Wideband IF Receiver Subsystem
AD6676
Data Sheet
FEATURES
APPLICATIONS
High instantaneous dynamic range
Noise figure (NF) as low as 13 dB
Noise spectral density (NSD) as low as −159 dBFS/Hz
IIP3 up to 36.9 dBm with spurious tones 85 dB
Rejection)
0.814
0.814
0.571
0.571
BW (>60 dB
Rejection)
0.834
0.834
0.617
0.617
Total Pipeline Latency
The digital filter path dominates the latency of the AD6676
whereas the JESD204B PHY adds a few samples of delay and
the ADC delay is a fraction of an output sample. The latency
between the ADC and digital filter output is fixed with the only
nondeterministic delay being associated with the JESD204B
PHY clock and lane FIFOs before synchronization. See the
Synchronization Using SYSREF section for additional
information. Table 15 provides the nominal pipeline delay
associated with each DEC_MODE. Note that although all
DEC_MODE settings provide similar delays relative to the
output data rate, fDATA_IQ, applications that require shorter
absolute time delays may consider using a lower decimation
factor to reduce the absolute delay by 2×.
Table 15. Nominal Pipeline Latency vs. DEC_MODE
(Sample Delay Relative to 1/fDATA_IQ)
DEC_MODE
1
2
3
4
All filter responses provide a linear phase response over its
pass band.
The usable IF bandwidth depends on the DEC_MODE as
well as the minimum acceptable pass band ripple and stop
band rejection requirements. Table 14 provides the
normalized usable complex bandwidth vs. DEC_MODE for
stop band rejections of greater than 85 dB and 60 dB.
The last filter stage sets the usable bandwidth and stop
band rejection because it has the most aggressive transition
band specifications. For this reason, the decimation factors
of 12 and 16 have the same normalized usable bandwidths
as does decimation factors of 24 and 32.
Rev. D | Page 40 of 90
Decimation
Factor
32
24
16
12
JESD204B
Lanes
1
1
2
2
IQ Data Output
Sample Delay
34.2
34.2
32.3
32.3
Data Sheet
AD6676
0.5
0
0.4
–20
0.3
MAGNITUDE (dB)
MAGNITUDE (dB)
0.2
0.1
0
–0.1
–0.2
–0.3
–40
–60
–80 MINIMUM ALIAS ATTENUATION 85.5dB
–100
–120
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
12348-056
–0.5
Figure 105. Pass Band Frequency Response of Decimate by 12
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
12348-059
–0.4
Figure 108. Folded Frequency Response of Decimate by 12 Shows Alias
Rejection
0.5
0
0.4
–20
0.3
MAGNITUDE (dB)
MAGNITUDE (dB)
0.2
0.1
0
–0.1
–0.2
–0.3
–40
–60
–80 MINIMUM ALIAS ATTENUATION 85.5dB
–100
–120
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
12348-057
–0.5
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
12348-060
–0.4
Figure 109. Folded Frequency Response of Decimate by 16 Shows Alias
Rejection
Figure 106. Pass Band Frequency Response of Decimate by 16
0.5
0
0.4
–20
0.3
MAGNITUDE (dB)
0.1
0
–0.1
–0.2
–0.3
–40
–60
–80 MINIMUM ALIAS ATTENUATION 85.5dB
–100
–0.5
–120
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
Figure 107. Pass Band Frequency Response of Decimate by 24
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
FREQUENCY NORMALIZED TO fDATA_IQ
0.45
0.50
12348-061
–0.4
12348-058
MAGNITUDE (dB)
0.2
Figure 110. Folded Frequency Response of Decimate by 24 Shows Alias
Rejection
Rev. D | Page 41 of 90
AD6676
Data Sheet
0.5
Peak Detection and AGC Flags
0.4
Peak detection occurs at the output of the second stage decimation
filter, as shown in Figure 103. Detection at this stage represents a
compromise between the accuracy of the peak detector, delay time
and ability to measure large out-of-band signals. At this stage, the
Σ-Δ ADC output signal has been frequency translated to dc and its
out-of-band noise sufficiently filtered for reasonable threshold
detection accuracy down to −12 dBFS peak signal levels. Note that
the peak detector monitors the peak power envelope response of
the IF input signal and calculates the peak power (that is, I2 + Q2)
expressed in dBFS with 12-bit resolution.
0.3
MAGNITUDE (dB)
0.2
0.1
0
–0.1
–0.2
–0.3
–0.5
0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FREQUENCY NORMALIZED TO fDATA_IQ
12348-062
–0.4
Figure 111. Pass Band Frequency Response of Decimate by 32
0
MAGNITUDE (dB)
–20
–40
–60
–80
MINIMUM ALIAS ATTENUATION 88dB
–100
Because the peak detector is monitoring the peak power at the
output of the second stage decimation filter, it provides a wider
frequency range than what can be observed in the final IQ data
output. The first stage filter is decimate by 3 or by 4; therefore,
the output of the second stage filter can be 1/6th or 1/8th of FADC.
Figure 113 shows the normalized measurement bandwidth relative
to the output rate of the second stage filter centered about its zero
IF. Table 16 references the measurement bandwidth to fDATA_IQ for
the different decimation factors such that its absolute bandwidth
can be easily determined. For example, the −1 dB bandwidth for an
fDATA_IQ of 100 MSPS with decimate by 24 or by 32 is 200 MHz and
remains at 200 MHz if the decimation factor is reduced to decimate
by 12 or by 16. Any droop occurring at the pass band edges, as well
as the Σ-Δ ADC STF, must be considered when setting thresholds.
0.10
0.15
0.20
0.25
0.30
0.35
0.40
FREQUENCY NORMALIZED TO fDATA_IQ
0.45
0.50
Figure 112. Folded Frequency Response of Decimate by 32 Shows Alias
Rejection
AGC FEATURES AND PEAK DETECTION
In receiver applications, it is desirable to have a mechanism to
reliably determine when the converter is about to be overdriven.
The AD6676 Σ-Δ ADC is based on a feedback loop that can be
overdriven into a nonlinear region, resulting in oscillation. This
oscillation persists until the Σ-Δ ADC is reset and the overload
condition is removed. Typically, a receiver lineup employs some
form of AGC that attempts to avoid this scenario.
The AD6676 pipeline latency along with any additional
overhead associated with the host processor (JESD204B Rx
PHY) may limit the ability to design a fast reacting digital-based
AGC required by some applications. For this reason, the
AD6676 includes the AGCx pins that serve as digital input/
outputs to facilitate the implementation of a fast AGC control
loop under the control of the host. The AGC4 and AGC3 pins
can be allocated to provide flag outputs after a programmable
threshold has been exceeded, including an ADC reset event,
while the AGC2 and AGC1 pins can be used to control the onchip attenuator. Register 0x18F and Register 0x193 through
Register 0x19E are used for AGC purposes.
0
–0.5
–1.0
–1.5
–2.0
–2.5
–3.0
–0.35
–0.25
–0.15
–0.05
0.05
0.15
0.25
0.35
FREQUENCY OFFSET FROM IF CENTER
(NORMALIZED TO FIQ@2ND STAGE FILTER)
12348-064
0.05
MAGNITUDE (dBFS)
0
12348-063
0.5
–120
Figure 113. Normalized Pass Band Filter Response Seen by the Peak Detector
Table 16. Normalized Measurement Bandwidth of Peak
Detector Relative to Output Data Rate, fDATA_IQ
DEC_
MODE
1
2
3
4
Rev. D | Page 42 of 90
Decimation
Factor
32
24
16
12
Normalized Measurement Bandwidth
Relative to fDATA_IQ
−0.5
−1.0
−2.0
−3.0
dBFS
dBFS
dBFS
dBFS
1.76
2.00
2.40
2.64
1.76
2.00
2.40
2.64
0.88
1.00
1.20
1.32
0.88
1.00
1.20
1.32
Data Sheet
AD6676
The AD6676 allows the user to set three threshold settings that
can trigger one of two possible flags. PKTHRH0 and PKTHRH1
are two upper threshold settings while LOWTHRH is a lower
threshold setting. The threshold settings are 12 bits with an
MSB and LSB register assigned to each threshold. The 12-bit
decimal equivalent value can be calculated using Equation 12.
Threshold = 3584 + (Threshold Setting in dBFS) × 256/3
(12)
where 0 dBFS corresponds to 3584 (0xE00) and −6 dBFS
corresponds to 3072 (0xC00).
In the time domain, a 0 dBFS setting corresponds to a signal
whose peaks observed at the I and Q outputs can reach plus or
minus full scale. Meaning, if the 16-bit I and Q output data are
normalized such that its peak values correspond to ±1, a 0 dBFS
setting corresponds to a signal whose peak can reach the unit
circle of a normalized I/Q constellation diagram.
The LOWTHRH_x register has an associated dwell time of
which the signal must remain below this threshold before a flag
can be set. The dwell time is represented in exponential form to
realize long dwell periods because the counter operates at
FADC/12 for decimate by 12 or 24 settings or FADC/16 for decimate
by 16 or 32 settings. The dwell time is set in the DWELL_TIME_
MANTISSA register and DWELL_TIME_EXP register using
Equation 13 relative to 1/FADC.
Dwell Time = N × [DWELL_TIME_MANTISSA] ×
2(DWELL_TIME_EXP)
(13)
where:
N = 12 for decimate by 12 or 24.
N = 16 for decimate by 16 or 32.
Note that the EN_FLAGx bits provide the additional option of
logically OR’ing an ADC reset event with an upper peak
threshold event to provide an even faster output flag to the host
processor indicating that the attenuation must be applied. This
scenario applies to the extreme case where the envelope
response of a blocker is exceedingly fast, such that the AGC
cannot react fast enough to the upper peak threshold setting
flag to prevent overloading the Σ-Δ ADC.
Figure 114 provides an example of how the Flag 0 and Flag 1
assigned pins behave to the envelope response of an arbitrary IF
input signal. Flag 1 is assigned an upper threshold set by
PKTHRH1_x, and Flag 0 is assigned a lower threshold and
dwell time set by LOWTHRH_x and DWELL_TIME_x. The
Flag 1 indicator goes high when the PKTTHR1_x threshold is
exceeded and returns low when the signal envelope falls below
this threshold. The Flag 0 indicator goes high only when the
envelope of the signal remains below the LOWTHRH_x
threshold for the designated dwell time. If the signal level
exceeds the LOWTHRH_x threshold before the dwell time
counter has expired, the dwell time counter resets again and the
Flag 0 indicator remains low until the conditions has been met.
By offsetting the PKTTHR1_x and LOWTHRH_x threshold
settings as well as optimizing the dwell time setting, it may be
possible to optimize the operation of an AGC so that it reacts to
signal strength variation due to fading conditions as opposed to the
peak to minimum response associated with digital modulated
signals.
IF Attenuator Control via the AGC2 and AGC1 Pins
A flag function can be assigned using the FLAG0_SEL register
and FLAG1_SEL register to indicate when any of the thresholds
have been exceeded or if an ADC reset event has occurred.
These flags must also be enabled via the EN_FLAG register
such that a CMOS level signal appears on the AGC4 and AGC3
pins where a logic high indicates when a threshold has been
exceeded.
The delay relative to the ADC input when an AGC threshold is
exceeded to when the flag signal goes high is dependent on the
DEC_MODE setting selected. For a DEC_MODE value of 1 or
2 (decimate by 32 or 24), the delay equates to 8 to 9 output
samples (1/fDATA_IQ). For DEC_MODE values of 3 or 4 (decimate
by 16 or 12), the delay is 16 to 18 samples. The delay associated
with an ADC reset event is much shorter because it avoids the
digital filter path. This delay is 1 sample for DEC_MODE values
of 1 or 2 and 2 samples for DEC_MODE values of 3 and 4.
Many AGC implementations require fast gain control if the
AGC threshold is exceeded. The AD6676 provides two modes in
which the IF attenuator can be quickly changed via the AGCx pins.
Use Register 0x180, Bit 0, to select the mode. The first mode uses
the AGC2 pin to switch between two attenuator settings that are
user defined in Register 0x181 and Register 0x182. The second
mode uses the AGC2 and AGC1 pins to decrement and increment
respectively the attenuation value in 1 dB steps with pulsed inputs.
The starting attenuator value is defined in Register 0x183. The
actual attenuator value can be read back via Register 0x184.
The first mode is used for the default AD6676 power-up setting
with both Register 0x181 and Register 0x182 set to 0x0C. For
applications that do not require IF attenuator control but
require a different attenuator setting, update both registers with
the desired attenuator setting value such that the attenuator
remains independent of the AGC2 pin state, if it is left floating.
Note that connecting the unused AGC2 and AGC1 pins to
VSSD via 100 kΩ pull-down resistors is still the preferred
method if these pins are unused.
Rev. D | Page 43 of 90
AD6676
Data Sheet
UPPER THRESHOLD (PKTTHR1)
DWELL TIME
TIMER RESET BY
RISE ABOVE
LOWER
THRESHOLD
MIDSCALE
LOWER THRESHOLD (LOWTHRH)
DWELL TIME
TIMER COMPLETES BEFORE
SIGNAL RISES ABOVE
LOWER THRESHOLD
12348-065
FLAG1
FLAG0
Figure 114. Example of Flag Behavior as Signal Envelope Crosses Upper and Lower Threshold Settings
Table 17. Power Saving for 3.0 GSPS Operation with CLKSYN Enabled and 125 MSPS IQ Rate (Single JESD204B Lane)
Power State at 3 GSPS
STDBY_SLOW
STDBY_FAST
Power Down
Power Up
IVDD2 (mA)
18
95
2.6
146
IVDD1+, IVDDC+ IVDDL (mA)
162
175
25
433
GPIO FUNCTIONALITY
The AGCx pins can also be configured for basic GPIO
functionality via Register 0x1B0 to Register 0x1B4. Register 0x1B0
determines which pins are used for GPIO functionality, whereas
Register 0x1B1 determines if an AGC pin serves as input or
output. If the pin serves as an output, Register 0x1B2 determines
the high or low state, and Register 0x1B3 reads back the state of
these designated output pins. Lastly, if an AGCx pin serves as an
input, Register 0x1B4 reads back the state of this pin.
POWER SAVING MODES
The AD6676 features two SPI configurable and selectable power
savings modes. The first mode is a sleep mode where the AD6676
is placed in a low power state for extended periods, and the
second mode is a standby mode where the AD6676 enters a
reduced power state but still keeps the JESD204B link and digital
clocks active to ensure multichip synchronization (or fixed
latency) during fast power cycling. Both sleep mode and standby
mode can be entered via a SPI write operation to the PD_MODE
bits in the DEVICE_CONFIG register (Register 0x002; Bits[1:0]).
Note that, depending on whether sleep or standby mode is
selected, various functional blocks within the Σ-Δ ADC itself
are either powered down, placed in a low bias state, or remain
powered.
The standby mode is also controllable via a user designated
AGCx pin for faster and more precise power cycling. This
feature is particularly useful for TDD-based communication
protocols, allowing the host processor to quickly power cycle
the AD6676 during transmit bursts. The PD_PIN_CTRL
register (Register 0x152) enables this feature as well as
designates the AGC pin.
IVDDD (mA)
216
221
29
310
PTOTAL (mW)
461
673
64
1182
Percent (%) Power Savings
61
43
Not applicable
Not applicable
The standby register (Register 0x150) powers down different
functional blocks during standby mode. However, all functional
blocks that affect the clock generation, distribution and the
JESD204B link remain enabled to maintain constant latency
while in standby. The only exception is STBY_VSS2GEN
(Register 0x150, Bit 6) where a trade-off exists in power savings
vs. wake-up time, depending on whether the negative voltage
generator is placed in standby.
Table 17 shows the realized power savings for the different
power savings modes at 3.0 GSPS operation with the AD6676
configured for 125 MSPS IQ output and the internal clock
synthesizer enabled. Note that STDBY_FAST and STDBY_SLOW
correspond to whether the STBY_VSS2GEN bit is enabled or
disabled during standby. Note that an additional 18% power
savings can be achieved when powering down the
STBY_VSS2GEN bit.
Although the AD6676 can enter into standby quickly, it does
require a few microseconds to exit standby. Figure 115 shows
that the AD6676 can achieve a low power state within 100 ns.
Figure 116 and Figure 117 show the wake-up time between the
STDBY_FAST and STDBY_SLOW cases to achieve 1% envelope
settling accuracy being around 2.5 μs and 11.5 μs, respectively.
The phase response is not shown because it settles faster than
the envelope response. Note that the digital data path is enabled
for these time domain figures such that the setting time
responses can be observed.
Rev. D | Page 44 of 90
Data Sheet
AD6676
1.0
INTRODUCTION TO THE JESD204B INTERFACE
NORMALIZED ENVELOPE OUTPUT
0.9
The JESD204B interface reduces the PCB area for data interface
routing yet enabling the use of smaller packages for converter
and logic devices. The AD6676 digital output complies with the
JEDEC Standard No. JESD204B, Serial Interface for Data
Converters. JESD204B is a protocol to link the AD6676 to a
digital processing device over a serial interface. The AD6676
supports link rates of up 5.333 Gbps while operating with two
output lanes in support of a maximum I/Q data rate (fDATA_IQ)
of 266.67 MSPS. Note that a two output lane configuration is
always required for decimation factors of 12 and 16.
0.8
0.7
AD6676
ENTERS STANDBY
WITHIN 100ns
0.6
0.5
0.4
0.3
0.2
0
10.0
10.0
10.2
10.3
10.4
10.5
10.6
10.7
10.8
10.9
11.0
TIME (µs)
12348-066
0.1
Figure 115. Fast Power-Down Response When the AD6676 Is Placed in
Standby
1.0
1% SETTLING TIME = 2.5µs
NORMALIZED ENVELOPE OUTPUT
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0
14.0
14.5
15.0
15.5
16.0
16.5
17.0
17.5
18.0
18.5
19.0
TIME (µs)
12348-067
0.1
Figure 116. Settling Time for STDBY_FAST with the STBY_VSS2GEN Enabled
for Fastest Recovery, Approximately 2.5 μs to 1 %
JESD204B data transmit block assembles the parallel data
from the ADC into frames and uses 8-bit/10-bit encoding as
well as optional scrambling to form serial output data. Lane
synchronization is supported through the use of special
characters during the initial establishment of the link and
additional synchronization is embedded in the data stream
thereafter. A JESD204B receiver is required to complete the
serial link. For additional details on the JESD204B interface,
refer to the JESD204B standard.
Because the AD6676 provides 16-bit complex IQ data, its
JESD204B transmit block effectively maps the output of two
virtual ADCs (M = 2) over a link. The link is configurable for
either single or dual lanes with each lane providing a serial data
stream via a differential output. The JESD204B specification
refers to a number of parameters to define the link and these
parameters must match between the AD6676 JESD204B
transmitter and receiver.
The following parameters describe a JESD204B link:
1.0
0.9
NORMALIZED ENVELOPE OUTPUT
JESD204B Overview
1% SETTLING TIME = 11.5µs
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0
0
2
4
6
8
10
TIME (µs)
12
14
16
18
20
12348-068
0.1
Figure 117. Settling Time for STDBY_SLOW with STBY_VSS2GEN in Standby
for Additional Power Savings, Approximately 11.5 μs to 1 %
Rev. D | Page 45 of 90
S = samples transmitted per single converter per frame
cycle (AD6676 value = 1)
M = number of converters per converter device (AD6676
value = 2)
L = number of lanes per converter device (AD6676 value
can be 1 or 2)
N = converter resolution (AD6676 value = 16)
N’ = total number of bits per sample (AD6676 value = 16)
CF = number of control words per frame clock cycle per
converter device (AD6676 value = 0)
CS = number of control bits per conversion sample
(AD6676 value = 0)
K = number of frames per multiframe (configurable on the
AD6676 up to 32)
HD = high density mode (AD6676 value = 0)
F = octets per frame (AD6676 value = 2 or 4, dependent on
L = 2 or 1)
T = tail bit (AD6676 value = 0)
SCR = scrambler enable or disable (configurable on the
AD6676)
AD6676
Data Sheet
The scrambler uses a self synchronizing polynomial-based
algorithm defined by the equation 1 + x14 + x15. The descrambler
in the receiver must be a self synchronizing version of the
scrambler polynomial.
Figure 118 shows a simplified block diagram of the AD6676
JESD204B link mapping the 16-bit I and Q outputs onto the two
separate lanes. Other configurations are also possible, such as
combining the I and Q outputs onto a single lane (fDATA_IQ ≤
153.6 MSPS) or changing the mapping of the I and Q output
paths. In any case, the 16-bit I and Q data are each broken into
two octets (eight bits of data). Bit 15 (MSB) through Bit 8 are in
the first octet. The second octet contains Bit 7 through Bit 0
(LSB). The four resulting octets (2 I octets and 2 Q octets) may
be scrambled. Scrambling is optional but is available to avoid
spectral peaks when transmitting similar digital data patterns.
The four octets are then encoded with an 8-bit/10-bit encoder. The
8-bit/10-bit encoder takes eight bits of data (an octet) and encodes
them into a 10-bit symbol. Figure 119 shows how the 16-bit I or
Q data is taken from the final decimation stage, formed into
octets, the two octets are scrambled, and how the octets are
encoded into two 10-bit symbols.
I
CONVERTER 0
JESD204B LINK
CONTROL
(L.M.F.M.S)
(SPI REG 0x1C3
TO 0x1C9)
FRAMER
SWAP
(SPI
REG 0x1E1)
ADC
SERDOUT0–,
SERDOUT0+
LANE SWAP
(SPI REG 0x1E1)
SERDOUT1–,
SERDOUT1+
12348-069
Q
CONVERTER 1
SYSREF±
SYNCINB±
Figure 118. Transmit Link Simplified Block Diagram
ADC TEST PATTERNS
(REG 0x1E5)
JESD204B SAMPLE
CONSTRUCTION
ADC
16-BIT
JESD204B TEST
PATTERNS
(REG 0x1E5,
REG 0x1F8 TO
REG 0x1FF)
FRAME
CONSTRUCTION
SCRAMBLER
1 + x14 + x15
(OPTIONAL
VIA 0x1C3)
JESD204B TEST
PATTERNS
(REG 0x1E5,
REG 0x1F8
TO REG 0x1FF)
SERIALIZER
8-BIT/10-BIT
ENCODER
(0x1E4)
a b
a b c d e f g h i j
SERDOUT0±
SERDOUT1±
i j a b
SYMBOL0
i j
SYMBOL1
a b c d e f g h i j
12348-070
JESD204B TEST
PATTERNS
(0x1E5)
Figure 119. Digital Formatting of JESD204B Lanes
PROCESSED
SAMPLES
FROM ADC
SAMPLE
CONSTRUCTION
FRAME
CONSTRUCTION
DATA LINK
LAYER
SCRAMBLER
ALIGNMENT
CHARACTER
GENERATION
8-BIT/10-BIT
ENCODER
PHYSICAL
LAYER
CROSSBAR
MUX
SERIALIZER
Tx
OUTPUT
12348-071
TRANSPORT
LAYER
SYSREF±
SYNCINB±
Figure 120. Data Flow
Rev. D | Page 46 of 90
Data Sheet
AD6676
FUNCTIONAL OVERVIEW
The flowchart in Figure 120 shows the flow of data through the
JESD204B hardware from the sample input to the physical output.
The processing is divided into layers that are derived from the
OSI model widely used to describe the abstraction layers of
communications systems. These are the transport layer, the data
link layer, and the physical layer (serializer and output driver).
Transport Layer
The transport layer packs the data into JESD204B frames,
which are mapped to 8-bit octets that are sent to the data link
layer. The transport layer mapping is controlled by rules derived
from the link parameters. The AD6676 uses no tail bits in the
transport layer because the output of its IQ digital data path is
considered two virtual 16-bit converters.
Data Link Layer
The data link layer is responsible for the low level functions of
passing data across the link. These include optional data
scrambling, inserting control characters for lane alignment/
monitoring, and encoding 8-bit octets into 10-bit symbols. The
data link layer also sends the initial lane alignment sequence
(ILAS), which contains the link configuration data, and is used
by the receiver to verify the settings in the transport layer.
The SYNCINB± pin operation options are controllable via SPI
registers. Although the SYNCINB input is configured for a
CMOS logic level on its positive pin by default, it can also be
configured for a differential LVDS input signal on its positive/
negative pins via Register 0x1E7. The polarity of the SYNCINB
input signal can also be inverted via Register 0x1E4.
Initial Lane Alignment Sequence (ILAS)
The ILAS phase follows the CGS phase and begins on the next
LMFC boundary. The ILAS consists of four multiframes, with a
/R/ character marking the beginning and an /A/ character
marking the end. The ILAS begins by sending an /R/ character
followed by a data ramp starting with the value, 0, over four
multiframes. On the second multiframe, the link configuration
data is sent, starting with the third character. The second character
in the second multiframe is a /Q/ character to confirm that the
link configuration data follows. All undefined data slots are
filled with ramp data. The ILAS sequence is never scrambled.
The ILAS sequence construction is shown in Figure 121. The
four multiframes include the following:
Physical Layer
The physical layer consists of the high speed circuitry clocked at
the serial clock rate. For the AD6676, the 16-bit I and Q data are
converted into one or two lanes of high speed differential serial data.
JESD204B LINK ESTABLISHMENT
The AD6676 JESD204B Tx interface operates in Subclass 0 or
Subclass 1 as defined in the JEDEC Standard No. 204B (July 2011)
specification. The link establishment process is divided into the
following steps: code group synchronization, ILAS, and user data.
Code Group Synchronization (CGS) and SYNCINB
Code group synchronization (CGS) is the process where the
JESD204B receiver finds the boundaries between the 10-bit
symbols in the stream of data. During the CGS phase, the
JESD204B transmit (JESD Tx) block transmits /K28.5/ characters.
The receiver must locate /K28.5/ characters in its input data
stream using clock and data recovery (CDR) techniques.
Multiframe 1: Begins with an /R/ character (/K28.0/) and
ends with an /A/ character (/K28.3/).
Multiframe 2: Begins with an /R/ character followed by a
/Q/ (/K28.4/) character, followed by link configuration
parameters over 14 configuration octets (see Table 18), and
ends with an /A/ character. Many of the parameter values
are of the notation of n − 1.
Multiframe 3: Begins with an /R/ character (/K28.0/) and
ends with an /A/ character (/K28.3/).
Multiframe 4: Begins with an /R/ character (/K28.0/) and
ends with an /A/ character (/K28.3/).
User Data and Error Detection
After the ILAS is complete, the user data is sent. Normally, in a
frame all characters are user data. However, to monitor the
frame clock and multiframe clock synchronization, there is a
mechanism for replacing characters with /F/ or /A/ alignment
characters when the data meets certain conditions. These
conditions are different for unscrambled and scrambled data.
The scrambling operation is disabled by default, but may be
enabled via Register 0x1C3.
The receiver issues a synchronization request by asserting a low
signal on the SYNCINB± pins of the AD6676. The JESD Tx
begins to send /K/ characters. After the receiver has synchronized,
it then deasserts its SYNCINB signal, causing it to go high. The
AD6676 then transmits an ILAS on the following LMFC boundary.
For more information on the CGS phase, see the JEDEC
Standard No. 204B (July 2011), Section 5.3.3.1.
Rev. D | Page 47 of 90
AD6676
Data Sheet
For scrambled data, any 0xFC character at the end of a frame is
replaced by an /F/ and any 0x7C character at the end of a
multiframe is replaced with an /A/. The JESD204B receiver
checks for /F/ and /A/ characters in the received data stream
and verifies that they only occur in the expected locations. If an
unexpected /F/ or /A/ character is found, the receiver uses
dynamic realignment or asserts the SYNCINB± signal for more
than four frames to initiate a resynchronization. For
unscrambled data, if the final character of two subsequent
frames is equal, the second character is replaced with an /F/ if it
is at the end of a frame, and an /A/ if it is at the end of a
multiframe.
K K R D
D A R Q C
C D
Insertion of alignment characters may be modified using SPI.
The frame alignment character insertion is enabled by default.
More information on the link controls is available in the SPI
register descriptions for Register 0x1E0 to Register 0x1E6.
8-Bit/10-Bit Encoder
The 8-bit/10-bit encoder converts 8-bit octets into 10-bit
symbols and inserts control characters into the stream when
needed. The control characters used in JESD204B are shown in
Table 18. The 8-bit/10-bit encoding ensures that the signal is dc
balanced by using the same number of ones and zeroes across
multiple symbols. Note that the 8-bit/10-bit interface has an
invert option available in Register 0x1E4 that has the same
effect of swapping the differential output data pins.
D A R D
D A R D
D A D
START OF
ILAS
START OF LINK
CONFIGURATION DATA
START OF
USER DATA
12348-072
END OF
MULTIFRAME
Figure 121. Initial Lane Alignment Sequence
Table 18. Control Characters used in JESD204B Including Running Disparity Values
Abbreviation
/R/
/A/
/Q/
/K/
/F/
Control Symbol
K28.0
K28.3
K28.4
K28.5
K28.7
8-Bit Value
000 11100
011 11100
100 11100
101 11100
111 11100
10-Bit Value (RD = −1)
001111 0100
001111 0011
001111 0010
001111 1010
001111 1000
Rev. D | Page 48 of 90
10-Bit Value (RD = +1)
110000 1011
110000 1100
110000 1101
110000 0101
110000 0111
Description
Start of multiframe
Lane alignment
Start of link configuration data
Group synchronization
Frame alignment
Data Sheet
AD6676
Digital Outputs, Timing and Controls
The AD6676 physical layer consists of digital drivers that are
defined in the JEDEC Standard No. 204B (July 2011). These
CML drivers are powered up by default via Register 0x1E2. The
drivers utilize a dynamic 100 Ω internal termination to reduce
unwanted reflections. A 100 Ω differential termination resistor
at each receiver input results in a nominal 300 mV p-p swing at
the receiver.
The AD6676 JESD204B differential outputs can interface with
custom ASICs and FPGA receivers, providing superior switching
performance in noisy environments. Single point-to-point
network topologies are recommended with receiver inputs having
a nominal differential 100 Ω termination. The common mode
of the digital output automatically biases itself to half the VDDHSI
supply of 1.1 V (VCM = 0.55 V), thus making ac coupling the
preferred coupling method to the receiver logic as shown
Figure 122. DC coupling can be considered if the receiver device
shares the same VDDHSI supply and input common-mode range.
SERDOUTx–
–100
–300
EYE: ALL BITS, OFFSET: –0.0055
UIs: 4000; 1059998, TOTAL: 4000; 1059998
–400
–150
–100
–50
0
50
100
150
TIME (ps)
Figure 123. Digital Outputs Data Eye with External 100 Ω Terminations at
5.333 Gbps in Accordance to LV-OIF-11G-SR Mask
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
–5
–4
–3
–2
–1
0
1
2
3
4
5
TIME (ps)
Figure 124. Digital Outputs Histogram with External 100 Ω Terminations at
5.333 Gbps
1
1–2
1–4
RECEIVER
1–6
0.1µF
OUTPUT SWING = 300mV
0
VCM = VDDHSI/2
1–8
1–10
Figure 122. AC-Coupled Digital Output Termination Example
1–12
Timing errors caused by a degraded eye diagram at the receiver
input can often be attributed to poor far end termination or
differential trace routing. These potential error sources can be
reduced by using well controlled differential 100 Ω traces with
lengths below six inches that connect to receivers with integrated
differential 100 Ω resistors.
1–14
1–16
–0.5
–0.4
–0.3
–0.2
–0.1
0
UIs
0.1
0.2
0.3
0.4
0.5
12348-505
100Ω
100
–200
BER
SERDOUTx+
200
100Ω
DIFFERENTIAL
TRACE
PAIR
0.1µF
12348-073
VDDHSI
300
12348-503
The optional SYSREF± input can be used for multichip
synchronization or establishing a repeatable latency between
the AD6676 and its host. The SYSREF± receiver circuit must be
disabled if not used (Register 0x1E7 = 0x04) to prevent
potential false triggering if the input pins are left open. The
SYSREF± input does not include an internal 100 Ω termination
resistor; thus, an external differential termination resistor must
be included if this input is used. The SYSREF± input is logic
complaint to LVPECL, LVDS, and CMOS.
400
12348-504
The AD6676 physical layer consists of consists of two digital
differential inputs, SYSREF± and SYNCINB±, whose equivalent
input circuits are shown in Figure 61 and Figure 64. These inputs
must be dc-coupled to their respective drivers because they are
or can be aperiodic. The SYNCINB± input is logic compliant to
both CMOS and LVDS via Register 0x1E7, Bit 2, with CMOS
being the default. Note that the SYNCINB± input includes an
internal 100 Ω termination resistor when LVDS is selected.
VOLTAGE (mV)
Digital Inputs
Figure 123, Figure 124, and Figure 125 show examples of the
digital output data eye, time interval error (TIE) jitter
histogram, and bathtub curve for one AD6676 lane running at
5.333 Gbps with Register 0x1EC set to 0xBD. The format of the
output data is twos complement by default. The output data
format can be changed via Register 0x146.
HITS
PHYSICAL LAYER INPUT/OUTPUTS
Figure 125. Digital Outputs Data Bathtub with External 100 Ω Terminations
at 5.333 Gbps
Rev. D | Page 49 of 90
AD6676
Data Sheet
Preemphasis
CONFIGURING THE JESD204B LINK
Preemphasis enables the receiver eye diagram mask to be met in
conditions where the interconnect insertion loss does not meet
the JESD204B specification. The preemphasis feature is controlled
via Register 0x1EF and must be used only when the receiver
cannot recover the clock due to excessive insertion loss. Under
normal conditions, it is disabled to conserve power. Additionally,
enabling and setting too high a deemphasis value on a short link
may cause the receiver eye diagram to fail or lead to potential
EMI issues. For these reasons, consider the use of preemphasis
only in instances where meeting the receiver eye diagram mask
is a challenge. See the Register Memory Map section for details.
The AD6676 has one JESD204B link. The serial outputs
(SERDOUT0± and SERDOUT1±) are part of one JESD204B
link. The basic parameters that determine the link setup are:
Serializer PLL
This PLL generates the serializer clock that is equal to the
JESD204B lane rate. The on-chip controller automatically
configures the PLL parameters based on the user specified IQ
data rate (FADC/M) and number of lanes. The status of the PLL
lock can be checked via the PLL_LCK status bit in Register 0x2DC.
This read only bit lets the user know if the PLL has achieved a
lock for the specific setup.
L is the number of lanes per link
M is the number of converters per link
F is the number of octets per frame
The maximum and minimum specified lane rates for the
AD6676 are 5.333 Gbps and 3.072 Gbps, respectively. For this
reason, the AD6676 supports a single lane interface for IQ data
rates (fDATA_IQ) from 76.8 MSPS to 133.3 MSPS and a two lane
interface from 153.6 MSPS to 266.7 MSPS.
The lane line rate is related to the JESD204B parameters using
the following equation:
Lane Line Rate
40 F
DATA_IQ
(14)
L
where:
FDATA_IQ
FADC
DEC
The decimation ratio (DEC) is the parameter programmed into
Register 0x140.
Table 19 shows the JESD204B output configurations supported
based on fDATA_IQ.
Table 19. JESD204B Output Configurations
No. Virtual Converters
Supported (same as M)
2
fDATA_IQ (MSPS)
76.8 to 133.3
153.6 to 266.7
JESD Serial Line Rate
40 × fDATA_IQ
20 × fDATA_IQ
L
1
2
Rev. D | Page 50 of 90
M
2
2
F
4
2
S
1
1
HD
0
0
N
16
16
N'
16
16
K
For F = 4, K ≥ 5
For F = 2, K ≥ 9
Data Sheet
AD6676
SYNCHRONIZATION USING SYSREF±
The AD6676 uses the SYSREF± input to provide synchronization
for the JESD204B serial output and to establish a fixed phase
reference for the decimation filters and the NCO within the QDDC.
Synchronization options are configurable via Register 0x1E8.
When initially synchronizing, the absolute phase offset relative
to the input clock applied to the CLK± pins depends on internal
clock phases and therefore has an uncertainty of ±1 ADC clock
cycles.
A clock tree diagram is shown in Figure 126 with an internal
clock signal, DIG_CLK, used to ultimately sample the SYSREF±
signal. Note that the SYSREF± setup and hold times are defined
with respect to the rising SYSREF± edge and rising CLK± (or
CLK+ with the clock synthesizer disabled) edge, as shown in
Figure 2. After the SYSREF± signal is sampled, the phase remains
locked to the same relative internal ADC_CLK phase offset
until the AD6676 is intentionally reset or its clock or power
interrupted.
Note the following considerations when using SYSREF± for
synchronization.
The SYSREF± pulse width must be at least two ADC_CLK
periods.
Bit 3 of Register 0x2BB must be set low when synchronizing
with the clock synthesizer enabled. In this case, that SYSREF±
is sampled on the rising edge of REF_CLK to allow for
significant margin in setup and hold time. This
synchronization signal is then sampled again with the
internally generated DIG_CLK.
Because SYSREF± is ultimately sampled with an internal
clock greater than 1 GHz, it can be difficult to maintain
synchronization of the clock and SYSREF± distribution in
a system over supply and temperature variations, as well as
REG 0x2A5
RF_CLK
CLK±
ADC_CLK
CLOCK
SYNTHESIZER
fOUT =
5.9GHz TO 6.4GHz
TO Σ-∆ ADC
AND DIGITAL
÷2
÷2
DIG_CLK
TO DIGITAL
REF_CLK
Q
REG 0x2BB
SYSREF±
Q
D
Q
D
Q
TO SYNCHRONIZATION
CIRCUITRY
12348-131
cumulative jitter affects. Use the one shot with the second
SYSREF pulse to avoid unnecessary resetting of the
JES204B link by setting Register 0x1E8 to 0x06. A
minimum of two SYSREF pulses are required.
The coarse and fine digital NCOs can be reset to an initial
phase defined in Register 0x143 through Register 0x145
upon receiving SYSREF±. For the recommended one shot
with the second SYSREF pulse, set Register 0x1E8 to 0x26
so that the same SYSREF pulse that is used to reset the
JESD204 internal dividers is used to reset the NCO phases.
If continuous SYSREF± is still preferred, it is recommended to
use the SYSREF_WIN_NEG and SYSREF_WIN_POS bits
in Register 0x1EA to allow for slight variation in SYSREF±
timing relative to DIG_CLK.
A phase variance of ±1 ADC clock cycles ultimately results
in fractions of a sample when referenced to the IQ output
data rate, fDATA_IQ, depending on the decimation factor. For
example, for a decimation factor of 32, the phase uncertainty
is expressed as ±1/32 samples relative to fDATA_IQ.
The course and fine digital NCOs are also set to an initial
phase up defined in Register 0x143 thru Register 0x145
upon receiving SYSREF±.
Figure 127 shows how the HMC7044 (or the AD9528) can
be used for mulichip synchronization. The HMC7044 is best
suited for delivering a low phase noise RF clock source for
each AD6676 (refer to Figure 135). In addition, its ability
to individually control the delays of both the CLK and
SYSREF signals to each AD6676 device allows
compensation of PCB skew delays.
Figure 126. Block Diagram Showing Options of Sampling the SYSREF Input Signal with the Clock Synthesizer Disabled or Enabled
Rev. D | Page 51 of 90
AD6676
Data Sheet
HMC7044
(OR AD9528)
1nF
CLKOUT0
CLKOUT
AD6676
CLK+
1nF
CLK–
CLKOUT0
240Ω
240Ω
1nF
SCLKOUT1
SYSREF+
SYSREF
1nF
SCLKOUT1
SYSREF–
240Ω
240Ω
1nF
CLKOUT2
AD6676
CLK+
1nF
CLK–
CLKOUT2
240Ω
240Ω
1nF
SCLKOUT3
SYSREF+
1nF
240Ω
240Ω
12348-727
SYSREF–
SCLKOUT3
Figure 127. Example of Multichip Synchronization of the AD6676 Using the HMC7044 or the AD9528
Rev. D | Page 52 of 90
Data Sheet
AD6676
APPLICATIONS INFORMATION
ANALOG INPUT CONSIDERATIONS
Table 20. Harmonic Levels When Dual Tones = −6 dBFS of
PIN_0dBFS Level is Situated at IF/2
Equivalent Input Impedance and S11
10
61
9
60
8
59
7
58
6
57
5
SHUNT
SHUNT
SHUNT
SHUNT
56
55
54
R WITH ATTENUATOR = 0dB
R WITH ATTENUATOR = 6dB
C WITH ATTENUATOR = 0dB
C WITH ATTENUATOR = 6dB
4
3
2
53
52
IF
(MHz)
200
250
300
350
400
LEXT
(nH)
43
19
19
10
10
PIN_0dBFS
(dBm)
−2.5
−2.2
−2.2
−2.2
−2.2
Dual Tone
Input Power
(dBm)
−8.5
−8.2
−8.2
−8.2
−8.2
f1 + f2
Spur
(dBc)
−69.5
−68.3
−73
−66.3
−68.5
Equivalent
IP2 (dBm)
61
60
65
58.5
60
Above the IF pass band, the AD6676 is sensitive to high frequency
blockers that can increase the noise floor due to jitter or generate an
image component that falls back into the pass band. The AD6676 is
also fairly insensitive to spurious tones falling in the alias regions
occurring at FADC ± FIF because the AD6676 provides over 50 dB
of alias rejection. Table 21 shows the typical alias rejection for
different FADC and IF combinations. Because mixers often produce
fixed large spurious at M × LO as well as its sum term of LO +
FRF, determine if any of these spurs can fall in the alias regions and
if so, add the appropriate level of filtering to suppress them below
the receivers required spurious level.
SHUNT C (pF)
62
1
0
100
200
300
400
500
600
700
FREQUENCY (MHz)
800
900
0
1000
12348-703
SHUNT R (Ω)
The AD6676 benign input structure along with its low drive
level requirements facilitates interfacing it to external driver
circuitry. Figure 128 shows the equivalent parallel impedance
for attenuator settings of 0 dB and 6 dB. Note that the slight
variation in impedance between the different attenuator settings
is an error source affecting the absolute accuracy of the attenuator
settings. The AD6676 input also displays excellent S11 return
loss over a wide frequency range, as shown in Figure 94.
Figure 128. Typical Equivalent Parallel Impedance of AIN for Attenuator = 0
and 6 dB Settings
Input Driver and Filter Considerations
The input driver requirements, along with any additional
filtering, are application dependent. Additional filtering maybe
considered if any large signal content or blockers falling above
or below the IF pass band of interest can cause desensitization
by either increasing the ADC noise or spur floor. Below the IF
pass band, the AD6676 is most sensitive to second harmonic
content that is typically induced by the driver stage itself due
to its limited IP2 performance. The AD6676 second-order
nonlinearity contribution is typically on par with a balanced
mixer and well below the contribution of a single-ended amplifier
stage (with output balun) used for VHF applications. Table 20
shows the measured f1 + f2 spurious level and equivalent IIP2
for different IFs when dual tones are injected at −6 dBFS levels
and at IF/2.
Table 21. Typical Alias Rejection for Different IF and ADC
Combinations
FADC (MHz)
2000
2400
2800
3200
IF (MHz)
150
200
300
400
FADC − IF Alias
Rejection (dBc)
58
53
51
51
FADC + IF Alias
Rejection (dBc)
59
54
59
59
Because the required attenuation of out-of-band signal signals is
application dependent, evaluate the AD6676 under the desired
application conditions to understand the effects and determine
what amount of filtering is required. In practice, a simple thirdorder low-pass roofing filter can provide adequate additional
suppression against spurs falling in the alias regions as well as
large signal signals falling a few 100 MHz above the IF pass
band. Note that the AD6676EBZ includes an optional 500 MHz
third-order low-pass filter (TDK MEA1210D501R) that may
suffice for many applications. This small, 0302 size differential
filter is also available with lower frequency options. Its effect on
the pass band flatness is mimimal but provides provides additional
suppression beyond 700 MHz. as shown in Figure 129 as well as
in the alias region as shown in Table 22.
Rev. D | Page 53 of 90
AD6676
Data Sheet
harmonics that are sensitive to balance fall outside the pass band.
Note that the second harmonic of the gain block still must fall
outside the VHF pass band so that it can also be digitally filtered.
0
–5
Some additional considerations pertaining to the analog input
are as follows:
–10
–15
WITHOUT
TDK LPF
–20
–25
WITH TDK LPF
–30
–35
0
100
200
300
400
500
600
700
800
900
1000
FREQUENCY (MHz)
12348-704
NORMALIZED STF RESPONSE (dBFS)
5
Figure 129. TDK Filter Has Minimal Effect on the Pass Band IF Response
Table 22. Typical Alias Rejection for Different IF and ADC
Combinations with TDK 500 MHz Low-Pass Filter Added
FADC (MHz)
2000
2400
2800
3200
IF (MHz)
150
200
300
400
FADC − IF Alias
Rejection (dBc)
82
77
71
74
FADC + IF Alias
Rejection (dBc)
83
85
83
81
A 1:1 balun is required in applications where the last amplification
stage is single-ended with a ZOUT of 50 Ω. This is typically the
case in a VHF receiver application where a gain block, such as
the ADL5541 to ADL5545 series, precedes the AD6676 for
preamplification.
ADL5541
TO
ADL5545
DEVICES
VHF
SIGNAL
1:1 BALUN
(MABA-007159)
1nF
AD6676
VIN+
12348-077
VIN–
1nF
Figure 130. RF Line-Up for Direct Sampling VHF Application
For many RF receiver applications, this differential signal may
originate from a RF-to-IF mixer whose output impedance often
falls within a 50 Ω to 200 Ω range. A low order matching network
that also serves as a low-pass roofing filter can compensate for
the mismatch impedance. It is worth noting that the impedance
mismatch between a source/load mismatch of 200 Ω/60 Ω and
100 Ω/60 Ω is approximately 1.5 dB and 0.3 dB, respectively. This
low mismatch loss may be tolerable for some applications
seeking a wide, low ripple IF pass band, especially considering
the loss of a higher order matching network with finite Q
components. Lastly, it is possible to reduce the ADC maximum
input power requirements slightly to compensate for this low
loss with minimal loss in dynamic range.
Other receiver applications in the VHF band may prefer that
the AD6676 directly digitize the signal. Typically, the radio lineup
may include a low NF gain block whose single-ended output is
converted to a differential output via an ac-coupled balun. The
amplitude/phase balance requirements of balun can be relaxed
(compared to traditional pipeline ADCs) because the even order
AC coupling with 10 nF or greater capacitors to the VIN±
input is required to a maintain 1 V common-mode voltage.
Note that this capacitor provides a high-pass response with
the AD6676 input impedance and thus must be sized
accordingly for low IF applications to prevent excessive
droop on the lower pass band response.
A series 10 Ω resistor and 0.1 μF decoupling capacitor is
recommended between the 2.5 V supply and first resonators
to provide additional filtering of supply induced noise and
ADC common-mode currents.
The feedback DAC (operating up to 3.2 GHz) also
generates high frequency content (that is, images, clock
feedthrough and shaped noise) that is ideally absorbed by
the internal source follower. Due to its finite impedance at
the higher frequencies, a small amount of this undesired
signal content leaks through the attenuator path back to
the VIN± input. Passive mixers are particularly susceptible
to this signal content due to poor isolation between the IF
and RF ports while passive mixers with on-chip IF amps
and active mixers provide a greater degree of reverse
isolation. A simple third-order roofing filter typically
provides sufficient rejection to suppress these ADC
artifacts while also suppressing the larger M × N artifacts
of the mixer. Note that this filter must be designed as two
single-ended, pi network filters with shunt capacitors
located next to the VIN± pins to steer this undesired signal
content to ground. Also, use care in component selection
and layout to reduce parasitics that can cause unanticipated
peaking in the stop-band region of the filter response.
CLOCK INPUT CONSIDERATIONS
The AD6676 Σ-Δ ADC operates with an internal ADC clock
rate (FADC) between 2.0 GSPS to 3.2 GSPS. The clock signal can
originate from an external clock source or, alternatively, from its
on-chip clock synthesizer. Consider an external clock source if
the on-chip synthesizer phase noise or spurious level is not
deemed sufficient or if the desired FADC falls below the 2.94 GHz
to 3.2 GHz range of the VCO. Referring to Figure 60, the selfbiased clock receiver is configured as either a differential or singleended receiver, depending on whether the clock synthesizer is
disabled. In either case, the external clock source must be ac
coupled to the AD6676 CLK± input and meet the minimum
specified input level and slew rate. Also, clock jitter and phase
noise must always be a concern in selecting the clock source.
When the clock synthesizer is enabled, the CLK± inputs are
connected to CMOS inverters as shown in Figure 60. These
inverters are self-biased at approximately 0.55 V and present an
input resistance exceeding 1.2 kΩ when Bit 2 of Register 0x2BB
is set.
Rev. D | Page 54 of 90
Data Sheet
AD6676
When the clock synthesizer is disabled, the CLK± inputs are
connected to a high speed differential clock receiver with on-chip
100 Ω termination to simplify interfacing to CML, LVPECL, or
sinusoidal clock sources. The clock signal is typically ac-coupled to
the CLK+ and CLK− pins via an RF balun or capacitors. These
pins are biased internally (see Figure 60) at approximately 700 mV
and require no external bias. The equivalent shunt impedance of
the CLK± input is shown in Figure 131. It is recommended to
use a 100 Ω differential transmission line to route the clock signal
to the CLK+ and CLK− pins due to the high frequency nature
of the signal.
130
INSTALL 100Ω
DIFFERENTIAL RESISTOR
ONLY WHEN CLK SYN
IS ENABLED
HMC7044,
AD9528 OR
ADCLK925
CLK+
10nF
PECL
DRIVER
AD6676
10nF
CLK–
240Ω
240Ω
1.0
110
0.6
100
0.4
90
0.2
80
0
70
Figure 133. Differential PECL Sample Clock Using the HMC7044, AD9528, and
ADCLK925
0.8
CAPACITANCE (pF)
REAL SHUNT
CAPACITANCE SHUNT
120
–0.2
60
50
2.0
2.5
3.0
–0.4
4.0
3.5
FREQUENCY (GHz)
12348-705
REAL SHUNT (Ω)
A single-ended CMOS or differential ac-coupled PECL/HSTL
clock signal can be delivered via clock generation and distribution
ICs such as the Analog Devices HMC7044, AD9528, and
ADCLK925. A PECL clock signal is recommended when providing
an RF clock input signal to the AD6676 or in applications that
require deterministic latency or synchronization while using the
internal clock synthesizer of the AD6676. Figure 133 shows a
simple differential interface in which the AD6676 interfaces to
the PECL output available from these ICs. The HMC7044 is an
excellent choice for JESD204B clock generation and multichip
synchronization because it also generates a very low phase noise
RF clock from 2.4 GHz to 3.2 GHz for multiple AD6676 devices.
12348-081
A single-ended clock source need only be ac coupled to the CLK+
input because the inverter output for CLK− input is not used.
For CMOS drivers, the addition of a 33 Ω series resistor is
recommended to dampen the response for long trace lengths.
For a differential clock source, such as an LVDS or PECL source,
the addition of a 100 Ω external termination resistor across the
CLK± pins is recommended to minimize any reflections that
result from distorting the clock input waveform.
Alternatively, PLL clock synthesizers with on-chip VCOs such
as the ADF4351, the ADF4355-2, and HMC1034 also make
excellent RF clock sources when multichip synchronization is
not required. The CML outputs of these devices allow a simple
interface as shown in Figure 134. Figure 135 compares the close
in phase noise between the ADF4351, the ADF4355-2, the
HMC7044, the AD6676 clock synthesizer, and the R&S SMA100A
for a near full-scale sine wave at 300 MHz. Note that the phase
noise improvement offered by the high quality RF generator only
becomes evident below 400 kHz when compared to the ADF4351.
Figure 131. Equivalent Shunt Differential Input Impedance of the CLK± Pins
with the Clock Synthesizer Disabled
3.9nH
PLL
FREF
VCO
REFOUTA+
AD6676
CLK+
CLK–
REFOUTA–
ADF4351
ADF4355-2
1nF
REFOUTB+
REFOUTB–
1nF
0.8V p-p
12348-079
VVCO
DIV-BY-2N
Figure 132 shows a single-ended clock solution for the AD6676
when its clock synthesizer is disabled. The low phase noise singleended source can be from an external VCXO. A ceramic RF chip
1:2 ratio balun creates the differential clock input signal. The
balun must be specified to have low loss (that is, less than 2 dB)
at the clock frequency of interest. The single-ended clock source
must be capable of 0 dBm drive capability to ensure adequate
signal swing into the clock input.
Figure 134. Differential CML Driver from the ADF4351 and the ADF4355-2
100pF
100pF
AD6676
CLK+
CLK–
JT 4000BL14100 100pF
12348-078
CLOCK
INPUT
Figure 132. Balun Coupled Differential Clock
Rev. D | Page 55 of 90
AD6676
Data Sheet
–110
Figure 136 shows a normalized image graph (relative to fDATA_IQ)
showing the image location relative for a given input frequency.
ADF4351
When N > 1, spurious content is often at lower magnitude than
other spurious thus often can be ignored. The exception is when
fIN falls below the IF pass band such that its lower order harmonics
may fall within the pass band (that is, IF/2 and IF/3).
–120
–125
AD6676 CLK SYN
–130
–135
ADF4355-2
R&S SMA100A
4.5
M=3
–140
M=2
M=1
4.0
HMC7044
0.1
FREQUENCY OFFSET (MHz)
1
Figure 135. Close In Phase Noise Comparison for Different Analog Devices
Clock Sources when Compared to the R&S SMA100A and the AD6676 Clock
Synthesizer (IF = 300 MHz, BW = 40 MHz, FADC = 3.2 GHz, L = 19 nH)
IF FREQUENCY PLANNING
3.0
M=7
2.5
2.0
M=6
1.5
1.0
The Σ-Δ ADC can achieve exceptional SFDR performance over
a wide IF frequency range because its high oversampling ratio
prevents low order harmonics from aliasing into the IF pass band.
Higher order harmonics that do alias back are typically of much
lower magnitude, with the shuffling option further reducing
their levels. However, finite isolation between the Σ-Δ ADC and
the digital block causes additional spurious signals that are a
function of the output data rate, fDATA_IQ, and input frequency,
fIN. Specifically, the feedback DACs in the Σ-Δ ADC suffer from
digital contamination of its clock signal. Therefore, the same
equation used to predict spurious locations on high speed DACs
with digital interpolation filters applies.
Equation 15 defines this relationship with the spur location
falling at fMN.
fMN = ±(M × fDATA_IQ) ± (N × fIN)
M=8
3.5
(15)
where:
M is the digital induced harmonic content from internal clocks.
N is the harmonics from the Σ-Δ ADC.
When N = 0, signal independent spurs fall at integer multiples
of fDATA_IQ. Table 23 shows the measured M × fDATA_IQ spurious
levels (dBFS) for different IF frequencies and decimation factors
with fDATA_IQ equal to 100 MSPS and 200 MSPS. All of the M ×
fDATA_IQ regions display low spurious with the exception of
200 MHz. This is because a large portion of digital circuitry is
clocked at FADC/16 for DEC_MODES of 1 and 3 or FADC/12 for
DEC_MODES of 2 and 4. As a result, the M = 2 spur is dominant
when operating at the higher decimation factors of 32 and 24
whereas the M = 1 spur is dominant when operating at the
lower decimation factors of 16 and 12.
When N = 1, signal dependent spurs falls at integer multiples of
fDATA_IQ. These M × N spurs are called images because they have
a 1:1 relationship in amplitude and frequency with the input signal,
fIN. Note that the magnitude of some images can also vary slightly
between power cycles, due to different phase relationships among
internal clock dividers upon device initialization.
M=2
0.5
0.5
1.0
M=3
1.5
2.0
M=4
2.5
3.0
M=5
3.5
4.0
4.5
NORMALIZED INPUT FREQUENCY
12348-082
–150
–155
0.01
NORMALIZED IMAGE LOCATION
–145
12348-080
PHASE NOISE @ IF = 300MHz (dBc/Hz)
AVG = 300
–115
Figure 136. Image Location for Different M Factors Normalized to fDATA_IQ
Table 23. Measured Spurious Levels at Different IFs Where
M × fDATA_IQ Falls On for fDATA_IQ of 100 MSPS and 200 MSPS
fDATA_IQ
100 MSPS
DEC_MODE = 1
DEC_MODE = 2
200 MSPS
DEC_MODE = 3
DEC_MODE = 4
1
IF =
100 MHz
Spurious Levels (dBFS)
IF =
IF =
IF =
200 MHz 300 MHz 400 MHz