DLYYB0(GND), .DLYYB1(GND), .DLYYB2(GND), .DLYYB3(GND), .DLYYB4(GND), .DLYGLB0(GND), .DLYGLB1(GND), .DLYGLB2(GND), .DLYGLB3(GND), .DLYGLB4(GND), .OCDIV0(GND), .OCDIV1(GND), .OCDIV2(GND), .OCDIV3(GND), .OCDIV4(GND), .OCMUX0(GND), .OCMUX1(GND), .OCMUX2(GND), .DLYYC0(GND), .DLYYC1(GND), .DLYYC2(GND), .DLYYC3(GND), .DLYYC4(GND), .DLYGLC0(GND), .DLYGLC1(GND), .DLYGLC2(GND), .DLYGLC3(GND) , .DLYGLC4(GND), .FINDIV0(VCC), .FINDIV1(GND), .FINDIV2( VCC), .FINDIV3(GND), .FINDIV4(GND), .FINDIV5(GND), .FINDIV6(GND), .FBDIV0(VCC), .FBDIV1(GND), .FBDIV2(VCC), .FBDIV3(GND), .FBDIV4(GND), .FBDIV5(GND), .FBDIV6(GND), .FBDLY0(GND), .FBDLY1(GND), .FBDLY2(GND), .FBDLY3(GND), .FBDLY4(GND), .FBSEL0(VCC), .FBSEL1(GND), .XDLYSEL(GND), .VCOSEL0(GND), .VCOSEL1(GND), .VCOSEL2(GND)); defparam Core.VCOFREQUENCY = 33.000; endmodule The "PLL Configuration Bits Description" section on page 106 provides descriptions of the PLL configuration bits for completeness. The configuration bits are shown as busses only for purposes of illustration. They will actually be broken up into individual pins in compilation libraries and all simulation models. For example, the FBSEL[1:0] bus will actually appear as pins FBSEL1 and FBSEL0. The setting of these select lines for the static PLL configuration is performed by the software and is completely transparent to the user. Revision 4 115 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Dynamic PLL Configuration To generate a dynamically reconfigurable CCC, the user should select Dynamic CCC in the configuration section of the SmartGen GUI (Figure 4-26). This will generate both the CCC core and the configuration shift register / control bit MUX. Figure 4-26 • SmartGen GUI Even if dynamic configuration is selected in SmartGen, the user must still specify the static configuration data for the CCC (Figure 4-27). The specified static configuration is used whenever the MODE signal is set to LOW and the CCC is required to function in the static mode. The static configuration data can be used as the default behavior of the CCC where required. Figure 4-27 • Dynamic CCC Configuration in SmartGen 116 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide When SmartGen is used to define the configuration that will be shifted in via the serial interface, SmartGen prints out the values of the 81 configuration bits. For ease of use, several configuration bits are automatically inferred by SmartGen when the dynamic PLL core is generated; however, (STATASEL, STATBSEL, STATCSEL) and (DYNASEL, DYNBSEL, DYNCSEL) depend on the input clock source of the corresponding CCC. Users must first run Layout in Designer to determine the exact setting for these ports. After Layout is complete, generate the "CCC_Configuration" report by choosing Tools > Reports > CCC_Configuration in the Designer software. Refer to "PLL Configuration Bits Description" on page 106 for descriptions of the PLL configuration bits. For simulation purposes, bits and are "don't care." Therefore, it is strongly suggested that SmartGen be used to generate the correct configuration bit settings for the dynamic PLL core. After setting all the required parameters, users can generate one or more PLL configurations with HDL or EDIF descriptions by clicking the Generate button. SmartGen gives the option of saving session results and messages in a log file: **************** Macro Parameters **************** Name Family Output Format Type Input Freq(MHz) CLKA Source Feedback Delay Value Index Feedback Mux Select XDLY Mux Select Primary Freq(MHz) Primary PhaseShift Primary Delay Value Index Primary Mux Select Secondary1 Freq(MHz) Use GLB Use YB GLB Delay Value Index YB Delay Value Index Secondary1 PhaseShift Secondary1 Mux Select Secondary1 Input Freq(MHz) CLKB Source Secondary2 Freq(MHz) Use GLC Use YC GLC Delay Value Index YC Delay Value Index Secondary2 PhaseShift Secondary2 Mux Select Secondary2 Input Freq(MHz) CLKC Source : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : dyn_pll_hardio ProASIC3E VERILOG Dynamic CCC 30.000 Hardwired I/O 1 1 No 33.000 0 1 4 40.000 YES NO 1 1 0 0 40.000 Hardwired I/O 50.000 YES NO 1 1 0 0 50.000 Hardwired I/O Configuration Bits: FINDIV[6:0] 0000101 FBDIV[6:0] 0100000 OADIV[4:0] 00100 OBDIV[4:0] 00000 OCDIV[4:0] 00000 OAMUX[2:0] 100 OBMUX[2:0] 000 OCMUX[2:0] 000 FBSEL[1:0] 01 FBDLY[4:0] 00000 XDLYSEL 0 DLYGLA[4:0] 00000 DLYGLB[4:0] 00000 Revision 4 117 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs DLYGLC[4:0] DLYYB[4:0] DLYYC[4:0] VCOSEL[2:0] 00000 00000 00000 100 Primary Clock Frequency 33.000 Primary Clock Phase Shift 0.000 Primary Clock Output Delay from CLKA 1.695 Secondary1 Clock Frequency 40.000 Secondary1 Clock Phase Shift 0.000 Secondary1 Clock Global Output Delay from CLKB 0.200 Secondary2 Clock Frequency 50.000 Secondary2 Clock Phase Shift 0.000 Secondary2 Clock Global Output Delay from CLKC 0.200 ###################################### # Dynamic Stream Data ###################################### -------------------------------------|NAME |SDIN |VALUE |TYPE | -------------------------------------|FINDIV |[6:0] |0000101 |EDIT | |FBDIV |[13:7] |0100000 |EDIT | |OADIV |[18:14] |00100 |EDIT | |OBDIV |[23:19] |00000 |EDIT | |OCDIV |[28:24] |00000 |EDIT | |OAMUX |[31:29] |100 |EDIT | |OBMUX |[34:32] |000 |EDIT | |OCMUX |[37:35] |000 |EDIT | |FBSEL |[39:38] |01 |EDIT | |FBDLY |[44:40] |00000 |EDIT | |XDLYSEL |[45] |0 |EDIT | |DLYGLA |[50:46] |00000 |EDIT | |DLYGLB |[55:51] |00000 |EDIT | |DLYGLC |[60:56] |00000 |EDIT | |DLYYB |[65:61] |00000 |EDIT | |DLYYC |[70:66] |00000 |EDIT | |STATASEL|[71] |X |MASKED | |STATBSEL|[72] |X |MASKED | |STATCSEL|[73] |X |MASKED | |VCOSEL |[76:74] |100 |EDIT | |DYNASEL |[77] |X |MASKED | |DYNBSEL |[78] |X |MASKED | |DYNCSEL |[79] |X |MASKED | |RESETEN |[80] |1 |READONLY | Below is the resultant Verilog HDL description of a legal dynamic PLL core configuration generated by SmartGen: module dyn_pll_macro(POWERDOWN, CLKA, LOCK, GLA, GLB, GLC, SDIN, SCLK, SSHIFT, SUPDATE, MODE, SDOUT, CLKB, CLKC); input POWERDOWN, CLKA; output LOCK, GLA, GLB, GLC; input SDIN, SCLK, SSHIFT, SUPDATE, MODE; output SDOUT; input CLKB, CLKC; wire VCC, GND; VCC VCC_1_net(.Y(VCC)); GND GND_1_net(.Y(GND)); 118 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide DYNCCC Core(.CLKA(CLKA), .EXTFB(GND), .POWERDOWN(POWERDOWN), .GLA(GLA), .LOCK(LOCK), .CLKB(CLKB), .GLB(GLB), .YB(), .CLKC(CLKC), .GLC(GLC), .YC(), .SDIN(SDIN), .SCLK(SCLK), .SSHIFT(SSHIFT), .SUPDATE(SUPDATE), .MODE(MODE), .SDOUT(SDOUT), .OADIV0(GND), .OADIV1(GND), .OADIV2(VCC), .OADIV3(GND), .OADIV4(GND), .OAMUX0(GND), .OAMUX1(GND), .OAMUX2(VCC), .DLYGLA0(GND), .DLYGLA1(GND), .DLYGLA2(GND), .DLYGLA3(GND), .DLYGLA4(GND), .OBDIV0(GND), .OBDIV1(GND), .OBDIV2(GND), .OBDIV3(GND), .OBDIV4(GND), .OBMUX0(GND), .OBMUX1(GND), .OBMUX2(GND), .DLYYB0(GND), .DLYYB1(GND), .DLYYB2(GND), .DLYYB3(GND), .DLYYB4(GND), .DLYGLB0(GND), .DLYGLB1(GND), .DLYGLB2(GND), .DLYGLB3(GND), .DLYGLB4(GND), .OCDIV0(GND), .OCDIV1(GND), .OCDIV2(GND), .OCDIV3(GND), .OCDIV4(GND), .OCMUX0(GND), .OCMUX1(GND), .OCMUX2(GND), .DLYYC0(GND), .DLYYC1(GND), .DLYYC2(GND), .DLYYC3(GND), .DLYYC4(GND), .DLYGLC0(GND), .DLYGLC1(GND), .DLYGLC2(GND), .DLYGLC3(GND), .DLYGLC4(GND), .FINDIV0(VCC), .FINDIV1(GND), .FINDIV2(VCC), .FINDIV3(GND), .FINDIV4(GND), .FINDIV5(GND), .FINDIV6(GND), .FBDIV0(GND), .FBDIV1(GND), .FBDIV2(GND), .FBDIV3(GND), .FBDIV4(GND), .FBDIV5(VCC), .FBDIV6(GND), .FBDLY0(GND), .FBDLY1(GND), .FBDLY2(GND), .FBDLY3(GND), .FBDLY4(GND), .FBSEL0(VCC), .FBSEL1(GND), .XDLYSEL(GND), .VCOSEL0(GND), .VCOSEL1(GND), .VCOSEL2(VCC)); defparam Core.VCOFREQUENCY = 165.000; endmodule Delayed Clock Configuration The CLKDLY macro can be generated with the desired delay and input clock source (Hardwired I/O, External I/O, or Core Logic), as in Figure 4-28. Figure 4-28 • Delayed Clock Configuration Dialog Box After setting all the required parameters, users can generate one or more PLL configurations with HDL or EDIF descriptions by clicking the Generate button. SmartGen gives the option of saving session results and messages in a log file: **************** Macro Parameters **************** Name Family Output Format Type Delay Index CLKA Source : : : : : : delay_macro ProASIC3 Verilog Delayed Clock 2 Hardwired I/O Total Clock Delay = 0.935 ns. The resultant CLKDLY macro Verilog netlist is as follows: module delay_macro(GL,CLK); output GL; input CLK; Revision 4 119 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs wire VCC, GND; VCC VCC_1_net(.Y(VCC)); GND GND_1_net(.Y(GND)); CLKDLY Inst1(.CLK(CLK), .GL(GL), .DLYGL0(VCC), .DLYGL1(GND), .DLYGL2(VCC), .DLYGL3(GND), .DLYGL4(GND)); endmodule Detailed Usage Information Clock Frequency Synthesis Deriving clocks of various frequencies from a single reference clock is known as frequency synthesis. The PLL has an input frequency range from 1.5 to 350 MHz. This frequency is automatically divided down to a range between 1.5 MHz and 5.5 MHz by input dividers (not shown in Figure 4-19 on page 100) between PLL macro inputs and PLL phase detector inputs. The VCO output is capable of an output range from 24 to 350 MHz. With dividers before the input to the PLL core and following the VCO outputs, the VCO output frequency can be divided to provide the final frequency range from 0.75 to 350 MHz. Using SmartGen, the dividers are automatically set to achieve the closest possible matches to the specified output frequencies. Users should be cautious when selecting the desired PLL input and output frequencies and the I/O buffer standard used to connect to the PLL input and output clocks. Depending on the I/O standards used for the PLL input and output clocks, the I/O frequencies have different maximum limits. Refer to the family datasheets for specifications of maximum I/O frequencies for supported I/O standards. Desired PLL input or output frequencies will not be achieved if the selected frequencies are higher than the maximum I/O frequencies allowed by the selected I/O standards. Users should be careful when selecting the I/O standards used for PLL input and output clocks. Performing post-layout simulation can help detect this type of error, which will be identified with pulse width violation errors. Users are strongly encouraged to perform post-layout simulation to ensure the I/O standard used can provide the desired PLL input or output frequencies. Users can also choose to cascade PLLs together to achieve the high frequencies needed for their applications. Details of cascading PLLs are discussed in the "Cascading CCCs" section on page 125. In SmartGen, the actual generated frequency (under typical operating conditions) will be displayed beside the requested output frequency value. This provides the ability to determine the exact frequency that can be generated by SmartGen, in real time. The log file generated by SmartGen is a useful tool in determining how closely the requested clock frequencies match the user specifications. For example, assume a user specifies 101 MHz as one of the secondary output frequencies. If the best output frequency that could be achieved were 100 MHz, the log file generated by SmartGen would indicate the actual generated frequency. Simulation Verification The integration of the generated PLL and CLKDLY modules is similar to any VHDL component or Verilog module instantiation in a larger design; i.e., there is no special requirement that users need to take into account to successfully synthesize their designs. For simulation purposes, users need to refer to the VITAL or Verilog library that includes the functional description and associated timing parameters. Refer to the Software Tools section of the Microsemi SoC Products Group website to obtain the family simulation libraries. If Designer is installed, these libraries are stored in the following locations: \lib\vtl\95\proasic3.vhd \lib\vtl\95\proasic3e.vhd \lib\vlog\proasic3.v \lib\vlog\proasic3e.v For Libero users, there is no need to compile the simulation libraries, as they are conveniently precompiled in the ModelSim® Microsemi simulation tool. 120 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide The following is an example of a PLL configuration utilizing the clock frequency synthesis and clock delay adjustment features. The steps include generating the PLL core with SmartGen, performing simulation for verification with ModelSim, and performing static timing analysis with SmartTime in Designer. Parameters of the example PLL configuration: Input Frequency – 20 MHz Primary Output Requirement – 20 MHz with clock advancement of 3.02 ns Secondary 1 Output Requirement – 40 MHz with clock delay of 2.515 ns Figure 4-29 shows the SmartGen settings. Notice that the overall delays are calculated automatically, allowing the user to adjust the delay elements appropriately to obtain the desired delays. Figure 4-29 • SmartGen Settings After confirming the correct settings, generate a structural netlist of the PLL and verify PLL core settings by checking the log file: Name Family Output Format Type Input Freq(MHz) CLKA Source Feedback Delay Value Index Feedback Mux Select XDLY Mux Select Primary Freq(MHz) Primary PhaseShift Primary Delay Value Index Primary Mux Select Secondary1 Freq(MHz) Use GLB Use YB … … … Primary Clock frequency 20.000 Primary Clock Phase Shift 0.000 : : : : : : : : : : : : : : : : test_pll_delays ProASIC3E VHDL Static PLL 20.000 Hardwired I/O 21 2 No 20.000 0 1 4 40.000 YES NO Revision 4 121 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Primary Clock Output Delay from CLKA -3.020 Secondary1 Clock frequency 40.000 Secondary1 Clock Phase Shift 0.000 Secondary1 Clock Global Output Delay from CLKA 2.515 Next, perform simulation in ModelSim to verify the correct delays. Figure 4-30 shows the simulation results. The delay values match those reported in the SmartGen PLL Wizard. Primary Clock Output Time Advancement from CLKA Secondary1 Clock Global Output Delay from CLKA Figure 4-30 • ModelSim Simulation Results The timing can also be analyzed using SmartTime in Designer. The user should import the synthesized netlist to Designer, perform Compile and Layout, and then invoke SmartTime. Go to Tools > Options and change the maximum delay operating conditions to Typical Case. Then expand the Clock-to-Out paths of GLA and GLB and the individual components of the path delays are shown. The path of GLA is shown in Figure 4-31 on page 123 displaying the same delay value. 122 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Figure 4-31 • Static Timing Analysis Using SmartTime Place-and-Route Stage Considerations Several considerations must be noted to properly place the CCC macros for layout. For CCCs with clock inputs configured with the Hardwired I/O–Driven option: • PLL macros must have the clock input pad coming from one of the GmA* locations. • CLKDLY macros must have the clock input pad coming from one of the Global I/Os. If a PLL with a Hardwired I/O input is used at a CCC location and a Hardwired I/O–Driven CLKDLY macro is used at the same CCC location, the clock input of the CLKDLY macro must be chosen from one of the GmB* or GmC* pin locations. If the PLL is not used or is an External I/O–Driven or Core Logic– Driven PLL, the clock input of the CLKDLY macro can be sourced from the GmA*, GmB*, or GmC* pin locations. For CCCs with clock inputs configured with the External I/O–Driven option, the clock input pad can be assigned to any regular I/O location (IO******** pins). Note that since global I/O pins can also be used as regular I/Os, regardless of CCC function (CLKDLY or PLL), clock inputs can also be placed in any of these I/O locations. By default, the Designer layout engine will place global nets in the design at one of the six chip globals. When the number of globals in the design is greater than six, the Designer layout engine will automatically assign additional globals to the quadrant global networks of the low power flash devices. If the user wishes to decide which global signals should be assigned to chip globals (six available) and which to the quadrant globals (three per quadrant for a total of 12 available), the assignment can be achieved with PinEditor, ChipPlanner, or by importing a placement constraint file. Layout will fail if the Revision 4 123 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs global assignments are not allocated properly. See the "Physical Constraints for Quadrant Clocks" section for information on assigning global signals to the quadrant clock networks. Promoted global signals will be instantiated with CLKINT macros to drive these signals onto the global network. This is automatically done by Designer when the Auto-Promotion option is selected. If the user wishes to assign the signals to the quadrant globals instead of the default chip globals, this can done by using ChipPlanner, by declaring a physical design constraint (PDC), or by importing a PDC file. Physical Constraints for Quadrant Clocks If it is necessary to promote global clocks (CLKBUF, CLKINT, PLL, CLKDLY) to quadrant clocks, the user can define PDCs to execute the promotion. PDCs can be created using PDC commands (pre-compile) or the MultiView Navigator (MVN) interface (post-compile). The advantage of using the PDC flow over the MVN flow is that the Compile stage is able to automatically promote any regular net to a global net before assigning it to a quadrant. There are three options to place a quadrant clock using PDC commands: • Place a clock core (not hardwired to an I/O) into a quadrant clock location. • Place a clock core (hardwired to an I/O) into an I/O location (set_io) or an I/O module location (set_location) that drives a quadrant clock location. • Assign a net driven by a regular net or a clock net to a quadrant clock using the following command: assign_local_clock -net -type quadrant where is the name of the net assigned to the local user clock region. defines which quadrant the net should be assigned to. Quadrant clock regions are defined as UL (upper left), UR (upper right), LL (lower left), and LR (lower right). Note: If the net is a regular net, the software inserts a CLKINT buffer on the net. For example: assign_local_clock -net localReset -type quadrant UR Keep in mind the following when placing quadrant clocks using MultiView Navigator: Hardwired I/O–Driven CCCs • Find the associated clock input port under the Ports tab, and place the input port at one of the Gmn* locations using PinEditor or I/O Attribute Editor, as shown in Figure 4-32. Figure 4-32 • Port Assignment for a CCC with Hardwired I/O Clock Input 124 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide • Use quadrant global region assignments by finding the clock net associated with the CCC macro under the Nets tab and creating a quadrant global region for the net, as shown in Figure 4-33. Figure 4-33 • Quadrant Clock Assignment for a Global Net External I/O–Driven CCCs The above-mentioned recommendation for proper layout techniques will ensure the correct assignment. It is possible that, especially with External I/O–Driven CCC macros, placement of the CCC macro in a desired location may not be achieved. For example, assigning an input port of an External I/O–Driven CCC near a particular CCC location does not guarantee global assignments to the desired location. This is because the clock inputs of External I/O–Driven CCCs can be assigned to any I/O location; therefore, it is possible that the CCC connected to the clock input will be routed to a location other than the one closest to the I/O location, depending on resource availability and placement constraints. Clock Placer The clock placer is a placement engine for low power flash devices that places global signals on the chip global and quadrant global networks. Based on the clock assignment constraints for the chip global and quadrant global clocks, it will try to satisfy all constraints, as well as creating quadrant clock regions when necessary. If the clock placer fails to create the quadrant clock regions for the global signals, it will report an error and stop Layout. The user must ensure that the constraints set to promote clock signals to quadrant global networks are valid. Cascading CCCs The CCCs in low power flash devices can be cascaded. Cascading CCCs can help achieve more accurate PLL output frequency results than those achievable with a single CCC. In addition, this technique is useful when the user application requires the output clock of the PLL to be a multiple of the reference clock by an integer greater than the maximum feedback divider value of the PLL (divide by 128) to achieve the desired frequency. For example, the user application may require a 280 MHz output clock using a 2 MHz input reference clock, as shown in Figure 4-34 on page 126. Revision 4 125 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Figure 4-34 • Cascade PLL Configuration Using internal feedback, we know from EQ 4-1 on page 102 that the maximum achievable output frequency from the primary output is fGLA = fCLKA × m / (n × u) = 2 MHz × 128 / (1 × 1) = 256 MHz EQ 4-5 Figure 4-35 shows the settings of the initial PLL. When configuring the initial PLL, specify the input to be either Hardwired I/O–Driven or External I/O–Driven. This generates a netlist with the initial PLL routed from an I/O. Do not specify the input to be Core Logic–Driven, as this prohibits the connection from the I/O pin to the input of the PLL. Figure 4-35 • First-Stage PLL Showing Input of 2 MHz and Output of 256 MHz A second PLL can be connected serially to achieve the required frequency. EQ 4-1 on page 102 to EQ 4-3 on page 102 are extended as follows: fGLA2 = fGLA × m2 / (n2 × u2) = fCLKA1 × m1 × m2 / (n1 × u1 × n2 × u2) – Primary PLL Output Clock EQ 4-6 fGLB2 = fYB2 = fCLKA1 × m1 × m2 / (n1 × n2 × v1 × v2) – Secondary 1 PLL Output Clock(s) EQ 4-7 fGLC2 = fYC2 = fCLKA1 × m1 × m2 / (n1 × n2 × w1 × w2) – Secondary 2 PLL Output Clock(s) EQ 4-8 In the example, the final output frequency (foutput) from the primary output of the second PLL will be as follows (EQ 4-9): foutput = fGLA2 = fGLA × m2 / (n2 × u2) = 256 MHz × 70 / (64 × 1) = 280 MHz EQ 4-9 Figure 4-36 on page 127 shows the settings of the second PLL. When configuring the second PLL (or any subsequent-stage PLLs), specify the input to be Core Logic–Driven. This generates a netlist with the second PLL routed internally from the core. Do not specify the input to be Hardwired I/O–Driven or External I/O–Driven, as these options prohibit the connection from the output of the first PLL to the input of the second PLL. 126 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Figure 4-36 • Second-Stage PLL Showing Input of 256 MHz from First Stage and Final Output of 280 MHz Figure 4-37 shows the simulation results, where the first PLL’s output period is 3.9 ns (~256 MHz), and the stage 2 (final) output period is 3.56 ns (~280 MHz). Stage 1 Output Clock Period Stage 2 Output Clock Period Figure 4-37 • ModelSim Simulation Results Revision 4 127 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Recommended Board-Level Considerations The power to the PLL core is supplied by VCCPLA/B/C/D/E/F (VCCPLx), and the associated ground connections are supplied by VCOMPLA/B/C/D/E/F (VCOMPLx). When the PLLs are not used, the Designer place-and-route tool automatically disables the unused PLLs to lower power consumption. The user should tie unused VCCPLx and VCOMPLx pins to ground. Optionally, the PLL can be turned on/off during normal device operation via the POWERDOWN port (see Table 4-3 on page 84). PLL Power Supply Decoupling Scheme The PLL core is designed to tolerate noise levels on the PLL power supply as specified in the datasheets. When operated within the noise limits, the PLL will meet the output peak-to-peak jitter specifications specified in the datasheets. User applications should always ensure the PLL power supply is powered from a noise-free or low-noise power source. However, in situations where the PLL power supply noise level is higher than the tolerable limits, various decoupling schemes can be designed to suppress noise to the PLL power supply. An example is provided in Figure 4-38. The VCCPLx and VCOMPLx pins correspond to the PLL analog power supply and ground. Microsemi strongly recommends that two ceramic capacitors (10 nF in parallel with 100 nF) be placed close to the power pins (less than 1 inch away). A third generic 10 µF electrolytic capacitor is recommended for low-frequency noise and should be placed farther away due to its large physical size. Microsemi recommends that a 6.8 µH inductor be placed between the supply source and the capacitors to filter out any low-/medium- and high-frequency noise. In addition, the PCB layers should be controlled so the VCCPLx and VCOMPLx planes have the minimum separation possible, thus generating a goodquality RF capacitor. For more recommendations, refer to the Board-Level Considerations application note. Recommended 100 nF capacitor: • Producer BC Components, type X7R, 100 nF, 16 V • BC Components part number: 0603B104K160BT • Digi-Key part number: BC1254CT-ND • Digi-Key part number: BC1254TR-ND Recommended 10 nF capacitor: • Surface-mount ceramic capacitor • Producer BC Components, type X7R, 10 nF, 50 V • BC Components part number: 0603B103K500BT • Digi-Key part number: BC1252CT-ND • Digi-Key part number: BC1252TR-ND VCCPLx IGLOO/e or ProASIC3/E Device 10 nF 100 nF 10 μF Power Supply VCOMPLx Figure 4-38 • Decoupling Scheme for One PLL (should be replicated for each PLL used) 128 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Conclusion The advanced CCCs of the IGLOO and ProASIC3 devices are ideal for applications requiring precise clock management. They integrate easily with the internal low-skew clock networks and provide flexible frequency synthesis, clock deskewing, and/or time-shifting operations. Related Documents Application Notes Board-Level Considerations http://www.microsemi.com/soc/documents/ALL_AC276_AN.pdf Datasheets Fusion Family of Mixed Signal FPGAs http://www.microsemi.com/soc/documents/Fusion_DS.pdf User’s Guides IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide http://www.microsemi.com/soc/documents/pa3_libguide_ug.pdf List of Changes The following table lists critical changes that were made in each revision of the chapter. Date August 2012 Changes Page The "Implementing EXTFB in ProASIC3/E Devices" section is new (SAR 36647). 86 Table 4-7 • Delay Values in Libero SoC Software per Device Family was added to the "Clock Delay Adjustment" section (SAR 22709). 102 The "Phase Adjustment" section was rewritten to explain better why the visual CCC shows both the actual phase and the actual delay that is equivalent to this phase 103 shift (SAR 29647). The hyperlink for the Board-Level Considerations application note was corrected (SAR 128, 129 36663) December 2011 Figure 4-20 • PLL Block Diagram, Figure 4-22 • CCC Block Control Bits – Graphical 101, Representation of Assignments, and Table 4-12 • MUXA, MUXB, MUXC were revised 105, 109 to change the phase shift assignments for PLLs 4 through 7 (SAR 33791). June 2011 The description for RESETEN in Table 4-8 • Configuration Bit Descriptions for the CCC Blocks was revised. The phrase "and should not be modified via dynamic configuration" was deleted because RESETEN is read only (SAR 25949). 106 July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A Notes were added where appropriate to point out that IGLOO nano and ProASIC3 nano devices do not support differential inputs (SAR 21449). N/A Revision 4 129 Clock Conditioning Circuits in Low Power Flash Devices and Mixed Signal FPGAs Date v1.4 (December 2008) v1.3 (October 2008) Changes Page The"CCC Support in Microsemi’s Flash Devices" section was updated to include IGLOO nano and ProASIC3 nano devices. 79 Figure 4-2 • CCC Options: Global Buffers with No Programmable Delay was revised to add the CLKBIBUF macro. 80 The description of the reference clock was revised in Table 4-2 • Input and Output Description of the CLKDLY Macro. 81 Figure 4-7 • Clock Input Sources (30 k gates devices and below) is new. Figure 4-8 • Clock Input Sources Including CLKBUF, CLKBUF_LVDS/LVPECL, and CLKINT (60 k gates devices and above) applies to 60 k gate devices and above. 88 The "IGLOO and ProASIC3" section was updated to include information for IGLOO nano devices. 89 A note regarding Fusion CCCs was added to Figure 4-9 • Illustration of Hardwired I/O (global input pins) Usage for IGLOO and ProASIC3 devices 60 k Gates and Larger and the name of the figure was changed from Figure 4-8 • Illustration of Hardwired I/O (global input pins) Usage. Figure 4-10 • Illustration of Hardwired I/O (global input pins) Usage for IGLOO and ProASIC3 devices 30 k Gates and Smaller is new. 90 Table 4-5 • Number of CCCs by Device Size and Package was updated to include IGLOO nano and ProASIC3 nano devices. Entries were added to note differences for the CS81, CS121, and CS201 packages. 94 The "Clock Conditioning Circuits without Integrated PLLs" section was rewritten. 95 The "IGLOO and ProASIC3 CCC Locations" section was updated for nano devices. 97 Figure 4-13 • CCC Locations in the 15 k and 30 k Gate Devices was deleted. 4-20 This document was updated to include Fusion and RT ProASIC3 device information. Please review the document very carefully. N/A The "CCC Support in Microsemi’s Flash Devices" section was updated. 79 In the "Global Buffer with Programmable Delay" section, the following sentence was changed from: 80 "In this case, the I/O must be placed in one of the dedicated global I/O locations." To "In this case, the software will automatically place the dedicated global I/O in the appropriate locations." 130 Figure 4-4 • CCC Options: Global Buffers with PLL was updated to include OADIVRST and OADIVHALF. 83 In Figure 4-6 • CCC with PLL Block "fixed delay" was changed to "programmable delay". 83 Table 4-3 • Input and Output Signals of the PLL Block was updated to include OADIVRST and OADIVHALF descriptions. 84 Table 4-8 • Configuration Bit Descriptions for the CCC Blocks was updated to include configuration bits 88 to 81. Note 2 is new. In addition, the description for bit was updated. 106 Table 4-16 • Fusion Dynamic CCC Clock Source Selection and Table 4-17 • Fusion Dynamic CCC NGMUX Configuration are new. 110 Table 4-18 • Fusion Dynamic CCC Division by Half Configuration and Table 4-19 • Configuration Bit / VCOSEL Selection for All Families are new. 111 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Date v1.2 (June 2008) Changes Page The following changes were made to the family descriptions in Figure 4-1 • Overview of the CCCs Offered in Fusion, IGLOO, and ProASIC3: 77 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. v1.1 (March 2008) v1.0 (January 2008) Table 4-1 • Flash-Based FPGAs and the associated text were updated to include the IGLOO PLUS family. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 79 The "Global Input Selections" section was updated to include 15 k gate devices as supported I/O types for globals, for CCC only. 87 Table 4-5 • Number of CCCs by Device Size and Package was revised to include ProASIC3L, IGLOO PLUS, A3P015, AGL015, AGLP030, AGLP060, and AGLP125. 94 The "IGLOO and ProASIC3 CCC Locations" section was revised to include 15 k gate devices in the exception statements, as they do not contain PLLs. 97 Information about unlocking the PLL was removed from the "Dynamic PLL Configuration" section. 103 In the "Dynamic PLL Configuration" section, information was added about running Layout and determining the exact setting of the ports. 116 In Table 4-8 • Configuration Bit Descriptions for the CCC Blocks, the following bits were updated to delete "transport to the user" and reference the footnote at the bottom of the table: 79 to 71. 106 Revision 4 131 5 – FlashROM in Microsemi’s Low Power Flash Devices Introduction The Fusion, IGLOO, and ProASIC3 families of low power flash-based devices have a dedicated nonvolatile FlashROM memory of 1,024 bits, which provides a unique feature in the FPGA market. The FlashROM can be read, modified, and written using the JTAG (or UJTAG) interface. It can be read but not modified from the FPGA core. Only low power flash devices contain on-chip user nonvolatile memory (NVM). Architecture of User Nonvolatile FlashROM Low power flash devices have 1 kbit of user-accessible nonvolatile flash memory on-chip that can be read from the FPGA core fabric. The FlashROM is arranged in eight banks of 128 bits (16 bytes) during programming. The 128 bits in each bank are addressable as 16 bytes during the read-back of the FlashROM from the FPGA core. Figure 5-1 shows the FlashROM logical structure. The FlashROM can only be programmed via the IEEE 1532 JTAG port. It cannot be programmed directly from the FPGA core. When programming, each of the eight 128-bit banks can be selectively reprogrammed. The FlashROM can only be reprogrammed on a bank boundary. Programming involves an automatic, on-chip bank erase prior to reprogramming the bank. The FlashROM supports synchronous read. The address is latched on the rising edge of the clock, and the new output data is stable after the falling edge of the same clock cycle. For more information, refer to the timing diagrams in the DC and Switching Characteristics chapter of the appropriate datasheet. The FlashROM can be read on byte boundaries. The upper three bits of the FlashROM address from the FPGA core define the bank being accessed. The lower four bits of the FlashROM address from the FPGA core define which of the 16 bytes in the bank is being accessed. Byte Number in Bank Bank Number 3 MSB of ADDR (READ) 15 14 13 12 11 4 LSB of ADDR (READ) 10 9 8 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 Figure 5-1 • FlashROM Architecture Revision 4 133 FlashROM in Microsemi’s Low Power Flash Devices FlashROM Support in Flash-Based Devices The flash FPGAs listed in Table 5-1 support the FlashROM feature and the functions described in this document. Table 5-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 5-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 5-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 134 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Bank 0 Bank 1 CCC SRAM Block 4,608-Bit Dual-Port SRAM or FIFO Block OSC I/Os CCC/PLL Bank 2 Bank 4 VersaTile User Nonvolatile FlashROM ISP AES Decryption Flash Memory Blocks Analog Quad Analog Quad Analog Quad Charge Pumps ADC Analog Quad Analog Quad SRAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Flash Memory Blocks Analog Quad Analog Quad Analog Quad Analog Quad Analog Quad CCC Bank 3 Figure 5-2 • Fusion Device Architecture Overview (AFS600) CCC RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block I/Os VersaTile RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block ISP AES Decryption Nonvolatile Memory FlashROM Charge Pumps Figure 5-3 • ProASIC3 and IGLOO Device Architecture Revision 4 135 FlashROM in Microsemi’s Low Power Flash Devices FlashROM Applications The SmartGen core generator is used to configure FlashROM content. You can configure each page independently. SmartGen enables you to create and modify regions within a page; these regions can be 1 to 16 bytes long (Figure 5-4). Page Number 15 14 13 12 Byte Number in Page 11 10 9 8 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 Figure 5-4 • FlashROM Configuration The FlashROM content can be changed independently of the FPGA core content. It can be easily accessed and programmed via JTAG, depending on the security settings of the device. The SmartGen core generator enables each region to be independently updated (described in the "Programming and Accessing FlashROM" section on page 138). This enables you to change the FlashROM content on a per-part basis while keeping some regions "constant" for all parts. These features allow the FlashROM to be used in diverse system applications. Consider the following possible uses of FlashROM: 136 • Internet protocol (IP) addressing (wireless or fixed) • System calibration settings • Restoring configuration after unpredictable system power-down • Device serialization and/or inventory control • Subscription-based business models (e.g., set-top boxes) • Secure key storage • Asset management tracking • Date stamping • Version management R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide FlashROM Security Low power flash devices have an on-chip Advanced Encryption Standard (AES) decryption core, combined with an enhanced version of the Microsemi flash-based lock technology (FlashLock®). Together, they provide unmatched levels of security in a programmable logic device. This security applies to both the FPGA core and FlashROM content. These devices use the 128-bit AES (Rijndael) algorithm to encrypt programming files for secure transmission to the on-chip AES decryption core. The same algorithm is then used to decrypt the programming file. This key size provides approximately 3.4 × 1038 possible 128-bit keys. A computing system that could find a DES key in a second would take approximately 149 trillion years to crack a 128-bit AES key. The 128-bit FlashLock feature in low power flash devices works via a FlashLock security Pass Key mechanism, where the user locks or unlocks the device with a user-defined key. Refer to the "Security in Low Power Flash Devices" section on page 301. If the device is locked with certain security settings, functions such as device read, write, and erase are disabled. This unique feature helps to protect against invasive and noninvasive attacks. Without the correct Pass Key, access to the FPGA is denied. To gain access to the FPGA, the device first must be unlocked using the correct Pass Key. During programming of the FlashROM or the FPGA core, you can generate the security header programming file, which is used to program the AES key and/or FlashLock Pass Key. The security header programming file can also be generated independently of the FlashROM and FPGA core content. The FlashLock Pass Key is not stored in the FlashROM. Low power flash devices with AES-based security allow for secure remote field updates over public networks such as the Internet, and ensure that valuable intellectual property (IP) remains out of the hands of IP thieves. Figure 5-5 shows this flow diagram. Programming Data Flash Device FlashROM FPGA Core AES Encryption Same AES Key AES-128 Decryption Core Untrusted Medium Encrypted Data Encrypted Data Figure 5-5 • Programming FlashROM Using AES Revision 4 137 FlashROM in Microsemi’s Low Power Flash Devices Programming and Accessing FlashROM The FlashROM content can only be programmed via JTAG, but it can be read back selectively through the JTAG programming interface, the UJTAG interface, or via direct FPGA core addressing. The pages of the FlashROM can be made secure to prevent read-back via JTAG. In that case, read-back on these secured pages is only possible by the FPGA core fabric or via UJTAG. A 7-bit address from the FPGA core defines which of the eight pages (three MSBs) is being read, and which of the 16 bytes within the selected page (four LSBs) are being read. The FlashROM content can be read on a random basis; the access time is 10 ns for a device supporting commercial specifications. The FPGA core will be powered down during writing of the FlashROM content. FPGA power-down during FlashROM programming is managed on-chip, and FPGA core functionality is not available during programming of the FlashROM. Table 5-2 summarizes various FlashROM access scenarios. Table 5-2 • FlashROM Read/Write Capabilities by Access Mode Access Mode FlashROM Read FlashROM Write JTAG Yes Yes UJTAG Yes No FPGA core Yes No Figure 5-6 shows the accessing of the FlashROM using the UJTAG macro. This is similar to FPGA core access, where the 7-bit address defines which of the eight pages (three MSBs) is being read and which of the 16 bytes within the selected page (four LSBs) are being read. Refer to the "UJTAG Applications in Microsemi’s Low Power Flash Devices" section on page 363 for details on using the UJTAG macro to read the FlashROM. Figure 5-7 on page 139 and Figure 5-8 on page 139 show the FlashROM access from the JTAG port. The FlashROM content can be read on a random basis. The three-bit address defines which page is being read or updated. UJTAG Address Generation and Data Serialization UIREG [7:0] Enable FlashROM RESET TDO TDI URSTB UDRUPD UDRCK TMS TCK TRST UDRCAP UDRSH Control CLK SDI Addr [6:0] Addr [6:0] Data [7:0] Data[7:0] SDO UTDI UTDO Figure 5-6 • Block Diagram of Using UJTAG to Read FlashROM Contents 138 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 7-Bit Address from Core 1110000 7 6 5 3 MSB of ADDR (READ) 3-Bit Page Address Word Number in Page Page Number 111 4-Bit Word Address 0000 4 3 2 1 0 15 14 13 12 11 4 LSB of ADDR (READ) 10 9 8 7 6 5 4 3 2 1 0 8-Bit Data 8-Bit Data to FPGA Core 8-Bit Data from Page 7 Word 0 Figure 5-7 • Accessing FlashROM Using FPGA Core ...........................00001:128 Bit Data To/From JTAG Interface 4-Bit Page Address from JTAG Interface Page Number Word Number in Page 15 14 13 12 11 4 LSB of ADDR (READ) 10 9 8 7 6 5 4 3 2 1 0 3 MSB of ADDR (READ) 7 6 5 4 3 2 1 0 Figure 5-8 • Accessing FlashROM Using JTAG Port Revision 4 139 FlashROM in Microsemi’s Low Power Flash Devices FlashROM Design Flow The Microsemi Libero System-on-Chip (SoC) software has extensive FlashROM support, including FlashROM generation, instantiation, simulation, and programming. Figure 5-9 shows the user flow diagram. In the design flow, there are three main steps: 1. FlashROM generation and instantiation in the design 2. Simulation of FlashROM design 3. Programming file generation for FlashROM design SmartGen UFC File FlashROM Netlist User Design MEM File Simulator Synthesis User Netlist BackAnnotated Netlist Designer Core Map Security Header Options FlashPoint Programmer Programming Files Figure 5-9 • FlashROM Design Flow 140 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide FlashROM Generation and Instantiation in the Design The SmartGen core generator, available in Libero SoC and Designer, is the only tool that can be used to generate the FlashROM content. SmartGen has several user-friendly features to help generate the FlashROM contents. Instead of selecting each byte and assigning values, you can create a region within a page, modify the region, and assign properties to that region. The FlashROM user interface, shown in Figure 5-10, includes the configuration grid, existing regions list, and properties field. The properties field specifies the region-specific information and defines the data used for that region. You can assign values to the following properties: 1. Static Fixed Data—Enables you to fix the data so it cannot be changed during programming time. This option is useful when you have fixed data stored in this region, which is required for the operation of the design in the FPGA. Key storage is one example. 2. Static Modifiable Data—Select this option when the data in a particular region is expected to be static data (such as a version number, which remains the same for a long duration but could conceivably change in the future). This option enables you to avoid changing the value every time you enter new data. 3. Read from File—This provides the full flexibility of FlashROM usage to the customer. If you have a customized algorithm for generating the FlashROM data, you can specify this setting. You can then generate a text file with data for as many devices as you wish to program, and load that into the FlashPoint programming file generation software to get programming files that include all the data. SmartGen will optionally pass the location of the file where the data is stored if the file is specified in SmartGen. Each text file has only one type of data format (binary, decimal, hex, or ASCII text). The length of each data file must be shorter than or equal to the selected region length. If the data is shorter than the selected region length, the most significant bits will be padded with 0s. For multiple text files for multiple regions, the first lines are for the first device. In SmartGen, Load Sim. Value From File allows you to load the first device data in the MEM file for simulation. 4. Auto Increment/Decrement—This scenario is useful when you specify the contents of FlashROM for a large number of devices in a series. You can specify the step value for the serial number and a maximum value for inventory control. During programming file generation, the actual number of devices to be programmed is specified and a start value is fed to the software. Figure 5-10 • SmartGen GUI of the FlashROM Revision 4 141 FlashROM in Microsemi’s Low Power Flash Devices SmartGen allows you to generate the FlashROM netlist in VHDL, Verilog, or EDIF format. After the FlashROM netlist is generated, the core can be instantiated in the main design like other SmartGen cores. Note that the macro library name for FlashROM is UFROM. The following is a sample FlashROM VHDL netlist that can be instantiated in the main design: library ieee; use ieee.std_logic_1164.all; library fusion; entity FROM_a is port( ADDR : in std_logic_vector(6 downto 0); DOUT : out std_logic_vector(7 downto 0)); end FROM_a; architecture DEF_ARCH of FROM_a is component UFROM generic (MEMORYFILE:string); port(DO0, DO1, DO2, DO3, DO4, DO5, DO6, DO7 : out std_logic; ADDR0, ADDR1, ADDR2, ADDR3, ADDR4, ADDR5, ADDR6 : in std_logic := 'U') ; end component; component GND port( Y : out std_logic); end component; signal U_7_PIN2 : std_logic ; begin GND_1_net : GND port map(Y => U_7_PIN2); UFROM0 : UFROM generic map(MEMORYFILE => "FROM_a.mem") port map(DO0 => DOUT(0), DO1 => DOUT(1), DO2 => DOUT(2), DO3 => DOUT(3), DO4 => DOUT(4), DO5 => DOUT(5), DO6 => DOUT(6), DO7 => DOUT(7), ADDR0 => ADDR(0), ADDR1 => ADDR(1), ADDR2 => ADDR(2), ADDR3 => ADDR(3), ADDR4 => ADDR(4), ADDR5 => ADDR(5), ADDR6 => ADDR(6)); end DEF_ARCH; SmartGen generates the following files along with the netlist. These are located in the SmartGen folder for the Libero SoC project. 1. MEM (Memory Initialization) file 2. UFC (User Flash Configuration) file 3. Log file The MEM file is used for simulation, as explained in the "Simulation of FlashROM Design" section on page 143. The UFC file, generated by SmartGen, has the FlashROM configuration for single or multiple devices and is used during STAPL generation. It contains the region properties and simulation values. Note that any changes in the MEM file will not be reflected in the UFC file. Do not modify the UFC to change FlashROM content. Instead, use the SmartGen GUI to modify the FlashROM content. See the "Programming File Generation for FlashROM Design" section on page 143 for a description of how the UFC file is used during the programming file generation. The log file has information regarding the file type and file location. 142 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Simulation of FlashROM Design The MEM file has 128 rows of 8 bits, each representing the contents of the FlashROM used for simulation. For example, the first row represents page 0, byte 0; the next row is page 0, byte 1; and so the pattern continues. Note that the three MSBs of the address define the page number, and the four LSBs define the byte number. So, if you send address 0000100 to FlashROM, this corresponds to the page 0 and byte 4 location, which is the fifth row in the MEM file. SmartGen defaults to 0s for any unspecified location of the FlashROM. Besides using the MEM file generated by SmartGen, you can create a binary file with 128 rows of 8 bits each and use this as a MEM file. Microsemi recommends that you use different file names if you plan to generate multiple MEM files. During simulation, Libero SoC passes the MEM file used as the generic file in the netlist, along with the design files and testbench. If you want to use different MEM files during simulation, you need to modify the generic file reference in the netlist. ………………… UFROM0: UFROM --generic map(MEMORYFILE => "F:\Appsnotes\FROM\test_designs\testa\smartgen\FROM_a.mem") --generic map(MEMORYFILE => "F:\Appsnotes\FROM\test_designs\testa\smartgen\FROM_b.mem") ……………………. The VITAL and Verilog simulation models accept the generics passed by the netlist, read the MEM file, and perform simulation with the data in the file. Programming File Generation for FlashROM Design FlashPoint is the programming software used to generate the programming files for flash devices. Depending on the applications, you can use the FlashPoint software to generate a STAPL file with different FlashROM contents. In each case, optional AES decryption is available. To generate a STAPL file that contains the same FPGA core content and different FlashROM contents, the FlashPoint software needs an Array Map file for the core and UFC file(s) for the FlashROM. This final STAPL file represents the combination of the logic of the FPGA core and FlashROM content. FlashPoint generates the STAPL files you can use to program the desired FlashROM page and/or FPGA core of the FPGA device contents. FlashPoint supports the encryption of the FlashROM content and/or FPGA Array configuration data. In the case of using the FlashROM for device serialization, a sequence of unique FlashROM contents will be generated. When generating a programming file with multiple unique FlashROM contents, you can specify in FlashPoint whether to include all FlashROM content in a single STAPL file or generate a different STAPL file for each FlashROM (Figure 5-11). The programming software (FlashPro) handles the single STAPL file that contains the FlashROM content from multiple devices. It enables you to program the FlashROM content into a series of devices sequentially (Figure 5-11). See the FlashPro User’s Guide for information on serial programming. UFC File for Single FlashROM Contents FPGA Array Map File UFC File for Multiple FlashROM Contents FPGA Array Map File FlashPoint Security Settings FlashPoint Security Settings Single STAPL File Single STAPL File Single STAPL File Figure 5-11 • Single or Multiple Programming File Generation Revision 4 143 FlashROM in Microsemi’s Low Power Flash Devices Figure 5-12 shows the programming file generator, which enables different STAPL file generation methods. When you select Program FlashROM and choose the UFC file, the FlashROM Settings window appears, as shown in Figure 5-13. In this window, you can select the FlashROM page you want to program and the data value for the configured regions. This enables you to use a different page for different programming files. Figure 5-12 • Programming File Generator Figure 5-13 • Setting FlashROM during Programming File Generation The programming hardware and software can load the FlashROM with the appropriate STAPL file. Programming software handles the single STAPL file that contains multiple FlashROM contents for multiple devices, and programs the FlashROM in sequential order (e.g., for device serialization). This feature is supported in the programming software. After programming with the STAPL file, you can run DEVICE_INFO to check the FlashROM content. 144 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide DEVICE_INFO displays the FlashROM content, serial number, Design Name, and checksum, as shown below: EXPORT IDCODE[32] = 123261CF EXPORT SILSIG[32] = 00000000 User information : CHECKSUM: 61A0 Design Name: TOP Programming Method: STAPL Algorithm Version: 1 Programmer: UNKNOWN ========================================= FlashROM Information : EXPORT Region_7_0[128] = FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF ========================================= Security Setting : Encrypted FlashROM Programming Enabled. Encrypted FPGA Array Programming Enabled. ========================================= The Libero SoC file manager recognizes the UFC and MEM files and displays them in the appropriate view. Libero SoC also recognizes the multiple programming files if you choose the option to generate multiple files for multiple FlashROM contents in Designer. These features enable a user-friendly flow for the FlashROM generation and programming in Libero SoC. Custom Serialization Using FlashROM You can use FlashROM for device serialization or inventory control by using the Auto Inc region or Read From File region. FlashPoint will automatically generate the serial number sequence for the Auto Inc region with the Start Value, Max Value, and Step Value provided. If you have a unique serial number generation scheme that you prefer, the Read From File region allows you to import the file with your serial number scheme programmed into the region. See the FlashPro User's Guide for custom serialization file format information. The following steps describe how to perform device serialization or inventory control using FlashROM: 1. Generate FlashROM using SmartGen. From the Properties section in the FlashROM Settings dialog box, select the Auto Inc or Read From File region. For the Auto Inc region, specify the desired step value. You will not be able to modify this value in the FlashPoint software. 2. Go through the regular design flow and finish place-and-route. 3. Select Programming File in Designer and open Generate Programming File (Figure 5-12 on page 144). 4. Click Program FlashROM, browse to the UFC file, and click Next. The FlashROM Settings window appears, as shown in Figure 5-13 on page 144. 5. Select the FlashROM page you want to program and the data value for the configured regions. The STAPL file generated will contain only the data that targets the selected FlashROM page. 6. Modify properties for the serialization. – For the Auto Inc region, specify the Start and Max values. – For the Read From File region, select the file name of the custom serialization file. 7. Select the FlashROM programming file type you want to generate from the two options below: – Single programming file for all devices: generates one programming file with all FlashROM values. – One programming file per device: generates a separate programming file for each FlashROM value. 8. Enter the number of devices you want to program and generate the required programming file. 9. Open the programming software and load the programming file. The programming software, FlashPro3 and Silicon Sculptor II, supports the device serialization feature. If, for some reason, the device fails to program a part during serialization, the software allows you to reuse or skip the serial data. Refer to the FlashPro User’s Guide for details. Revision 4 145 FlashROM in Microsemi’s Low Power Flash Devices Conclusion The Fusion, IGLOO, and ProASIC3 families are the only FPGAs that offer on-chip FlashROM support. This document presents information on the FlashROM architecture, possible applications, programming, access through the JTAG and UJTAG interface, and integration into your design. In addition, the Libero tool set enables easy creation and modification of the FlashROM content. The nonvolatile FlashROM block in the FPGA can be customized, enabling multiple applications. Additionally, the security offered by the low power flash devices keeps both the contents of FlashROM and the FPGA design safe from system over-builders, system cloners, and IP thieves. Related Documents User’s Guides FlashPro User’s Guide http://www.microsemi.com/documents/FlashPro_UG.pdf List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 5-1 • Flash-Based FPGAs. 134 v1.3 (October 2008) The "FlashROM Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 134 Figure 5-2 • Fusion Device Architecture Overview (AFS600) was replaced. 135, 137 Figure 5-5 • Programming FlashROM Using AES was revised to change "Fusion" to "Flash Device." v1.2 (June 2008) v1.1 (March 2008) 146 The FlashPoint User’s Guide was removed from the "User’s Guides" section, as its content is now part of the FlashPro User’s Guide. 146 The following changes were made to the family descriptions in Table 5-1 • FlashBased FPGAs: 134 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The chapter was updated to include the IGLOO PLUS family and information regarding 15 k gate devices. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. R e vi s i o n 4 N/A 6 – SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Introduction As design complexity grows, greater demands are placed upon an FPGA's embedded memory. Fusion, IGLOO, and ProASIC3 devices provide the flexibility of true dual-port and two-port SRAM blocks. The embedded memory, along with built-in, dedicated FIFO control logic, can be used to create cascading RAM blocks and FIFOs without using additional logic gates. IGLOO, IGLOO PLUS, and ProASIC3L FPGAs contain an additional feature that allows the device to be put in a low power mode called Flash*Freeze. In this mode, the core draws minimal power (on the order of 2 to 127 µW) and still retains values on the embedded SRAM/FIFO and registers. Flash*Freeze technology allows the user to switch to Active mode on demand, thus simplifying power management and the use of SRAM/FIFOs. Device Architecture The low power flash devices feature up to 504 kbits of RAM in 4,608-bit blocks (Figure 6-1 on page 148 and Figure 6-2 on page 149). The total embedded SRAM for each device can be found in the datasheets. These memory blocks are arranged along the top and bottom of the device to allow better access from the core and I/O (in some devices, they are only available on the north side of the device). Every RAM block has a flexible, hardwired, embedded FIFO controller, enabling the user to implement efficient FIFOs without sacrificing user gates. In the IGLOO and ProASIC3 families of devices, the following memories are supported: • 30 k gate devices and smaller do not support SRAM and FIFO. • 60 k and 125 k gate devices support memories on the north side of the device only. • 250 k devices and larger support memories on the north and south sides of the device. In Fusion devices, the following memories are supported: • AFS090 and AFS250 support memories on the north side of the device only. • AFS600 and AFS1500 support memories on the north and south sides of the device. Revision 4 147 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Bank 0 Bank 1 Bank 3 CCC RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block I/Os Bank 1 Bank 3 VersaTile ISP AES Decryption 1 User Nonvolatile FlashRom Flash*Freeze Technology 2 Charge Pumps Bank 2 Notes: 1. AES decryption not supported in 30 k gate devices and smaller. 2. Flash*Freeze is supported in all IGLOO devices and the ProASIC3L devices. Figure 6-1 • IGLOO and ProASIC3 Device Architecture Overview 148 R e vi s i o n 4 RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block ProASIC3L FPGA Fabric User’s Guide Bank 1 Bank 0 CCC/PLL RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block OSC I/Os CCC Bank 2 Bank 4 VersaTile ISP AES Decryption User Nonvolatile FlashROM (FROM) Flash Array Analog Quad Analog Quad Analog Quad Charge Pumps ADC Analog Quad Analog Quad Analog Quad RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Flash Array Analog Quad Analog Quad Analog Quad Analog Quad Bank 3 Figure 6-2 • Fusion Device Architecture Overview (AFS600) Revision 4 149 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices SRAM/FIFO Support in Flash-Based Devices The flash FPGAs listed in Table 6-1 support SRAM and FIFO blocks and the functions described in this document. Table 6-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 6-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 6-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 150 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide SRAM and FIFO Architecture To meet the needs of high-performance designs, the memory blocks operate strictly in synchronous mode for both read and write operations. The read and write clocks are completely independent, and each can operate at any desired frequency up to 250 MHz. • 4k×1, 2k×2, 1k×4, 512×9 (dual-port RAM—2 read / 2 write or 1 read / 1 write) • 512×9, 256×18 (2-port RAM—1 read / 1 write) • Sync write, sync pipelined / nonpipelined read Automotive ProASIC3 devices support single-port SRAM capabilities or dual-port SRAM only under specific conditions. Dual-port mode is supported if the clocks to the two SRAM ports are the same and 180° out of phase (i.e., the port A clock is the inverse of the port B clock). The Libero SoC software macro libraries support a dual-port macro only. For use of this macro as a single-port SRAM, the inputs and clock of one port should be tied off (grounded) to prevent errors during design compile. For use in dual-port mode, the same clock with an inversion between the two clock pins of the macro should be used in the design to prevent errors during compile. The memory block includes dedicated FIFO control logic to generate internal addresses and external flag logic (FULL, EMPTY, AFULL, AEMPTY). Simultaneous dual-port read/write and write/write operations at the same address are allowed when certain timing requirements are met. During RAM operation, addresses are sourced by the user logic, and the FIFO controller is ignored. In FIFO mode, the internal addresses are generated by the FIFO controller and routed to the RAM array by internal MUXes. The low power flash device architecture enables the read and write sizes of RAMs to be organized independently, allowing for bus conversion. For example, the write size can be set to 256×18 and the read size to 512×9. Both the write width and read width for the RAM blocks can be specified independently with the WW (write width) and RW (read width) pins. The different D×W configurations are 256×18, 512×9, 1k×4, 2k×2, and 4k×1. When widths of one, two, or four are selected, the ninth bit is unused. For example, when writing nine-bit values and reading four-bit values, only the first four bits and the second four bits of each nine-bit value are addressable for read operations. The ninth bit is not accessible. Conversely, when writing four-bit values and reading nine-bit values, the ninth bit of a read operation will be undefined. The RAM blocks employ little-endian byte order for read and write operations. Memory Blocks and Macros Memory blocks can be configured with many different aspect ratios, but are generically supported in the macro libraries as one of two memory elements: RAM4K9 or RAM512X18. The RAM4K9 is configured as a true dual-port memory block, and the RAM512X18 is configured as a two-port memory block. Dualport memory allows the RAM to both read from and write to either port independently. Two-port memory allows the RAM to read from one port and write to the other using a common clock or independent read and write clocks. If needed, the RAM4K9 blocks can be configured as two-port memory blocks. The memory block can be configured as a FIFO by combining the basic memory block with dedicated FIFO controller logic. The FIFO macro is named FIFO4KX18 (Figure 6-3 on page 152). Clocks for the RAM blocks can be driven by the VersaNet (global resources) or by regular nets. When using local clock segments, the clock segment region that encompasses the RAM blocks can drive the RAMs. In the dual-port configuration (RAM4K9), each memory block port can be driven by either risingedge or falling-edge clocks. Each port can be driven by clocks with different edges. Though only a risingedge clock can drive the physical block itself, the Microsemi Designer software will automatically bubblepush the inversion to properly implement the falling-edge trigger for the RAM block. Revision 4 151 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices RAM512x18 RAM4K9 ADDRA11 ADDRA10 DOUTA8 DOUTA7 RADDR8 RADDR7 RD17 RD16 ADDRA0 DINA8 DINA7 DOUTA0 RADDR0 RD0 RW1 RW0 DINA0 WIDTHA1 WIDTHA0 PIPEA WMODEA BLKA WENA CLKA ADDRB0 DOUTB0 DINB8 DINB7 WIDTHB1 WIDTHB0 PIPEB WMODEB BLKB WENB CLKB RESET RD0 FULL AFULL EMPTY AEMPTY AFVAL11 AFVAL10 WADDR8 WADDR7 WADDR0 WD17 WD16 WD0 DINB0 RD17 RD16 AEVAL0 REN RCLK DOUTB8 DOUTB7 RW2 RW1 RW0 WW2 WW1 WW0 ESTOP FSTOP AEVAL11 AEVAL10 PIPE ADDRB11 ADDRB10 FIFO4K18 AFVAL0 REN RBLK RCLK WD17 WD16 WD0 WW1 WW0 WEN WBLK WCLK RPIPE WEN WCLK RESET RESET Notes: 1. Automotive ProASIC3 devices restrict RAM4K9 to a single port or to dual ports with the same clock 180° out of phase (inverted) between clock pins. In single-port mode, inputs to port B should be tied to ground to prevent errors during compile. This warning applies only to automotive ProASIC3 parts of certain revisions and earlier. Contact Technical Support at soc_tech@microsemi.com for information on the revision number for a particular lot and date code. 2. For FIFO4K18, the same clock 180° out of phase (inverted) between clock pins should be used. Figure 6-3 • Supported Basic RAM Macros 152 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide SRAM Features RAM4K9 Macro RAM4K9 is the dual-port configuration of the RAM block (Figure 6-4). The RAM4K9 nomenclature refers to both the deepest possible configuration and the widest possible configuration the dual-port RAM block can assume, and does not denote a possible memory aspect ratio. The RAM block can be configured to the following aspect ratios: 4,096×1, 2,048×2, 1,024×4, and 512×9. RAM4K9 is fully synchronous and has the following features: • Two ports that allow fully independent reads and writes at different frequencies • Selectable pipelined or nonpipelined read • Active-low block enables for each port • Toggle control between read and write mode for each port • Active-low asynchronous reset • Pass-through write data or hold existing data on output. In pass-through mode, the data written to the write port will immediately appear on the read port. • Designer software will automatically facilitate falling-edge clocks by bubble-pushing the inversion to previous stages. DINA DOUTA ADDRA BLKA WENA CLKA Write Data Write Data Read Data Read Data Address BLK Address RAM4K9 BLK WEN WEN CLK CLK DINB DOUTB ADDRB BLKB WENB CLKB Reset Note: For timing diagrams of the RAM signals, refer to the appropriate family datasheet. Figure 6-4 • RAM4K9 Simplified Configuration Signal Descriptions for RAM4K9 Note: Automotive ProASIC3 devices support single-port SRAM capabilities, or dual-port SRAM only under specific conditions. Dual-port mode is supported if the clocks to the two SRAM ports are the same and 180° out of phase (i.e., the port A clock is the inverse of the port B clock). Since Libero SoC macro libraries support a dual-port macro only, certain modifications must be made. These are detailed below. The following signals are used to configure the RAM4K9 memory element: WIDTHA and WIDTHB These signals enable the RAM to be configured in one of four allowable aspect ratios (Table 6-2 on page 154). Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, WIDTHB should be tied to ground. Revision 4 153 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Table 6-2 • Allowable Aspect Ratio Settings for WIDTHA[1:0] WIDTHA[1:0] WIDTHB[1:0] D×W 00 00 4k×1 01 01 2k×2 10 10 1k×4 11 11 512×9 Note: The aspect ratio settings are constant and cannot be changed on the fly. BLKA and BLKB These signals are active-low and will enable the respective ports when asserted. When a BLKx signal is deasserted, that port’s outputs hold the previous value. Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, BLKB should be tied to ground. WENA and WENB These signals switch the RAM between read and write modes for the respective ports. A LOW on these signals indicates a write operation, and a HIGH indicates a read. Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, WENB should be tied to ground. CLKA and CLKB These are the clock signals for the synchronous read and write operations. These can be driven independently or with the same driver. Note: For Automotive ProASIC3 devices, dual-port mode is supported if the clocks to the two SRAM ports are the same and 180° out of phase (i.e., the port A clock is the inverse of the port B clock). For use of this macro as a single-port SRAM, the inputs and clock of one port should be tied off (grounded) to prevent errors during design compile. PIPEA and PIPEB These signals are used to specify pipelined read on the output. A LOW on PIPEA or PIPEB indicates a nonpipelined read, and the data appears on the corresponding output in the same clock cycle. A HIGH indicates a pipelined read, and data appears on the corresponding output in the next clock cycle. Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, PIPEB should be tied to ground. For use in dual-port mode, the same clock with an inversion between the two clock pins of the macro should be used in the design to prevent errors during compile. WMODEA and WMODEB These signals are used to configure the behavior of the output when the RAM is in write mode. A LOW on these signals makes the output retain data from the previous read. A HIGH indicates pass-through behavior, wherein the data being written will appear immediately on the output. This signal is overridden when the RAM is being read. Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, WMODEB should be tied to ground. RESET This active-low signal resets the control logic, forces the output hold state registers to zero, disables reads and writes from the SRAM block, and clears the data hold registers when asserted. It does not reset the contents of the memory array. While the RESET signal is active, read and write operations are disabled. As with any asynchronous reset signal, care must be taken not to assert it too close to the edges of active read and write clocks. ADDRA and ADDRB These are used as read or write addresses, and they are 12 bits wide. When a depth of less than 4 k is specified, the unused high-order bits must be grounded (Table 6-3 on page 155). 154 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, ADDRB should be tied to ground. Table 6-3 • Address Pins Unused/Used for Various Supported Bus Widths ADDRx D×W Unused Used 4k×1 None [11:0] 2k×2 [11] [10:0] 1k×4 [11:10] [9:0] 512×9 [11:9] [8:0] Note: The "x" in ADDRx implies A or B. DINA and DINB These are the input data signals, and they are nine bits wide. Not all nine bits are valid in all configurations. When a data width less than nine is specified, unused high-order signals must be grounded (Table 6-4). Note: When using the SRAM in single-port mode for Automotive ProASIC3 devices, DINB should be tied to ground. DOUTA and DOUTB These are the nine-bit output data signals. Not all nine bits are valid in all configurations. As with DINA and DINB, high-order bits may not be used (Table 6-4). The output data on unused pins is undefined. Table 6-4 • Unused/Used Input and Output Data Pins for Various Supported Bus Widths DINx/DOUTx D×W Unused Used 4k×1 [8:1] [0] 2k×2 [8:2] [1:0] 1k×4 [8:4] [3:0] 512×9 None [8:0] Note: The "x" in DINx or DOUTx implies A or B. RAM512X18 Macro RAM512X18 is the two-port configuration of the same RAM block (Figure 6-5 on page 156). Like the RAM4K9 nomenclature, the RAM512X18 nomenclature refers to both the deepest possible configuration and the widest possible configuration the two-port RAM block can assume. In two-port mode, the RAM block can be configured to either the 512×9 aspect ratio or the 256×18 aspect ratio. RAM512X18 is also fully synchronous and has the following features: • Dedicated read and write ports • Active-low read and write enables • Selectable pipelined or nonpipelined read • Active-low asynchronous reset • Designer software will automatically facilitate falling-edge clocks by bubble-pushing the inversion to previous stages. Revision 4 155 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Write Data WD WADDR Read Data Write Address WEN WCLK Read Address RD RADDR RAM512X18 Write Enable Read Enable Write CLK Read CLK REN RCLK Reset Note: For timing diagrams of the RAM signals, refer to the appropriate family datasheet. Figure 6-5 • 512X18 Two-Port RAM Block Diagram Signal Descriptions for RAM512X18 RAM512X18 has slightly different behavior from RAM4K9, as it has dedicated read and write ports. WW and RW These signals enable the RAM to be configured in one of the two allowable aspect ratios (Table 6-5). Table 6-5 • Aspect Ratio Settings for WW[1:0] WW[1:0] RW[1:0] D×W 01 01 512×9 10 10 256×18 00, 11 Reserved 00, 11 WD and RD These are the input and output data signals, and they are 18 bits wide. When a 512×9 aspect ratio is used for write, WD[17:9] are unused and must be grounded. If this aspect ratio is used for read, RD[17:9] are undefined. WADDR and RADDR These are read and write addresses, and they are nine bits wide. When the 256×18 aspect ratio is used for write or read, WADDR[8] and RADDR[8] are unused and must be grounded. WCLK and RCLK These signals are the write and read clocks, respectively. They can be clocked on the rising or falling edge of WCLK and RCLK. WEN and REN These signals are the write and read enables, respectively. They are both active-low by default. These signals can be configured as active-high. RESET This active-low signal resets the control logic, forces the output hold state registers to zero, disables reads and writes from the SRAM block, and clears the data hold registers when asserted. It does not reset the contents of the memory array. While the RESET signal is active, read and write operations are disabled. As with any asynchronous reset signal, care must be taken not to assert it too close to the edges of active read and write clocks. PIPE This signal is used to specify pipelined read on the output. A LOW on PIPE indicates a nonpipelined read, and the data appears on the output in the same clock cycle. A HIGH indicates a pipelined read, and data appears on the output in the next clock cycle. 156 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide SRAM Usage The following descriptions refer to the usage of both RAM4K9 and RAM512X18. Clocking The dual-port SRAM blocks are only clocked on the rising edge. SmartGen allows falling-edge-triggered clocks by adding inverters to the netlist, hence achieving dual-port SRAM blocks that are clocked on either edge (rising or falling). For dual-port SRAM, each port can be clocked on either edge and by separate clocks by port. Note that for Automotive ProASIC3, the same clock, with an inversion between the two clock pins of the macro, should be used in design to prevent errors during compile. Low power flash devices support inversion (bubble-pushing) throughout the FPGA architecture, including the clock input to the SRAM modules. Inversions added to the SRAM clock pin on the design schematic or in the HDL code will be automatically accounted for during design compile without incurring additional delay in the clock path. The two-port SRAM can be clocked on the rising or falling edge of WCLK and RCLK. If negative-edge RAM and FIFO clocking is selected for memory macros, clock edge inversion management (bubble-pushing) is automatically used within the development tools, without performance penalty. Modes of Operation There are two read modes and one write mode: • Read Nonpipelined (synchronous—1 clock edge): In the standard read mode, new data is driven onto the RD bus in the same clock cycle following RA and REN valid. The read address is registered on the read port clock active edge, and data appears at RD after the RAM access time. Setting PIPE to OFF enables this mode. • Read Pipelined (synchronous—2 clock edges): The pipelined mode incurs an additional clock delay from address to data but enables operation at a much higher frequency. The read address is registered on the read port active clock edge, and the read data is registered and appears at RD after the second read clock edge. Setting PIPE to ON enables this mode. • Write (synchronous—1 clock edge): On the write clock active edge, the write data is written into the SRAM at the write address when WEN is HIGH. The setup times of the write address, write enables, and write data are minimal with respect to the write clock. RAM Initialization Each SRAM block can be individually initialized on power-up by means of the JTAG port using the UJTAG mechanism. The shift register for a target block can be selected and loaded with the proper bit configuration to enable serial loading. The 4,608 bits of data can be loaded in a single operation. FIFO Features The FIFO4KX18 macro is created by merging the RAM block with dedicated FIFO logic (Figure 6-6 on page 158). Since the FIFO logic can only be used in conjunction with the memory block, there is no separate FIFO controller macro. As with the RAM blocks, the FIFO4KX18 nomenclature does not refer to a possible aspect ratio, but rather to the deepest possible data depth and the widest possible data width. FIFO4KX18 can be configured into the following aspect ratios: 4,096×1, 2,048×2, 1,024×4, 512×9, and 256×18. In addition to being fully synchronous, the FIFO4KX18 also has the following features: • Four FIFO flags: Empty, Full, Almost-Empty, and Almost-Full • Empty flag is synchronized to the read clock • Full flag is synchronized to the write clock • Both Almost-Empty and Almost-Full flags have programmable thresholds • Active-low asynchronous reset • Active-low block enable • Active-low write enable • Active-high read enable • Ability to configure the FIFO to either stop counting after the empty or full states are reached or to allow the FIFO counters to continue Revision 4 157 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices • Designer software will automatically facilitate falling-edge clocks by bubble-pushing the inversion to previous stages. Write Data WD Full Flag FULL AFULL Read Data Almost-Full Flag FIFO4KX18 AEMPTY Read Enable Write Clock WCLK EMPTY Almost-Empty Flag Write Enable WEN RD Empty Flag REN Read Clock RCLK Reset Figure 6-6 • FIFO4KX18 Block Diagram RD[17:0] RCLK WCLK RCLK WCLK RADD[J:0] WADD[J:0] FREN RBLK REN REN WEN FWEN RD RAM RW[2:0] WW[2:0] WD[17:0] RPIPE WD CNT 12 E = ESTOP AFVAL FULL AFULL AEMPTY WBLK WEN CNT 12 SUB 12 AEVAL E = FSTOP EMPTY Reset Figure 6-7 • RAM Block with Embedded FIFO Controller The FIFOs maintain a separate read and write address. Whenever the difference between the write address and the read address is greater than or equal to the almost-full value (AFVAL), the Almost-Full flag is asserted. Similarly, the Almost-Empty flag is asserted whenever the difference between the write address and read address is less than or equal to the almost-empty value (AEVAL). Due to synchronization between the read and write clocks, the Empty flag will deassert after the second read clock edge from the point that the write enable asserts. However, since the Empty flag is synchronized to the read clock, it will assert after the read clock reads the last data in the FIFO. Also, since the Full flag is dependent on the actual hardware configuration, it will assert when the actual physical implementation of the FIFO is full. For example, when a user configures a 128×18 FIFO, the actual physical implementation will be a 256×18 FIFO element. Since the actual implementation is 256×18, the Full flag will not trigger until the 158 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 256×18 FIFO is full, even though a 128×18 FIFO was requested. For this example, the Almost-Full flag can be used instead of the Full flag to signal when the 128th data word is reached. To accommodate different aspect ratios, the almost-full and almost-empty values are expressed in terms of data bits instead of data words. SmartGen translates the user’s input, expressed in data words, into data bits internally. SmartGen allows the user to select the thresholds for the Almost-Empty and AlmostFull flags in terms of either the read data words or the write data words, and makes the appropriate conversions for each flag. After the empty or full states are reached, the FIFO can be configured so the FIFO counters either stop or continue counting. For timing numbers, refer to the appropriate family datasheet. Signal Descriptions for FIFO4K18 The following signals are used to configure the FIFO4K18 memory element: WW and RW These signals enable the FIFO to be configured in one of the five allowable aspect ratios (Table 6-6). Table 6-6 • Aspect Ratio Settings for WW[2:0] WW[2:0] RW[2:0] D×W 000 000 4k×1 001 001 2k×2 010 010 1k×4 011 011 512×9 100 100 256×18 101, 110, 111 Reserved 101, 110, 111 WBLK and RBLK These signals are active-low and will enable the respective ports when LOW. When the RBLK signal is HIGH, that port’s outputs hold the previous value. WEN and REN Read and write enables. WEN is active-low and REN is active-high by default. These signals can be configured as active-high or -low. WCLK and RCLK These are the clock signals for the synchronous read and write operations. These can be driven independently or with the same driver. Note: For the Automotive ProASIC3 FIFO4K18, for the same clock, 180° out of phase (inverted) between clock pins should be used. RPIPE This signal is used to specify pipelined read on the output. A LOW on RPIPE indicates a nonpipelined read, and the data appears on the output in the same clock cycle. A HIGH indicates a pipelined read, and data appears on the output in the next clock cycle. RESET This active-low signal resets the control logic and forces the output hold state registers to zero when asserted. It does not reset the contents of the memory array (Table 6-7 on page 160). While the RESET signal is active, read and write operations are disabled. As with any asynchronous RESET signal, care must be taken not to assert it too close to the edges of active read and write clocks. WD This is the input data bus and is 18 bits wide. Not all 18 bits are valid in all configurations. When a data width less than 18 is specified, unused higher-order signals must be grounded (Table 6-7 on page 160). Revision 4 159 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices RD This is the output data bus and is 18 bits wide. Not all 18 bits are valid in all configurations. Like the WD bus, high-order bits become unusable if the data width is less than 18. The output data on unused pins is undefined (Table 6-7). Table 6-7 • Input Data Signal Usage for Different Aspect Ratios D×W WD/RD Unused 4k×1 WD[17:1], RD[17:1] 2k×2 WD[17:2], RD[17:2] 1k×4 WD[17:4], RD[17:4] 512×9 WD[17:9], RD[17:9] 256×18 – ESTOP, FSTOP ESTOP is used to stop the FIFO read counter from further counting once the FIFO is empty (i.e., the EMPTY flag goes HIGH). A HIGH on this signal inhibits the counting. FSTOP is used to stop the FIFO write counter from further counting once the FIFO is full (i.e., the FULL flag goes HIGH). A HIGH on this signal inhibits the counting. For more information on these signals, refer to the "ESTOP and FSTOP Usage" section. FULL, EMPTY When the FIFO is full and no more data can be written, the FULL flag asserts HIGH. The FULL flag is synchronous to WCLK to inhibit writing immediately upon detection of a full condition and to prevent overflows. Since the write address is compared to a resynchronized (and thus time-delayed) version of the read address, the FULL flag will remain asserted until two WCLK active edges after a read operation eliminates the full condition. When the FIFO is empty and no more data can be read, the EMPTY flag asserts HIGH. The EMPTY flag is synchronous to RCLK to inhibit reading immediately upon detection of an empty condition and to prevent underflows. Since the read address is compared to a resynchronized (and thus time-delayed) version of the write address, the EMPTY flag will remain asserted until two RCLK active edges after a write operation removes the empty condition. For more information on these signals, refer to the "FIFO Flag Usage Considerations" section on page 161. AFULL, AEMPTY These are programmable flags and will be asserted on the threshold specified by AFVAL and AEVAL, respectively. When the number of words stored in the FIFO reaches the amount specified by AEVAL while reading, the AEMPTY output will go HIGH. Likewise, when the number of words stored in the FIFO reaches the amount specified by AFVAL while writing, the AFULL output will go HIGH. AFVAL, AEVAL The AEVAL and AFVAL pins are used to specify the almost-empty and almost-full threshold values. They are 12-bit signals. For more information on these signals, refer to the "FIFO Flag Usage Considerations" section on page 161. FIFO Usage ESTOP and FSTOP Usage The ESTOP pin is used to stop the read counter from counting any further once the FIFO is empty (i.e., the EMPTY flag goes HIGH). Likewise, the FSTOP pin is used to stop the write counter from counting any further once the FIFO is full (i.e., the FULL flag goes HIGH). The FIFO counters in the device start the count at zero, reach the maximum depth for the configuration (e.g., 511 for a 512×9 configuration), and then restart at zero. An example application for ESTOP, where the read counter keeps counting, would be writing to the FIFO once and reading the same content over and over without doing another write. 160 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide FIFO Flag Usage Considerations The AEVAL and AFVAL pins are used to specify the 12-bit AEMPTY and AFULL threshold values. The FIFO contains separate 12-bit write address (WADDR) and read address (RADDR) counters. WADDR is incremented every time a write operation is performed, and RADDR is incremented every time a read operation is performed. Whenever the difference between WADDR and RADDR is greater than or equal to AFVAL, the AFULL output is asserted. Likewise, whenever the difference between WADDR and RADDR is less than or equal to AEVAL, the AEMPTY output is asserted. To handle different read and write aspect ratios, AFVAL and AEVAL are expressed in terms of total data bits instead of total data words. When users specify AFVAL and AEVAL in terms of read or write words, the SmartGen tool translates them into bit addresses and configures these signals automatically. SmartGen configures the AFULL flag to assert when the write address exceeds the read address by at least a predefined value. In a 2k×8 FIFO, for example, a value of 1,500 for AFVAL means that the AFULL flag will be asserted after a write when the difference between the write address and the read address reaches 1,500 (there have been at least 1,500 more writes than reads). It will stay asserted until the difference between the write and read addresses drops below 1,500. The AEMPTY flag is asserted when the difference between the write address and the read address is less than a predefined value. In the example above, a value of 200 for AEVAL means that the AEMPTY flag will be asserted when a read causes the difference between the write address and the read address to drop to 200. It will stay asserted until that difference rises above 200. Note that the FIFO can be configured with different read and write widths; in this case, the AFVAL setting is based on the number of write data entries, and the AEVAL setting is based on the number of read data entries. For aspect ratios of 512×9 and 256×18, only 4,096 bits can be addressed by the 12 bits of AFVAL and AEVAL. The number of words must be multiplied by 8 and 16 instead of 9 and 18. The SmartGen tool automatically uses the proper values. To avoid halfwords being written or read, which could happen if different read and write aspect ratios were specified, the FIFO will assert FULL or EMPTY as soon as at least one word cannot be written or read. For example, if a two-bit word is written and a four-bit word is being read, the FIFO will remain in the empty state when the first word is written. This occurs even if the FIFO is not completely empty, because in this case, a complete word cannot be read. The same is applicable in the full state. If a four-bit word is written and a two-bit word is read, the FIFO is full and one word is read. The FULL flag will remain asserted because a complete word cannot be written at this point. Variable Aspect Ratio and Cascading Variable aspect ratio and cascading allow users to configure the memory in the width and depth required. The memory block can be configured as a FIFO by combining the basic memory block with dedicated FIFO controller logic. The FIFO macro is named FIFO4KX18. Low power flash device RAM can be configured as 1, 2, 4, 9, or 18 bits wide. By cascading the memory blocks, any multiple of those widths can be created. The RAM blocks can be from 256 to 4,096 bits deep, depending on the aspect ratio, and the blocks can also be cascaded to create deeper areas. Refer to the aspect ratios available for each macro cell in the "SRAM Features" section on page 153. The largest continuous configurable memory area is equal to half the total memory available on the device, because the RAM is separated into two groups, one on each side of the device. The SmartGen core generator will automatically configure and cascade both RAM and FIFO blocks. Cascading is accomplished using dedicated memory logic and does not consume user gates for depths up to 4,096 bits deep and widths up to 18, depending on the configuration. Deeper memory will utilize some user gates to multiplex the outputs. Generated RAM and FIFO macros can be created as either structural VHDL or Verilog for easy instantiation into the design. Users of Libero SoC can create a symbol for the macro and incorporate it into a design schematic. Table 6-10 on page 163 shows the number of memory blocks required for each of the supported depth and width memory configurations, and for each depth and width combination. For example, a 256-bit deep by 32-bit wide two-port RAM would consist of two 256×18 RAM blocks. The first 18 bits would be stored in the first RAM block, and the remaining 14 bits would be implemented in the other 256×18 RAM block. This second RAM block would have four bits of unused storage. Similarly, a dual-port memory block that is 8,192 bits deep and 8 bits wide would be implemented using 16 memory blocks. The dualport memory would be configured in a 4,096×1 aspect ratio. These blocks would then be cascaded two deep to achieve 8,192 bits of depth, and eight wide to achieve the eight bits of width. Revision 4 161 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Table 6-8 and Table 6-9 show the maximum potential width and depth configuration for each device. Note that 15 k and 30 k gate devices do not support RAM or FIFO. Table 6-8 • Memory Availability per IGLOO and ProASIC3 Device Maximum Potential Width1 Device IGLOO IGLOO nano IGLOO PLUS RAM ProASIC3 ProASIC3 nano Block s ProASIC3L Maximum Potential Depth2 Depth Width Depth Width AGL060 AGLN060 AGLP060 A3P060 A3PN060 4 256 72 (4×18) 16,384 (4,096×4) 1 AGL125 AGLN125 AGLP125 A3P125 A3PN125 8 256 144 (8×18) 32,768 (4,094×8) 1 AGL250 AGLN250 A3P250/L A3PN250 8 256 144 (8×18) 32,768 (4,096×8) 1 AGL400 A3P400 12 256 216 (12×18) 49,152 (4,096×12) 1 AGL600 A3P600/L 24 256 432 (24×18) 98,304 (4,096×24) 1 AGL1000 A3P1000/L 32 256 576 (32×18) 131,072 (4,096×32) 1 AGLE600 AGLE3000 A3PE600 24 256 432 (24×18) 98,304 (4,096×24) 1 A3PE1500 60 256 1,080 (60×18) 245,760 (4,096×60) 1 A3PE3000/L 112 256 2,016 (112×18) 458,752 (4,096×112) 1 Notes: 1. Maximum potential width uses the two-port configuration. 2. Maximum potential depth uses the dual-port configuration. Table 6-9 • Memory Availability per Fusion Device Maximum Potential Width1 Device RAM Blocks Depth Width Depth Width AFS090 6 256 108 (6×18) 24,576 (4,094×6) 1 AFS250 8 256 144 (8×18) 32,768 (4,094×8) 1 AFS600 24 256 432 (24×18) 98,304 (4,096×24) 1 AFS1500 60 256 1,080 (60×18) 245,760 (4,096×60) 1 Notes: 1. Maximum potential width uses the two-port configuration. 2. Maximum potential depth uses the dual-port configuration. 162 Maximum Potential Depth2 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 6-10 • RAM and FIFO Memory Block Consumption Depth 256 Two-Port 1 2 4 8 9 Width 16 18 32 36 64 72 Note: Dual-Port 512 1,024 2,048 4,096 8,192 16,384 32,768 65,536 Dual-Port Dual-Port Dual-Port Dual-Port Dual-Port Dual-Port Dual-Port Dual-Port Number Block 1 1 1 1 1 1 2 4 8 16 × 1 Configuration Any Any Any 1,024 × 4 2,048 × 2 4,096 × 1 2 × (4,096 × 1) Cascade Deep 4 × (4,096 × 1) Cascade Deep 8 × (4,096 × 1) Cascade Deep 16 × (4,096 × 1) Cascade Deep 4 8 16 32 Number Block 1 1 1 1 1 2 Configuration Any Any Any 1,024×4 2,048 × 2 2 × (4,096 × 1) Cascaded Wide Number Block 1 1 1 1 2 4 Configuration Any Any Any 1,024 × 4 2 × (2,048 × 2) Cascaded Wide 4 × (4,096 × 1) Cascaded Wide Number Block 1 1 1 2 4 8 Configuration Any Any Any 2 × (1,024 × 4) Cascaded Wide 4 × (2,048 × 2) Cascaded Wide 8 × (4,096 × 1) Cascaded Wide 4 × (4,096 × 1) 8 × (4,096 × 1) 16 × (4,096 × 1) 32 × (4,096 × 1) Cascaded 2 Deep Cascaded 4 Deep Cascaded 8 Deep Cascaded 16 and 2 Wide and 2 Wide and 2 Wide Deep and 2 Wide 8 16 16 32 1 1 1 2 4 8 16 32 Configuration Any Any Any 2 × (512 × 9) Cascaded Deep 4 × (512 × 9) Cascaded Deep 8 × (512 × 9) Cascaded Deep 16 × (512 × 9) Cascaded Deep 32 × (512 × 9) Cascaded Deep 16 32 64 1 1 1 4 8 Configuration 256 × 18 256 × 18 256 × 18 4 × (1,024 × 4) Cascaded Wide 8 × (2,048 × 2) Cascaded Wide 4 8 Number Block 1 2 2 Configuration 256 × 8 2 × (512 × 9) Cascaded Wide 2 × (512 × 9) Cascaded Wide 2 4 4 8 16 Configuration 2 × (256 × 18) Cascaded Wide 4 × (512 × 9) Cascaded Wide 4 × (512 × 9) Cascaded Wide 8 × (1,024 × 4) Cascaded Wide 16 × (2,048 × 2) Cascaded Wide 8 16 2 4 4 Configuration 2 × (256 × 18) Cascaded Wide 4 × (512 × 9) Cascaded Wide 4 × (512 × 9) Cascaded Wide 64 16 × (4,096 × 1) 32 × (4,096 × 1) 32 × (4,096 × 1) Cascaded Wide Cascaded 2 Deep Cascaded 4 Deep and 16 Wide and 16 Wide 18 32 4 × (512 × 9) 8 × (512 × 9) 16 × (512 × 9) 16 × (512 × 9) Cascaded 2 Deep Cascaded 4 Deep Cascaded 8 Deep Cascaded 16 and 2 Wide and 2 Wide and 2 Wide Deep and 2 Wide Number Block Number Block 64 16 × (4,096 × 1) 32 × (4,096 × 1) 64 × (4,096 × 1) Cascaded 2 Deep Cascaded 4 Deep Cascaded 8 Deep and 8 Wide and 8 Wide and 8 Wide Number Block Number Block 32 4 × (4,096 × 1) 16 × (4,096 × 1) 32 × (4,096 × 1) 64 × (4,096 × 1) Cascaded 2 Deep Cascaded 4 Deep Cascaded 8 Deep Cascaded 16 and 4 Wide and 4 Wide and 4 Wide Deep and 4 Wide 32 64 32 × (4,096 × 1) 64 × (4,096 × 1) Cascaded Wide Cascaded 2 Deep and 32 Wide 32 4 × (512 × 9) 16 × (512 × 9) 16 × (512 × 9) Cascaded 2 Deep Cascaded 4 Deep Cascaded 8 Deep and 4 Wide and 4 Wide and 4 Wide Number Block 4 8 8 16 32 64 Configuration 4 × (256 × 18) Cascaded Wide 8 × (512 × 9) Cascaded Wide 8 × (512 × 9) Cascaded Wide 16 × (1,024 × 4) Cascaded Wide 32 × (2,048 × 2) Cascaded Wide 64 × (4,096 × 1) Cascaded Wide Number Block 4 8 8 16 32 Configuration 4 × (256 × 18) Cascaded Wide 8 × (512 × 9) Cascaded Wide 8 × (512 × 9) Cascaded Wide 16 × (512 × 9) Cascaded Wide 16 × (512 × 9) Cascaded 4 Deep and 8 Wide Memory configurations represented by grayed cells are not supported. R e visi on 4 163 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Initializing the RAM/FIFO The SRAM blocks can be initialized with data to use as a lookup table (LUT). Data initialization can be accomplished either by loading the data through the design logic or through the UJTAG interface. The UJTAG macro is used to allow access from the JTAG port to the internal logic in the device. By sending the appropriate initialization string to the JTAG Test Access Port (TAP) Controller, the designer can put the JTAG circuitry into a mode that allows the user to shift data into the array logic through the JTAG port using the UJTAG macro. For a more detailed explanation of the UJTAG macro, refer to the "FlashROM in Microsemi’s Low Power Flash Devices" section on page 133. A user interface is required to receive the user command, initialization data, and clock from the UJTAG macro. The interface must synchronize and load the data into the correct RAM block of the design. The main outputs of the user interface block are the following: • Memory block chip select: Selects a memory block for initialization. The chip selects signals for each memory block that can be generated from different user-defined pockets or simple logic, such as a ring counter (see below). • Memory block write address: Identifies the address of the memory cell that needs to be initialized. • Memory block write data: The interface block receives the data serially from the UTDI port of the UJTAG macro and loads it in parallel into the write data ports of the memory blocks. • Memory block write clock: Drives the WCLK of the memory block and synchronizes the write data, write address, and chip select signals. Figure 6-8 shows the user interface between UJTAG and the memory blocks. RAM1 WD WADDR WCLK UJTAG TRST TDO TDI TRST TDO TDI TMS TMS TCK TCK User Interface IR[7:0] Reset UIREG[7:0] URSTB UDRUPD DR_UPDATE UDRSH UDRCAP UDRCK UTDI UTDO DR_SHIFT DR_CAPTURE DR_CLK DIN DOUT WDATA WADDR WCLK WEN1 WEN RAM2 WD WADDR WCLK WEN2 WEN WEN3 RAM3 WD WADDR WCLK WEN Figure 6-8 • Interfacing TAP Ports and SRAM Blocks An important component of the interface between the UJTAG macro and the RAM blocks is a serialin/parallel-out shift register. The width of the shift register should equal the data width of the RAM blocks. The RAM data arrives serially from the UTDI output of the UJTAG macro. The data must be shifted into a shift register clocked by the JTAG clock (provided at the UDRCK output of the UJTAG macro). Then, after the shift register is fully loaded, the data must be transferred to the write data port of the RAM block. To synchronize the loading of the write data with the write address and write clock, the output of the shift register can be pipelined before driving the RAM block. The write address can be generated in different ways. It can be imported through the TAP using a different instruction opcode and another shift register, or generated internally using a simple counter. Using a counter to generate the address bits and sweep through the address range of the RAM blocks is 164 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide recommended, since it reduces the complexity of the user interface block and the board-level JTAG driver. Moreover, using an internal counter for address generation speeds up the initialization procedure, since the user only needs to import the data through the JTAG port. The designer may use different methods to select among the multiple RAM blocks. Using counters along with demultiplexers is one approach to set the write enable signals. Basically, the number of RAM blocks needing initialization determines the most efficient approach. For example, if all the blocks are initialized with the same data, one enable signal is enough to activate the write procedure for all of them at the same time. Another alternative is to use different opcodes to initialize each memory block. For a small number of RAM blocks, using counters is an optimal choice. For example, a ring counter can be used to select from multiple RAM blocks. The clock driver of this counter needs to be controlled by the address generation process. Once the addressing of one block is finished, a clock pulse is sent to the (ring) counter to select the next memory block. Figure 6-9 illustrates a simple block diagram of an interface block between UJTAG and RAM blocks. Serial-to-Port Shift Register UTDI UDRSH UDRCK SIN POUT Data Reg. D Q n n Enable CLK SOUT CLK UTDO UDRUPDI UIREG WDATA WCLK In Compare Result with Defined Opcode En Reset CLK URSTB Chip Select WEN1 Ring Counter WEN2 WENi m En Reset CLK Addr Counter Q m WADDR Binary Counter Figure 6-9 • Block Diagram of a Sample User Interface In the circuit shown in Figure 6-9, the shift register is enabled by the UDRSH output of the UJTAG macro. The counters and chip select outputs are controlled by the value of the TAP Instruction Register. The comparison block compares the UIREG value with the "start initialization" opcode value (defined by the user). If the result is true, the counters start to generate addresses and activate the WEN inputs of appropriate RAM blocks. The UDRUPD output of the UJTAG macro, also shown in Figure 6-9, is used for generating the write clock (WCLK) and synchronizing the data register and address counter with WCLK. UDRUPD is HIGH when the TAP Controller is in the Data Register Update state, which is an indication of completing the loading of one data word. Once the TAP Controller goes into the Data Register Update state, the UDRUPD output of the UJTAG macro goes HIGH. Therefore, the pipeline register and the address counter place the proper data and address on the outputs of the interface block. Meanwhile, WCLK is defined as the inverted UDRUPD. This will provide enough time (equal to the UDRUPD HIGH time) for the data and address to be placed at the proper ports of the RAM block before the rising edge of WCLK. The inverter is not required if the RAM blocks are clocked at the falling edge of the write clock. An example of this is described in the "Example of RAM Initialization" section on page 166. Revision 4 165 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices Example of RAM Initialization This section of the document presents a sample design in which a 4×4 RAM block is being initialized through the JTAG port. A test feature has been implemented in the design to read back the contents of the RAM after initialization to verify the procedure. The interface block of this example performs two major functions: initialization of the RAM block and running a test procedure to read back the contents. The clock output of the interface is either the write clock (for initialization) or the read clock (for reading back the contents). The Verilog code for the interface block is included in the "Sample Verilog Code" section on page 167. For simulation purposes, users can declare the input ports of the UJTAG macro for easier assignment in the testbench. However, the UJTAG input ports should not be declared on the top level during synthesis. If the input ports of the UJTAG are declared during synthesis, the synthesis tool will instantiate input buffers on these ports. The input buffers on the ports will cause Compile to fail in Designer. Figure 6-10 shows the simulation results for the initialization step of the example design. The CLK_OUT signal, which is the clock output of the interface block, is the inverted DR_UPDATE output of the UJTAG macro. It is clear that it gives sufficient time (while the TAP Controller is in the Data Register Update state) for the write address and data to become stable before loading them into the RAM block. Figure 6-11 presents the test procedure of the example. The data read back from the memory block matches the written data, thus verifying the design functionality. Figure 6-10 • Simulation of Initialization Step Figure 6-11 • Simulation of the Test Procedure of the Example 166 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide The ROM emulation application is based on RAM block initialization. If the user's main design has access only to the read ports of the RAM block (RADDR, RD, RCLK, and REN), and the contents of the RAM are already initialized through the TAP, then the memory blocks will emulate ROM functionality for the core design. In this case, the write ports of the RAM blocks are accessed only by the user interface block, and the interface is activated only by the TAP Instruction Register contents. Users should note that the contents of the RAM blocks are lost in the absence of applied power. However, the 1 kbit of flash memory, FlashROM, in low power flash devices can be used to retain data after power is removed from the device. Refer to the "SRAM and FIFO Memories in Microsemi's Low Power Flash Devices" section on page 147 for more information. Sample Verilog Code Interface Block `define Initialize_start 8'h22 //INITIALIZATION START COMMAND VALUE `define Initialize_stop 8'h23 //INITIALIZATION START COMMAND VALUE module interface(IR, rst_n, data_shift, clk_in, data_update, din_ser, dout_ser, test, test_out,test_clk,clk_out,wr_en,rd_en,write_word,read_word,rd_addr, wr_addr); input [7:0] IR; input [3:0] read_word; //RAM DATA READ BACK input rst_n, data_shift, clk_in, data_update, din_ser; //INITIALIZATION SIGNALS input test, test_clk; //TEST PROCEDURE CLOCK AND COMMAND INPUT output [3:0] test_out; //READ DATA output [3:0] write_word; //WRITE DATA output [1:0] rd_addr; //READ ADDRESS output [1:0] wr_addr; //WRITE ADDRESS output dout_ser; //TDO DRIVER output clk_out, wr_en, rd_en; wire wire wire wire wire [3:0] write_word; [1:0] rd_addr; [1:0] wr_addr; [3:0] Q_out; enable, test_active; reg clk_out; //SELECT CLOCK FOR INITIALIZATION OR READBACK TEST always @(enable or test_clk or data_update) begin case ({test_active}) 1 : clk_out = test_clk ; 0 : clk_out = !data_update; default : clk_out = 1'b1; endcase end assign assign assign assign assign assign test_active = test && (IR == 8'h23); enable = (IR == 8'h22); wr_en = !enable; rd_en = !test_active; test_out = read_word; dout_ser = Q_out[3]; //4-bit SIN/POUT SHIFT REGISTER shift_reg data_shift_reg (.Shiften(data_shift), .Shiftin(din_ser), .Clock(clk_in), .Q(Q_out)); //4-bit PIPELINE REGISTER D_pipeline pipeline_reg (.Data(Q_out), .Clock(data_update), .Q(write_word)); Revision 4 167 SRAM and FIFO Memories in Microsemi's Low Power Flash Devices // addr_counter counter_1 (.Clock(data_update), .Q(wr_addr), .Aset(rst_n), .Enable(enable)); addr_counter counter_2 (.Clock(test_clk), .Q(rd_addr), .Aset(rst_n), .Enable( test_active)); endmodule Interface Block / UJTAG Wrapper This example is a sample wrapper, which connects the interface block to the UJTAG and the memory blocks. // WRAPPER module top_init (TDI, TRSTB, TMS, TCK, TDO, test, test_clk, test_ out); input TDI, TRSTB, TMS, TCK; output TDO; input test, test_clk; output [3:0] test_out; wire wire wire wire wire [7:0] IR; reset, DR_shift, DR_cap, init_clk, DR_update, data_in, data_out; clk_out, wen, ren; [3:0] word_in, word_out; [1:0] write_addr, read_addr; UJTAG UJTAG_U1 (.UIREG0(IR[0]), .UIREG1(IR[1]), .UIREG2(IR[2]), .UIREG3(IR[3]), .UIREG4(IR[4]), .UIREG5(IR[5]), .UIREG6(IR[6]), .UIREG7(IR[7]), .URSTB(reset), .UDRSH(DR_shift), .UDRCAP(DR_cap), .UDRCK(init_clk), .UDRUPD(DR_update), .UT-DI(data_in), .TDI(TDI), .TMS(TMS), .TCK(TCK), .TRSTB(TRSTB), .TDO(TDO), .UT-DO(data_out)); mem_block RAM_block (.DO(word_out), .RCLOCK(clk_out), .WCLOCK(clk_out), .DI(word_in), .WRB(wen), .RDB(ren), .WAD-DR(write_addr), .RADDR(read_addr)); interface init_block (.IR(IR), .rst_n(reset), .data_shift(DR_shift), .clk_in(init_clk), .data_update(DR_update), .din_ser(data_in), .dout_ser(data_out), .test(test), .test_out(test_out), .test_clk(test_clk), .clk_out(clk_out), .wr_en(wen), .rd_en(ren), .write_word(word_in), .read_word(word_out), .rd_addr(read_addr), .wr_addr(write_addr)); endmodule Address Counter module addr_counter (Clock, Q, Aset, Enable); input Clock; output [1:0] Q; input Aset; input Enable; reg [1:0] Qaux; always @(posedge Clock or negedge Aset) begin if (!Aset) Qaux IN1, Y =>x); I2 : INBUF_LVCMOS5 port map(PAD => IN2, Y => y); I3 : OUTBUF_SSTL3_II port map(D => z, PAD => OUT1); other port mapping… end DEF_ARCH; Synthesizing the Design Libero SoC integrates with the Synplify® synthesis tool. Other synthesis tools can also be used with Libero SoC. Refer to the Libero SoC User’s Guide or Libero online help for details on how to set up the Libero tool profile with synthesis tools from other vendors. During synthesis, the following rules apply: • • Generic macros: – Users can instantiate generic INBUF, OUTBUF, TRIBUF, and BIBUF macros. – Synthesis will automatically infer generic I/O macros. – The default I/O technology for these macros is LVTTL. – Users will need to use the I/O Attribute Editor in Designer to change the default I/O standard if needed (see Figure 9-6 on page 259). Technology-specific I/O macros: – 258 Technology-specific I/O macros, such as INBUF_LVCMO25 and OUTBUF_GTL25, can be instantiated in the design. Synthesis will infer these I/O macros in the netlist. R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide – The I/O standard of technology-specific I/O macros cannot be changed in the I/O Attribute Editor (see Figure 9-6). – The user MUST instantiate differential I/O macros (LVDS/LVPECL) in the design. This is the only way to use these standards in the design (IGLOO nano and ProASIC3 nano devices do not support differential inputs). – To implement the DDR I/O function, the user must instantiate a DDR_REG or DDR_OUT macro. This is the only way to use a DDR macro in the design. Figure 9-6 • Assigning a Different I/O Standard to the Generic I/O Macro Performing Place-and-Route on the Design The netlist created by the synthesis tool should now be imported into Designer and compiled. During Compile, the user can specify the I/O placement and attributes by importing the PDC file. The user can also specify the I/O placement and attributes using ChipPlanner and the I/O Attribute Editor under MVN. Defining I/O Assignments in the PDC File A PDC file is a Tcl script file specifying physical constraints. This file can be imported to and exported from Designer. Table 9-3 shows I/O assignment constraints supported in the PDC file. Table 9-3 • PDC I/O Constraints Command Action Example Comment I/O Banks Setting Constraints set_iobank set_vref Sets the I/O supply voltage, VCCI, and the input reference voltage, VREF, for the specified I/O bank. set_iobank bankname [-vcci vcci_voltage] [-vref vref_voltage] Must use in case of mixed I/O voltage (VCCI) design set_iobank Bank7 -vcci 1.50 -vref 0.75 Assigns a VREF pin to a set_vref -bank [bankname] [pinnum] bank. Must use if voltagereferenced I/Os are used set_vref -bank Bank0 685 704 723 742 761 set_vref_defaults Sets the default VREF set_vref_defaults bankname pins for the specified set_vref_defaults bank2 bank. This command is ignored if the bank does not need a VREF pin. Note: Refer to the Libero SoC User’s Guide for detailed rules on PDC naming and syntax conventions. Revision 4 259 I/O Software Control in Low Power Flash Devices Table 9-3 • PDC I/O Constraints (continued) Command Action Example Comment I/O Attribute Constraint set_io Sets the attributes of an set_io portname [-pinname value] I/O [-fixed value] [-iostd value] [-out_drive value] [-slew value] [-res_pull value] [-schmitt_trigger value] [-in_delay value] [-skew value] [-out_load value] [-register value] set_io IN2 -pinname 28 -fixed yes -iostd LVCMOS15 -out_drive 12 -slew high -RES_PULL None -SCHMITT_TRIGGER Off -IN_DELAY Off –skew off -REGISTER No If the I/O macro is generic (e.g., INBUF) or technologyspecific (INBUF_LVCMOS25), then all I/O attributes can be assigned using this constraint. If the netlist has an I/O macro that specifies one of its attributes, that attribute cannot be changed using this constraint, though other attributes can be changed. Example: OUTBUF_S_24 (low slew, output drive 24 mA) Slew and output drive cannot be changed. I/O Region Placement Constraints define_region If any number of I/Os must be Defines either a define_region assigned to a particular I/O rectangular region or a -name [region_name] -type [region_type] x1 y1 x2 y2 region, such a region can be rectilinear region created with this constraint. define_region -name test -type inclusive 0 15 2 29 assign_region Assigns a set of macros assign_region [region name] [macro_name...] to a specified region assign_region test U12 This constraint assigns I/O macros to the I/O regions. When assigning an I/O macro, PDC naming conventions must be followed if the macro name contains special characters; e.g., if the macro name is \\$1I19\\, the correct use of escape characters is \\\\\$1I19\\\\. Note: Refer to the Libero SoC User’s Guide for detailed rules on PDC naming and syntax conventions. 260 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Compiling the Design During Compile, a PDC I/O constraint file can be imported along with the netlist file. If only the netlist file is compiled, certain I/O assignments need to be completed before proceeding to Layout. All constraints that can be entered in PDC can also be entered using ChipPlanner, I/O Attribute Editor, and PinEditor. There are certain rules that must be followed in implementing I/O register combining and the I/O DDR macro (refer to the I/O Registers section of the handbook for the device that you are using and the "DDR" section on page 256 for details). Provided these rules are met, the user can enable or disable I/O register combining by using the PDC command set_io portname –register yes|no in the I/O Attribute Editor or selecting a check box in the Compile Options dialog box (see Figure 9-7). The Compile Options dialog box appears when the design is compiled for the first time. It can also be accessed by choosing Options > Compile during successive runs. I/O register combining is off by default. The PDC command overrides the setting in the Compile Options dialog box. Figure 9-7 • Setting Register Combining During Compile Understanding the Compile Report The I/O bank report is generated during Compile and displayed in the log window. This report lists the I/O assignments necessary before Layout can proceed. When Designer is started, the I/O Bank Assigner tool is run automatically if the Layout command is executed. The I/O Bank Assigner takes care of the necessary I/O assignments. However, these assignments can also be made manually with MVN or by importing the PDC file. Refer to the "Assigning Technologies and VREF to I/O Banks" section on page 264 for further description. The I/O bank report can also be extracted from Designer by choosing Tools > Report and setting the Report Type to IOBank. This report has the following tables: I/O Function, I/O Technology, I/O Bank Resource Usage, and I/O Voltage Usage. This report is useful if the user wants to do I/O assignments manually. Revision 4 261 I/O Software Control in Low Power Flash Devices I/O Function Figure 9-8 shows an example of the I/O Function table included in the I/O bank report: Figure 9-8 • I/O Function Table This table lists the number of input I/Os, output I/Os, bidirectional I/Os, and differential input and output I/O pairs that use I/O and DDR registers. Note: IGLOO nano and ProASIC3 nano devices do not support differential inputs. Certain rules must be met to implement registered and DDR I/O functions (refer to the I/O Structures section of the handbook for the device you are using and the "DDR" section on page 256). I/O Technology The I/O Technology table (shown in Figure 9-9) gives the values of VCCI and VREF (reference voltage) for all the I/O standards used in the design. The user should assign these voltages appropriately. Figure 9-9 • I/O Technology Table 262 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide I/O Bank Resource Usage This is an important portion of the report. The user must meet the requirements stated in this table. Figure 9-10 shows the I/O Bank Resource Usage table included in the I/O bank report: Figure 9-10 • I/O Bank Resource Usage Table The example in Figure 9-10 shows that none of the I/O macros is assigned to the bank because more than one VCCI is detected. I/O Voltage Usage The I/O Voltage Usage table provides the number of VREF (E devices only) and VCCI assignments required in the design. If the user decides to make I/O assignments manually (PDC or MVN), the issues listed in this table must be resolved before proceeding to Layout. As stated earlier, VREF assignments must be made if there are any voltage-referenced I/Os. Figure 9-11 shows the I/O Voltage Usage table included in the I/O bank report. Figure 9-11 • I/O Voltage Usage Table The table in Figure 9-11 indicates that there are two voltage-referenced I/Os used in the design. Even though both of the voltage-referenced I/O technologies have the same VCCI voltage, their VREF voltages are different. As a result, two I/O banks are needed to assign the VCCI and VREF voltages. In addition, there are six single-ended I/Os used that have the same VCCI voltage. Since two banks are already assigned with the same VCCI voltage and there are enough unused bonded I/Os in Revision 4 263 I/O Software Control in Low Power Flash Devices those banks, the user does not need to assign the same VCCI voltage to another bank. The user needs to assign the other three VCCI voltages to three more banks. Assigning Technologies and VREF to I/O Banks Low power flash devices offer a wide variety of I/O standards, including voltage-referenced standards. Before proceeding to Layout, each bank must have the required VCCI voltage assigned for the corresponding I/O technologies used for that bank. The voltage-referenced standards require the use of a reference voltage (VREF). This assignment can be done manually or automatically. The following sections describe this in detail. Manually Assigning Technologies to I/O Banks The user can import the PDC at this point and resolve this requirement. The PDC command is set_iobank [bank name] –vcci [vcci value] Another method is to use the I/O Bank Settings dialog box (MVN > Edit > I/O Bank Settings) to set up the VCCI voltage for the bank (Figure 9-12). Figure 9-12 • Setting VCCI for a Bank 264 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide The procedure is as follows: 1. Select the bank to which you want VCCI to be assigned from the Choose Bank list. 2. Select the I/O standards for that bank. If you select any standard, the tool will automatically show all compatible standards that have a common VCCI voltage requirement. 3. Click Apply. 4. Repeat steps 1–3 to assign VCCI voltages to other banks. Refer to Figure 9-11 on page 263 to find out how many I/O banks are needed for VCCI bank assignment. Manually Assigning VREF Pins Voltage-referenced inputs require an input reference voltage (VREF). The user must assign VREF pins before running Layout. Before assigning a VREF pin, the user must set a VREF technology for the bank to which the pin belongs. VREF Rules for the Implementation of Voltage-Referenced I/O Standards The VREF rules are as follows: 1. Any I/O (except JTAG I/Os) can be used as a VREF pin. 2. One VREF pin can support up to 15 I/Os. It is recommended, but not required, that eight of them be on one side and seven on the other side (in other words, all 15 can still be on one side of VREF). 3. SSTL3 (I) and (II): Up to 40 I/Os per north or south bank in any position 4. LVPECL / GTL+ 3.3 V / GTL 3.3 V: Up to 48 I/Os per north or south bank in any position (not applicable for IGLOO nano and ProASIC3 nano devices) 5. SSTL2 (I) and (II) / GTL+ 2.5 V / GTL 2.5 V: Up to 72 I/Os per north or south bank in any position 6. VREF minibanks partition rule: Each I/O bank is physically partitioned into VREF minibanks. The VREF pins within a VREF minibank are interconnected internally, and consequently, only one VREF voltage can be used within each VREF minibank. If a bank does not require a VREF signal, the VREF pins of that bank are available as user I/Os. 7. The first VREF minibank includes all I/Os starting from one end of the bank to the first power triple and eight more I/Os after the power triple. Therefore, the first VREF minibank may contain (0 + 8), (2 + 8), (4 + 8), (6 + 8), or (8 + 8) I/Os. The second VREF minibank is adjacent to the first VREF minibank and contains eight I/Os, a power triple, and eight more I/Os after the triple. An analogous rule applies to all other VREF minibanks but the last. The last VREF minibank is adjacent to the previous one but contains eight I/Os, a power triple, and all I/Os left at the end of the bank. This bank may also contain (8 + 0), (8 + 2), (8 + 4), (8 + 6), or (8 + 8) available I/Os. Example: 4 I/Os → Triple → 8 I/Os, 8 I/Os → Triple → 8 I/Os, 8 I/Os → Triple → 2 I/Os That is, minibank A = (4 + 8) I/Os, minibank B = (8 + 8) I/Os, minibank C = (8 + 2) I/Os. 8. Only minibanks that contain input or bidirectional I/Os require a VREF. A VREF is not needed for minibanks composed of output or tristated I/Os. Assigning the VREF Voltage to a Bank When importing the PDC file, the VREF voltage can be assigned to the I/O bank. The PDC command is as follows: set_iobank –vref [value] Another method for assigning VREF is by using MVN > Edit > I/O Bank Settings (Figure 9-13 on page 266). Revision 4 265 I/O Software Control in Low Power Flash Devices VREF for GTL+ 3.3 V Figure 9-13 • Selecting VREF Voltage for the I/O Bank Assigning VREF Pins for a Bank The user can use default pins for VREF. In this case, select the Use default pins for VREFs check box (Figure 9-13). This option guarantees full VREF coverage of the bank. The equivalent PDC command is as follows: set_vref_default [bank name] To be able to choose VREF pins, adequate VREF pins must be created to allow legal placement of the compatible voltage-referenced I/Os. To assign VREF pins manually, the PDC command is as follows: set_vref –bank [bank name] [package pin numbers] For ChipPlanner/PinEditor to show the range of a VREF pin, perform the following steps: 1. Assign VCCI to a bank using MVN > Edit > I/O Bank Settings. 2. Open ChipPlanner. Zoom in on an I/O package pin in that bank. 3. Highlight the pin and then right-click. Choose Use Pin for VREF. 266 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 4. Right-click and then choose Highlight VREF range. All the pins covered by that VREF pin will be highlighted (Figure 9-14). Figure 9-14 • VREF Range Using PinEditor or ChipPlanner, VREF pins can also be assigned (Figure 9-15). Figure 9-15 • Assigning VREF from PinEditor To unassign a VREF pin: 1. Select the pin to unassign. 2. Right-click and choose Use Pin for VREF. The check mark next to the command disappears. The VREF pin is now a regular pin. Resetting the pin may result in unassigning I/O cores, even if they are locked. In this case, a warning message appears so you can cancel the operation. After you assign the VREF pins, right-click a VREF pin and choose Highlight VREF Range to see how many I/Os are covered by that pin. To unhighlight the range, choose Unhighlight All from the Edit menu. Revision 4 267 I/O Software Control in Low Power Flash Devices Automatically Assigning Technologies to I/O Banks The I/O Bank Assigner (IOBA) tool runs automatically when you run Layout. You can also use this tool from within the MultiView Navigator (Figure 9-17). The IOBA tool automatically assigns technologies and VREF pins (if required) to every I/O bank that does not currently have any technologies assigned to it. This tool is available when at least one I/O bank is unassigned. To automatically assign technologies to I/O banks, choose I/O Bank Assigner from the Tools menu (or click the I/O Bank Assigner's toolbar button, shown in Figure 9-16). Figure 9-16 • I/O Bank Assigner’s Toolbar Button Messages will appear in the Output window informing you when the automatic I/O bank assignment begins and ends. If the assignment is successful, the message "I/O Bank Assigner completed successfully" appears in the Output window, as shown in Figure 9-17. Figure 9-17 • I/O Bank Assigner Displays Messages in Output Window 268 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide If the assignment is not successful, an error message appears in the Output window. To undo the I/O bank assignments, choose Undo from the Edit menu. Undo removes the I/O technologies assigned by the IOBA. It does not remove the I/O technologies previously assigned. To redo the changes undone by the Undo command, choose Redo from the Edit menu. To clear I/O bank assignments made before using the Undo command, manually unassign or reassign I/O technologies to banks. To do so, choose I/O Bank Settings from the Edit menu to display the I/O Bank Settings dialog box. Conclusion Fusion, IGLOO, and ProASIC3 support for multiple I/O standards minimizes board-level components and makes possible a wide variety of applications. The Microsemi Designer software, integrated with Libero SoC, presents a clear visual display of I/O assignments, allowing users to verify I/O and board-level design requirements before programming the device. The device I/O features and functionalities ensure board designers can produce low-cost and low power FPGA applications fulfilling the complexities of contemporary design needs. Related Documents User’s Guides Libero SoC User’s Guide http://www.microsemi.com/soc/documents/libero_ug.pdf IGLOO, ProASIC3, SmartFusion, and Fusion Macro Library Guide http://www.microsemi.com/soc/documents/pa3_libguide_ug.pdf SmartGen Core Reference Guide http://www.microsemi.com/soc/documents/genguide_ug.pdf Revision 4 269 I/O Software Control in Low Power Flash Devices List of Changes The following table lists critical changes that were made in each revision of the document. Date Changes Page August 2012 The notes in Table 9-2 • Designer State (resulting from I/O attribute modification) were revised to clarify which device families support programmable input delay (SAR 39666). 253 June 2011 Figure 9-2 • SmartGen Catalog was updated (SAR 24310). Figure 8-3 • Expanded I/O Section and the step associated with it were deleted to reflect changes in the software. 254 The following rule was added to the "VREF Rules for the Implementation of Voltage-Referenced I/O Standards" section: 265 Only minibanks that contain input or bidirectional I/Os require a VREF. A VREF is not needed for minibanks composed of output or tristated I/Os (SAR 24310). July 2010 Notes were added where appropriate to point out that IGLOO nano and ProASIC3 nano devices do not support differential inputs (SAR 21449). N/A v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 9-1 • Flash-Based FPGAs. 252 The notes for Table 9-2 • Designer State (resulting from I/O attribute modification) were revised to indicate that skew control and input delay do not apply to nano devices. 253 v1.3 (October 2008) The "Flash FPGAs I/O Support" section was revised to include new families and make the information more concise. 252 v1.2 (June 2008) The following changes were made to the family descriptions in Table 9-1 • FlashBased FPGAs: 252 v1.1 (March 2008) 270 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. This document was previously part of the I/O Structures in IGLOO and ProASIC3 Devices document. The content was separated and made into a new document. N/A Table 9-2 • Designer State (resulting from I/O attribute modification) was updated to include note 2 for IGLOO PLUS. 253 R e vi s i o n 4 10 – DDR for Microsemi’s Low Power Flash Devices Introduction The I/Os in Fusion, IGLOO, and ProASIC3 devices support Double Data Rate (DDR) mode. In this mode, new data is present on every transition (or clock edge) of the clock signal. This mode doubles the data transfer rate compared with Single Data Rate (SDR) mode, where new data is present on one transition (or clock edge) of the clock signal. Low power flash devices have DDR circuitry built into the I/O tiles. I/Os are configured to be DDR receivers or transmitters by instantiating the appropriate special macros (examples shown in Figure 10-4 on page 276 and Figure 10-5 on page 277) and buffers (DDR_OUT or DDR_REG) in the RTL design. This document discusses the options the user can choose to configure the I/Os in this mode and how to instantiate them in the design. Double Data Rate (DDR) Architecture Low power flash devices support 350 MHz DDR inputs and outputs. In DDR mode, new data is present on every transition of the clock signal. Clock and data lines have identical bandwidths and signal integrity requirements, making them very efficient for implementing very high-speed systems. High-speed DDR interfaces can be implemented using LVDS (not applicable for IGLOO nano and ProASIC3 nano devices). In IGLOOe, ProASIC3E, AFS600, and AFS1500 devices, DDR interfaces can also be implemented using the HSTL, SSTL, and LVPECL I/O standards. The DDR feature is primarily implemented in the FPGA core periphery and is not tied to a specific I/O technology or limited to any I/O standard. PAD CLK INBUF_SSTL2_I DDR_REG Y PAD D QR DataR DDR_OUT DR DF Q OUTBUF_SSTL3_I D PAD DataF QF CLR CLR CLR Figure 10-1 • DDR Support in Low Power Flash Devices Revision 4 271 DDR for Microsemi’s Low Power Flash Devices DDR Support in Flash-Based Devices The flash FPGAs listed in Table 10-1 support the DDR feature and the functions described in this document. Table 10-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 10-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 10-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 272 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide I/O Cell Architecture Low power flash devices support DDR in the I/O cells in four different modes: Input, Output, Tristate, and Bidirectional pins. For each mode, different I/O standards are supported, with most I/O standards having special sub-options. For the ProASIC3 nano and IGLOO nano devices, DDR is supported only in the 60 k, 125 k, and 250 k logic densities. Refer to Table 10-2 for a sample of the available I/O options. Additional I/O options can be found in the relevant family datasheet. Table 10-2 • DDR I/O Options DDR Register Type Receive Register Transmit Register I/O Type I/O Standard Sub-Options Input Normal None LVCMOS Voltage 1.5 V, 1.8 V, 2.5 V, 5 V (1.5 V default) Pull-Up None (default) Output Comments 3.3 V TTL (default) PCI/PCI-X None GTL/GTL+ Voltage HSTL Class I / II (I default) SSTL2/SSTL3 Class I / II (I default) LVPECL None LVDS None Normal None LVTTL Output Drive Slew Rate 2.5 V, 3.3 V (3.3 V default) 3.3 V TTL (default) 2, 4, 6, 8, 12, 16, 24, 36 mA (8 mA default) Low/high (high default) LVCMOS Voltage 1.5 V, 1.8 V, 2.5 V, 5 V (1.5 V default) PCI/PCI-X None GTL/GTL+ Voltage HSTL Class I / II (I default) SSTL2/SSTL3 Class I / II (I default) LVPECL* None LVDS* None 1.8 V, 2.5 V, 3.3 V (3.3 V default) Note: *IGLOO nano and ProASIC3 nano devices do not support differential inputs. Revision 4 273 DDR for Microsemi’s Low Power Flash Devices Table 10-2 • DDR I/O Options (continued) DDR Register Type I/O Type I/O Standard Transmit Register (continued) Tristate Buffer Normal LVTTL Sub-Options Comments Enable Polarity Low/high (low default) Output Drive Slew Rate 2, 4, 6, 8, 12,16, 24, 36 mA (8 mA default) Low/high (high default) Enable Polarity Low/high (low default) Pull-Up/-Down None (default) LVCMOS Voltage 1.5 V, 1.8 V, 2.5 V, 5 V (1.5 V default) Output Drive 2, 4, 6, 8, 12, 16, 24, 36 mA (8 mA default) Slew Rate Low/high (high default) Enable Polarity Low/high (low default) Pull-Up/-Down None (default) PCI/PCI-X GTL/GTL+ Enable Polarity Low/high (low default) Voltage 1.8 V, 2.5 V, 3.3 V (3.3 V default) Enable Polarity Low/high (low default) HSTL Class I / II (I default) Enable Polarity Low/high (low default) SSTL2/SSTL3 Class I / II (I default) Enable Polarity Low/high (low default) Bidirectional Buffer Normal LVTTL Enable Polarity Low/high (low default) Output Drive Slew Rate 2, 4, 6, 8, 12, 16, 24, 36 mA (8 mA default) Low/high (high default) Enable Polarity Low/high (low default) Pull-Up/-Down None (default) LVCMOS Voltage 1.5 V, 1.8 V, default) 2.5 V, 5V (1.5 V Enable Polarity Low/high (low default) Pull-Up PCI/PCI-X None (default) None Enable Polarity Low/high (low default) GTL/GTL+ Voltage 1.8 V, 2.5 V, 3.3 V (3.3 V default) Enable Polarity Low/high (low default) HSTL Class I / II (I default) Enable Polarity Low/high (low default) SSTL2/SSTL3 Class I / II (I default) Enable Polarity Low/high (low default) Note: *IGLOO nano and ProASIC3 nano devices do not support differential inputs. 274 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Input Support for DDR The basic structure to support a DDR input is shown in Figure 10-2. Three input registers are used to capture incoming data, which is presented to the core on each rising edge of the I/O register clock. Each I/O tile supports DDR inputs. INBUF_SSTL2_I DDR_REG PAD Y PAD CLK D QR QR QF CLR QF CLR Figure 10-2 • DDR Input Register Support in Low Power Flash Devices Output Support for DDR The basic DDR output structure is shown in Figure 10-1 on page 271. New data is presented to the output every half clock cycle. Note: DDR macros and I/O registers do not require additional routing. The combiner automatically recognizes the DDR macro and pushes its registers to the I/O register area at the edge of the chip. The routing delay from the I/O registers to the I/O buffers is already taken into account in the DDR macro. DDR_OUT DataR DataF DR DF Q OUTBUF_SSTL3_I D PAD CLK CLR CLR Figure 10-3 • DDR Output Register (SSTL3 Class I) Revision 4 275 DDR for Microsemi’s Low Power Flash Devices Instantiating DDR Registers Using SmartGen is the simplest way to generate the appropriate RTL files for use in the design. Figure 10-4 shows an example of using SmartGen to generate a DDR SSTL2 Class I input register. SmartGen provides the capability to generate all of the DDR I/O cells as described. The user, through the graphical user interface, can select from among the many supported I/O standards. The output formats supported are Verilog, VHDL, and EDIF. Figure 10-5 on page 277 through Figure 10-8 on page 280 show the I/O cell configured for DDR using SSTL2 Class I technology. For each I/O standard, the I/O pad is buffered by a special primitive that indicates the I/O standard type. Figure 10-4 • Example of Using SmartGen to Generate a DDR SSTL2 Class I Input Register 276 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide DDR Input Register INBUF_SSTL2_I DDR_REG PAD PAD Y D CLK QR QR QF CLR QF CLR Figure 10-5 • DDR Input Register (SSTL2 Class I) The corresponding structural representations, as generated by SmartGen, are shown below: Verilog module DDR_InBuf_SSTL2_I(PAD,CLR,CLK,QR,QF); input output PAD, CLR, CLK; QR, QF; wire Y; INBUF_SSTL2_I INBUF_SSTL2_I_0_inst(.PAD(PAD),.Y(Y)); DDR_REG DDR_REG_0_inst(.D(Y),.CLK(CLK),.CLR(CLR),.QR(QR),.QF(QF)); endmodule VHDL library ieee; use ieee.std_logic_1164.all; --The correct library will be inserted automatically by SmartGen library proasic3; use proasic3.all; --library fusion; use fusion.all; --library igloo; use igloo.all; entity DDR_InBuf_SSTL2_I is port(PAD, CLR, CLK : in std_logic; end DDR_InBuf_SSTL2_I; architecture DEF_ARCH of QR, QF : out std_logic) ; DDR_InBuf_SSTL2_I is component INBUF_SSTL2_I port(PAD : in std_logic := 'U'; Y : out std_logic) ; end component; component DDR_REG port(D, CLK, CLR : in std_logic := 'U'; QR, QF : out std_logic) ; end component; signal Y : std_logic ; begin INBUF_SSTL2_I_0_inst : INBUF_SSTL2_I port map(PAD => PAD, Y => Y); DDR_REG_0_inst : DDR_REG port map(D => Y, CLK => CLK, CLR => CLR, QR => QR, QF => QF); end DEF_ARCH; Revision 4 277 DDR for Microsemi’s Low Power Flash Devices DDR Output Register DDR_OUT DR DF DataR DataF OUTBUF_SSTL3_I Q D PAD CLK CLR CLR Figure 10-6 • DDR Output Register (SSTL3 Class I) Verilog module DDR_OutBuf_SSTL3_I(DataR,DataF,CLR,CLK,PAD); input output DataR, DataF, CLR, CLK; PAD; wire Q, VCC; VCC VCC_1_net(.Y(VCC)); DDR_OUT DDR_OUT_0_inst(.DR(DataR),.DF(DataF),.CLK(CLK),.CLR(CLR),.Q(Q)); OUTBUF_SSTL3_I OUTBUF_SSTL3_I_0_inst(.D(Q),.PAD(PAD)); endmodule VHDL library ieee; use ieee.std_logic_1164.all; library proasic3; use proasic3.all; entity DDR_OutBuf_SSTL3_I is port(DataR, DataF, CLR, CLK : in std_logic; end DDR_OutBuf_SSTL3_I; architecture DEF_ARCH of PAD : out std_logic) ; DDR_OutBuf_SSTL3_I is component DDR_OUT port(DR, DF, CLK, CLR : in std_logic := 'U'; Q : out std_logic) ; end component; component OUTBUF_SSTL3_I port(D : in std_logic := 'U'; PAD : out std_logic) ; end component; component VCC port( Y : out std_logic); end component; signal Q, VCC_1_net : std_logic ; begin VCC_2_net : VCC port map(Y => VCC_1_net); DDR_OUT_0_inst : DDR_OUT port map(DR => DataR, DF => DataF, CLK => CLK, CLR => CLR, Q => Q); OUTBUF_SSTL3_I_0_inst : OUTBUF_SSTL3_I port map(D => Q, PAD => PAD); end DEF_ARCH; 278 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide DDR Tristate Output Register INV A Y TrienAux Trien DDR_OUT DataR DataF DR DF Q D PAD TRIBUFF_F_8U CLK CLR CLR Figure 10-7 • DDR Tristate Output Register, LOW Enable, 8 mA, Pull-Up (LVTTL) Verilog module DDR_TriStateBuf_LVTTL_8mA_HighSlew_LowEnb_PullUp(DataR, DataF, CLR, CLK, Trien, PAD); input output DataR, DataF, CLR, CLK, Trien; PAD; wire TrienAux, Q; INV Inv_Tri(.A(Trien),.Y(TrienAux)); DDR_OUT DDR_OUT_0_inst(.DR(DataR),.DF(DataF),.CLK(CLK),.CLR(CLR),.Q(Q)); TRIBUFF_F_8U TRIBUFF_F_8U_0_inst(.D(Q),.E(TrienAux),.PAD(PAD)); endmodule VHDL library ieee; use ieee.std_logic_1164.all; library proasic3; use proasic3.all; entity DDR_TriStateBuf_LVTTL_8mA_HighSlew_LowEnb_PullUp is port(DataR, DataF, CLR, CLK, Trien : in std_logic; PAD : out std_logic) ; end DDR_TriStateBuf_LVTTL_8mA_HighSlew_LowEnb_PullUp; architecture DEF_ARCH of DDR_TriStateBuf_LVTTL_8mA_HighSlew_LowEnb_PullUp is component INV port(A : in std_logic := 'U'; Y : out std_logic) ; end component; component DDR_OUT port(DR, DF, CLK, CLR : in std_logic := 'U'; Q : out std_logic) ; end component; component TRIBUFF_F_8U port(D, E : in std_logic := 'U'; PAD : out std_logic) ; end component; signal TrienAux, Q : std_logic ; begin Inv_Tri : INV port map(A => Trien, Y => TrienAux); Revision 4 279 DDR for Microsemi’s Low Power Flash Devices DDR_OUT_0_inst : DDR_OUT port map(DR => DataR, DF => DataF, CLK => CLK, CLR => CLR, Q => Q); TRIBUFF_F_8U_0_inst : TRIBUFF_F_8U port map(D => Q, E => TrienAux, PAD => PAD); end DEF_ARCH; DDR Bidirectional Buffer Trien INV A Y E DDR_OUT DataR DataF DR DF Q CLK CLR D PAD BIBUF_HSTL_I CLR QR QF DDR_REG QR D Y QF CLR Figure 10-8 • DDR Bidirectional Buffer, LOW Output Enable (HSTL Class II) Verilog module DDR_BiDir_HSTL_I_LowEnb(DataR,DataF,CLR,CLK,Trien,QR,QF,PAD); input output inout DataR, DataF, CLR, CLK, Trien; QR, QF; PAD; wire TrienAux, D, Q; INV Inv_Tri(.A(Trien), .Y(TrienAux)); DDR_OUT DDR_OUT_0_inst(.DR(DataR),.DF(DataF),.CLK(CLK),.CLR(CLR),.Q(Q)); DDR_REG DDR_REG_0_inst(.D(D),.CLK(CLK),.CLR(CLR),.QR(QR),.QF(QF)); BIBUF_HSTL_I BIBUF_HSTL_I_0_inst(.PAD(PAD),.D(Q),.E(TrienAux),.Y(D)); endmodule 280 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide VHDL library ieee; use ieee.std_logic_1164.all; library proasic3; use proasic3.all; entity DDR_BiDir_HSTL_I_LowEnb is port(DataR, DataF, CLR, CLK, Trien : in std_logic; QR, QF : out std_logic; PAD : inout std_logic) ; end DDR_BiDir_HSTL_I_LowEnb; architecture DEF_ARCH of DDR_BiDir_HSTL_I_LowEnb is component INV port(A : in std_logic := 'U'; Y : out std_logic) ; end component; component DDR_OUT port(DR, DF, CLK, CLR : in std_logic := 'U'; Q : out std_logic) ; end component; component DDR_REG port(D, CLK, CLR : in std_logic := 'U'; QR, QF : out std_logic) ; end component; component BIBUF_HSTL_I port(PAD : inout std_logic := 'U'; D, E : in std_logic := 'U'; Y : out std_logic) ; end component; signal TrienAux, D, Q : std_logic ; begin Inv_Tri : INV port map(A => Trien, Y => TrienAux); DDR_OUT_0_inst : DDR_OUT port map(DR => DataR, DF => DataF, CLK => CLK, CLR => CLR, Q => Q); DDR_REG_0_inst : DDR_REG port map(D => D, CLK => CLK, CLR => CLR, QR => QR, QF => QF); BIBUF_HSTL_I_0_inst : BIBUF_HSTL_I port map(PAD => PAD, D => Q, E => TrienAux, Y => D); end DEF_ARCH; Revision 4 281 DDR for Microsemi’s Low Power Flash Devices Design Example Figure 10-9 shows a simple example of a design using both DDR input and DDR output registers. The user can copy the HDL code in Libero SoC software and go through the design flow. Figure 10-10 and Figure 10-11 on page 283 show the netlist and ChipPlanner views of the ddr_test design. Diagrams may vary slightly for different families. PAD INBUF_SSTL2_I DDR_REG Y PAD D QR CLK QF DataR DDR_OUT DR DF DataF CLR CLR CLR Figure 10-9 • Design Example Figure 10-10 • DDR Test Design as Seen by NetlistViewer for IGLOO/e Devices 282 R e vi s i o n 4 Q OUTBUF_SSTL3_I D PAD ProASIC3L FPGA Fabric User’s Guide Figure 10-11 • DDR Input/Output Cells as Seen by ChipPlanner for IGLOO/e Devices Verilog module Inbuf_ddr(PAD,CLR,CLK,QR,QF); input PAD, CLR, CLK; output QR, QF; wire Y; DDR_REG DDR_REG_0_inst(.D(Y), .CLK(CLK), .CLR(CLR), .QR(QR), .QF(QF)); INBUF INBUF_0_inst(.PAD(PAD), .Y(Y)); endmodule module Outbuf_ddr(DataR,DataF,CLR,CLK,PAD); input DataR, DataF, CLR, CLK; output PAD; wire Q, VCC; VCC VCC_1_net(.Y(VCC)); DDR_OUT DDR_OUT_0_inst(.DR(DataR), .DF(DataF), .CLK(CLK), .CLR(CLR), .Q(Q)); OUTBUF OUTBUF_0_inst(.D(Q), .PAD(PAD)); endmodule Revision 4 283 DDR for Microsemi’s Low Power Flash Devices module ddr_test(DIN, CLK, CLR, DOUT); input DIN, CLK, CLR; output DOUT; Inbuf_ddr Inbuf_ddr (.PAD(DIN), .CLR(clr), .CLK(clk), .QR(qr), .QF(qf)); Outbuf_ddr Outbuf_ddr (.DataR(qr),.DataF(qf), .CLR(clr), .CLK(clk),.PAD(DOUT)); INBUF INBUF_CLR (.PAD(CLR), .Y(clr)); INBUF INBUF_CLK (.PAD(CLK), .Y(clk)); endmodule Simulation Consideration Microsemi DDR simulation models use inertial delay modeling by default (versus transport delay modeling). As such, pulses that are shorter than the actual gate delays should be avoided, as they will not be seen by the simulator and may be an issue in post-routed simulations. The user must be aware of the default delay modeling and must set the correct delay model in the simulator as needed. Conclusion Fusion, IGLOO, and ProASIC3 devices support a wide range of DDR applications with different I/O standards and include built-in DDR macros. The powerful capabilities provided by SmartGen and its GUI can simplify the process of including DDR macros in designs and minimize design errors. Additional considerations should be taken into account by the designer in design floorplanning and placement of I/O flip-flops to minimize datapath skew and to help improve system timing margins. Other system-related issues to consider include PLL and clock partitioning. 284 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide List of Changes The following table lists critical changes that were made in each revision of the chapter. Date July 2010 v1.4 (December 2008) Changes Page This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A Notes were added where appropriate to point out that IGLOO nano and ProASIC3 nano devices do not support differential inputs (SAR 21449). N/A IGLOO nano and ProASIC3 nano devices were added to Table 10-1 • Flash-Based FPGAs. 272 The "I/O Cell Architecture" section was updated with information applicable to nano devices. 273 The output buffer (OUTBUF_SSTL3_I) input was changed to D, instead of Q, in 271, Figure 10-1 • DDR Support in Low Power Flash Devices, Figure 10-3 • DDR Output 275, Register (SSTL3 Class I), Figure 10-6 • DDR Output Register (SSTL3 Class I), 278, 279 Figure 10-7 • DDR Tristate Output Register, LOW Enable, 8 mA, Pull-Up (LVTTL), and the output from the DDR_OUT macro was connected to the input of the TRIBUFF macro in Figure 10-7 • DDR Tristate Output Register, LOW Enable, 8 mA, Pull-Up (LVTTL). v1.3 (October 2008) v1.2 (June 2008) The "Double Data Rate (DDR) Architecture" section was updated to include mention of the AFS600 and AFS1500 devices. 271 The "DDR Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 272 The following changes were made to the family descriptions in Table 10-1 • FlashBased FPGAs: 272 • • v1.1 (March 2008) ProASIC3L was updated to include 1.5 V. The number of PLLs for ProASIC3E was changed from five to six. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. Revision 4 272 285 11 – Programming Flash Devices Introduction This document provides an overview of the various programming options available for the Microsemi flash families. The electronic version of this document includes active links to all programming resources, which are available at http://www.microsemi.com/soc/products/hardware/default.aspx. For Microsemi antifuse devices, refer to the Programming Antifuse Devices document. Summary of Programming Support FlashPro4 and FlashPro3 are high-performance in-system programming (ISP) tools targeted at the latest generation of low power flash devices offered by the SmartFusion,® Fusion, IGLOO,® and ProASIC®3 families, including ARM-enabled devices. FlashPro4 and FlashPro3 offer extremely high performance through the use of USB 2.0, are high-speed compliant for full use of the 480 Mbps bandwidth, and can program ProASIC3 devices in under 30 seconds. Powered exclusively via USB, FlashPro4 and FlashPro3 provide a VPUMP voltage of 3.3 V for programming these devices. FlashPro4 replaced FlashPro3 in 2010. FlashPro4 supports SmartFusion, Fusion, ProASIC3,and IGLOO devices as well as future generation flash devices. FlashPro4 also adds 1.2 V programming for IGLOO nano V2 devices. FlashPro4 is compatible with FlashPro3; however it adds a programming mode (PROG_MODE) signal to the previously unused pin 4 of the JTAG connector. The PROG_MODE goes high during programming and can be used to turn on a 1.5 V external supply for those devices that require 1.5 V for programming. If both FlashPro3 and FlashPro4 programmers are used for programming the same boards, pin 4 of the JTAG connector must not be connected to anything on the board because FlashPro4 uses pin 4 for PROG_MODE. FlashPro3 or FlashPro4 FlashPro Software JTAG ProASIC3/E Programming File: PDB, STP, or FDB Figure 11-1 • FlashPro Programming Setup Revision 4 287 Programming Flash Devices Programming Support in Flash Devices The flash FPGAs listed in Table 11-1 support flash in-system programming and the functions described in this document. Table 11-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution, supporting 1.2 V to 1.5 V core voltage with Flash*Freeze technology IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V core voltage with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications SmartFusion SmartFusion Mixed-signal FPGA integrating FPGA fabric, programmable microcontroller subsystem (MSS), including programmable analog and ARM® Cortex™-M3 hard processor and flash memory in a monolithic device Fusion Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device ProASIC ProASIC First generation ProASIC devices ProASICPLUS Second generation ProASIC devices Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 11-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 11-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 288 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide General Flash Programming Information Programming Basics When choosing a programming solution, there are a number of options available. This section provides a brief overview of those options. The next sections provide more detail on those options as they apply to Microsemi FPGAs. Reprogrammable or One-Time-Programmable (OTP) Depending on the technology chosen, devices may be reprogrammable or one-time-programmable. As the name implies, a reprogrammable device can be programmed many times. Generally, the contents of such a device will be completely overwritten when it is reprogrammed. All Microsemi flash devices are reprogrammable. An OTP device is programmable one time only. Once programmed, no more changes can be made to the contents. Microsemi flash devices provide the option of disabling the reprogrammability for security purposes. This combines the convenience of reprogrammability during design verification with the security of an OTP technology for highly sensitive designs. Device Programmer or In-System Programming There are two fundamental ways to program an FPGA: using a device programmer or, if the technology permits, using in-system programming. A device programmer is a piece of equipment in a lab or on the production floor that is used for programming FPGA devices. The devices are placed into a socket mounted in a programming adapter module, and the appropriate electrical interface is applied. The programmed device can then be placed on the board. A typical programmer, used during development, programs a single device at a time and is referred to as a single-site engineering programmer. With ISP, the device is already mounted onto the system printed circuit board when programming occurs. Typically, ISD programming is performed via a JTAG interface on the FPGA. The JTAG pins can be controlled either by an on-board resource, such as a microprocessor, or by an off-board programmer through a header connection. Once mounted, it can be programmed repeatedly and erased. If the application requires it, the system can be designed to reprogram itself using a microprocessor, without the use of any external programmer. If multiple devices need to be programmed with the same program, various multi-site programming hardware is available in order to program many devices in parallel. Microsemi In House Programming is also available for this purpose. Programming Features for Microsemi Devices Flash Devices The flash devices supplied by Microsemi are reprogrammable by either a generic device programmer or ISP. Microsemi supports ISP using JTAG, which is supported by the FlashPro4 and FlashPro3, FlashPro Lite, Silicon Sculptor 3, and Silicon Sculptor II programmers. Levels of ISP support vary depending on the device chosen: • All SmartFusion, Fusion, IGLOO, and ProASIC3 devices support ISP. • IGLOO, IGLOOe, IGLOO nano V5, and IGLOO PLUS devices can be programmed in-system when the device is using a 1.5 V supply voltage to the FPGA core. • IGLOO nano V2 devices can be programmed at 1.2 V core voltage (when using FlashPro4 only) or 1.5 V. IGLOO nano V5 devices are programmed with a VCC core voltage of 1.5 V. Revision 4 289 Programming Flash Devices Types of Programming for Flash Devices The number of devices to be programmed will influence the optimal programming methodology. Those available are listed below: • • • In-system programming – Using a programmer – Using a microprocessor or microcontroller Device programmers – Single-site programmers – Multi-site programmers, batch programmers, or gang programmers – Automated production (robotic) programmers Volume programming services – Microsemi in-house programming – Programming centers In-System Programming Device Type Supported: Flash ISP refers to programming the FPGA after it has been mounted on the system printed circuit board. The FPGA may be preprogrammed and later reprogrammed using ISP. The advantage of using ISP is the ability to update the FPGA design many times without any changes to the board. This eliminates the requirement of using a socket for the FPGA, saving cost and improving reliability. It also reduces programming hardware expenses, as the ISP methodology is die-/packageindependent. There are two methods of in-system programming: external and internal. • Programmer ISP—Refer to the "In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X" section on page 327 for more information. Using an external programmer and a cable, the device can be programmed through a header on the system board. In Microsemi SoC Products Group documentation, this is referred to as external ISP. Microsemi provides FlashPro4, FlashPro3, FlashPro Lite, or Silicon Sculptor 3 to perform external ISP. Note that Silicon Sculptor II and Silicon Sculptor 3 can only provide ISP for ProASIC and ProASICPLUS® families, not for SmartFusion, Fusion, IGLOO, or ProASIC3. Silicon Sculptor II and Silicon Sculptor 3 can be used for programming ProASIC and ProASICPLUS devices by using an adapter module (part number SMPA-ISP-ACTEL-3). • – Advantages: Allows local control of programming and data files for maximum security. The programming algorithms and hardware are available from Microsemi. The only hardware required on the board is a programming header. – Limitations: A negligible board space requirement for the programming header and JTAG signal routing Microprocessor ISP—Refer to the "Microprocessor Programming of Microsemi’s Low Power Flash Devices" chapter of an appropriate FPGA fabric user’s guide for more information. Using a microprocessor and an external or internal memory, you can store the program in memory and use the microprocessor to perform the programming. In Microsemi documentation, this is referred to as internal ISP. Both the code for the programming algorithm and the FPGA programming file must be stored in memory on the board. Programming voltages must also be generated on the board. 290 – Advantages: The programming code is stored in the system memory. An external programmer is not required during programming. – Limitations: This is the approach that requires the most design work, since some way of getting and/or storing the data is needed; a system interface to the device must be designed; and the low-level API to the programming firmware must be written and linked into the code provided by Microsemi. While there are benefits to this methodology, serious thought and planning should go into the decision. R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Device Programmers Single Device Programmer Single device programmers are used to program a device before it is mounted on the system board. The advantage of using device programmers is that no programming hardware is required on the system board. Therefore, no additional components or board space are required. Adapter modules are purchased with single device programmers to support the FPGA packages used. The FPGA is placed in the adapter module and the programming software is run from a PC. Microsemi supplies the programming software for all of the Microsemi programmers. The software allows for the selection of the correct die/package and programming files. It will then program and verify the device. • Single-site programmers A single-site programmer programs one device at a time. Microsemi offers Silicon Sculptor 3, built by BP Microsystems, as a single-site programmer. Silicon Sculptor 3 and associated software are available only from Microsemi. • – Advantages: Lower cost than multi-site programmers. No additional overhead for programming on the system board. Allows local control of programming and data files for maximum security. Allows on-demand programming on-site. – Limitations: Only programs one device at a time. Multi-site programmers Often referred to as batch or gang programmers, multi-site programmers can program multiple devices at the same time using the same programming file. This is often used for large volume programming and by programming houses. The sites often have independent processors and memory enabling the sites to operate concurrently, meaning each site may start programming the same file independently. This enables the operator to change one device while the other sites continue programming, which increases throughput. Multiple adapter modules for the same package are required when using a multi-site programmer. Silicon Sculptor I, II, and 3 programmers can be cascaded to program multiple devices in a chain. Multi-site programmers, such as the BP2610 and BP2710, can also be purchased from BP Microsystems. When using BP Microsystems multi-site programmers, users must use programming adapter modules available only from Microsemi. Visit the Microsemi SoC Products Group website to view the part numbers of the desired adapter module: http://www.microsemi.com/soc/products/hardware/program_debug/ss/modules.aspx. Also when using BP Microsystems programmers, customers must use Microsemi programming software to ensure the best programming result will occur. • – Advantages: Provides the capability of programming multiple devices at the same time. No additional overhead for programming on the system board. Allows local control of programming and data files for maximum security. – Limitations: More expensive than a single-site programmer Automated production (robotic) programmers Automated production programmers are based on multi-site programmers. They consist of a large input tray holding multiple parts and a robotic arm to select and place parts into appropriate programming sockets automatically. When the programming of the parts is complete, the parts are removed and placed in a finished tray. The automated programmers are often used in volume programming houses to program parts for which the programming time is small. BP Microsystems part number BP4710, BP4610, BP3710 MK2, and BP3610 are available for this purpose. Auto programmers cannot be used to program RTAX-S devices. Where an auto-programmer is used, the appropriate open-top adapter module from BP Microsystems must be used. Revision 4 291 Programming Flash Devices Volume Programming Services Device Type Supported: Flash and Antifuse Once the design is stable for applications with large production volumes, preprogrammed devices can be purchased. Table 11-2 describes the volume programming services. Table 11-2 • Volume Programming Services Programmer In-House Programming Distributor Programming Centers Independent Programming Centers Vendor Availability Microsemi Contact Microsemi Sales Memec Unique Contact Distribution Various Contact Vendor Advantages: As programming is outsourced, this solution is easier to implement than creating a substantial in-house programming capability. As programming houses specialize in large-volume programming, this is often the most cost-effective solution. Limitations: There are some logistical issues with the use of a programming service provider, such as the transfer of programming files and the approval of First Articles. By definition, the programming file must be released to a third-party programming house. Nondisclosure agreements (NDAs) can be signed to help ensure data protection; however, for extremely security-conscious designs, this may not be an option. • Microsemi In-House Programming When purchasing Microsemi devices in volume, IHP can be requested as part of the purchase. If this option is chosen, there is a small cost adder for each device programmed. Each device is marked with a special mark to distinguish it from blank parts. Programming files for the design will be sent to Microsemi. Sample parts with the design programmed, First Articles, will be returned for customer approval. Once approval of First Articles has been received, Microsemi will proceed with programming the remainder of the order. To request Microsemi IHP, contact your local Microsemi representative. • Distributor Programming Centers If purchases are made through a distributor, many distributors will provide programming for their customers. Consult with your preferred distributor about this option. 292 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Programming Solutions Details for the available programmers can be found in the programmer user's guides listed in the "Related Documents" section on page 297. All the programmers except FlashPro4, FlashPro3, FlashPro Lite, and FlashPro require adapter modules, which are designed to support device packages. All modules are listed on the Microsemi SoC Products Group website at http://www.microsemi.com/soc/products/hardware/program_debug/ss/modules.aspx. They are not listed in this document, since this list is updated frequently with new package options and any upgrades required to improve programming yield or support new families. Table 11-3 • Programming Solutions Programmer FlashPro4 FlashPro3 FlashPro Lite2 Vendor Microsemi Microsemi Microsemi ISP Single Device Only Only Only Multi-Device Availability Yes Yes 1 Available Yes Yes1 Available Yes Yes1 Available Discontinued Microsemi Only Yes Yes1 Silicon Sculptor 3 Microsemi Yes3 Yes Cascade option (up to two) Available Silicon Sculptor II Microsemi Yes3 Yes Cascade option (up to two) Available Silicon Sculptor Microsemi Yes Yes Cascade option (up to four) Discontinued Sculptor 6X Microsemi No Yes Yes Discontinued BP Microsystems No Yes Yes Contact BP Microsystems at www.bpmicro.com FlashPro BP MicroProgrammers Notes: 1. Multiple devices can be connected in the same JTAG chain for programming. 2. If FlashPro Lite is used for programming, the programmer derives all of its power from the target pc board's VDD supply. The FlashPro Lite's VPP and VPN power supplies use the target pc board's VDD as a power source. The target pc board must supply power to both the VDDP and VDD power pins of the ProASICPLUS device in addition to supplying VDD to the FlashPro Lite. The target pc board needs to provide at least 500 mA of current to the FlashPro Lite VDD connection for programming. 3. Silicon Sculptor II and Silicon Sculptor 3 can only provide ISP for ProASIC and ProASICPLUS families, not for Fusion, IGLOO, or ProASIC3 devices. Revision 4 293 Programming Flash Devices Programmer Ordering Codes The products shown in Table 11-4 can be ordered through Microsemi sales and will be shipped directly from Microsemi. Products can also be ordered from Microsemi distributors, but will still be shipped directly from Microsemi. Table 11-4 includes ordering codes for the full kit, as well as codes for replacement items and any related hardware. Some additional products can be purchased from external suppliers for use with the programmers. Ordering codes for adapter modules used with Silicon Sculptor are available at http://www.microsemi.com/soc/products/hardware/program_debug/ss/modules.aspx. Table 11-4 • Programming Ordering Codes Description Vendor Ordering Code Comment FlashPro4 ISP programmer Microsemi FLASHPRO 4 Uses a 2×5, RA male header connector FlashPro Lite ISP programmer Microsemi FLASHPRO LITE Supports small programming header or large header through header converter (not included) Silicon Sculptor 3 Microsemi SILICON-SCULPTOR 3 USB 2.0 high-speed production programmer Silicon Sculptor II Microsemi SILICON-SCULPTOR II Requires add-on adapter modules to support devices Silicon Sculptor ISP module Microsemi SMPA-ISP-ACTEL-3-KIT Ships with both large and small header support ISP cable for small header Microsemi ISP-CABLE-S Supplied with SMPA-ISP-ACTEL-3-KIT ISP cable for large header Microsemi PA-ISP-CABLE Supplied with SMPA-ISP-ACTEL-3-KIT Programmer Device Support Refer to www.microsemi.com/soc for the current information on programmer and device support. Certified Programming Solutions The Microsemi-certified programmers for flash devices are FlashPro4, FlashPro3, FlashPro Lite, FlashPro, Silicon Sculptor II, Silicon Sculptor 3, and any programmer that is built by BP Microsystems. All other programmers are considered noncertified programmers. • FlashPro4, FlashPro3, FlashPro Lite, FlashPro The Microsemi family of FlashPro device programmers provides in-system programming in an easy-to-use, compact system that supports all flash families. Whether programming a board containing a single device or multiple devices connected in a chain, the Microsemi line of FlashPro programmers enables fast programming and reprogramming. Programming with the FlashPro series of programmers saves board space and money as it eliminates the need for sockets on the board. There are no built-in algorithms, so there is no delay between product release and programming support. The FlashPro programmer is no longer available. • Silicon Sculptor 3, Silicon Sculptor II Silicon Sculptor 3 and Silicon Sculptor II are robust, compact, single-device programmers with standalone software for the PC. They are designed to enable concurrent programming of multiple units from the same PC with speeds equivalent to or faster than previous Microsemi programmers. • Noncertified Programmers Microsemi does not test programming solutions from other vendors, and DOES NOT guarantee programming yield. Also, Microsemi will not perform any failure analysis on devices programmed on non-certified programmers. Please refer to the Programming and Functional Failure Guidelines document for more information. 294 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide • Programming Centers Microsemi programming hardware policy also applies to programming centers. Microsemi expects all programming centers to use certified programmers to program Microsemi devices. If a programming center uses noncertified programmers to program Microsemi devices, the "Noncertified Programmers" policy applies. Important Programming Guidelines Preprogramming Setup Before programming, several steps are required to ensure an optimal programming yield. Use Proper Handling and Electrostatic Discharge (ESD) Precautions Microsemi FPGAs are sensitive electronic devices that are susceptible to damage from ESD and other types of mishandling. For more information about ESD, refer to the Quality and Reliability Guide, beginning with page 41. Use the Latest Version of the Designer Software to Generate Your Programming File (recommended) The files used to program Microsemi flash devices (*.bit, *.stp, *.pdb) contain important information about the switches that will be programmed in the FPGA. Find the latest version and corresponding release notes at http://www.microsemi.com/soc/download/software/designer/. Also, programming files must always be zipped during file transfer to avoid the possibility of file corruption. Use the Latest Version of the Programming Software The programming software is frequently updated to accommodate yield enhancements in FPGA manufacturing. These updates ensure maximum programming yield and minimum programming times. Before programming, always check the version of software being used to ensure it is the most recent. Depending on the programming software, refer to one of the following: • FlashPro: http://www.microsemi.com/soc/download/program_debug/flashpro/ • Silicon Sculptor: http://www.microsemi.com/soc/download/program_debug/ss/ Use the Most Recent Adapter Module with Silicon Sculptor Occasionally, Microsemi makes modifications to the adapter modules to improve programming yields and programming times. To identify the latest version of each module before programming, visit http://www.microsemi.com/soc/products/hardware/program_debug/ss/modules.aspx. Perform Routine Hardware Self-Diagnostic Test • Adapter modules must be regularly cleaned. Adapter modules need to be inserted carefully into the programmer to make sure the DIN connectors (pins at the back side) are not damaged. • FlashPro The self-test is only applicable when programming with FlashPro and FlashPro3 programmers. It is not supported with FlashPro4 or FlashPro Lite. To run the self-diagnostic test, follow the instructions given in the "Performing a Self-Test" section of http://www.microsemi.com/soc/documents/FlashPro_UG.pdf. • Silicon Sculptor The self-diagnostic test verifies correct operation of the pin drivers, power supply, CPU, memory, and adapter module. This test should be performed with an adapter module installed and before every programming session. At minimum, the test must be executed every week. To perform selfdiagnostic testing using the Silicon Sculptor software, perform the following steps, depending on the operating system: – DOS: From anywhere in the software, type ALT + D. – Windows: Click Device > choose Actel Diagnostic > select the Test tab > click OK. Silicon Sculptor programmers must be verified annually for calibration. Refer to the Silicon Sculptor Verification of Calibration Work Instruction document on the website. Revision 4 295 Programming Flash Devices Signal Integrity While Using ISP For ISP of flash devices, customers are expected to follow the board-level guidelines provided on the Microsemi SoC Products Group website. These guidelines are discussed in the datasheets and application notes (refer to the “Related Documents” section of the datasheet for application note links). Customers are also expected to troubleshoot board-level signal integrity issues by measuring voltages and taking oscilloscope plots. Programming Failure Allowances Microsemi has strict policies regarding programming failure allowances. Please refer to Programming and Functional Failure Guidelines on the Microsemi SoC Products Group website for details. Contacting the Customer Support Group Highly skilled engineers staff the Customer Applications Center from 7:00 A.M. to 6:00 P.M., Pacific time, Monday through Friday. You can contact the center by one of the following methods: Electronic Mail You can communicate your technical questions to our email address and receive answers back by email, fax, or phone. Also, if you have design problems, you can email your design files to receive assistance. Microsemi monitors the email account throughout the day. When sending your request to us, please be sure to include your full name, company name, and contact information for efficient processing of your request. The technical support email address is soc_tech@microsemi.com. Telephone Our Technical Support Hotline answers all calls. The center retrieves information, such as your name, company name, telephone number, and question. Once this is done, a case number is assigned. Then the center forwards the information to a queue where the first available applications engineer receives the data and returns your call. The phone hours are from 7:00 A.M. to 6:00 P.M., Pacific time, Monday through Friday. The Customer Applications Center number is (800) 262-1060. European customers can call +44 (0) 1256 305 600. 296 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Related Documents Below is a list of related documents, their location on the Microsemi SoC Products Group website, and a brief summary of each document. Application Notes Programming Antifuse Devices http://www.microsemi.com/soc/documents/AntifuseProgram_AN.pdf Implementation of Security in Actel's ProASIC and ProASICPLUS Flash-Based FPGAs http://www.microsemi.com/soc/documents/Flash_Security_AN.pdf User’s Guides FlashPro Programmers FlashPro4,1 FlashPro3, FlashPro Lite, and FlashPro2 http://www.microsemi.com/soc/products/hardware/program_debug/flashpro/default.aspx FlashPro User's Guide http://www.microsemi.com/soc/documents/FlashPro_UG.pdf The FlashPro User’s Guide includes hardware and software setup, self-test instructions, use instructions, and a troubleshooting / error message guide. Silicon Sculptor 3 and Silicon Sculptor II http://www.microsemi.com/soc/products/hardware/program_debug/ss/default.aspx Other Documents http://www.microsemi.com/soc/products/solutions/security/default.aspx#flashlock The security resource center describes security in Microsemi Flash FPGAs. Quality and Reliability Guide http://www.microsemi.com/soc/documents/RelGuide.pdf Programming and Functional Failure Guidelines http://www.microsemi.com/soc/documents/FA_Policies_Guidelines_5-06-00002.pdf 1. 2. FlashPro4 replaced FlashPro3 in Q1 2010. FlashPro is no longer available. Revision 4 297 Programming Flash Devices List of Changes The following table lists critical changes that were made in each revision of the chapter. Date July 2010 Changes Page FlashPro4 is a replacement for FlashPro3 and has been added to this chapter. FlashPro is no longer available. N/A The chapter was updated to include SmartFusion devices. N/A The following were deleted: N/A "Live at Power-Up (LAPU) or Boot PROM" section "Design Security" section Table 14-2 • Programming Features for Actel Devices and much of the text in the "Programming Features for Microsemi Devices" section "Programming Flash FPGAs" section "Return Material Authorization (RMA) Policies" section 298 The "Device Programmers" section was revised. 291 The Independent Programming Centers information was removed from the "Volume Programming Services" section. 292 Table 11-3 • Programming Solutions was revised to add FlashPro4 and note that FlashPro is discontinued. A note was added for FlashPro Lite regarding power supply requirements. 293 Most items were removed from Table 11-4 • Programming Ordering Codes, including FlashPro3 and FlashPro. 294 The "Programmer Device Support" section was deleted and replaced with a reference to the Microsemi SoC Products Group website for the latest information. 294 The "Certified Programming Solutions" section was revised to add FlashPro4 and remove Silicon Sculptor I and Silicon Sculptor 6X. Reference to Programming and Functional Failure Guidelines was added. 294 The file type *.pdb was added to the "Use the Latest Version of the Designer Software to Generate Your Programming File (recommended)" section. 295 Instructions on cleaning and careful insertion were added to the "Perform Routine Hardware Self-Diagnostic Test" section. Information was added regarding testing Silicon Sculptor programmers with an adapter module installed before every programming session verifying their calibration annually. 295 The "Signal Integrity While Using ISP" section is new. 296 The "Programming Failure Allowances" section was revised. 296 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Date v1.3 (December 2008) v1.2 (October 2008) Changes The "Programming Support in Flash Devices" section was updated to include IGLOO nano and ProASIC3 nano devices. 288 The "Flash Devices" section was updated to include information for IGLOO nano devices. The following sentence was added: IGLOO PLUS devices can also be operated at any voltage between 1.2 V and 1.5 V; the Designer software allows 50 mV increments in the voltage. 289 Table 11-4 · Programming Ordering Codes was updated to replace FP3-26PINADAPTER with FP3-10PIN-ADAPTER-KIT. 294 Table 14-6 · Programmer Device Support was updated to add IGLOO nano and ProASIC3 nano devices. AGL400 was added to the IGLOO portion of the table. 317 The "Programming Support in Flash Devices" section was revised to include new families and make the information more concise. 288 Figure 11-1 · FlashPro Programming Setup and the "Programming Support in Flash Devices" section are new. v1.1 (March 2008) Page 287, 288 Table 14-6 · Programmer Device Support was updated to include A3PE600L with the other ProASIC3L devices, and the RT ProASIC3 family was added. 317 The "Flash Devices" section was updated to include the IGLOO PLUS family. The text, "Voltage switching is required in-system to switch from a 1.2 V core to 1.5 V core for programming," was revised to state, "Although the device can operate at 1.2 V core voltage, the device can only be reprogrammed when the core voltage is 1.5 V. Voltage switching is required in-system to switch from a 1.2 V supply (VCC, VCCI, and VJTAG) to 1.5 V for programming." 289 The ProASIC3L family was added to Table 14-6 · Programmer Device Support as a separate set of rows rather than combined with ProASIC3 and ProASIC3E devices. The IGLOO PLUS family was included, and AGL015 and A3P015 were added. 317 Revision 4 299 12 – Security in Low Power Flash Devices Security in Programmable Logic The need for security on FPGA programmable logic devices (PLDs) has never been greater than today. If the contents of the FPGA can be read by an external source, the intellectual property (IP) of the system is vulnerable to unauthorized copying. Fusion, IGLOO, and ProASIC3 devices contain state-of-the-art circuitry to make the flash-based devices secure during and after programming. Low power flash devices have a built-in 128-bit Advanced Encryption Standard (AES) decryption core (except for 30 k gate devices and smaller). The decryption core facilitates secure in-system programming (ISP) of the FPGA core array fabric, the FlashROM, and the Flash Memory Blocks (FBs) in Fusion devices. The FlashROM, Flash Blocks, and FPGA core fabric can be programmed independently of each other, allowing the FlashROM or Flash Blocks to be updated without the need for change to the FPGA core fabric. Microsemi has incorporated the AES decryption core into the low power flash devices and has also included the Microsemi flash-based lock technology, FlashLock.® Together, they provide leading-edge security in a programmable logic device. Configuration data loaded into a device can be decrypted prior to being written to the FPGA core using the AES 128-bit block cipher standard. The AES encryption key is stored in on-chip, nonvolatile flash memory. This document outlines the security features offered in low power flash devices, some applications and uses, as well as the different software settings for each application. Figure 12-1 • Overview on Security Revision 4 301 Security in Low Power Flash Devices Security Support in Flash-Based Devices The flash FPGAs listed in Table 12-1 support the security feature and the functions described in this document. Table 12-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 12-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 12-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 302 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Security Architecture Fusion, IGLOO, and ProASIC3 devices have been designed with the most comprehensive programming logic design security in the industry. In the architecture of these devices, security has been designed into the very fabric. The flash cells are located beneath seven metal layers, and the use of many device design and layout techniques makes invasive attacks difficult. Since device layers cannot be removed without disturbing the charge on the programmed (or erased) flash gates, devices cannot be easily deconstructed to decode the design. Low power flash devices are unique in being reprogrammable and having inherent resistance to both invasive and noninvasive attacks on valuable IP. Secure, remote ISP is now possible with AES encryption capability for the programming file during electronic transfer. Figure 12-2 shows a view of the AES decryption core inside an IGLOO device; Figure 12-3 on page 304 shows the AES decryption core inside a Fusion device. The AES core is used to decrypt the encrypted programming file when programming. Bank 0 Bank 1 Bank 3 CCC RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block I/Os Bank 1 Bank 3 VersaTile ISP AES Decryption* User Nonvolatile FlashRom Flash*Freeze Technology Charge Pumps RAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Bank 2 Note: *ISP AES Decryption is not supported by 30 k gate devices and smaller. For details of other architecture features by device, refer to the appropriate family datasheet. Figure 12-2 • Block Representation of the AES Decryption Core in IGLOO and ProASIC3 Devices Revision 4 303 Security in Low Power Flash Devices Bank 0 Bank 1 CCC SRAM Block 4,608-Bit Dual-Port SRAM or FIFO Block OSC I/Os CCC/PLL Bank 2 Bank 4 VersaTile ISP AES Decryption User Nonvolatile FlashROM Flash Memory Blocks Analog Quad Analog Quad Analog Quad Analog Quad ADC Analog Quad Charge Pumps SRAM Block 4,608-Bit Dual-Port SRAM or FIFO Block Flash Memory Blocks Analog Quad Analog Quad Analog Quad Analog Quad Analog Quad CCC Bank 3 Figure 12-3 • Block Representation of the AES Decryption Core in a Fusion AFS600 FPGA Security Features IGLOO and ProASIC3 devices have two entities inside: FlashROM and the FPGA core fabric. Fusion devices contain three entities: FlashROM, FBs, and the FPGA core fabric. The parts can be programmed or updated independently with a STAPL programming file. The programming files can be AES-encrypted or plaintext. This allows maximum flexibility in providing security to the entire device. Refer to the "Programming Flash Devices" section on page 287 for information on the FlashROM structure. Unlike SRAM-based FPGA devices, which require a separate boot PROM to store programming data, low power flash devices are nonvolatile, and the secured configuration data is stored in on-chip flash cells that are part of the FPGA fabric. Once programmed, this data is an inherent part of the FPGA array and does not need to be loaded at system power-up. SRAM-based FPGAs load the configuration bitstream upon power-up; therefore, the configuration is exposed and can be read easily. The built-in FPGA core, FBs, and FlashROM support programming files encrypted with the 128-bit AES (FIPS-192) block ciphers. The AES key is stored in dedicated, on-chip flash memory and can be programmed before the device is shipped to other parties (allowing secure remote field updates). Security in ARM-Enabled Low Power Flash Devices There are slight differences between the regular flash devices and the ARM®-enabled flash devices, which have the M1 and M7 prefix. The AES key is used by Microsemi and preprogrammed into the device to protect the ARM IP. As a result, the design is encrypted along with the ARM IP, according to the details below. 304 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Cortex-M1 Device Security Cortex-M1–enabled devices are shipped with the following security features: • FPGA array enabled for AES-encrypted programming and verification • FlashROM enabled for AES-encrypted Write and Verify • Fusion Embedded Flash Memory enabled for AES-encrypted Write AES Encryption of Programming Files Low power flash devices employ AES as part of the security mechanism that prevents invasive and noninvasive attacks. The mechanism entails encrypting the programming file with AES encryption and then passing the programming file through the AES decryption core, which is embedded in the device. The file is decrypted there, and the device is successfully programmed. The AES master key is stored in on-chip nonvolatile memory (flash). The AES master key can be preloaded into parts in a secure programming environment (such as the Microsemi In-House Programming center), and then "blank" parts can be shipped to an untrusted programming or manufacturing center for final personalization with an AES-encrypted bitstream. Late-stage product changes or personalization can be implemented easily and securely by simply sending a STAPL file with AES-encrypted data. Secure remote field updates over public networks (such as the Internet) are possible by sending and programming a STAPL file with AESencrypted data. The AES key protects the programming data for file transfer into the device with 128-bit AES encryption. If AES encryption is used, the AES key is stored or preprogrammed into the device. To program, you must use an AES-encrypted file, and the encryption used on the file must match the encryption key already in the device. The AES key is protected by a FlashLock security Pass Key that is also implemented in each device. The AES key is always protected by the FlashLock Key, and the AES-encrypted file does NOT contain the FlashLock Key. This FlashLock Pass Key technology is exclusive to the Microsemi flash-based device families. FlashLock Pass Key technology can also be implemented without the AES encryption option, providing a choice of different security levels. In essence, security features can be categorized into the following three options: • AES encryption with FlashLock Pass Key protection • FlashLock protection only (no AES encryption) • No protection Each of the above options is explained in more detail in the following sections with application examples and software implementation options. Advanced Encryption Standard The 128-bit AES standard (FIPS-192) block cipher is the NIST (National Institute of Standards and Technology) replacement for DES (Data Encryption Standard FIPS46-2). AES has been designed to protect sensitive government information well into the 21st century. It replaces the aging DES, which NIST adopted in 1977 as a Federal Information Processing Standard used by federal agencies to protect sensitive, unclassified information. The 128-bit AES standard has 3.4 × 1038 possible 128-bit key variants, and it has been estimated that it would take 1,000 trillion years to crack 128-bit AES cipher text using exhaustive techniques. Keys are stored (securely) in low power flash devices in nonvolatile flash memory. All programming files sent to the device can be authenticated by the part prior to programming to ensure that bad programming data is not loaded into the part that may possibly damage it. All programming verification is performed on-chip, ensuring that the contents of low power flash devices remain secure. Microsemi has implemented the 128-bit AES (Rijndael) algorithm in low power flash devices. With this key size, there are approximately 3.4 × 1038 possible 128-bit keys. DES has a 56-bit key size, which provides approximately 7.2 × 1016 possible keys. In their AES fact sheet, the National Institute of Standards and Technology uses the following hypothetical example to illustrate the theoretical security provided by AES. If one were to assume that a computing system existed that could recover a DES key in a second, it would take that same machine approximately 149 trillion years to crack a 128-bit AES key. NIST continues to make their point by stating the universe is believed to be less than 20 billion years old.1 Revision 4 305 Security in Low Power Flash Devices The AES key is securely stored on-chip in dedicated low power flash device flash memory and cannot be read out. In the first step, the AES key is generated and programmed into the device (for example, at a secure or trusted programming site). The Microsemi Designer software tool provides AES key generation capability. After the key has been programmed into the device, the device will only correctly decrypt programming files that have been encrypted with the same key. If the individual programming file content is incorrect, a Message Authentication Control (MAC) mechanism inside the device will fail in authenticating the programming file. In other words, when an encrypted programming file is being loaded into a device that has a different programmed AES key, the MAC will prevent this incorrect data from being loaded, preventing possible device damage. See Figure 12-3 on page 304 and Figure 12-4 on page 306 for graphical representations of this process. It is important to note that the user decides what level of protection will be implemented for the device. When AES protection is desired, the FlashLock Pass Key must be set. The AES key is a content protection mechanism, whereas the FlashLock Pass Key is a device protection mechanism. When the AES key is programmed into the device, the device still needs the Pass Key to protect the FPGA and FlashROM contents and the security settings, including the AES key. Using the FlashLock Pass Key prevents modification of the design contents by means of simply programming the device with a different AES key. AES Decryption and MAC Authentication Low power flash devices have a built-in 128-bit AES decryption core, which decrypts the encrypted programming file and performs a MAC check that authenticates the file prior to programming. MAC authenticates the entire programming data stream. After AES decryption, the MAC checks the data to make sure it is valid programming data for the device. This can be done while the device is still operating. If the MAC validates the file, the device will be erased and programmed. If the MAC fails to validate, then the device will continue to operate uninterrupted. This will ensure the following: • Correct decryption of the encrypted programming file • Prevention of erroneous or corrupted data being programmed during the programming file transfer • Correct bitstream passed to the device for decryption IGLOO and ProASIC3 MAC Validation Designer Software Programming File Generation with AES Encryption Decrypted Bitstream AES Key AES Decryption Core FlashROM FPGA Core Transmit Medium / Public Network Encrypted Bitstream Figure 12-4 • Example Application Scenario Using AES in IGLOO and ProASIC3 Devices 1. 306 National Institute of Standards and Technology, “ADVANCED ENCRYPTION STANDARD (AES) Questions and Answers,” 28 January 2002 (10 January 2005). See http://csrc.nist.gov/archive/aes/index1.html for more information. R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Fusion MAC Validation Designer Software Programming File Generation with AES Encryption Decrypted Bitstream AES Key AES Decryption Core FlashROM FPGA Core FBs Transmit Medium / Public Network Encrypted Bitstream Figure 12-5 • Example Application Scenario Using AES in Fusion Devices FlashLock Additional Options for IGLOO and ProASIC3 Devices The user also has the option of prohibiting Write operations to the FPGA array but allowing Verify operations on the FPGA array and/or Read operations on the FlashROM without the use of the FlashLock Pass Key. This option provides the user the freedom of verifying the FPGA array and/or reading the FlashROM contents after the device is programmed, without having to provide the FlashLock Pass Key. The user can incorporate AES encryption on the programming files to better enhance the level of security used. Permanent Security Setting Options In applications where a permanent lock is not desired, yet the security settings should not be modifiable, IGLOO and ProASIC3 devices can accommodate this requirement. This application is particularly useful in cases where a device is located at a remote location and must be reprogrammed with a design or data update. Refer to the "Application 3: Nontrusted Environment—Field Updates/Upgrades" section on page 310 for further discussion and examples of how this can be achieved. The user must be careful when considering the Permanent FlashLock or Permanent Security Settings option. Once the design is programmed with the permanent settings, it is not possible to reconfigure the security settings already employed on the device. Therefore, exercise careful consideration before programming permanent settings. Permanent FlashLock The purpose of the permanent lock feature is to provide the benefits of the highest level of security to IGLOO and ProASIC3 devices. If selected, the permanent FlashLock feature will create a permanent barrier, preventing any access to the contents of the device. This is achieved by permanently disabling Write and Verify access to the array, and Write and Read access to the FlashROM. After permanently locking the device, it has been effectively rendered one-time-programmable. This feature is useful if the intended applications do not require design or system updates to the device. Revision 4 307 Security in Low Power Flash Devices Security in Action This section illustrates some applications of the security advantages of Microsemi’s devices (Figure 12-6). . Plaintext Source File AES Encryption Cipher Text Source File Application 3 Application 2 Application 1 Public Domain AES Decryption Core FlashROM FPGA Core Flash Device Note: Flash blocks are only used in Fusion devices Figure 12-6 • Security Options 308 R e vi s i o n 4 Flash Blocks ProASIC3L FPGA Fabric User’s Guide Application 1: Trusted Environment As illustrated in Figure 12-7, this application allows the programming of devices at design locations where research and development take place. Therefore, encryption is not necessary and is optional to the user. This is often a secure way to protect the design, since the design program files are not sent elsewhere. In situations where production programming is not available at the design location, programming centers (such as Microsemi In-House Programming) provide a way of programming designs at an alternative, secure, and trusted location. In this scenario, the user generates a STAPL programming file from the Designer software in plaintext format, containing information on the entire design or the portion of the design to be programmed. The user can choose to employ the FlashLock Pass Key feature with the design. Once the design is programmed to unprogrammed devices, the design is protected by this FlashLock Pass Key. If no future programming is needed, the user can consider permanently securing the IGLOO and ProASIC3 device, as discussed in the "Permanent FlashLock" section on page 307. Application 2: Nontrusted Environment—Unsecured Location Often, programming of devices is not performed in the same location as actual design implementation, to reduce manufacturing cost. Overseas programming centers and contract manufacturers are examples of this scenario. To achieve security in this case, the AES key and the FlashLock Pass Key can be initially programmed in-house (trusted environment). This is done by generating a programming file with only the security settings and no design contents. The design FPGA core, FlashROM, and (for Fusion) FB contents are generated in a separate programming file. This programming file must be set with the same AES key that was used to program to the device previously so the device will correctly decrypt this encrypted programming file. As a result, the encrypted design content programming file can be safely sent off-site to nontrusted programming locations for design programming. Figure 12-7 shows a more detailed flow for this application. OEM Trusted Environment Generates and Programs Security Settings Only (programming of the security keys) Generates Design Contents Encrypted with AES Flash Device Security Settings* Security Settings FPGA/FlashROM/FBs FPGA/FlashROM/FBs Flash Device AES and/or Pass Key Protected Programming File Sends File(s) to Manufacturer Ships Devices to Manufacturer Returns Programmed Devices to Vendor Ships Programmed Devices to End Customer Security Settings Programs Design Contents to Devices FPGA/FlashROM/FBs Contents Flash Device OEM Customers Nontrusted Manufacturing Environment Notes: 1. Programmed portion indicated with dark gray. 2. Programming of FBs applies to Fusion only. Figure 12-7 • Application 2: Device Programming in a Nontrusted Environment Revision 4 309 Security in Low Power Flash Devices Application 3: Nontrusted Environment—Field Updates/Upgrades Programming or reprogramming of devices may occur at remote locations. Reconfiguration of devices in consumer products/equipment through public networks is one example. Typically, the remote system is already programmed with particular design contents. When design update (FPGA array contents update) and/or data upgrade (FlashROM and/or FB contents upgrade) is necessary, an updated programming file with AES encryption can be generated, sent across public networks, and transmitted to the remote system. Reprogramming can then be done using this AES-encrypted programming file, providing easy and secure field upgrades. Low power flash devices support this secure ISP using AES. The detailed flow for this application is shown in Figure 12-8. Refer to the "Microprocessor Programming of Microsemi’s Low Power Flash Devices" chapter of an appropriate FPGA fabric user’s guide for more information. To prepare devices for this scenario, the user can initially generate a programming file with the available security setting options. This programming file is programmed into the devices before shipment. During the programming file generation step, the user has the option of making the security settings permanent or not. In situations where no changes to the security settings are necessary, the user can select this feature in the software to generate the programming file with permanent security settings. Microsemi recommends that the programming file use encryption with an AES key, especially when ISP is done via public domain. For example, if the designer wants to use an AES key for the FPGA array and the FlashROM, Permanent needs to be chosen for this setting. At first, the user chooses the options to use an AES key for the FPGA array and the FlashROM, and then chooses Permanently lock the security settings. A unique AES key is chosen. Once this programming file is generated and programmed to the devices, the AES key is permanently stored in the on-chip memory, where it is secured safely. The devices are sent to distant locations for the intended application. When an update is needed, a new programming file must be generated. The programming file must use the same AES key for encryption; otherwise, the authentication will fail and the file will not be programmed in the device. Trusted Environment OEM Generates Updated Design Contents Encrypted with AES AES Encrypted Programming File Transmits to Remote System Update/Upgrade Flash Device Original Design Contents AES Encrypted and FlashLock Pass Key Protected Remote Environment / System Figure 12-8 • Application 3: Nontrusted Environment—Field Updates/Upgrades 310 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide FlashROM Security Use Models Each of the subsequent sections describes in detail the available selections in Microsemi Designer as an aid to understanding security applications and generating appropriate programming files for those applications. Before proceeding, it is helpful to review Figure 12-7 on page 309, which gives a general overview of the programming file generation flow within the Designer software as well as what occurs during the device programming stage. Specific settings are discussed in the following sections. In Figure 12-7 on page 309, the flow consists of two sub-flows. Sub-flow 1 describes programming security settings to the device only, and sub-flow 2 describes programming the design contents only. In Application 1, described in the "Application 1: Trusted Environment" section on page 309, the user does not need to generate separate files but can generate one programming file containing both security settings and design contents. Then programming of the security settings and design contents is done in one step. Both sub-flow 1 and sub-flow 2 are used. In Application 2, described in the "Application 2: Nontrusted Environment—Unsecured Location" section on page 309, the trusted site should follow sub-flows 1 and 2 separately to generate two separate programming files. The programming file from sub-flow 1 will be used at the trusted site to program the device(s) first. The programming file from sub-flow 2 will be sent off-site for production programming. In Application 3, described in the "Application 3: Nontrusted Environment—Field Updates/Upgrades" section on page 310, typically only sub-flow 2 will be used, because only updates to the design content portion are needed and no security settings need to be changed. In the event that update of the security settings is necessary, see the "Reprogramming Devices" section on page 321 for details. For more information on programming low power flash devices, refer to the "InSystem Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X" section on page 327. Revision 4 311 Security in Low Power Flash Devices User Designer Software 1 Program Security Settings Programming Software User Assigns Desired Security Settings To FPGA/FlashROM/FB/All: – AES Key and FlashLock Pass Key – FlashLock Pass Key Only Software Generates Programming File with Desired Security Settings: – Encrypted with AES and Protected with FlashLock Pass Key – Protected with FlashLock Pass Key Only Device Previously Programmed? No Software Programs Selected Security Settings into Device Yes Yes Software Performs Comparison of FlashLock Pass Key between Programming File and Device Does FlashLock Pass Key Match? No Returns Error 2 Program Design Contents Programming Previously Secured Device(s)? No Software Generates Programming File with Desired Design Contents (FPGA Array, FlashROM, FB, or All) Yes AES Key Used Previously? No User Must Reassign Exact FlashLock Pass Key Previously Programmed into the Device Design Content Programmed into Device Yes Software Performs Comparison of FlashLock Pass Key between Programming File and Device No Yes User Must Reassign Exact AES Key Previously Programmed into the Device Does FlashLock Pass Key Match? Returns Error Software Generates Programming File with FlashLock Pass Key and Design Contents No Encrypted Design Content Passes through MAC for Authentication Correct? Yes Software Generates Programming File with Encrypted Design Contents Design Content Decrypted and Programmed into Device Note: If programming the Security Header only, just perform sub-flow 1. If programming design content only, just perform sub-flow 2. Figure 12-9 • Security Programming Flows 312 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Generating Programming Files Generation of the Programming File in a Trusted Environment— Application 1 As discussed in the "Application 1: Trusted Environment" section on page 309, in a trusted environment, the user can choose to program the device with plaintext bitstream content. It is possible to use plaintext for programming even when the FlashLock Pass Key option has been selected. In this application, it is not necessary to employ AES encryption protection. For AES encryption settings, refer to the next sections. The generated programming file will include the security setting (if selected) and the plaintext programming file content for the FPGA array, FlashROM, and/or FBs. These options are indicated in Table 12-2 and Table 12-3. Table 12-2 • IGLOO and ProASIC3 Plaintext Security Options, No AES FlashROM Only FPGA Core Only Both FlashROM and FPGA No AES / no FlashLock ✓ ✓ ✓ FlashLock only ✓ ✓ ✓ AES and FlashLock – – – Security Protection Table 12-3 • Fusion Plaintext Security Options Security Protection FlashROM Only FPGA Core Only FB Core Only All No AES / no FlashLock ✓ ✓ ✓ ✓ FlashLock ✓ ✓ ✓ ✓ AES and FlashLock – – – – Note: For all instructions, the programming of Flash Blocks refers to Fusion only. For this scenario, generate the programming file as follows: 1. Select the Silicon features to be programmed (Security Settings, FPGA Array, FlashROM, Flash Memory Blocks), as shown in Figure 12-10 on page 314 and Figure 12-11 on page 314. Click Next. If Security Settings is selected (i.e., the FlashLock security Pass Key feature), an additional dialog will be displayed to prompt you to select the security level setting. If no security setting is selected, you will be directed to Step 3. Revision 4 313 Security in Low Power Flash Devices Figure 12-10 • All Silicon Features Selected for IGLOO and ProASIC3 Devices Figure 12-11 • All Silicon Features Selected for Fusion 314 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 2. Choose the appropriate security level setting and enter a FlashLock Pass Key. The default is the Medium security level (Figure 12-12). Click Next. If you want to select different options for the FPGA and/or FlashROM, this can be set by clicking Custom Level. Refer to the "Advanced Options" section on page 322 for different custom security level options and descriptions of each. Figure 12-12 • Medium Security Level Selected for Low Power Flash Devices Revision 4 315 Security in Low Power Flash Devices 3. Choose the desired settings for the FlashROM configurations to be programmed (Figure 12-13). Click Finish to generate the STAPL programming file for the design. Figure 12-13 • FlashROM Configuration Settings for Low Power Flash Devices Generation of Security Header Programming File Only— Application 2 As mentioned in the "Application 2: Nontrusted Environment—Unsecured Location" section on page 309, the designer may employ FlashLock Pass Key protection or FlashLock Pass Key with AES encryption on the device before sending it to a nontrusted or unsecured location for device programming. To achieve this, the user needs to generate a programming file containing only the security settings desired (Security Header programming file). Note: If AES encryption is configured, FlashLock Pass Key protection must also be configured. The available security options are indicated in Table 12-4 and Table 12-5 on page 317. Table 12-4 • FlashLock Security Options for IGLOO and ProASIC3 FlashROM Only FPGA Core Only Both FlashROM and FPGA No AES / no FlashLock – – – FlashLock only ✓ ✓ ✓ AES and FlashLock ✓ ✓ ✓ Security Option 316 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 12-5 • FlashLock Security Options for Fusion Security Option FlashROM Only FPGA Core Only FB Core Only All No AES / no FlashLock – – – – FlashLock ✓ ✓ ✓ ✓ AES and FlashLock ✓ ✓ ✓ ✓ For this scenario, generate the programming file as follows: 1. Select only the Security settings option, as indicated in Figure 12-14 and Figure 12-15 on page 318. Click Next. Figure 12-14 • Programming IGLOO and ProASIC3 Security Settings Only Revision 4 317 Security in Low Power Flash Devices Figure 12-15 • Programming Fusion Security Settings Only 2. Choose the desired security level setting and enter the key(s). – The High security level employs FlashLock Pass Key with AES Key protection. – The Medium security level employs FlashLock Pass Key protection only. Figure 12-16 • High Security Level to Implement FlashLock Pass Key and AES Key Protection 318 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Table 12-6 and Table 12-7 show all available options. If you want to implement custom levels, refer to the "Advanced Options" section on page 322 for information on each option and how to set it. 3. When done, click Finish to generate the Security Header programming file. Table 12-6 • All IGLOO and ProASIC3 Header File Security Options FlashROM Only FPGA Core Only Both FlashROM and FPGA No AES / no FlashLock ✓ ✓ ✓ FlashLock only ✓ ✓ ✓ AES and FlashLock ✓ ✓ ✓ Security Option Note: ✓ = options that may be used Table 12-7 • All Fusion Header File Security Options Security Option FlashROM Only FPGA Core Only FB Core Only All No AES / No FlashLock ✓ ✓ ✓ ✓ FlashLock ✓ ✓ ✓ ✓ AES and FlashLock ✓ ✓ ✓ ✓ Generation of Programming Files with AES Encryption— Application 3 This section discusses how to generate design content programming files needed specifically at unsecured or remote locations to program devices with a Security Header (FlashLock Pass Key and AES key) already programmed ("Application 2: Nontrusted Environment—Unsecured Location" section on page 309 and "Application 3: Nontrusted Environment—Field Updates/Upgrades" section on page 310). In this case, the encrypted programming file must correspond to the AES key already programmed into the device. If AES encryption was previously selected to encrypt the FlashROM, FBs, and FPGA array, AES encryption must be set when generating the programming file for them. AES encryption can be applied to the FlashROM only, the FBs only, the FPGA array only, or all. The user must ensure both the FlashLock Pass Key and the AES key match those already programmed to the device(s), and all security settings must match what was previously programmed. Otherwise, the encryption and/or device unlocking will not be recognized when attempting to program the device with the programming file. The generated programming file will be AES-encrypted. In this scenario, generate the programming file as follows: 1. Deselect Security settings and select the portion of the device to be programmed (Figure 12-17 on page 320). Select Programming previously secured device(s). Click Next. Revision 4 319 Security in Low Power Flash Devices Note: The settings in this figure are used to show the generation of an AES-encrypted programming file for the FPGA array, FlashROM, and FB contents. One or all locations may be selected for encryption. Figure 12-17 • Settings to Program a Device Secured with FlashLock and using AES Encryption Choose the High security level to reprogram devices using both the FlashLock Pass Key and AES key protection (Figure 12-18 on page 321). Enter the AES key and click Next. A device that has already been secured with FlashLock and has an AES key loaded must recognize the AES key to program the device and generate a valid bitstream in authentication. The FlashLock Key is only required to unlock the device and change the security settings. This is what makes it possible to program in an untrusted environment. The AES key is protected inside the device by the FlashLock Key, so you can only program if you have the correct AES key. In fact, the AES key is not in the programming file either. It is the key used to encrypt the data in the file. The same key previously programmed with the FlashLock Key matches to decrypt the file. An AES-encrypted file programmed to a device without FlashLock would not be secure, since without FlashLock to protect the AES key, someone could simply reprogram the AES key first, then program with any AES key desired or no AES key at all. This option is therefore not available in the software. 320 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Figure 12-18 • Security Level Set High to Reprogram Device with AES Key Programming with this file is intended for an unsecured environment. The AES key encrypts the programming file with the same AES key already used in the device and utilizes it to program the device. Reprogramming Devices Previously programmed devices can be reprogrammed using the steps in the "Generation of the Programming File in a Trusted Environment—Application 1" section on page 313 and "Generation of Security Header Programming File Only—Application 2" section on page 316. In the case where a FlashLock Pass Key has been programmed previously, the user must generate the new programming file with a FlashLock Pass Key that matches the one previously programmed into the device. The software will check the FlashLock Pass Key in the programming file against the FlashLock Pass Key in the device. The keys must match before the device can be unlocked to perform further programming with the new programming file. Figure 12-10 on page 314 and Figure 12-11 on page 314 show the option Programming previously secured device(s), which the user should select before proceeding. Upon going to the next step, the user will be notified that the same FlashLock Pass Key needs to be entered, as shown in Figure 12-19 on page 322. Revision 4 321 Security in Low Power Flash Devices Figure 12-19 • FlashLock Pass Key, Previously Programmed Devices It is important to note that when the security settings need to be updated, the user also needs to select the Security settings check box in Step 1, as shown in Figure 12-10 on page 314 and Figure 12-11 on page 314, to modify the security settings. The user must consider the following: • If only a new AES key is necessary, the user must re-enter the same Pass Key previously programmed into the device in Designer and then generate a programming file with the same Pass Key and a different AES key. This ensures the programming file can be used to access and program the device and the new AES key. • If a new Pass Key is necessary, the user can generate a new programming file with a new Pass Key (with the same or a new AES key if desired). However, for programming, the user must first load the original programming file with the Pass Key that was previously used to unlock the device. Then the new programming file can be used to program the new security settings. Advanced Options As mentioned, there may be applications where more complicated security settings are required. The “Custom Security Levels” section in the FlashPro User's Guide describes different advanced options available to aid the user in obtaining the best available security settings. 322 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Programming File Header Definition In each STAPL programming file generated, there will be information about how the AES key and FlashLock Pass Key are configured. Table 12-8 shows the header definitions in STAPL programming files for different security levels. Table 12-8 • STAPL Programming File Header Definitions by Security Level Security Level STAPL File Header Definition No security (no FlashLock Pass Key or AES key) NOTE "SECURITY" "Disable"; FlashLock Pass Key with no AES key NOTE "SECURITY" "KEYED "; FlashLock Pass Key with AES key NOTE "SECURITY" "KEYED ENCRYPT "; Permanent Security Settings option enabled NOTE "SECURITY" "PERMLOCK ENCRYPT "; AES-encrypted FPGA array (for programming updates) NOTE "SECURITY" "ENCRYPT CORE "; AES-encrypted FlashROM (for programming updates) NOTE "SECURITY" "ENCRYPT FROM "; AES-encrypted FPGA programming updates) array and FlashROM (for NOTE "SECURITY" "ENCRYPT FROM CORE "; Example File Headers STAPL Files Generated with FlashLock Key and AES Key Containing Key Information • FlashLock Key / AES key indicated in STAPL file header definition • Intended ONLY for secured/trusted environment programming applications ============================================= NOTE "CREATOR" "Designer Version: 6.1.1.108"; NOTE "DEVICE" "A3PE600"; NOTE "PACKAGE" "208 PQFP"; NOTE "DATE" "2005/04/08"; NOTE "STAPL_VERSION" "JESD71"; NOTE "IDCODE" "$123261CF"; NOTE "DESIGN" "counter32"; NOTE "CHECKSUM" "$EDB9"; NOTE "SAVE_DATA" "FRomStream"; NOTE "SECURITY" "KEYED ENCRYPT "; NOTE "ALG_VERSION" "1"; NOTE "MAX_FREQ" "20000000"; NOTE "SILSIG" "$00000000"; NOTE "PASS_KEY" "$00123456789012345678901234567890"; NOTE "AES_KEY" "$ABCDEFABCDEFABCDEFABCDEFABCDEFAB"; ============================================== Revision 4 323 Security in Low Power Flash Devices STAPL File with AES Encryption • Does not contain AES key / FlashLock Key information • Intended for transmission through web or service to unsecured locations for programming ============================================= NOTE "CREATOR" "Designer Version: 6.1.1.108"; NOTE "DEVICE" "A3PE600"; NOTE "PACKAGE" "208 PQFP"; NOTE "DATE" "2005/04/08"; NOTE "STAPL_VERSION" "JESD71"; NOTE "IDCODE" "$123261CF"; NOTE "DESIGN" "counter32"; NOTE "CHECKSUM" "$EF57"; NOTE "SAVE_DATA" "FRomStream"; NOTE "SECURITY" "ENCRYPT FROM CORE "; NOTE "ALG_VERSION" "1"; NOTE "MAX_FREQ" "20000000"; NOTE "SILSIG" "$00000000"; Conclusion The new and enhanced security features offered in Fusion, IGLOO, and ProASIC3 devices provide stateof-the-art security to designs programmed into these flash-based devices. Microsemi low power flash devices employ the encryption standard used by NIST and the U.S. government—AES using the 128-bit Rijndael algorithm. The combination of an on-chip AES decryption engine and FlashLock technology provides the highest level of security against invasive attacks and design theft, implementing the most robust and secure ISP solution. These security features protect IP within the FPGA and protect the system from cloning, wholesale “black box” copying of a design, invasive attacks, and explicit IP or data theft. Glossary Term Explanation Security Header programming file Programming file used to program the FlashLock Pass Key and/or AES key into the device to secure the FPGA, FlashROM, and/or FBs. AES (encryption) key 128-bit key defined by the user when the AES encryption option is set in the Microsemi Designer software when generating the programming file. FlashLock Pass Key 128-bit key defined by the user when the FlashLock option is set in the Microsemi Designer software when generating the programming file. The FlashLock Key protects the security settings programmed to the device. Once a device is programmed with FlashLock, whatever settings were chosen at that time are secure. FlashLock The combined security features that protect the device content from attacks. These features are the following: • Flash technology that does not require an external bitstream to program the device • FlashLock Pass Key that secures device content by locking the security settings and preventing access to the device as defined by the user • AES key that allows secure, encrypted device reprogrammability References National Institute of Standards and Technology. “ADVANCED ENCRYPTION STANDARD (AES) Questions and Answers.” 28 January 2002 (10 January 2005). See http://csrc.nist.gov/archive/aes/index1.html for more information. 324 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Related Documents User’s Guides FlashPro User's Guide http://www.microsemi.com/soc/documents/flashpro_ug.pdf List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.5 (August 2009) The "CoreMP7 Device Security" section was removed from "Security in ARMEnabled Low Power Flash Devices", since M7-enabled devices are no longer supported. 304 v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 12-1 • Flash-Based FPGAs. 302 v1.3 (October 2008) The "Security Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 302 v1.2 (June 2008) The following changes were made to the family descriptions in Table 12-1 • FlashBased FPGAs: 302 v1.1 (March 2008) • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The chapter was updated to include the IGLOO PLUS family and information regarding 15 k gate devices. N/A The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 302 Revision 4 325 13 – In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X Introduction Microsemi’s low power flash devices are all in-system programmable. This document describes the general requirements for programming a device and specific requirements for the FlashPro4/3/3X programmers1. IGLOO, ProASIC3, SmartFusion, and Fusion devices offer a low power, single-chip, live-at-power-up solution with the ASIC advantages of security and low unit cost through nonvolatile flash technology. Each device contains 1 kbit of on-chip, user-accessible, nonvolatile FlashROM. The FlashROM can be used in diverse system applications such as Internet Protocol (IP) addressing, user system preference storage, device serialization, or subscription-based business models. IGLOO, ProASIC3, SmartFusion, and Fusion devices offer the best in-system programming (ISP) solution, FlashLock® security features, and AES-decryption-based ISP. ISP Architecture Low power flash devices support ISP via JTAG and require a single VPUMP voltage of 3.3 V during programming. In addition, programming via a microcontroller in a target system is also supported. Refer to the "Microprocessor Programming of Microsemi’s Low Power Flash Devices" chapter of an appropriate FPGA fabric user’s guide. Family-specific support: • ProASIC3, ProASIC3E, SmartFusion, and Fusion devices support ISP. • ProASIC3L devices operate using a 1.2 V core voltage; however, programming can be done only at 1.5 V. Voltage switching is required in-system to switch from a 1.2 V core to 1.5 V core for programming. • IGLOO and IGLOOe V5 devices can be programmed in-system when the device is using a 1.5 V supply voltage to the FPGA core. • IGLOO nano V2 devices can be programmed at 1.2 V core voltage (when using FlashPro4 only) or 1.5 V. IGLOO nano V5 devices are programmed with a VCC core voltage of 1.5 V. Voltage switching is required in-system to switch from a 1.2 V supply (VCC,VCCI, and VJTAG) to 1.5 V for programming. The exception is that V2 devices can be programmed at 1.2 V VCC with FlashPro4. IGLOO devices cannot be programmed in-system when the device is in Flash*Freeze mode. The device should exit Flash*Freeze mode and be in normal operation for programming to start. Programming operations in IGLOO devices can be achieved when the device is in normal operating mode and a 1.5 V core voltage is used. JTAG 1532 IGLOO, ProASIC3, SmartFusion, and Fusion devices support the JTAG-based IEEE 1532 standard for ISP. To start JTAG operations, the IGLOO device must exit Flash*Freeze mode and be in normal operation before starting to send JTAG commands to the device. As part of this support, when a device is in an unprogrammed state, all user I/O pins are disabled. This is achieved by keeping the global IO_EN 1. FlashPro4 replaced FlashPro3/3X in 2010 and is backward compatible with FlashPro3/3X as long as there is no connection to pin 4 on the JTAG header on the board. On FlashPro3/3X, there is no connection to pin 4 on the JTAG header; however, pin 4 is used for programming mode (Prog_Mode) on FlashPro4. When converting from FlashPro3/3X to FlashPro4, users should make sure that JTAG connectors on system boards do not have any connection to pin 4. FlashPro3X supports discrete TCK toggling that is needed to support non-JTAG compliant devices in the chain. This feature is included in FlashPro4. Revision 4 327 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X signal deactivated, which also has the effect of disabling the input buffers. The SAMPLE/PRELOAD instruction captures the status of pads in parallel and shifts them out as new data is shifted in for loading into the Boundary Scan Register (BSR). When the device is in an unprogrammed state, the OE and output BSR will be undefined; however, the input BSR will be defined as long as it is connected and being used. For JTAG timing information on setup, hold, and fall times, refer to the FlashPro User’s Guide. ISP Support in Flash-Based Devices The flash FPGAs listed in Table 13-1 support the ISP feature and the functions described in this document. Table 13-1 • Flash-Based FPGAs Supporting ISP Family* Series IGLOO ProASIC3 Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications SmartFusion SmartFusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable microcontroller subsystem (MSS) which includes programmable analog and an ARM® Cortex™-M3 hard processor and flash memory in a monolithic device Fusion Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device ProASIC ProASIC First generation ProASIC devices ProASICPLUS Second generation ProASIC devices Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 13-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 13-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 328 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Programming Voltage (VPUMP) and VJTAG Low-power flash devices support on-chip charge pumps, and therefore require only a single 3.3 V programming voltage for the VPUMP pin during programming. When the device is not being programmed, the VPUMP pin can be left floating or can be tied (pulled up) to any voltage between 0 V and 3.6 V2. During programming, the target board or the FlashPro4/3/3X programmer can provide VPUMP. FlashPro4/3/3X is capable of supplying VPUMP to a single device. If more than one device is to be programmed using FlashPro4/3/3X on a given board, FlashPro4/3/3X should not be relied on to supply the VPUMP voltage. A FlashPro4/3/3X programmer is not capable of providing reliable VJTAG voltage. The board must supply VJTAG voltage to the device and the VJTAG pin of the programmer header must be connected to the device VJTAG pin. Microsemi recommends that VPUMP3 and VJTAG power supplies be kept separate with independent filtering capacitors rather than supplying them from a common rail. Refer to the "Board-Level Considerations" section on page 337 for capacitor requirements. Low power flash device I/Os support a bank-based, voltage-supply architecture that simultaneously supports multiple I/O voltage standards (Table 13-2). By isolating the JTAG power supply in a separate bank from the user I/Os, low power flash devices provide greater flexibility with supply selection and simplify power supply and printed circuit board (PCB) design. The JTAG pins can be run at any voltage from 1.5 V to 3.3 V (nominal). Microsemi recommends that TCK be tied to GND through a 200 ohm to 1 Kohm resistor. This prevents a possible totempole current on the input buffer stage. For TDI, TMS, and TRST pins, the devices provide an internal nominal 10 Kohm pull-up resistor. During programming, all I/O pins, except for JTAG interface pins, are tristated and weakly pulled up to VCCI. This isolates the part and prevents the signals from floating. The JTAG interface pins are driven by the FlashPro4/3/3X during programming, including the TRST pin, which is driven HIGH. Table 13-2 • Power Supplies Programming Mode Current during Programming VCC 1.2 V / 1.5 V < 70 mA VCCI 1.2 V / 1.5 V / 1.8 V / 2.5 V / 3.3 V (bank-selectable) I/Os are weakly pulled up. VJTAG 1.2 V / 1.5 V / 1.8 V / 2.5 V / 3.3 V < 20 mA VPUMP 3.15 V to 3.45 V < 80 mA Power Supply Note: All supply voltages should be at 1.5 V or higher, regardless of the setting during normal operation, except for IGLOO nano, where 1.2 V VCC and VJTAG programming is allowed. Nonvolatile Memory (NVM) Programming Voltage SmartFusion and Fusion devices need stable VCCNVM/VCCENVM3 (1.5 V power supply to the embedded nonvolatile memory blocks) and VCCOSC/VCCROSC4 (3.3 V power supply to the integrated RC oscillator). The tolerance of VCCNVM/VCCENVM is ± 5% and VCCOSC/VCCROSC is ± 5%. Unstable supply voltage on these pins can cause an NVM programming failure due to NVM page corruption. The NVM page can also be corrupted if the NVM reset pin has noise. This signal must be tied off properly. Microsemi recommends installing the following capacitors5 on the VCCNVM/VCCENVM and VCCOSC/VCCROSC pins: 2. 3. 4. 5. • Add one bypass capacitor of 10 µF for each power supply plane followed by an array of decoupling capacitors of 0.1 µF. • Add one 0.1 µF capacitor near each pin. During sleep mode in IGLOO devices connect VPUMP to GND. VPUMP has to be quiet for successful programming. Therefore VPUMP must be separate and required capacitors must be installed close to the FPGA VPUMP pin. VCCROSC is for SmartFusion. The capacitors cannot guarantee reliable operation of the device if the board layout is not done properly. Revision 4 329 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X IEEE 1532 (JTAG) Interface The supported industry-standard IEEE 1532 programming interface builds on the IEEE 1149.1 (JTAG) standard. IEEE 1532 defines the standardized process and methodology for ISP. Both silicon and software issues are addressed in IEEE 1532 to create a simplified ISP environment. Any IEEE 1532 compliant programmer can be used to program low power flash devices. Device serialization is not supported when using the IEEE1532 standard. Refer to the standard for detailed information about IEEE 1532. Security Unlike SRAM-based FPGAs that require loading at power-up from an external source such as a microcontroller or boot PROM, Microsemi nonvolatile devices are live at power-up, and there is no bitstream required to load the device when power is applied. The unique flash-based architecture prevents reverse engineering of the programmed code on the device, because the programmed data is stored in nonvolatile memory cells. Each nonvolatile memory cell is made up of small capacitors and any physical deconstruction of the device will disrupt stored electrical charges. Each low power flash device has a built-in 128-bit Advanced Encryption Standard (AES) decryption core, except for the 30 k gate devices and smaller. Any FPGA core or FlashROM content loaded into the device can optionally be sent as encrypted bitstream and decrypted as it is loaded. This is particularly suitable for applications where device updates must be transmitted over an unsecured network such as the Internet. The embedded AES decryption core can prevent sensitive data from being intercepted (Figure 13-1 on page 331). A single 128-bit AES Key (32 hex characters) is used to encrypt FPGA core programming data and/or FlashROM programming data in the Microsemi tools. The low power flash devices also decrypt with a single 128-bit AES Key. In addition, low power flash devices support a Message Authentication Code (MAC) for authentication of the encrypted bitstream on-chip. This allows the encrypted bitstream to be authenticated and prevents erroneous data from being programmed into the device. The FPGA core, FlashROM, and Flash Memory Blocks (FBs), in Fusion only, can be updated independently using a programming file that is AES-encrypted (cipher text) or uses plain text. 330 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Security in ARM-Enabled Low Power Flash Devices There are slight differences between the regular flash device and the ARM-enabled flash devices, which have the M1 prefix. The AES key is used by Microsemi and preprogrammed into the device to protect the ARM IP. As a result, the design will be encrypted along with the ARM IP, according to the details below. Cortex-M1 and Cortex-M3 Device Security Cortex-M1–enabled and Cortex-M3 devices are shipped with the following security features: • FPGA array enabled for AES-encrypted programming and verification • FlashROM enabled for AES-encrypted write and verify • Embedded Flash Memory enabled for AES encrypted write Flash Device Designer Software MAC Validation User Encryption AES Key Decrypted Bitstream Programming File Generation with AES Encryption AES Decryption FPGA Core, FlashROM, FBs Transmit Medium / Public Network Encrypted Bistream Figure 13-1 • AES-128 Security Features Revision 4 331 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X Figure 13-2 shows different applications for ISP programming. 1. In a trusted programming environment, you can program the device using the unencrypted (plaintext) programming file. 2. You can program the AES Key in a trusted programming environment and finish the final programming in an untrusted environment using the AES-encrypted (cipher text) programming file. 3. For the remote ISP updating/reprogramming, the AES Key stored in the device enables the encrypted programming bitstream to be transmitted through the untrusted network connection. Microsemi low power flash devices also provide the unique Microsemi FlashLock feature, which protects the Pass Key and AES Key. Unless the original FlashLock Pass Key is used to unlock the device, security settings cannot be modified. Microsemi does not support read-back of FPGA core-programmed data; however, the FlashROM contents can selectively be read back (or disabled) via the JTAG port based on the security settings established by the Microsemi Designer software. Refer to the "Security in Low Power Flash Devices" section on page 301 for more information. Source Plain Text AES Encryption Source Encrypted Bitstream FlashROM Option 3 Option 2 Option 1 TCP/IP AES Decryption FPGA Core IGLOO or ProASIC3 Device Figure 13-2 • Different ISP Use Models 332 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide FlashROM and Programming Files Each low power flash device has 1 kbit of on-chip, nonvolatile flash memory that can be accessed from the FPGA core. This nonvolatile FlashROM is arranged in eight pages of 128 bits (Figure 13-3). Each page can be programmed independently, with or without the 128-bit AES encryption. The FlashROM can only be programmed via the IEEE 1532 JTAG port and cannot be programmed from the FPGA core. In addition, during programming of the FlashROM, the FPGA core is powered down automatically by the on-chip programming control logic. Page Number 15 14 13 12 Byte Number in Page 11 10 9 8 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 Figure 13-3 • FlashROM Architecture When using FlashROM combined with AES, many subscription-based applications or device serialization applications are possible. The FROM configurator found in the Libero SoC Catalog supports easy management of the FlashROM contents, even over large numbers of devices. The FROM configurator can support FlashROM contents that contain the following: • Static values • Random numbers • Values read from a file • Independent updates of each page In addition, auto-incrementing of fields is possible. In applications where the FlashROM content is different for each device, you have the option to generate a single STAPL file for all the devices or individual serialization files for each device. For more information on how to generate the FlashROM content for device serialization, refer to the "FlashROM in Microsemi’s Low Power Flash Devices" section on page 133. Libero SoC includes a unique tool to support the generation and management of FlashROM and FPGA programming files. This tool is called FlashPoint. Depending on the applications, designers can use the FlashPoint software to generate a STAPL file with different contents. In each case, optional AES encryption and/or different security settings can be set. In Designer, when you click the Programming File icon, FlashPoint launches, and you can generate STAPL file(s) with four different cases (Figure 13-4 on page 334). When the serialization feature is used during the configuration of FlashROM, you can generate a single STAPL file that will program all the devices or an individual STAPL file for each device. The following cases present the FPGA core and FlashROM programming file combinations that can be used for different applications. In each case, you can set the optional security settings (FlashLock Pass Key and/or AES Key) depending on the application. 1. A single STAPL file or multiple STAPL files with multiple FlashROM contents and the FPGA core content. A single STAPL file will be generated if the device serialization feature is not used. You can program the whole FlashROM or selectively program individual pages. 2. A single STAPL file for the FPGA core content Revision 4 333 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X 3. A single STAPL file or multiple STAPL files with multiple FlashROM contents. A single STAPL file will be generated if the device serialization feature is not used. You can program the whole FlashROM or selectively program individual pages. 4. A single STAPL file to configure the security settings for the device, such as the AES Key and/or Pass Key. Libero SoC Catalog Designer Software Suite Netlist Programming File (FlashPoint) 1 2 FlashROM Configuration File (*.ufc) 3 Security Settings Security Settings Security Settings Single/Multiple FlashROM Content(s) FPGA Core Content Single/Multiple FlashROM Content(s) 4 Security Settings FPGA Core Content Figure 13-4 • Flexible Programming File Generation for Different Applications Programming Solution For device programming, any IEEE 1532–compliant programmer can be used; however, the FlashPro4/3/3X programmer must be used to control the low power flash device's rich security features and FlashROM programming options. The FlashPro4/3/3X programmer is a low-cost portable programmer for the Microsemi flash families. It can also be used with a powered USB hub for parallel programming. General specifications for the FlashPro4/3/3X programmer are as follows: 334 • Programming clock – TCK is used with a maximum frequency of 20 MHz, and the default frequency is 4 MHz. • Programming file – STAPL • Daisy chain – Supported. You can use the ChainBuilder software to build the programming file for the chain. • Parallel programming – Supported. Multiple FlashPro4/3/3X programmers can be connected together using a powered USB hub or through the multiple USB ports on the PC. • Power supply – The target board must provide VCC, VCCI, VPUMP, and VJTAG during programming. However, if there is only one device on the target board, the FlashPro4/3/3X programmer can generate the required VPUMP voltage from the USB port. R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide ISP Programming Header Information The FlashPro4/3/3X programming cable connector can be connected with a 10-pin, 0.1"-pitch programming header. The recommended programming headers are manufactured by AMP (103310-1) and 3M (2510-6002UB). If you have limited board space, you can use a compact programming header manufactured by Samtec (FTSH-105-01-L-D-K). Using this compact programming header, you are required to order an additional header adapter manufactured by Microsemi SoC Products Group (FP310PIN-ADAPTER-KIT). Existing ProASICPLUS family customers who are using the Samtec Small Programming Header (FTSH-113-01-L-D-K) and are planning to migrate to IGLOO or ProASIC3 devices can also use FP3-10PIN-ADAPTER-KIT. Table 13-3 • Programming Header Ordering Codes Manufacturer Part Number Description 103310-1 10-pin, 0.1"-pitch cable header (right-angle PCB mount angle) 2510-6002UB 10-pin, 0.1"-pitch cable header (straight PCB mount angle) Samtec FTSH-113-01-L-D-K Small programming header supported by FlashPro and Silicon Sculptor Samtec FTSH-105-01-L-D-K Compact programming header Samtec FFSD-05-D-06.00-01-N AMP 3M Microsemi 10-pin cable with 50 mil pitch sockets; included in FP310PIN-ADAPTER-KIT. FP3-10PIN-ADAPTER-KIT Transition adapter kit to allow FP3 to be connected to a micro 10-pin header (50 mil pitch). Includes a 6 inch Samtec FFSD-05-D-06.00-01-N cable in the kit. The transition adapter board was previously offered as FP3-26PIN-ADAPTER and includes a 26-pin adapter for design transitions from ProASICPLUS based boards to ProASIC3 based boards. TCK TDO TMS VPUMP TDI 1 2 3 4 5 6 7 8 9 10 GND NC (FlashPro3/3X); Prog_Mode* (FlashPro4) VJTAG TRST GND Note: *Prog_Mode on FlashPro4 is an output signal that goes High during device programming and returns to Low when programming is complete. This signal can be used to drive a system to provide a 1.5 V programming signal to IGLOO nano, ProASIC3L, and RT ProASIC3 devices that can run with 1.2 V core voltage but require 1.5 V for programming. IGLOO nano V2 devices can be programmed at 1.2 V core voltage (when using FlashPro4 only), but IGLOO nano V5 devices are programmed with a VCC core voltage of 1.5 V. Figure 13-5 • Programming Header (top view) Revision 4 335 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X Table 13-4 • Programming Header Pin Numbers and Description Pin Signal Source 1 TCK Programmer 2 GND1 – Signal Reference 3 TDO Target Board Test Data Output 4 NC – 5 TMS Programmer Test Mode Select 6 VJTAG Target Board JTAG Supply Voltage 2 Description JTAG Clock No Connect (FlashPro3/3X); Prog_Mode (FlashPro4). See note associated with Figure 13-5 on page 335 regarding Prog_Mode on FlashPro4. 7 VPUMP Programmer/Target Board 8 nTRST Programmer JTAG Test Reset (Hi-Z with 10 kΩ pull-down, HIGH, LOW, or toggling) 9 TDI Programmer Test Data Input 10 1 GND – Programming Supply Voltage Signal Reference Notes: 1. Both GND pins must be connected. 2. FlashPro4/3/3X can provide VPUMP if there is only one device on the target board. 336 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Board-Level Considerations A bypass capacitor is required from VPUMP to GND for all low power flash devices during programming. This bypass capacitor protects the devices from voltage spikes that may occur on the VPUMP supplies during the erase and programming cycles. Refer to the "Pin Descriptions and Packaging" chapter of the appropriate device datasheet for specific recommendations. For proper programming, 0.01 µF and 0.33 µF capacitors (both rated at 16 V) are to be connected in parallel across VPUMP and GND, and positioned as close to the FPGA pins as possible. The bypass capacitor must be placed within 2.5 cm of the device pins. VJTAG from the target board VCCI from the target board VCC from the target board VCC VCCI VJTAG Low Power Flash Device Polarizing Notch GND TCK TDO TMS VPUMP 1 TCK 3 TDO 5 TMS 7 VPUMP TDI 9 TDI 2 GND 4 NC* 6 VJTAG 8 TRST 10 GND TRST C1 C2 R R Note: *NC (FlashPro3/3X); Prog_Mode (FlashPro4). Prog_Mode on FlashPro4 is an output signal that goes High during device programming and returns to Low when programming is complete. This signal can be used to drive a system to provide a 1.5 V programming signal to IGLOO nano, ProASIC3L, and RT ProASIC3 devices that can run with 1.2 V core voltage but require 1.5 V for programming. IGLOO nano V2 devices can be programmed at 1.2 V core voltage (when using FlashPro4 only), but IGLOO nano V5 devices are programmed with a VCC core voltage of 1.5 V. Figure 13-6 • Board Layout and Programming Header Top View Troubleshooting Signal Integrity Symptoms of a Signal Integrity Problem A signal integrity problem can manifest itself in many ways. The problem may show up as extra or dropped bits during serial communication, changing the meaning of the communication. There is a normal variation of threshold voltage and frequency response between parts even from the same lot. Because of this, the effects of signal integrity may not always affect different devices on the same board in the same way. Sometimes, replacing a device appears to make signal integrity problems go away, but this is just masking the problem. Different parts on identical boards will exhibit the same problem sooner or later. It is important to fix signal integrity problems early. Unless the signal integrity problems are severe enough to completely block all communication between the device and the programmer, they may show up as subtle problems. Some of the FlashPro4/3/3X exit codes that are caused by signal integrity problems are listed below. Signal integrity problems are not the only possible cause of these Revision 4 337 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X errors, but this list is intended to show where problems can occur. FlashPro4/3/3X allows TCK to be lowered from 6 MHz down to 1 MHz to allow you to address some signal integrity problems that may occur with impedance mismatching at higher frequencies. Customers are expected to troubleshoot board-level signal integrity issues by measuring voltages and taking scope plots. Scan Chain Failure Normally, the FlashPro4/3/3X Scan Chain command expects to see 0x1 on the TDO pin. If the command reports reading 0x0 or 0x3, it is seeing the TDO pin stuck at 0 or 1. The only time the TDO pin comes out of tristate is when the JTAG TAP state machine is in the Shift-IR or Shift-DR state. If noise or reflections on the TCK or TMS lines have disrupted the correct state transitions, the device's TAP state controller might not be in one of these two states when the programmer tries to read the device. When this happens, the output is floating when it is read and does not match the expected data value. This can also be caused by a broken TDO net. Only a small amount of data is read from the device during the Scan Chain command, so marginal problems may not always show up during this command. Occasionally a faulty programmer can cause intermittent scan chain failures. Exit 11 This error occurs during the verify stage of programming a device. After programming the design into the device, the device is verified to ensure it is programmed correctly. The verification is done by shifting the programming data into the device. An internal comparison is performed within the device to verify that all switches are programmed correctly. Noise induced by poor signal integrity can disrupt the writes and reads or the verification process and produce a verification error. While technically a verification error, the root cause is often related to signal integrity. Refer to the FlashPro User's Guide for other error messages and solutions. For the most up-to-date known issues and solutions, refer to http://www.microsemi.com/soc/support. Conclusion IGLOO, ProASIC3, SmartFusion, and Fusion devices offer a low-cost, single-chip solution that is live at power-up through nonvolatile flash technology. The FlashLock Pass Key and 128-bit AES Key security features enable secure ISP in an untrusted environment. On-chip FlashROM enables a host of new applications, including device serialization, subscription-based applications, and IP addressing. Additionally, as the FlashROM is nonvolatile, all of these services can be provided without battery backup. Related Documents User’s Guides FlashPro User's Guide http://www.microsemi.com/soc/documents/flashpro_ug.pdf 338 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide List of Changes The following table lists critical changes that were made in each revision of the chapter. Date August 2012 June 2011 Changes Page This chapter will now be published standalone as an application note in addition to being part of the IGLOO/ProASIC3/Fusion FPGA fabric user’s guides (SAR 38769). N/A The "ISP Programming Header Information" section was revised to update the description of FP3-10PIN-ADAPTER-KIT in Table 13-3 • Programming Header Ordering Codes, clarifying that it is the adapter kit used for ProASICPLUS based boards, and also for ProASIC3 based boards where a compact programming header is being used (SAR 36779). 335 The VPUMP programming mode voltage was corrected in Table 13-2 • Power Supplies. The correct value is 3.15 V to 3.45 V (SAR 30668). 329 The notes associated with Figure 13-5 • Programming Header (top view) and 335, 337 Figure 13-6 • Board Layout and Programming Header Top View were revised to make clear the fact that IGLOO nano V2 devices can be programmed at 1.2 V (SAR 30787). July 2010 Figure 13-6 • Board Layout and Programming Header Top View was revised to include resistors tying TCK and TRST to GND. Microsemi recommends tying off TCK and TRST to GND if JTAG is not used (SAR 22921). RT ProASIC3 was added to the list of device families. 337 In the "ISP Programming Header Information" section, the kit for adapting ProASICPLUS devices was changed from FP3-10PIN-ADAPTER-KIT to FP3-26PINADAPTER-KIT (SAR 20878). 335 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A References to FlashPro4 and FlashPro3X were added to this chapter, giving distinctions between them. References to SmartGen were deleted and replaced with Libero IDE Catalog. N/A The "ISP Architecture" section was revised to indicate that V2 devices can be programmed at 1.2 V VCC with FlashPro4. 327 SmartFusion was added to Table 13-1 • Flash-Based FPGAs Supporting ISP. 328 The "Programming Voltage (VPUMP) and VJTAG" section was revised and 1.2 V was added to Table 13-2 • Power Supplies. 329 The "Nonvolatile Memory (NVM) Programming Voltage" section is new. 329 Cortex-M3 was added to the "Cortex-M1 and Cortex-M3 Device Security" section. 331 In the "ISP Programming Header Information" section, the additional header adapter ordering number was changed from FP3-26PIN-ADAPTER to FP3-10PINADAPTER-KIT, which contains 26-pin migration capability. 335 The description of NC was updated in Figure 13-5 • Programming Header (top 335, 336 view), Table 13-4 • Programming Header Pin Numbers and Description and Figure 13-6 • Board Layout and Programming Header Top View. The "Symptoms of a Signal Integrity Problem" section was revised to add that customers are expected to troubleshoot board-level signal integrity issues by measuring voltages and taking scope plots. "FlashPro4/3/3X allows TCK to be lowered from 6 MHz down to 1 MHz to allow you to address some signal integrity problems" formerly read, "from 24 MHz down to 1 MHz." "The Scan Chain command expects to see 0x2" was changed to 0x1. Revision 4 337 339 In-System Programming (ISP) of Microsemi’s Low Power Flash Devices Using FlashPro4/3/3X Date Changes Page July 2010 (continued) The "Chain Integrity Test Error Analyze Chain Failure" section was renamed to the "Scan Chain Failure" section, and the Analyze Chain command was changed to Scan Chain. It was noted that occasionally a faulty programmer can cause scan chain failures. 338 v1.5 (August 2009) The "CoreMP7 Device Security" section was removed from "Security in ARMEnabled Low Power Flash Devices", since M7-enabled devices are no longer supported. 331 v1.4 (December 2008) The "ISP Architecture" section was revised to include information about core voltage for IGLOO V2 and ProASIC3L devices, as well as 50 mV increments allowable in Designer software. 327 IGLOO nano and ProASIC3 nano devices were added to Table 13-1 • Flash-Based FPGAs Supporting ISP. 328 A second capacitor was added to Figure 13-6 • Board Layout and Programming Header Top View. 337 v1.3 (October 2008) The "ISP Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 328 v1.2 (June 2008) The following changes were made to the family descriptions in Table 13-1 • FlashBased FPGAs Supporting ISP: 328 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The "ISP Architecture" section was updated to included the IGLOO PLUS family in the discussion of family-specific support. The text, "When 1.2 V is used, the device can be reprogrammed in-system at 1.5 V only," was revised to state, "Although the device can operate at 1.2 V core voltage, the device can only be reprogrammed when all supplies (VCC, VCCI, and VJTAG) are at 1.5 V." 327 The "ISP Support in Flash-Based Devices" section and Table 13-1 • Flash-Based FPGAs Supporting ISP were updated to include the IGLOO PLUS family. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 328 The "Security" section was updated to mention that 15 k gate devices do not have a built-in 128-bit decryption core. 330 Table 13-2 • Power Supplies was revised to remove the Normal Operation column and add a table note stating, "All supply voltages should be at 1.5 V or higher, regardless of the setting during normal operation." 329 The "ISP Programming Header Information" section was revised to change FP3-26PIN-ADAPTER to FP3-10PIN-ADAPTER-KIT. Table 13-3 • Programming Header Ordering Codes was updated with the same change, as well as adding the part number FFSD-05-D-06.00-01-N, a 10-pin cable with 50-mil-pitch sockets. 335 The "Board-Level Considerations" section was updated to describe connecting two capacitors in parallel across VPUMP and GND for proper programming. 337 v1.0 (January 2008) Information was added to the "Programming Voltage (VPUMP) and VJTAG" section about the JTAG interface pin. 329 51900055-2/7.06 ACTgen was changed to SmartGen. N/A In Figure 13-6 • Board Layout and Programming Header Top View, the order of the text was changed to: 337 v1.1 (March 2008) VJTAG from the target board VCCI from the target board VCC from the target board 340 R e vi s i o n 4 14 – Core Voltage Switching Circuit for IGLOO and ProASIC3L In-System Programming Introduction The IGLOO® and ProASIC®3L families offer devices that can be powered by either 1.5 V or, in the case of V2 devices, a core supply voltage anywhere in the range of 1.2 V to 1.5 V, in 50 mV increments. Since IGLOO and ProASIC3L devices are flash-based, they can be programmed and reprogrammed multiple times in-system using Microsemi FlashPro3. FlashPro3 uses the JTAG standard interface (IEEE 1149.1) and STAPL file (defined in JESD 71 to support programming of programmable devices using IEEE 1149.1) for in-system configuration/programming (IEEE 1532) of a device. Programming can also be executed by other methods, such as an embedded microcontroller that follows the same standards above. All IGLOO and ProASIC3L devices must be programmed with the VCC core voltage at 1.5 V. Therefore, applications using IGLOO or ProASIC3L devices powered by a 1.2 V supply must switch the core supply to 1.5 V for in-system programming. The purpose of this document is to describe an easy-to-use and cost-effective solution for switching the core supply voltage from 1.2 V to 1.5 V during in-system programming for IGLOO and ProASIC3L devices. Revision 4 341 Core Voltage Switching Circuit for IGLOO and ProASIC3L In-System Programming Microsemi’s Flash Families Support Voltage Switching Circuit The flash FPGAs listed in Table 14-1 support the voltage switching circuit feature and the functions described in this document. Table 14-1 • Flash-Based FPGAs Supporting Voltage Switching Circuit Family* Series IGLOO ProASIC3 Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 14-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 14-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 342 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Circuit Description All IGLOO devices as well as the ProASIC3L product family are available in two versions: V5 devices, which are powered by a 1.5 V supply and V2 devices, which are powered by a supply anywhere in the range of 1.2 V to 1.5 V in 50 mV increments. Applications that use IGLOO or ProASIC3L devices powered by a 1.2 V core supply must have a mechanism that switches the core voltage from 1.2 V (or other voltage below 1.5 V) to 1.5 V during in-system programming (ISP). There are several possible techniques to meet this requirement. Microsemi recommends utilizing a linear voltage regulator, a resistor voltage divider, and an N-Channel Digital FET to set the appropriate VCC voltage, as shown in Figure 14-1. Where 1.2 V is mentioned in the following text, the meaning applies to any voltage below the 1.5 V range. Resistor values in the figures have been calculated for 1.2 V, so refer to power regulator datasheets if a different core voltage is required. The main component of Microsemi's recommended circuit is the LTC3025 linear voltage regulator from LinearTech. The output voltage of the LTC3025 on the OUT pin is set by the ratio of two external resistors, R37 and R38, in a voltage divider. The linear voltage regulator adjusts the voltage on the OUT pin to maintain the ADJ pin voltage at 0.4 V (referenced to ground). By using an R38 value of 40.2 kΩ and an R37 value of 80.6 kΩ, the output voltage on the OUT pin is 1.2 V. To achieve 1.5 V on the OUT pin, R44 can be used in parallel with R38. The OUT pin can now be used as a switchable source for the VCC supply. Refer to the LTC3025 Linear Voltage Regulator datasheet for more information. In Figure 14-1, the N-Channel Digital FET is used to enable and disable R44. This FET is controlled by the JTAG TRST signal driven by the FlashPro3 programmer. During programming of the device, the TRST signal is driven HIGH by the FlashPro3, and turns the N-Channel Digital FET ON. When the FET is ON, R44 becomes enabled as a parallel resistance to R38, which forces the regulator to set OUT to 1.5 V. When the FlashPro3 is connected and not in programming mode or when it is not connected, the pulldown resistor, R10, will pull the TRST signal LOW. When this signal is LOW, the N-Channel Digital FET is "open" and R44 is not part of the resistance seen by the LTC3025. The new resistance momentarily changes the voltage value on the ADJ pin, which in turn causes the output of the LTC3025 to compensate by setting OUT to 1.2 V. Now the device will run in regular active mode at the regular 1.2 V core voltage. Figure 14-1 • Circuit Diagram Revision 4 343 Core Voltage Switching Circuit for IGLOO and ProASIC3L In-System Programming Circuit Verification The power switching circuit recommended above is implemented on Microsemi's Icicle board (Figure 14-2). On the Icicle board, VJTAGENB is used to control the N-Channel Digital FET; however, this circuit was modified to use TRST instead of VJTAGENB in this application. There are three important aspects of this circuit that were verified: 1. The rise on VCC from 1.2 V to 1.5 V when TRST is HIGH 2. VCC rises to 1.5 V before programming begins. 3. VCC switches from 1.5 V to 1.2 V when TRST is LOW. Verification Steps 1. The rise on VCC from 1.2 V to 1.5 V when TRST is HIGH. VCC Signal TRST Signal Figure 14-2 • Core Voltage on the IGLOO AGL125-QNG132 Device In the oscilloscope plots (Figure 14-2), the TRST from FlashPro3 and the VCC core voltage of the IGLOO device are labeled. This plot shows the rise characteristic of the TRST signal from FlashPro3. Once the TRST signal is asserted HIGH, the LTC3025 shown in Figure 14-1 on page 343 senses the increase in voltage and changes the output from 1.2 V to 1.5 V. It takes the circuit approximately 100 µs to respond to TRST and change the voltage to 1.5 V on the VCC core. 344 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 2. VCC rises to 1.5 V before programming begins. TMS Signal Green Floating Signal TDI/TMS VCC Core Voltage TRST Signal (purple) TDI Signal (yellow) Figure 14-3 • Programming Algorithm The oscilloscope plot in Figure 14-3 shows a wider time interval for the programming algorithm and includes the TDI and TMS signals from the FlashPro3. These signals carry the programming information that is programmed into the device and should only start toggling after the VCC core voltage reaches 1.5 V. Again, TRST from FlashPro3 and the VCC core voltage of the IGLOO device are labeled. As shown in Figure 14-3, TDI and TMS are floating initially, and the core voltage is 1.2 V. When a programming command on the FlashPro3 is executed, TRST is driven HIGH and TDI is momentarily driven to ground. In response to the HIGH TRST signal, the circuit responds and pulls the core voltage to 1.5 V. After 100 ms, TRST is briefly driven LOW by the FlashPro software. This is expected behavior that ensures the device JTAG state machine is in Reset prior to programming. TRST remains HIGH for the duration of the programming. It can be seen in Figure 14-3 that the VCC core voltage signal remains at 1.5 V for approximately 50 ms before information starts passing through on TDI and TMS. This confirms that the voltage switching circuit drives the VCC core supply voltage to 1.5 V prior to programming. Revision 4 345 Core Voltage Switching Circuit for IGLOO and ProASIC3L In-System Programming 3. VCC switches from 1.5 V to 1.2 V when TRST is LOW. TRST Signal VCC Core Signal Figure 14-4 • TRST Toggled LOW In Figure 14-4, the TRST signal and the VCC core voltage signal are labeled. As TRST is pulled to ground, the core voltage is observed to switch from 1.5 V to 1.2 V. The observed fall time is approximately 2 ms. DirectC The above analysis is based on FlashPro3, but there are other solutions to ISP, such as DirectC. DirectC is a microprocessor program that can be run in-system to program Microsemi flash devices. For FlashPro3, TRST is the most convenient control signal to use for the recommended circuit. However, for DirectC, users may use any signal to control the FET. For example, the DirectC code can be edited so that a separate non-JTAG signal can be asserted from the microcontroller that signals the board that it is about to start programming the device. After asserting the N-Channel Digital FET control signal, the programming algorithm must allow sufficient time for the supply to rise to 1.5 V before initiating DirectC programming. As seen in Figure 14-3 on page 345, 50 ms is adequate time. Depending on the size of the PCB and the capacitance on the VCC supply, results may vary from system to system. Microsemi recommends using a conservative value for the wait time to make sure that the VCC core voltage is at the right level. Conclusion For applications using IGLOO and ProASIC3L low power FPGAs and taking advantage of the low core voltage power supplies with less than 1.5 V operation, there must be a way for the core voltage to switch from 1.2 V (or other voltage) to 1.5 V, which is required during in-system programming. The circuit explained in this document illustrates one simple, cost-effective way of handling this requirement. A JTAG signal from the FlashPro3 programmer allows the circuit to sense when programming is in progress, enabling it to switch to the correct core voltage. 346 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.1 (October 2008) The "Introduction" was revised to include information about the core supply voltage range of operation in V2 devices. 341 IGLOO nano device support was added to Table 14-1 • Flash-Based FPGAs Supporting Voltage Switching Circuit. 342 The "Circuit Description" section was updated to include IGLOO PLUS core operation from 1.2 V to 1.5 V in 50 mV increments. 343 The "Microsemi’s Flash Families Support Voltage Switching Circuit" section was revised to include new families and make the information more concise. 342 v1.0 (August 2008) Revision 4 347 15 – Microprocessor Programming of Microsemi’s Low Power Flash Devices Introduction The Fusion, IGLOO, and ProASIC3 families of flash FPGAs support in-system programming (ISP) with the use of a microprocessor. Flash-based FPGAs store their configuration information in the actual cells within the FPGA fabric. SRAM-based devices need an external configuration memory, and hybrid nonvolatile devices store the configuration in a flash memory inside the same package as the SRAM FPGA. Since the programming of a true flash FPGA is simpler, requiring only one stage, it makes sense that programming with a microprocessor in-system should be simpler than with other SRAM FPGAs. This reduces bill-of-materials costs and printed circuit board (PCB) area, and increases system reliability. Nonvolatile flash technology also gives the low power flash devices the advantage of a secure, low power, live-at-power-up, and single-chip solution. Low power flash devices are reprogrammable and offer time-to-market benefits at an ASIC-level unit cost. These features enable engineers to create highdensity systems using existing ASIC or FPGA design flows and tools. This document is an introduction to microprocessor programming only. To explain the difference between the options available, user's guides for DirectC and STAPL provide more detail on implementing each style. Microprocessor Internal/External Memory Running DirectC Internal RAM On-Board Memory Device .dat file I/O Functions JTAG Bus Flash Device Figure 15-1 • ISP Using Microprocessor Revision 4 349 Microprocessor Programming of Microsemi’s Low Power Flash Devices Microprocessor Programming Support in Flash Devices The flash-based FPGAs listed in Table 15-1 support programming with a microprocessor and the functions described in this document. Table 15-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 15-1. Where the information applies to only one device or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 15-1. Where the information applies to only one device or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 350 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Programming Algorithm JTAG Interface The low power flash families are fully compliant with the IEEE 1149.1 (JTAG) standard. They support all the mandatory boundary scan instructions (EXTEST, SAMPLE/PRELOAD, and BYPASS) as well as six optional public instructions (USERCODE, IDCODE, HIGHZ, and CLAMP). IEEE 1532 The low power flash families are also fully compliant with the IEEE 1532 programming standard. The IEEE 1532 standard adds programming instructions and associated data registers to devices that comply with the IEEE 1149.1 standard (JTAG). These instructions and registers extend the capabilities of the IEEE 1149.1 standard such that the Test Access Port (TAP) can be used for configuration activities. The IEEE 1532 standard greatly simplifies the programming algorithm, reducing the amount of time needed to implement microprocessor ISP. Implementation Overview To implement device programming with a microprocessor, the user should first download the C-based STAPL player or DirectC code from the Microsemi SoC Products Group website. Refer to the website for future updates regarding the STAPL player and DirectC code. http://www.microsemi.com/soc/download/program_debug/stapl/default.aspx http://www.microsemi.com/soc/download/program_debug/directc/default.aspx Using the easy-to-follow user's guide, create the low-level application programming interface (API) to provide the necessary basic functions. These API functions act as the interface between the programming software and the actual hardware (Figure 15-2). Programming Algorithm and Data STAPL File STAPL Player or DirectC Programming Software I/O and Memory Functions API Figure 15-2 • Device Programming Code Relationship The API is then linked with the STAPL player or DirectC and compiled using the microprocessor's compiler. Once the entire code is compiled, the user must download the resulting binary into the MCU system's program memory (such as ROM, EEPROM, or flash). The system is now ready for programming. To program a design into the FPGA, the user creates a bitstream or STAPL file using the Microsemi Designer software, downloads it into the MCU system's volatile memory, and activates the stored programming binary file (Figure 15-3 on page 352). Once the programming is completed, the bitstream or STAPL file can be removed from the system, as the configuration profile is stored in the flash FPGA fabric and does not need to be reloaded at every system power-on. Revision 4 351 Microprocessor Programming of Microsemi’s Low Power Flash Devices Programming Software Source Code Programming File Microprocessor Compiler BIN File Download to System Program Device Figure 15-3 • MCU FPGA Programming Model FlashROM Microsemi low power flash devices have 1 kbit of user-accessible, nonvolatile, FlashROM on-chip. This nonvolatile FlashROM can be programmed along with the core or on its own using the standard IEEE 1532 JTAG programming interface. The FlashROM is architected as eight pages of 128 bits. Each page can be individually programmed (erased and written). Additionally, on-chip AES security decryption can be used selectively to load data securely into the FlashROM (e.g., over public or private networks, such as the Internet). Refer to the "FlashROM in Microsemi’s Low Power Flash Devices" section on page 133. 352 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide STAPL vs. DirectC Programming the low power flash devices is performed using DirectC or the STAPL player. Both tools use the STAPL file as an input. DirectC is a compiled language, whereas STAPL is an interpreted language. Microprocessors will be able to load the FPGA using DirectC much more quickly than STAPL. This speed advantage becomes more apparent when lower clock speeds of 8- or 16-bit microprocessors are used. DirectC also requires less memory than STAPL, since the programming algorithm is directly implemented. STAPL does have one advantage over DirectC—the ability to upgrade. When a new programming algorithm is required, the STAPL user simply needs to regenerate a STAPL file using the latest version of the Designer software and download it to the system. The DirectC user must download the latest version of DirectC from Microsemi, compile everything, and download the result into the system (Figure 15-4). STAPL Flow DirectC Flow Generate the New STAPL File DirectC Source Code Download to System Microprocessor Compiler Program Device BIN File Input STAPL File Download to System Program Device Figure 15-4 • STAPL vs. DirectC Revision 4 353 Microprocessor Programming of Microsemi’s Low Power Flash Devices Remote Upgrade via TCP/IP Transmission Control Protocol (TCP) provides a reliable bitstream transfer service between two endpoints on a network. TCP depends on Internet Protocol (IP) to move packets around the network on its behalf. TCP protects against data loss, data corruption, packet reordering, and data duplication by adding checksums and sequence numbers to transmitted data and, on the receiving side, sending back packets and acknowledging the receipt of data. The system containing the low power flash device can be assigned an IP address when deployed in the field. When the device requires an update (core or FlashROM), the programming instructions along with the new programming data (AES-encrypted cipher text) can be sent over the Internet to the target system via the TCP/IP protocol. Once the MCU receives the instruction and data, it can proceed with the FPGA update. Low power flash devices support Message Authentication Code (MAC), which can be used to validate data for the target device. More details are given in the "Message Authentication Code (MAC) Validation/Authentication" section. Hardware Requirement To facilitate the programming of the low power flash families, the system must have a microprocessor (with access to the device JTAG pins) to process the programming algorithm, memory to store the programming algorithm, programming data, and the necessary programming voltage. Refer to the relevant datasheet for programming voltages. Security Encrypted Programming As an additional security measure, the devices are equipped with AES decryption. AES works in two steps. The first step is to program a key into the devices in a secure or trusted programming center (such as Microsemi SoC Products Group In-House Programming (IHP) center). The second step is to encrypt any programming files with the same encryption key. The encrypted programming file will only work with the devices that have the same key. The AES used in the low power flash families is the 128-bit AES decryption engine (Rijndael algorithm). Message Authentication Code (MAC) Validation/Authentication As part of the AES decryption flow, the devices are equipped with a MAC validation/authentication system. MAC is an authentication tag, also called a checksum, derived by applying an on-chip authentication scheme to a STAPL file as it is loaded into the FPGA. MACs are computed and verified with the same key so they can only be verified by the intended recipient. When the MCU system receives the AES-encrypted programming data (cipher text), it can validate the data by loading it into the FPGA and performing a MAC verification prior to loading the data, via a second programming pass, into the FPGA core cells. This prevents erroneous or corrupt data from getting into the FPGA. Low power flash devices with AES and MAC are superior to devices with only DES or 3DES encryption. Because the MAC verifies the correctness of the data, the FPGA is protected from erroneous loading of invalid programming data that could damage a device (Figure 15-5 on page 355). The AES with MAC enables field updates over public networks without fear of having the design stolen. An encrypted programming file can only work on devices with the correct key, rendering any stolen files 354 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide useless to the thief. To learn more about the low power flash devices’ security features, refer to the "Security in Low Power Flash Devices" section on page 301. ProASIC3 MAC Validation Designer Software AES KEY AES Encryption Decrypted Stream AES Decryption Programming Control TCP/IP Public Network Encrypted Stream Encrypted Stream Figure 15-5 • ProASIC3 Device Encryption Flow Conclusion The Fusion, IGLOO, and ProASIC3 FPGAs are ideal for applications that require field upgrades. The single-chip devices save board space by eliminating the need for EEPROM. The built-in AES with MAC enables transmission of programming data over any network without fear of design theft. Fusion, IGLOO, and ProASIC3 FPGAs are IEEE 1532–compliant and support STAPL, making the target programming software easy to implement. Revision 4 355 Microprocessor Programming of Microsemi’s Low Power Flash Devices List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page September 2012 The "Security" section was modified to clarify that Microsemi does not support read-back of FPGA core-programmed data (SAR 41235). 354 July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 15-1 • FlashBased FPGAs. 350 v1.3 (October 2008) The "Microprocessor Programming Support in Flash Devices" section was revised to include new families and make the information more concise. 350 v1.2 (June 2008) The following changes were made to the family descriptions in Table 15-1 • Flash-Based FPGAs: 350 v1.1 (March 2008) 356 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The "Microprocessor Programming Support in Flash Devices" section was updated to include information on the IGLOO PLUS family. The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. R e vi s i o n 4 350 16 – Boundary Scan in Low Power Flash Devices Boundary Scan Low power flash devices are compatible with IEEE Standard 1149.1, which defines a hardware architecture and the set of mechanisms for boundary scan testing. JTAG operations are used during boundary scan testing. The basic boundary scan logic circuit is composed of the TAP controller, test data registers, and instruction register (Figure 16-2 on page 360). Low power flash devices support three types of test data registers: bypass, device identification, and boundary scan. The bypass register is selected when no other register needs to be accessed in a device. This speeds up test data transfer to other devices in a test data path. The 32-bit device identification register is a shift register with four fields (LSB, ID number, part number, and version). The boundary scan register observes and controls the state of each I/O pin. Each I/O cell has three boundary scan register cells, each with serial-in, serial-out, parallel-in, and parallel-out pins. TAP Controller State Machine The TAP controller is a 4-bit state machine (16 states) that operates as shown in Figure 16-1. The 1s and 0s represent the values that must be present on TMS at a rising edge of TCK for the given state transition to occur. IR and DR indicate that the instruction register or the data register is operating in that state. The TAP controller receives two control inputs (TMS and TCK) and generates control and clock signals for the rest of the test logic architecture. On power-up, the TAP controller enters the Test-Logic-Reset state. To guarantee a reset of the controller from any of the possible states, TMS must remain HIGH for five TCK cycles. The TRST pin can also be used to asynchronously place the TAP controller in the TestLogic-Reset state. 1 TEST_LOGIC_RESET 0 0 RUN_TEST_IDLE 1 SELECT_DR 1 SELECT_IR 0 0 CAPTURE_DR 1 1 CAPTURE_IR 0 0 0 SHIFT_DR 1 EXIT1_DR EXIT1_IR 0 0 0 PAUSE_DR 1 0 PAUSE_IR 1 1 EXIT2_DR 0 0 EXIT2_IR 1 1 UPDATE_DR 1 0 SHIFT_IR 1 1 1 0 UPDATE_IR 1 0 Figure 16-1 • TAP Controller State Machine Revision 4 357 Boundary Scan in Low Power Flash Devices Microsemi’s Flash Devices Support the JTAG Feature The flash-based FPGAs listed in Table 16-1 support the JTAG feature and the functions described in this document. Table 16-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC®3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 16-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 16-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 358 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Boundary Scan Support in Low Power Devices The information in this document applies to all Fusion, IGLOO, and ProASIC3 devices. For IGLOO, IGLOO PLUS, and ProASIC3L devices, the Flash*Freeze pin must be deasserted for successful boundary scan operations. Devices cannot enter JTAG mode directly from Flash*Freeze mode. Boundary Scan Opcodes Low power flash devices support all mandatory IEEE 1149.1 instructions (EXTEST, SAMPLE/PRELOAD, and BYPASS) and the optional IDCODE instruction (Table 16-2). Table 16-2 • Boundary Scan Opcodes Hex Opcode EXTEST 00 HIGHZ 07 USERCODE 0E SAMPLE/PRELOAD 01 IDCODE 0F CLAMP 05 BYPASS FF Boundary Scan Chain The serial pins are used to serially connect all the boundary scan register cells in a device into a boundary scan register chain (Figure 16-2 on page 360), which starts at the TDI pin and ends at the TDO pin. The parallel ports are connected to the internal core logic I/O tile and the input, output, and control ports of an I/O buffer to capture and load data into the register to control or observe the logic state of each I/O. Each test section is accessed through the TAP, which has five associated pins: TCK (test clock input), TDI, TDO (test data input and output), TMS (test mode selector), and TRST (test reset input). TMS, TDI, and TRST are equipped with pull-up resistors to ensure proper operation when no input data is supplied to them. These pins are dedicated for boundary scan test usage. Refer to the "JTAG Pins" section in the "Pin Descriptions and Packaging" chapter of the appropriate device datasheet for pull-up/-down recommendations for TCK and TRST pins. Pull-down recommendations are also given in Table 16-3 on page 360 Revision 4 359 Boundary Scan in Low Power Flash Devices I/O I/O I/O I/O I/O TDI Test Data Registers Instruction Register TAP Controller Device Logic TDO I/O TRST I/O TMS I/O TCK I/O Bypass Register I/O I/O I/O I/O I/O Figure 16-2 • Boundary Scan Chain Board-Level Recommendations Table 16-3 gives pull-down recommendations for the TRST and TCK pins. Table 16-3 • TRST and TCK Pull-Down Recommendations VJTAG Tie-Off Resistance* VJTAG at 3.3 V 200 Ω to 1 kΩ VJTAG at 2.5 V 200 Ω to 1 kΩ VJTAG at 1.8 V 500 Ω to 1 kΩ VJTAG at 1.5 V 500 Ω to 1 kΩ VJTAG at 1.2 V TBD Note: Equivalent parallel resistance if more than one device is on JTAG chain (Figure 16-3) 360 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 1.5 V JTAG Header VJTAG TRST GND TCK 2 kΩ TDO TDI Microsemi FPGA 1 TDO TDI Microsemi FPGA 2 1.5 kΩ 2 kΩ 1.5 kΩ 2 kΩ TDO TDI Microsemi FPGA3 TDO TDI Microsemi FPGA 4 1.5 kΩ 2 kΩ 1.5 kΩ Note: TCK is correctly wired with an equivalent tie-off resistance of 500 Ω, which satisfies the table for VJTAG of 1.5 V. The resistor values for TRST are not appropriate in this case, as the tie-off resistance of 375 Ω is below the recommended minimum for VJTAG = 1.5 V, but would be appropriate for a VJTAG setting of 2.5 V or 3.3 V. Figure 16-3 • Parallel Resistance on JTAG Chain of Devices Advanced Boundary Scan Register Settings You will not be able to control the order in which I/Os are released from boundary scan control. Testing has produced cases where, depending on I/O placement and FPGA routing, a 5 ns glitch has been seen on exiting programming mode. The following setting is recommended to prevent such I/O glitches: 1. In the FlashPro software, configure the advanced BSR settings for Specify I/O Settings During Programming. 2. Set the input BSR cell to Low for the input I/O. Revision 4 361 Boundary Scan in Low Power Flash Devices List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page In the "Boundary Scan Chain" section, the reference made to the datasheet for pull-up/-down recommendations was changed to mention TCK and TRST pins rather than TDO and TCK pins. TDO is an output, so no pull resistor is needed (SAR 35937). 359 The "Advanced Boundary Scan Register Settings" section is new (SAR 38432). 361 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A Table 16-3 • TRST and TCK Pull-Down Recommendations was revised to add VJTAG at 1.2 V. 360 v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 16-1 • Flash-Based FPGAs. 358 v1.3 (October 2008) The "Boundary Scan Support in Low Power Devices" section was revised to include new families and make the information more concise. 359 v1.2 (June 2008) The following changes were made to the family descriptions in Table 16-1 • FlashBased FPGAs: 358 August 2012 July 2010 v1.1 (March 2008) 362 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The chapter was updated to include the IGLOO PLUS family and information regarding 15 k gate devices. N/A The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 358 R e vi s i o n 4 17 – UJTAG Applications in Microsemi’s Low Power Flash Devices Introduction In Fusion, IGLOO, and ProASIC3 devices, there is bidirectional access from the JTAG port to the core VersaTiles during normal operation of the device (Figure 17-1). User JTAG (UJTAG) is the ability for the design to use the JTAG ports for access to the device for updates, etc. While regular JTAG is used, the UJTAG tiles, located at the southeast area of the die, are directly connected to the JTAG Test Access Port (TAP) Controller in normal operating mode. As a result, all the functional blocks of the device, such as Clock Conditioning Circuits (CCCs) with PLLs, SRAM blocks, embedded FlashROM, flash memory blocks, and I/O tiles, can be reached via the JTAG ports. The UJTAG functionality is available by instantiating the UJTAG macro directly in the source code of a design. Access to the FPGA core VersaTiles from the JTAG ports enables users to implement different applications using the TAP Controller (JTAG port). This document introduces the UJTAG tile functionality and discusses a few application examples. However, the possible applications are not limited to what is presented in this document. UJTAG can serve different purposes in many designs as an elementary or auxiliary part of the design. For detailed usage information, refer to the "Boundary Scan in Low Power Flash Devices" section on page 357. UJTAG Address Generation and Data Serlialization UIREG[7:0] TDO TDI TMS TCK TRST Enable FROM RESET URSTB UDRUPD Control UDRCK CLK UDRCAP SDI UDRSH Addr[6:0] Addr [6:0] Data[7:0] Data[7:0] SDO UTDI UTDO Figure 17-1 • Block Diagram of Using UJTAG to Read FlashROM Contents Revision 4 363 UJTAG Applications in Microsemi’s Low Power Flash Devices UJTAG Support in Flash-Based Devices The flash-based FPGAs listed in Table 17-1 support the UJTAG feature and the functions described in this document. Table 17-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Fusion Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Fusion Mixed signal FPGA integrating ProASIC3 FPGA fabric, programmable analog block, support for ARM® Cortex™-M1 soft processors, and flash memory into a monolithic device Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 17-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 17-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 364 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide UJTAG Macro The UJTAG tiles can be instantiated in a design using the UJTAG macro from the Fusion, IGLOO, or ProASIC3 macro library. Note that "UJTAG" is a reserved name and cannot be used for any other userdefined blocks. A block symbol of the UJTAG tile macro is presented in Figure 17-2. In this figure, the ports on the left side of the block are connected to the JTAG TAP Controller, and the right-side ports are accessible by the FPGA core VersaTiles. The TDI, TMS, TDO, TCK, and TRST ports of UJTAG are only provided for design simulation purposes and should be treated as external signals in the design netlist. However, these ports must NOT be connected to any I/O buffer in the netlist. Figure 17-3 on page 366 illustrates the correct connection of the UJTAG macro to the user design netlist. Microsemi Designer software will automatically connect these ports to the TAP during place-and-route. Table 17-2 gives the port descriptions for the rest of the UJTAG ports: Table 17-2 • UJTAG Port Descriptions Port Description UIREG [7:0] This 8-bit bus carries the contents of the JTAG Instruction Register of each device. Instruction Register values 16 to 127 are not reserved and can be employed as user-defined instructions. URSTB URSTB is an active-low signal and will be asserted when the TAP Controller is in Test-Logic-Reset mode. URSTB is asserted at power-up, and a power-on reset signal resets the TAP Controller. URSTB will stay asserted until an external TAP access changes the TAP Controller state. UTDI This port is directly connected to the TAP's TDI signal. UTDO This port is the user TDO output. Inputs to the UTDO port are sent to the TAP TDO output MUX when the IR address is in user range. UDRSH Active-high signal enabled in the ShiftDR TAP state UDRCAP Active-high signal enabled in the CaptureDR TAP state UDRCK This port is directly connected to the TAP's TCK signal. UDRUPD Active-high signal enabled in the UpdateDR TAP state TDO TDI TMS UIREG0 UIREG1 UIREG2 UIREG3 UIREG4 UIREG5 UIREG6 UIREG7 URSTB UDRUPD TCK TRST UDRCK UDRCAP UDRSH UTDI UTDO Figure 17-2 • UJTAG Tile Block Symbol Revision 4 365 UJTAG Applications in Microsemi’s Low Power Flash Devices a) CORRECT Instantiation UIREG[7:0] TDO URSTB TDI INPUTS UDRUPD UDRCK TMS UDRCAP TCK FPGA VersaTiles UDRSH TRST UTDI UTDO OUTPUTS b) INCORRECT Instantiation TDO UIREG[7:0] URSTB TDI INPUTS UDRUPD TMS TCK UDRCK UDRCAP FPGA VersaTiles UDRSH TRST UTDI UTDO OUTPUTS Note: Do not connect JTAG pins (TDO, TDI, TMS, TCK, or TRST) to I/Os in the design. Figure 17-3 • Connectivity Method of UJTAG Macro UJTAG Operation There are a few basic functions of the UJTAG macro that users must understand before designing with it. The most important fundamental concept of the UJTAG design is its connection with the TAP Controller state machine. TAP Controller State Machine The 16 states of the TAP Controller state machine are shown in Figure 17-4 on page 367. The 1s and 0s, shown adjacent to the state transitions, represent the TMS values that must be present at the time of a rising TCK edge for a state transition to occur. In the states that include the letters "IR," the instruction register operates; in the states that contain the letters "DR," the test data register operates. The TAP Controller receives two control inputs, TMS and TCK, and generates control and clock signals for the rest of the test logic. On power-up (or the assertion of TRST), the TAP Controller enters the Test-Logic-Reset state. To reset the controller from any other state, TMS must be held HIGH for at least five TCK cycles. After reset, the TAP state changes at the rising edge of TCK, based on the value of TMS. 366 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide 1 Test_Logic_Reset 0 0 Run_Test/ Idle 1 Select_ DR_Scan 0 1 1 Select_ IR_Scan 0 1 Capture_DR Capture_IR 0 0 0 Shift_DR 1 Exit1_DR 1 Exit1_IR 0 1 0 Pause_DR 0 Pause_IR 1 0 1 Exit2_DR 0 Exit2_IR 1 1 Update_DR 1 0 Shift_IR 1 0 1 0 Update_IR 1 0 Figure 17-4 • TAP Controller State Diagram UJTAG Port Usage UIREG[7:0] hold the contents of the JTAG instruction register. The UIREG vector value is updated when the TAP Controller state machine enters the Update_IR state. Instructions 16 to 127 are user-defined and can be employed to encode multiple applications and commands within an application. Loading new instructions into the UIREG vector requires users to send appropriate logic to TMS to put the TAP Controller in a full IR cycle starting from the Select IR_Scan state and ending with the Update_IR state. UTDI, UTDO, and UDRCK are directly connected to the JTAG TDI, TDO, and TCK ports, respectively. The TDI input can be used to provide either data (TAP Controller in the Shift_DR state) or the new contents of the instruction register (TAP Controller in the Shift_IR state). UDRSH, UDRUPD, and UDRCAP are HIGH when the TAP Controller state machine is in the Shift_DR, Update_DR, and Capture_DR states, respectively. Therefore, they act as flags to indicate the stages of the data shift process. These flags are useful for applications in which blocks of data are shifted into the design from JTAG pins. For example, an active UDRSH can indicate that UTDI contains the data bitstream, and UDRUPD is a candidate for the end-of-data-stream flag. As mentioned earlier, users should not connect the TDI, TDO, TCK, TMS, and TRST ports of the UJTAG macro to any port or net of the design netlist. The Designer software will automatically handle the port connection. Revision 4 367 UJTAG Applications in Microsemi’s Low Power Flash Devices Typical UJTAG Applications Bidirectional access to the JTAG port from VersaTiles—without putting the device into test mode— creates flexibility to implement many different applications. This section describes a few of these. All are based on importing/exporting data through the UJTAG tiles. Clock Conditioning Circuitry—Dynamic Reconfiguration In low power flash devices, CCCs, which include PLLs, can be configured dynamically through either an 81-bit embedded shift register or static flash programming switches. These 81 bits control all the characteristics of the CCC: routing MUX architectures, delay values, divider values, etc. Table 17-3 lists the 81 configuration bits in the CCC. Table 17-3 • Configuration Bits of Fusion, IGLOO, and ProASIC3 CCC Blocks Bit Number(s) Control Function 80 RESET ENABLE 79 DYNCSEL 78 DYNBSEL 77 DYNASEL VCOSEL [2:0] 73 STATCSEL 72 STATBSEL 71 STATASEL DLYC [4:0] DLYB {4:0] DLYGLC [4:0] DLYGLB [4:0] DLYGLA [4:0] 45 XDLYSEL FBDLY [4:0] FBSEL OCMUX [2:0] OBMUX [2:0] OAMUX [2:0] OCDIV [4:0] OBDIV [4:0] OADIV [4:0] FBDIV [6:0] FINDIV [6:0] The embedded 81-bit shift register (for the dynamic configuration of the CCC) is accessible to the VersaTiles, which, in turn, have access to the UJTAG tiles. Therefore, the CCC configuration shift register can receive and load the new configuration data stream from JTAG. Dynamic reconfiguration eliminates the need to reprogram the device when reconfiguration of the CCC functional blocks is needed. The CCC configuration can be modified while the device continues to operate. Employing the UJTAG core requires the user to design a module to provide the configuration data and control the CCC configuration shift register. In essence, this is a user-designed TAP Controller requiring chip resources. Similar reconfiguration capability exists in the ProASICPLUS® family. The only difference is the number of shift register bits controlling the CCC (27 in ProASICPLUS and 81 in IGLOO, ProASIC3, and Fusion). 368 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Fine Tuning In some applications, design constants or parameters need to be modified after programming the original design. The tuning process can be done using the UJTAG tile without reprogramming the device with new values. If the parameters or constants of a design are stored in distributed registers or embedded SRAM blocks, the new values can be shifted onto the JTAG TAP Controller pins, replacing the old values. The UJTAG tile is used as the “bridge” for data transfer between the JTAG pins and the FPGA VersaTiles or SRAM logic. Figure 17-5 shows a flow chart example for fine-tuning application steps using the UJTAG tile. In Figure 17-5, the TMS signal sets the TAP Controller state machine to the appropriate states. The flow mainly consists of two steps: a) shifting the defined instruction and b) shifting the new data. If the target parameter is constantly used in the design, the new data can be shifted into a temporary shift register from UTDI. The UDRSH output of UJTAG can be used as a shift-enable signal, and UDRCK is the shift clock to the shift register. Once the shift process is completed and the TAP Controller state is moved to the Update_DR state, the UDRUPD output of the UJTAG can latch the new parameter value from the temporary register into a permanent location. This avoids any interruption or malfunctioning during the serial shift of the new value. TAP Controller in Test_Logic_Reset State Set TAP state to SHIFT_DR Set TAP state to SHIFT_IR Shift data into TDI and record UTDI in a shift register Shift the user-defined instruction of tuning application Set TAP state in Update_DR Set TAP state to Update_IR UIREG Equal to the user-defined instruction Yes Latch the recorded data onto the location of stored parameter No Figure 17-5 • Flow Chart Example of Fine-Tuning an Application Using UJTAG Revision 4 369 UJTAG Applications in Microsemi’s Low Power Flash Devices Silicon Testing and Debugging In many applications, the design needs to be tested, debugged, and verified on real silicon or in the final embedded application. To debug and test the functionality of designs, users may need to monitor some internal logic (or nets) during device operation. The approach of adding design test pins to monitor the critical internal signals has many disadvantages, such as limiting the number of user I/Os. Furthermore, adding external I/Os for test purposes may require additional or dedicated board area for testing and debugging. The UJTAG tiles of low power flash devices offer a flexible and cost-effective solution for silicon test and debug applications. In this solution, the signals under test are shifted out to the TDO pin of the TAP Controller. The main advantage is that all the test signals are monitored from the TDO pin; no pins or additional board-level resources are required. Figure 17-6 illustrates this technique. Multiple test nets are brought into an internal MUX architecture. The selection of the MUX is done using the contents of the TAP Controller instruction register, where individual instructions (values from 16 to 127) correspond to different signals under test. The selected test signal can be synchronized with the rising or falling edge of TCK (optional) and sent out to UTDO to drive the TDO output of JTAG. For flash devices, TDO (the output) is configured as low slew and the highest drive strength available in the technology and/or device. Here are some examples: 1. If the device is A3P1000 and VCCI is 3.3 V, TDO will be configured as LVTTL 3.3 V output, 24 mA, low slew. 2. If the device is AGLN020 and VCCI is 1.8 V, TDO will be configured as LVCMOS 1.8 V output, 4 mA, low slew. 3. If the device is AGLE300 and VCCI is 2.5 V, TDO will be configured as LVCMOS 2.5 V output, 24 mA, low slew. The test and debug procedure is not limited to the example in Figure 17-5 on page 369. Users can customize the debug and test interface to make it appropriate for their applications. For example, multiple test signals can be registered and then sent out through UTDO, each at a different edge of TCK. In other words, n signals are sampled with an FTCK / n sampling rate. The bandwidth of the information sent out to TDO is always proportional to the frequency of TCK. Internal Test Nets Instruction Decode UIREG[7:0] To Scope Channel TDO TDI URSTB UDRUPD UDRCK TMS TCK TRST D UDRCAP CLK UDRSH UTDI UTDO Figure 17-6 • UJTAG Usage Example in Test and Debug Applications 370 R e vi s i o n 4 Q ProASIC3L FPGA Fabric User’s Guide SRAM Initialization Users can also initialize embedded SRAMs of the low power flash devices. The initialization of the embedded SRAM blocks of the design can be done using UJTAG tiles, where the initialization data is imported using the TAP Controller. Similar functionality is available in ProASICPLUS devices using JTAG. The guidelines for implementation and design examples are given in the RAM Initialization and ROM Emulation in ProASICPLUS Devices application note. SRAMs are volatile by nature; data is lost in the absence of power. Therefore, the initialization process should be done at each power-up if necessary. FlashROM Read-Back Using JTAG The low power flash architecture contains a dedicated nonvolatile FlashROM block, which is formatted into eight 128-bit pages. For more information on FlashROM, refer to the "FlashROM in Microsemi’s Low Power Flash Devices" section on page 133. The contents of FlashROM are available to the VersaTiles during normal operation through a read operation. As a result, the UJTAG macro can be used to provide the FlashROM contents to the JTAG port during normal operation. Figure 17-7 illustrates a simple block diagram of using UJTAG to read the contents of FlashROM during normal operation. The FlashROM read address can be provided from outside the FPGA through the TDI input or can be generated internally using the core logic. In either case, data serialization logic is required (Figure 17-7) and should be designed using the VersaTile core logic. FlashROM contents are read asynchronously in parallel from the flash memory and shifted out in a synchronous serial format to TDO. Shifting the serial data out of the serialization block should be performed while the TAP is in UDRSH mode. The coordination between TCK and the data shift procedure can be done using the TAP state machine by monitoring UDRSH, UDRCAP, and UDRUPD. UJTAG Address Generation and Data Serlialization UIREG[7:0] TDO TDI TMS TCK TRST Enable FROM RESET URSTB UDRUPD Control UDRCK CLK UDRCAP SDI UDRSH Addr[6:0] Addr [6:0] Data[7:0] Data[7:0] SDO UTDI UTDO Figure 17-7 • Block Diagram of Using UJTAG to Read FlashROM Contents Revision 4 371 UJTAG Applications in Microsemi’s Low Power Flash Devices Conclusion Microsemi low power flash FPGAs offer many unique advantages, such as security, nonvolatility, reprogrammablity, and low power—all in a single chip. In addition, Fusion, IGLOO, and ProASIC3 devices provide access to the JTAG port from core VersaTiles while the device is in normal operating mode. A wide range of available user-defined JTAG opcodes allows users to implement various types of applications, exploiting this feature of these devices. The connection between the JTAG port and core tiles is implemented through an embedded and hardwired UJTAG tile. A UJTAG tile can be instantiated in designs using the UJTAG library cell. This document presents multiple examples of UJTAG applications, such as dynamic reconfiguration, silicon test and debug, fine-tuning of the design, and RAM initialization. Each of these applications offers many useful advantages. Related Documents Application Notes RAM Initialization and ROM Emulation in ProASICPLUS Devices http://www.microsemi.com/soc/documents/APA_RAM_Initd_AN.pdf List of Changes The following table lists critical changes that were made in each revision of the chapter. Date Changes Page December 2011 Information on the drive strength and slew rate of TDO pins was added to the "Silicon Testing and Debugging" section (SAR 31749). 370 July 2010 This chapter is no longer published separately with its own part number and version but is now part of several FPGA fabric user’s guides. N/A v1.4 (December 2008) IGLOO nano and ProASIC3 nano devices were added to Table 17-1 • Flash-Based FPGAs. 364 v1.3 (October 2008) The "UJTAG Support in Flash-Based Devices" section was revised to include new families and make the information more concise. 364 The title of Table 17-3 • Configuration Bits of Fusion, IGLOO, and ProASIC3 CCC Blocks was revised to include Fusion. 368 The following changes were made to the family descriptions in Table 17-1 • FlashBased FPGAs: 364 v1.2 (June 2008) v1.1 (March 2008) 372 • ProASIC3L was updated to include 1.5 V. • The number of PLLs for ProASIC3E was changed from five to six. The chapter was updated to include the IGLOO PLUS family and information regarding 15 k gate devices. N/A The "IGLOO Terminology" section and "ProASIC3 Terminology" section are new. 364 R e vi s i o n 4 18 – Power-Up/-Down Behavior of Low Power Flash Devices Introduction Microsemi’s low power flash devices are flash-based FPGAs manufactured on a 0.13 µm process node. These devices offer a single-chip, reprogrammable solution and support Level 0 live at power-up (LAPU) due to their nonvolatile architecture. Microsemi's low power flash FPGA families are optimized for logic area, I/O features, and performance. IGLOO® devices are optimized for power, making them the industry's lowest power programmable solution. IGLOO PLUS FPGAs offer enhanced I/O features beyond those of the IGLOO ultra-low power solution for I/O-intensive low power applications. IGLOO nano devices are the industry's lowest-power cost-effective solution. ProASIC3®L FPGAs balance low power with high performance. The ProASIC3 family is Microsemi's high-performance flash FPGA solution. ProASIC3 nano devices offer the lowestcost solution with enhanced I/O capabilities. Microsemi’s low power flash devices exhibit very low transient current on each power supply during power-up. The peak value of the transient current depends on the device size, temperature, voltage levels, and power-up sequence. The following devices can have inputs driven in while the device is not powered: • IGLOO (AGL015 and AGL030) • IGLOO nano (all devices) • IGLOO PLUS (AGLP030, AGLP060, AGLP125) • IGLOOe (AGLE600, AGLE3000) • ProASIC3L (A3PE3000L) • ProASIC3 (A3P015, A3P030) • ProASIC3 nano (all devices) • ProASIC3E (A3PE600, A3PE1500, A3PE3000) • Military ProASIC3EL (A3PE600L, A3PE3000L, but not A3P1000) • RT ProASIC3 (RT3PE600L, RT3PE3000L) The driven I/Os do not pull up power planes, and the current draw is limited to very small leakage current, making them suitable for applications that require cold-sparing. These devices are hot-swappable, meaning they can be inserted in a live power system.1 1. For more details on the levels of hot-swap compatibility in Microsemi’s low power flash devices, refer to the "Hot-Swap Support" section in the I/O Structures chapter of the FPGA fabric user’s guide for the device you are using. Revision 4 373 Power-Up/-Down Behavior of Low Power Flash Devices Flash Devices Support Power-Up Behavior The flash FPGAs listed in Table 18-1 support power-up behavior and the functions described in this document. Table 18-1 • Flash-Based FPGAs Family* Series IGLOO ProASIC3 Description IGLOO Ultra-low power 1.2 V to 1.5 V FPGAs with Flash*Freeze technology IGLOOe Higher density IGLOO FPGAs with six PLLs and additional I/O standards IGLOO nano The industry’s lowest-power, smallest-size solution IGLOO PLUS IGLOO FPGAs with enhanced I/O capabilities ProASIC3 Low power, high-performance 1.5 V FPGAs ProASIC3E Higher density ProASIC3 FPGAs with six PLLs and additional I/O standards ProASIC3 nano Lowest-cost solution with enhanced I/O capabilities ProASIC3L ProASIC3 FPGAs supporting 1.2 V to 1.5 V with Flash*Freeze technology RT ProASIC3 Radiation-tolerant RT3PE600L and RT3PE3000L Military ProASIC3/EL Military temperature A3PE600L, A3P1000, and A3PE3000L Automotive ProASIC3 ProASIC3 FPGAs qualified for automotive applications Note: *The device names link to the appropriate datasheet, including product brief, DC and switching characteristics, and packaging information. IGLOO Terminology In documentation, the terms IGLOO series and IGLOO devices refer to all of the IGLOO devices as listed in Table 18-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. ProASIC3 Terminology In documentation, the terms ProASIC3 series and ProASIC3 devices refer to all of the ProASIC3 devices as listed in Table 18-1. Where the information applies to only one product line or limited devices, these exclusions will be explicitly stated. To further understand the differences between the IGLOO and ProASIC3 devices, refer to the Industry’s Lowest Power FPGAs Portfolio. 374 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide Power-Up/-Down Sequence and Transient Current Microsemi's low power flash devices use the following main voltage pins during normal operation:2 • VCCPLX • VJTAG • VCC: Voltage supply to the FPGA core – VCC is 1.5 V ± 0.075 V for IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3 devices operating at 1.5 V. – VCC is 1.2 V ± 0.06 V for IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3L devices operating at 1.2 V. – V5 devices will require a 1.5 V VCC supply, whereas V2 devices can utilize either a 1.2 V or 1.5 V VCC. • VCCIBx: Supply voltage to the bank's I/O output buffers and I/O logic. Bx is the I/O bank number. • VMVx: Quiet supply voltage to the input buffers of each I/O bank. x is the bank number. (Note: IGLOO nano, IGLOO PLUS, and ProASIC3 nano devices do not have VMVx supply pins.) The I/O bank VMV pin must be tied to the VCCI pin within the same bank. Therefore, the supplies that need to be powered up/down during normal operation are VCC and VCCI. These power supplies can be powered up/down in any sequence during normal operation of IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3L, ProASIC3, and ProASIC3 nano FPGAs. During power-up, I/Os in each bank will remain tristated until the last supply (either VCCIBx or VCC) reaches its functional activation voltage. Similarly, during power-down, I/Os of each bank are tristated once the first supply reaches its brownout deactivation voltage. Although Microsemi's low power flash devices have no power-up or power-down sequencing requirements, Microsemi identifies the following power conditions that will result in higher than normal transient current. Use this information to help maximize power savings: Microsemi recommends tying VCCPLX to VCC and using proper filtering circuits to decouple VCC noise from the PLL. a. If VCCPLX is powered up before VCC, a static current of up to 5 mA (typical) per PLL may be measured on VCCPLX. The current vanishes as soon as VCC reaches the VCCPLX voltage level. The same current is observed at power-down (VCC before VCCPLX). b. If VCCPLX is powered up simultaneously or after VCC: i. Microsemi's low power flash devices exhibit very low transient current on VCC. For ProASIC3 devices, the maximum transient current on VCC does not exceed the maximum standby current specified in the device datasheet. The source of transient current, also known as inrush current, varies depending on the FPGA technology. Due to their volatile technology, the internal registers in SRAM FPGAs must be initialized before configuration can start. This initialization is the source of significant inrush current in SRAM FPGAs during power-up. Due to the nonvolatile nature of flash technology, low power flash devices do not require any initialization at power-up, and there is very little or no crossbar current through PMOS and NMOS devices. Therefore, the transient current at power-up is significantly less than for SRAM FPGAs. Figure 18-1 on page 376 illustrates the types of power consumption by SRAM FPGAs compared to Microsemi's antifuse and flash FPGAs. 2. For more information on Microsemi FPGA voltage supplies, refer to the appropriate datasheet located at http://www.microsemi.com/soc/techdocs/ds. Revision 4 375 Power-Up/-Down Behavior of Low Power Flash Devices Current Power-On Inrush SRAM FPGAs SRAM Microsemi FPGAs Active Frequency Dependent System Supply Voltage Configuration SRAM FPGAs Static Time (or frequency) Figure 18-1 • Types of Power Consumption in SRAM FPGAs and Microsemi Nonvolatile FPGAs Transient Current on VCC The characterization of the transient current on VCC is performed on nearly all devices within the IGLOO, ProASIC3L, and ProASIC3 families. A sample size of five units is used from each device family member. All the device I/Os are internally pulled down while the transient current measurements are performed. For ProASIC3 devices, the measurements at typical conditions show that the maximum transient current on VCC, when the power supply is powered at ramp-rates ranging from 15 V/ms to 0.15 V/ms, does not exceed the maximum standby current specified in the device datasheets. Refer to the DC and Switching Characteristics chapters of the ProASIC3 Flash Family FPGAS datasheet and ProASIC3E Flash Family FPGAs datasheet for more information. Similarly, IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3L devices exhibit very low transient current on VCC. The transient current does not exceed the typical operating current of the device while in active mode. For example, the characterization of AGL600-FG256 V2 and V5 devices has shown that the transient current on VCC is typically in the range of 1–5 mA. Transient Current on VCCI The characterization of the transient current on VCCI is performed on devices within the IGLOO, IGLOO nano, IGLOO PLUS, ProASIC3, ProASIC3 nano, and ProASIC3L groups of devices, similarly to VCC transient current measurements. For ProASIC3 devices, the measurements at typical conditions show that the maximum transient current on VCCI, when the power supply is powered at ramp-rates ranging from 33 V/ms to 0.33 V/ms, does not exceed the maximum standby current specified in the device datasheet. Refer to the DC and Switching Characteristics chapters of the ProASIC3 Flash Family FPGAS datasheet and ProASIC3E Flash Family FPGAs datasheet for more information. Similarly, IGLOO, IGLOO PLUS, and ProASIC3L devices exhibit very low transient current on VCCI. The transient current does not exceed the typical operating current of the device while in active mode. For example, the characterization of AGL600-FG256 V2 and V5 devices has shown that the transient current on VCCI is typically in the range of 1–2 mA. 376 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide I/O Behavior at Power-Up/-Down This section discusses the behavior of device I/Os, used and unused, during power-up/-down of VCC and VCCI. As mentioned earlier, VMVx and VCCIBx are tied together, and therefore, inputs and outputs are powered up/down at the same time. I/O State during Power-Up/-Down This section discusses the characteristics of I/O behavior during device power-up and power-down. Before the start of power-up, all I/Os are in tristate mode. The I/Os will remain tristated during power-up until the last voltage supply (VCC or VCCI) is powered to its functional level (power supply functional levels are discussed in the "Power-Up to Functional Time" section on page 378). After the last supply reaches the functional level, the outputs will exit the tristate mode and drive the logic at the input of the output buffer. Similarly, the input buffers will pass the external logic into the FPGA fabric once the last supply reaches the functional level. The behavior of user I/Os is independent of the VCC and VCCI sequence or the state of other voltage supplies of the FPGA (VPUMP and VJTAG). Figure 18-2 shows the output buffer driving HIGH and its behavior during power-up with 10 kΩ external pull-down. In Figure 18-2, VCC is powered first, and VCCI is powered 5 ms after VCC. Figure 18-3 on page 378 shows the state of the I/O when VCCI is powered about 5 ms before VCC. In the circuitry shown in Figure 18-3 on page 378, the output is externally pulled down. During power-down, device I/Os become tristated once the first power supply (VCC or VCCI) drops below its brownout voltage level. The I/O behavior during power-down is also independent of voltage supply sequencing. Figure 18-2 • I/O State when VCC Is Powered before VCCI Revision 4 377 Power-Up/-Down Behavior of Low Power Flash Devices Figure 18-3 • I/O State when VCCI Is Powered before VCC Power-Up to Functional Time At power-up, device I/Os exit the tristate mode and become functional once the last voltage supply in the power-up sequence (VCCI or VCC) reaches its functional activation level. The power-up–to–functional time is the time it takes for the last supply to power up from zero to its functional level. Note that the functional level of the power supply during power-up may vary slightly within the specification at different ramp-rates. Refer to Table 18-2 for the functional level of the voltage supplies at power-up. Typical I/O behavior during power-up–to–functional time is illustrated in Figure 18-2 on page 377 and Figure 18-3. Table 18-2 • Power-Up Functional Activation Levels for VCC and VCCI VCC Functional Activation Level (V) VCCI Functional Activation Level (V) ProASIC3, ProASIC3 nano, IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3L devices running at VCC = 1.5 V* 0.85 V ± 0.25 V 0.9 V ± 0.3 V IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3L devices running at VCC = 1.2 V* 0.85 V ± 0.2 V 0.9 V ± 0.15 V Device Note: *V5 devices will require a 1.5 V VCC supply, whereas V2 devices can utilize either a 1.2 V or 1.5 V VCC. Microsemi’s low power flash devices meet Level 0 LAPU; that is, they can be functional prior to VCC reaching the regulated voltage required. This important advantage distinguishes low power flash devices from their SRAM-based counterparts. SRAM-based FPGAs, due to their volatile technology, require hundreds of milliseconds after power-up to configure the design bitstream before they become functional. Refer to Figure 18-4 on page 379 and Figure 18-5 on page 380 for more information. 378 R e vi s i o n 4 ProASIC3L FPGA Fabric User’s Guide VCC = VCCI + VT Where VT can be from 0.58 V to 0.9 V (typically 0.75 V) VCC VCC = 1.575 V Region 4: I/O buffers are ON. I/Os are functional (except differential inputs) but slower because VCCI is Region 1: I/O Buffers are OFF below specifcation. For the same reason, input buffers do not Region 5: I/O buffers are ON and power supplies are within specification. I/Os meet the entire datasheet and timer specifications for speed, VIH/VIL , VOH /VOL , etc. meet VIH/VIL levels, and output buffers do not meet VOH/VOL levels. VCC = 1.425 V Region 2: I/O buffers are ON. I/Os are functional (except differential inputs) but slower because VCCI / VCC are below specification. For the same reason, input buffers do not meet VIH / VIL levels, and output buffers do not meet VOH / VOL levels. Activation trip point: Va = 0.85 V ± 0.25 V Deactivation trip point: Vd = 0.75 V ± 0.25 V Region 3: I/O buffers are ON. I/Os are functional; I/O DC specifications are met, but I/Os are slower because the VCC is below specification Region 1: I/O buffers are OFF Activation trip point: Va = 0.9 V ± 0.3 V Deactivation trip point: Vd = 0.8 V ± 0.3 V Min VCCI datasheet specification voltage at a selected I/O standard; i.e., 1.425 V or 1.7 V or 2.3 V or 3.0 V VCCI Figure 18-4 • I/O State as a Function of VCCI and VCC Voltage Levels for IGLOO V5, IGLOO nano V5, IGLOO PLUS V5, ProASIC3L, and ProASIC3 Devices Running at VCC = 1.5 V ± 0.075 V Revision 4 379 Power-Up/-Down Behavior of Low Power Flash Devices VCC = VCCI + VT where VT can be from 0.58 V to 0.9 V (typically 0.75 V) VCC VCC = 1.575 V Region 4: I/O buffers are ON. I/Os are functional (except differential inputs) but slower because VCCI is below specification. For the same reason, input buffers do not meet VIH / VIL levels, and output buffers do not meet VOH / VOL levels. Region 1: I/O Buffers are OFF Region 5: I/O buffers are ON and power supplies are within specification. I/Os meet the entire datasheet and timer specifications for speed, VIH / VIL , VOH / VOL , etc. VCC = 1.14 V Region 2: I/O buffers are ON. I/Os are functional (except differential inputs) but slower because VCCI/VCC are below specification. For the same reason, input buffers do not meet VIH/VIL levels, and output buffers do not meet VOH/VOL levels. Activation trip point: Va = 0.85 V ± 0.2 V Deactivation trip point: Vd = 0.75 V ± 0.2 V Region 3: I/O buffers are ON. I/Os are functional; I/O DC specifications are met, but I/Os are slower because the VCC is below specification. Region 1: I/O buffers are OFF Activation trip point: Va = 0.9 V ± 0.15 V Deactivation trip point: Vd = 0.8 V ± 0.15 V Min VCCI datasheet specification voltage at a selected I/O standard; i.e., 1.14 V,1.425 V, 1.7 V, 2.3 V, or 3.0 V Figure 18-5 • I/O State as a Function of VCCI and VCC Voltage Levels for IGLOO V2, IGLOO nano V2, IGLOO PLUS V2, and ProASIC3L Devices Running at VCC = 1.2 V ± 0.06 V 380 R e vi s i o n 4 VCCI ProASIC3L FPGA Fabric User’s Guide Brownout Voltage Brownout is a condition in which the voltage supplies are lower than normal, causing the device to malfunction as a result of insufficient power. In general, Microsemi does not guarantee the functionality of the design inside the flash FPGA if voltage supplies are below their minimum recommended operating condition. Microsemi has performed measurements to characterize the brownout levels of FPGA power supplies. Refer to Table 18-3 for device-specific brownout deactivation levels. For the purpose of characterization, a direct path from the device input to output is monitored while voltage supplies are lowered gradually. The brownout point is defined as the voltage level at which the output stops following the input. Characterization tests performed on several IGLOO, ProASIC3L, and ProASIC3 devices in typical operating conditions showed the brownout voltage levels to be within the specification. During device power-down, the device I/Os become tristated once the first supply in the power-down sequence drops below its brownout deactivation voltage. Table 18-3 • Brownout Deactivation Levels for VCC and VCCI VCC Brownout VCCI Brownout Deactivation Level (V) Deactivation Level (V) Devices ProASIC3, ProASIC3 nano, IGLOO, IGLOO nano, IGLOO PLUS and ProASIC3L devices running at VCC = 1.5 V 0.75 V ± 0.25 V 0.8 V ± 0.3 V IGLOO, IGLOO nano, IGLOO PLUS, and ProASIC3L devices running at VCC = 1.2 V 0.75 V ± 0.2 V 0.8 V ± 0.15 V PLL Behavior at Brownout Condition When PLL power supply voltage and/or VCC levels drop below the VCC brownout levels mentioned above for 1.5 V and 1.2 V devices, the PLL output lock signal goes LOW and/or the output clock is lost. The following sections explain PLL behavior during and after the brownout condition. VCCPLL and VCC Tied Together In this condition, both VCC and VCCPLL drop below the 0.75 V (± 0.25 V or ± 0.2 V) brownout level. During the brownout recovery, once VCCPLL and VCC reach the activation point (0.85 ± 0.25 V or ± 0.2 V) again, the PLL output lock signal may still remain LOW with the PLL output clock signal toggling. If this condition occurs, there are two ways to recover the PLL output lock signal: 1. Cycle the power supplies of the PLL (power off and on) by using the PLL POWERDOWN signal. 2. Turn off the input reference clock to the PLL and then turn it back on. Only VCCPLL Is at Brownout In this case, only VCCPLL drops below the 0.75 V (± 0.25 V or ± 0.2 V) brownout level and the VCC supply remains at nominal recommended operating voltage (1.5 V ± 0.075 V for 1.5 V devices and 1.2 V ± 0.06 V for 1.2 V devices). In this condition, the PLL behavior after brownout recovery is similar to initial power-up condition, and the PLL will regain lock automatically after VCCPLL is ramped up above the activation level (0.85 ± 0.25 V or ± 0.2 V). No intervention is necessary in this case. Only VCC Is at Brownout In this condition, VCC drops below the 0.75 V (± 0.25 V or ± 0.2 V) brownout level and VCCPLL remains at nominal recommended operating voltage (1.5 V ± 0.075 V for 1.5 V devices and 1.2 V ± 0.06 V for 1.2 V devices). During the brownout recovery, once VCC reaches the activation point again (0.85 ± 0.25 V or ± 0.2 V), the PLL output lock signal may still remain LOW with the PLL output clock signal toggling. If this condition occurs, there are two ways to recover the PLL output lock signal: 1. Cycle the power supplies of the PLL (power off and on) by using the PLL POWERDOWN signal. 2. Turn off the input reference clock to the PLL and then turn it back on. It is important to note that Microsemi recommends using a monotonic power supply or voltage regulator to ensure proper power-up behavior. Revision 4 381 Power-Up/-Down Behavior of Low Power Flash Devices Internal Pull-Up and Pull-Down Low power flash device I/Os are equipped with internal weak pull-up/-down resistors that can be used by designers. If used, these internal pull-up/-down resistors will be activated during power-up, once both VCC and VCCI are above their functional activation level. Similarly, during power-down, these internal pull-up/-down resistors will turn off once the first supply voltage falls below its brownout deactivation level. Cold-Sparing In cold-sparing applications, voltage can be applied to device I/Os before and during power-up. Coldsparing applications rely on three important characteristics of the device: 1. I/Os must be tristated before and during power-up. 2. Voltage applied to the I/Os must not power up any part of the device. 3. VCCI should not exceed 3.6 V, per datasheet specifications. As described in the "Power-Up to Functional Time" section on page 378, Microsemi’s low power flash I/Os are tristated before and during power-up until the last voltage supply (VCC or VCCI) is powered up past its functional level. Furthermore, applying voltage to the FPGA I/Os does not pull up VCC or VCCI and, therefore, does not partially power up the device. Table 18-4 includes the cold-sparing test results on A3PE600-PQ208 devices. In this test, leakage current on the device I/O and residual voltage on the power supply rails were measured while voltage was applied to the I/O before power-up. Table 18-4 • Cold-Sparing Test Results for A3PE600 Devices Residual Voltage (V) Device I/O VCC VCCI Leakage Current Input 0 0.003
A3P1000L-1FG144
1. 物料型号 2. 器件简介 3. 引脚分配 4. 参数特性 5. 功能详解 6. 应用信息 7. 封装信息

由于文档内容较多,我将提供概要信息。如果需要更详细的内容,请告知。

1. 物料型号:文档中未明确列出具体的物料型号,可能需要根据上下文或文档的其他部分来确定。 2. 器件简介:文档提供了Microsemi的低功耗闪存FPGA家族的概述,包括IGLOO、ProASIC3L、ProASIC3等系列。 3. 引脚分配:文档中没有提供具体的引脚分配图或表格,但通常这些信息可以在器件的数据手册中找到。 4. 参数特性:文档讨论了低功耗闪存设备的电源特性,如瞬态电流、电源上升时间等,但没有列出具体的参数值。 5. 功能详解:文档详细讨论了低功耗闪存设备的功能,包括其非易失性特性、ISP(在系统编程)能力、安全性特性等。 6. 应用信息:文档提供了低功耗闪存设备的应用场景,如冷备份和热插拔支持,适用于需要这些特性的应用。 7. 封装信息:文档中没有提供具体的封装信息,但通常这些信息也可以在器件的数据手册中找到。
A3P1000L-1FG144 价格&库存

很抱歉,暂时无法提供与“A3P1000L-1FG144”相匹配的价格&库存,您可以联系我们找货

免费人工找货