Application Note: 7 Series FPGAs and Zynq-7000 AP SoCs

LVDS 4x Asynchronous Oversampling Using

7 Series FPGAs and Zynq-7000 AP SoCs
Author: Marc Defossez
XAPP523 (v1.1) May 17, 2017

Summary This application note describes a method of capturing asynchronous communication using
LVDS with SelectIO™ interface primitives. The method consists of oversampling the data with
a clock of similar frequency (±100 ppm). This oversampling technique involves taking multiple
samples of the data at different clock phases to get a sample of the data at the most ideal point.
The SelectIO interface in 7 series FPGAs and Zynq®-7000 All Programmable SoCs can
perform 4x asynchronous oversampling at 1.25 Gb/s. Oversampling is performed by using
ISERDESE2 primitives. Clocks are generated from a mixed-mode clock manager
(MMCME2_ADV) through dedicated high-performance paths between the components.

Introduction Synchronizing the clock and data is the most common method of achieving communication
between devices using low-voltage differential signaling (LVDS). This means that the clock is
transmitted on one differential channel and the data on one or several other differential pairs. At
the receiver, the clock (after synchronization) is used to capture the data. This is known as
source-synchronous communication.
When transmitting data without a separate accompanying clock signal, the clock used to
capture the data must be recovered at the receiver side from the incoming data stream. This is
called asynchronous communication, also known as data and/or clock recovery. Xilinx®
transceivers use this principle. Data recovery allows a receiver to extract data from the
incoming clock/data stream and then move the data into a new clock domain. Sometimes, the
recovered clock is used for onward data treatment or transmission.
The circuit described in this application note provides a “partial solution” in that no clock is
actually recovered, but the arriving data is fully extracted. Figure 1 shows a typical use case.

© Copyright 2012–2017 Xilinx, Inc. Xilinx, the Xilinx logo, Artix, ISE, Kintex, Spartan, Virtex, Zynq, and other designated brands included herein are trademarks of Xilinx in the
United States and other countries. All other trademarks are the property of their respective owners.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 1

Asynchronous Oversampling

X-Ref Target - Figure 1

I/O Bank CMT FPGA Logic

1.25 Gb/s Link Data

Data Recovery
Capture

1.25 Gb/s Link Data CLK90 Data Recovery

Capture
CLK

Clock System Clock (125 MHz)

Adjust

1.25 Gb/s Link Data

Data Recovery
Capture

MMCM

1.25 Gb/s Link Data

Data Recovery
Capture

X523_01_012012

Figure 1: Typical Data Recovery Application

Asynchronous For signal processing, “oversampling” means sampling a signal using a sampling frequency
Oversampling significantly higher than twice the bandwidth (or highest frequency) of the signal being
sampled. For the communication interface described in this application note, the “significantly
higher” sampling frequency is obtained using different edges of multiple phase-shifted clocks. It
is called asynchronous oversampling because the clocks used to create the sampling
frequency are nominally equal to the data stream frequency.
The circuit discussed here uses a clock (local oscillator) running at the same nominal frequency
as the data stream being captured. “Nominal” here means that the local oscillator is either
slightly faster or slightly slower than the incoming clock/data stream.
Through the use of a clock manager (MMCME2), high-speed phase-shifted clocks are
generated from a slow system clock typically provided by a local clock oscillator (see Figure 2).

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 2

Asynchronous Oversampling

X-Ref Target - Figure 2

OVERSAMPLE Mode
BUFIO CLK

To All ISERDES in
BUFIO 125 MHz
CLK90

BUFG IntClk MMCM

BUFG IntClkDiv

To All Clocked FPGA

Pattern Logic in this Clock Area

OSERDES
To Data Recovery
Logic in this Clock Area
ISERDES

State
Machine

X523_02_041817

Figure 2: Clock Generation Using MMCME2

The function of the two extra clocks and ISERDES/OSERDES combination shown in Figure 2
is explained in Clocking and Data Flow, page 10. The generated CLK and CLK90 clocks make
it possible to oversample an incoming data stream on four edges, meaning that each bit of the
DDR data stream can be sampled twice, as shown in Figure 3.
X-Ref Target - Figure 3

0 90 180 270

CLK

CLK90

Different
positions of
the data with
respect to the
clocks.

Two Data Sample Edges Per Bit X523_03_012012

Figure 3: Data Oversampling on Four Clock Edges

If the incoming data stream is split into two branches and one branch is delayed by 45°, it is
possible to 4x oversample every data bit. The details of how this circuit is constructed using
MMCME2, IODELAYE2, and ISERDESE2 is shown in Figure 4.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 3

Asynchronous Oversampling

X-Ref Target - Figure 4

BUFIO CLK
CLKIN
BUFIO CLK90

BUFG IntClk MMCM

BUFG IntClkDiv

ISERDES
IDELAY
IBUFDS_DIFF_OUT
0 Shift

Data Recovery Unit

IDELAY

ISERDES
45 Shift

X523_04_012512

Figure 4: MMCM Phase Clock and Phased Data Generation

The MMCME2 generates two clock phases (CLK0 and CLK90). These are routed to the
ISERDESE2, where both the positive-transitioning and negative-transitioning edges of the two
clocks are used, creating four clock phases. Two copies of the incoming data are created by
means of an IBUFDS_DIFFOUT. One branch of the data gets a shift of 45° and the other
branch gets no phase shift. The phase shift is obtained by passing the data of both branches
through IODELAYE2. This phase-shifted version of the data is passed into a slave
ISERDESE2, effectively doubling the sample rate.
Eight clock sample phases for bit oversampling are thus created by using a combination of four
clock phases and two data sample phases, as shown in Figure 5.
X-Ref Target - Figure 5

0 90 180 270

CLK

CLK90

DATA

DATA

45 deg

DATA
135
45
90
0

X523_05_012012

Figure 5: Sample Edges

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 4

7 Series ISERDESE2 Oversampling Mode

IDELAY Tap Setting Calculation Example

The following list outlines the logical flow of timing assumptions and calculations leading to the
IDELAY tap settings:
1. Assume that the incoming data stream runs at 1.25 Gb/s; the bit time is therefore 800 ps.
2. The CLK and CLK90 clocks therefore run at 625 MHz, or 1.6 ns.
3. The edges of both clocks arriving at 0°, 90°, 180°, and 270° have positions at 0, 400, 800,
and 1200 ps.
4. To shift 45°, one branch of the data must be delayed by 200 ps.
5. The tap delay of an IDELAY component is controlled by an IDELAYCTRL component. The
IDELAYCTRL component in this design is clocked at 310 MHz, so a single-tap delay is
52 ps. (Refer to Note 1 in the Input/Output Delay Switching Characteristics table in DS182,
Kintex-7 FPGAs Data Sheet: DC and Switching Characteristics.)
6. The 200 ps desired delay for 45° phase shift is divided by 52 ps per tap to give 3.8 or
4 taps.
Therefore, the IDELAY_VALUE of the first IDELAY must be set to 0, and the IDELAY_VALUE of
the second (slave) IDELAY must be set to 4.

7 Series The ISERDESE2 component in 7 series FPGAs is an improved version of similar components
ISERDESE2 in previous FPGA families (ISERDES in Virtex-5 FPGAs and ISERDESE1 in Virtex-6 FPGAs).
The ISERDESE2 component can implement (i.e., be configured as) different functions:
Oversampling
• In its most basic function, the ISERDESE2 provides the functionality of an IDDR flip-flop.
Mode
• A second and more complex function is that of a dedicated serial-to-parallel converter with
specific clocking and logic features designed to facilitate the implementation of high-speed
source-synchronous applications (NETWORKING mode).
• A third function is MEMORY mode, where the ISERDES is configured as a dedicated
interface for different types of memories (QDR, DDR3, etc.).
• In its fourth and final function, the ISERDESE2 can be used in OVERSAMPLING mode.
Here, ISERDESE2 is used to capture two phases of DDR data. In this mode, the
ISERDESE2 is thus used as a dual set of IDDR flip-flops.
For detailed descriptions of the ISERDESE2 functionality, see UG471, 7 Series FPGAs
SelectIO Resources User Guide.
For convenience, Figure 6 shows the ISERDESE2, with oversampling mode configuration. In
earlier implementations, the oversampling design was implemented in FPGA logic using SLICE
flip-flops. With 7 series FPGAs, this functionality is implemented in the ISERDESE2.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 5

7 Series ISERDESE2 Oversampling Mode

X-Ref Target - Figure 6

SHIFTIN1 SHIFTOUT1
SHIFTIN2 SHIFTOUT2
INTERFACE_TYPE : string := “OVERSAMPLE”;
SERDES_MODE : string := “MASTER”; OFB
DATA_WIDTH : interger := 4; D O
DATA_RATE : string := “DDR”; DDLY Q1
OFB_USED : string := “FALSE”; CE1 Q2
IOBDELAY : string := “IFD”;
NUM_CE : integer := 1; CE2 Q3
DYN_CLKDIV_INV_EN : string := “FALSE”; RST Q4 Q1
DDLY
DYN_CLK_INV_EN : string := “FALSE”; BITSLIP Q5 D Q D Q D Q
INIT_Q1 : bit := ‘0’;
CLK Q6 CLK
INIT_Q2 : bit := ‘0’;
INIT_Q3 : bit := ‘0’; CLKB Q7
INIT_Q4 : bit := ‘0’; CLKDIV Q8 Q2
SRVAL_Q1 : bit := ‘0’; CLKDIVP D Q D Q D Q
SRVAL_Q2 : bit := ‘0’;
DYNCLKDIVSEL CLKB
SRVAL_Q3 : bit := ‘0’;
SRVAL_Q4 : bit := ‘0’; DYNCLKSEL
OCLK Q3
OCLKB ISERDESE2 D Q D Q D Q

OCLK

Q4
D Q D Q D Q

OCLKB

X523_15_021012

Figure 6: ISERDESE2 in OVERSAMPLING Mode Configuration

Data Recovery Unit

An ISERDESE2 used in networking mode requires a high-speed sampling clock (CLK) to
capture the serial data stream in the ISERDESE2. It also needs a low-speed function of CLK
(CLKDIV) to present the captured data in parallel format at the outputs of the ISERDESE2. The
internal circuits in the ISERDESE2 ensure that this conversion from the CLK level to the
CLKDIV level works as a clock-domain crossing (CDC) circuit.
The output of the ISERDESE2 components in oversampling mode is generated from the
high-speed sampling clocks (CLK/CLKB and OCLK/OCLKB). These clocks are only available
for clocking of ISERDESE2 and/or OSERDESE2.
The CDC operation must thus be implemented in registers in the FPGA logic. The details of
how this is done are discussed in Clocking and Data Flow, page 10. The CDC registers and
some comparing logic are implemented in the data recovery unit (DRU) and are clocked by the
CLK input. The low-speed clock (CLKP or CLKDIV) clocks the rest of the DRU.
The locations of the sampling and comparison points relative to the data stream coming into the
FPGA are shown in Figure 7. There are two streams of data, one offset from the other by
200 ps (4 IDELAY taps). For this application, the speed of the arriving data stream is1.25 Gb/s.
Through use of the IBUFDS_DIFFOUT primitive, one data stream is the complement of the
other data stream, similar to the input data of IBUFDS_DIFF_OUT (i.e., differential signaling).

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 6

7 Series ISERDESE2 Oversampling Mode

X-Ref Target - Figure 7

CLK0 CLK90 CLK180 CLK270 CLK0 CLK90 CLK180 CLK270

400 ps 400 ps

Q1 M0 Q3 M0 Q2 M0 Q4 M0 Q1 M1 Q3 M1 Q2 M1 Q4 M1
Master Data (Tap 0,
Delay = 0 ps) R0 F0 R1 F1

E4[3] E4[0] E4[1] E4[2] E4[3] E4[0] E4[2]

Slave Data (Tap 4, Q1 S0 Q3 S0 Q2 S0 Q4 S0 Q1 S1 Q3 S1 Q2 S1 Q4 S1

~R0 ~F0 ~R1 ~F1
Delay = 200 ps)

200 ps
X523_07_032112

Figure 7: Data Stream Sample and Comparison Points

The data is sampled through four clock phases 400 ps or 90° apart and named CLK0, CLK90,
CLK180, and CLK270, as shown in Figure 3, page 3. Sampling points occur where the clocks
intersect the data streams. These points are named according to the format:
Qx [M or S]x
Where:
Qx = the ISERDESE2 outputs Q1, Q2, Q3, or Q4
Mx or Sx = the source ISERDESE2 (M = master, S = slave) of the data outputs (Qx)
For example, sample point Q1M1 shows where CLK0 samples the data and creates an output
at port Q1 of the master ISERDESE2.
The lines labeled E4[0] through E4[3] that connect the sample points show where the DRU is
comparing data and looking for a data edge. The formulas for the four comparisons are shown
in Equation 1 through Equation 4.
E4[0] = [Q1M1 xor Q1S1] or [Q2M1 xor Q2S1] Equation 1

E4[1] = [Q3M1 xor Q1S1] or [Q4M1 xor Q2S1] Equation 2

E4[2] = [Q2M1 xor Q3S1] or [Q4M1 xor Q4S1] Equation 3

E4[3] = [Q1M1 xor Q4S0] or [Q2M1 xor Q3S1] Equation 4

These comparison points, relative to the original data stream, are actually 200 ps apart. For
example, Equation 1 (E4[0]) xor-compares Q1M1 against Q1S1 and Q2M1 against Q2S1.
These comparisons are shown by two gray dashed lines each labeled E4[0].
Referring to Figure 7 and looking first at the Q1M1 xor Q1S1 comparison, it can be seen that
both points are sampled by CLK0. However, the Q1S1 sample is delayed 200 ps (by the action
of IDELAYE2) relative to Q1M1, thus allowing comparison of two samples 200 ps apart.
Similarly, Q2M1 and Q2S1 are both sampled by CLK180, but again, the sample points are
separated by 200 ps due to the action of IDELAYE2 on the slave data stream.
If either the CLK0 or CLK180 sample points produces an xor result of 1—that is, the levels of
the sampled data do not match—it can be concluded that there is an edge (a level transition)
between those two sample points. The first E4[0] sample point comparison occurs in
rising-edge zones R1 and R1, while the second E4[0] sample point occurs in falling-edge zones
F1 and F1. Thus, both comparisons would match, and the xor outputs for these tests would
both be 0. The DRU state machine would know that there are no data transition edges there.
A contrasting example is shown by Equation 4, which xor-compares Q1M1 against Q4S0 and
Q2M1 against Q3S1. Q1M1 is sampled by CLK0 from the master data stream, while Q4S0 is
sampled by CLK270 from the slave (phase-delayed) data stream and is then stored for an extra
cycle in the DRU. CLK270 and CLK0 are 400 ps (90°) apart, but because the slave data is
delayed by 200 ps, the Q1M1 and Q4S0 sample points are really only 200 ps apart, again
relative to the original data stream. Similarly, Q2M1 is sampled by CLK180, and Q3S1 is

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 7

7 Series ISERDESE2 Oversampling Mode

sampled by CLK90. Again, the sample points are 200 ps apart relative to the original data
stream. For each comparison, one sample point falls in a rising-edge zone and the other falls in
a falling-edge zone. These two comparisons would produce an xor result of 1, indicating that an
edge (level transition) exists somewhere between the two sample points in each comparison.
Figure 8 shows what Equation 1 through Equation 4 look like in logic and how the data flows
out of the ISERDESE2 and into that logic. A stage of registers between the ISERDESE2 and
the logic facilitates the timing. This also shows how the Q4 output of the slave ISERDESE2 is
stored from the previous sample set to be compared with the new sample set.
X-Ref Target - Figure 8

Data Capture DRU

Q(1) II(1)
Q1
E4(0)
Q(5) II(5)
Q2
Master
ISERDESE2 Q3 Q(3) II(3)

Q(7) II(7)
Q4 E4(1)

Q(0) II(0)
Q1

Q(4) II(4) E4(2)

Q2
Slave
ISERDESE2 Q(2) II(2)
Q3

Q(6) II(6)
Q4
E4(3)

BUFG
625 MHz
X523_08_0203012

Figure 8: Edge Detection Circuit

At this point, it should be clear how the data comes into the FPGA and is then fed into the DRU
for edge detection. The next step in the DRU is to process the comparison data. This simple
state machine, based upon where the data edge was and where it moves to, then chooses a
sample point away from the data edge.
The ideal sample point can be expected to move around because of voltage and temperature
variations, jitter, and offset between the source and receiver clocks. This means that the
comparison point equations are always changing value, and the state machine is always
updating based on these changing results. Figure 9 and Table 1 describe the flow of the state
machine from one set of data to the next.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 8

7 Series ISERDESE2 Oversampling Mode

X-Ref Target - Figure 9

00
E4(2) E4(3)

E4(0) E4(1)
10 01
E4(3) E4(2)

E4(1) E4(0)
11
X523_09_020312

Figure 9: Data Select with Edge Information

Table 1: Data Select Map

EQ DO
00 Use samples Q(0) and Q(4)
01 Use samples Q(1) and Q(5)
11 Use samples Q(2) and Q(6)
10 Use samples Q(3) and Q(7)

In Table 1, the EQ column shows the current position of the state machine with input from
Equation 1 through Equation 4. The DO column shows what sample set is used in the
interconnect logic. Remembering that each ISERDESE1 in oversampling mode acts as two
sets of IDDR flip-flops, DO indicates which IDDR flip-flops should be used as the ideal sample
points.
Figure 9 shows, for each given state or sample set, where the state machine would go next. For
example, assume the state machine starts in state 01, which uses the Q(1) and Q(5) signals.
This maps to Q1 and Q2 out of the ISERDESE1 master, which would be CLK0 and CLK180,
respectively.
Then, if the data edge were to move to the left, the center point would be shifted from
CLK0/CLK180 to CLK90/CLK270. This would be seen by E4(3) changing its value from 0 to 1.
When this happens, the state machine moves from state 00 to state 01.

Bit Skip
When an edge moves to the left of the first data bit sample, bit skip occurs. Bit skip also occurs
when an edge moves to the right of the last data bit sample.
When an edge is detected to the left of the last sample, the new current sample is moved from
the last sample to the right corresponding to the first sample of the next data. As shown in
Table 1, when the state machine is in state 10, it samples Q(3) and Q(7). The state machine
then changes to state 00, sampling Q(0) and Q(4). However, a sample of data was already
taken in the previous state when the state machine was 10. As a result, during the first 00 state
of the state machine, one bit of the sampled bits is dropped. This is called a negative bit skip.
Negative bit skip outputs five bits per clock.
When the edge is detected to the right of the first sample, the new current sample is moved to
the left, which corresponds to the last sample of the next data. As shown in Table 1, when the
state machine is at 00, it samples Q(0) and Q(4). The state machine then changes to state 10.
In this state, it samples Q(3) and Q(7). However, no sample of data was taken during states 00

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 9

7 Series ISERDESE2 Oversampling Mode

and 10. As a result, during state 10 of the state machine, the last sample is taken along with
current samples, causing seven bits to be output. This is called a positive bit skip. Positive bit
skip outputs seven bits per clock.
From Figure 9, therefore, it can be seen that bit skip occurs when there is a transition between
states 00 and 10. When there are no bit-skip conditions, the sampled data output has one bit
per clock in the SDR mode and two bits per clock in the DDR mode.
Therefore, for six bits of parallel data:
• The number of bits for a negative bit skip condition is five.
• The number of bits for a positive bit skip condition is seven.
• The number of bits for no bit skip condition is six.

Clocking and Data Flow

The data flow from the pins of the FPGA to the 10-bit-wide interface presented to the FPGA
interconnect logic takes several steps, all under the control of several clock sources. The clock
sources are generated through an MMCME2 component. The setup of this design is shown in
Figure 10. The numbers above each section represent clocking transfer points discussed in the
numbered sections following Figure 10.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 10

7 Series ISERDESE2 Oversampling Mode

X-Ref Target - Figure 10

2A 2B

ISERDES

Register
IBUFDS_DIFF_OUT
DRU Logic

IntClkDiv
ISERDES

DRU

BUFG ClkRef
IDELAYCTRL
BUFIO CLK

BUFIO CLK90
125 MHz
MMCM
BUFG IntClk

BUFG IntClkDiv
OSERDES

Pattern

4
ISERDES

State
Machine

X523_10_041817

Figure 10: MMCM Clock Resources

1. Data enters FPGA at 1.25 Gb/s

The ISERDESE2 uses the CLK and CLK90 clocks running at 625 MHz. These are applied
through single-region BUFIO clocks, which is a requirement for this design.
Capturing of the ISERDESE2 outputs is done using SLICE flip-flops in the BUFG clock
domain. This cross-domain clocking forces a 600 ps mandatory constraint between the
outputs of the ISERDESE2 running in the BUFIO clock domain and the registers capturing
the data in the BUFG clock domain. (These constraints are supplied in the HDL source
code.)
2. DRU
a. The CLK and CLK90 clocks applied to the ISERDESE2 are available only for clocking
of devices in the I/O banks. Data must be transferred from the ISERDESE2 into SLICE
flip-flops. The SLICE flip-flops capturing the data must run at a separate 625 MHz
clock.
To move from the BUFIO clock network to the BUFG clock network, the use of clock
domain crossing (CDC) techniques is mandatory. In this case, the clocks for the

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 11

7 Series ISERDESE2 Oversampling Mode

ISERDESE2 and the clocks for the DRU logic are phase-aligned using a state
machine. It is also very important that the delay from ISERDESE2 to the registers in
the DRU does not exceed 600 ps.
b. In this stage of the DRU, data from the 625 MHz BUFG clock to the 312.5 MHz BUFG
clock is handled. These clocks are in phase with each other, and thus there are no
special requirements.
3. Data/clock presentation
The 10-bit data from the DRU is presented to the user interconnect logic at a 312.5 MHz
clock rate with a clock enable.
4. Clock alignment state machine
BUFIO and BUFG have an undefined phase relationship with each other (length of routing,
delay of clock buffers, etc.). To transfer data between both clock domains, CDC logic is
needed or, as used here, the clocks must be phase-aligned. To perform clock phase
alignment, a calibration scheme is set up. The clock alignment circuit uses the knowledge
that all I/Os in an FPGA I/O bank are identical and therefore have identical timing.
• An OSERDESE2 is loaded with a fixed data pattern and is clocked by the clocks from
the BUFG clock tree (IntClk, IntClkDiv). The output of the OSERDESE2 is a clock
pattern at the IntClk (625 MHz, BUFG) clock rate. Through the feedback path, the
clock pattern is captured by the neighbor ISERDESE2 running from the BUFIO clock
tree. The data-capturing ISERDESE2 runs on this same clock tree.
• Using this technique, it is possible to measure the phase relationship between the two
clocks. Using the independent phase-shift capability of the MMCM with a small state
machine, the BUFG clocks are phase-shifted to phase-match the BUFIO clocks. Along
with the CLK clock (625 MHz), the CLK90 clock (625 MHz) is phase-shifted, and with
the IntClk clock (625 MHz), the IntClkDiv clock (312.5 MHz) is phase-shifted.
The phase calibration process is illustrated in Figure 11.
X-Ref Target - Figure 11

625 MHz BUFIO

Initial Alignment of
625 MHz BUFG
(1)
Initial Phase
Adjustment Alignment
of 625 MHz BUFG
(2)
Final Alignment of
625 MHz BUFG
(3)
1 2 3
Output of ISERDESE1
at Stages of Phase 1010 0101 1010
Alignment
X523_11_041817

Figure 11: BUFG to BUFIO Phase Alignment Waveform Example

To allow for this clocking scheme, correct configuration of the MMCME2 is necessary. An
example of the MMCM configuration is provided in the next section.

MMCME2 Clock Generation

The calculations in this section are done assuming that the input for the MMCME2 is 125 MHz
and the output clocks need to sample a 1.25 Gb/s (625 MHz DDR) data stream. The
component used for this calculation example is a Kintex™-7 FPGA, -2 speed grade.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 12

7 Series ISERDESE2 Oversampling Mode

As shown in Figure 11, one MMCME2 is used as clock source for the interface. This means that
the MMCME2 must deliver these clocks:
• ClkRef ideally running at 312.5 MHz, but limited to 310 MHz by the IDELAY_CTRL
component parameters (REFCLK frequency = 300 MHz ±10 MHz)
• CLK at 625 MHz via BUFIO
• CLK90 inverted CLK at 625 MHz via BUFIO
• IntClk at 625 MHz via BUFG
• IntClkDiv at 312.5 MHz via BUFG
From DS182, Kintex-7 FPGAs Data Sheet: DC and Switching Characteristics, the MMCM
switching characteristics for the -2 speed grade are:
• MMCM_FIN_MIN = 10 MHz
• MMCM_FIN_MAX = 933 MHz
• MMCM_FVCO_MIN = 600 MHz
• MMCM_FVCO_MAX = 1440 MHz
• MMCM_FOUT_MIN = 4.69 MHz
• MMCM_FOUT_MAX = 933 MHz
• MMCM_FPFD_MIN = 10 MHz
• MMCM_FPFD_MAX = 500 MHz (Bandwidth set to High or Optimized)

Formulas:
 F IN 
D MIN = RoundUp  ----------------------------  = 1 Equation 5
 F PFD_MAX 

 F IN 
D MAX = RoundDown  ----------------------------  = 12 Equation 6
 F PFD_MAX 

  F VCO_MIN  
M MIN =  RoundUp  ----------------------------   • D MIN = 5 Equation 7
  F IN 

  D MAX  
M MAX = RoundDown   ----------------------------  ⁄ F IN  = 138 Equation 8
  F VCO_MIN  

 D MIN • F VCO_MAX 
M IDEAL =  -----------------------------------------------------  = 11.52 Equation 9
 F IN 

To let the PLL inside the MMCME2 run in optimal conditions, FVCO must be maximized while
the dynamic phase detector (FPFD_MAX) cannot be exceeded.
The VCO frequency is given by Equation 10.

F VCO = F IN •  M
-----  Equation 10
D 
D is calculated as 1 and the M value must be between 5 and 138. With an input clock of
125 MHz, if M is taken as 10, the VCO frequency is 1250 MHz. (Taking M as 12 provides a VCO
frequency of 1500 MHz, which is too high).

F VCO  F IN •  M
-----   125 MHz • ------ = 1250 MHz
10
Equation 11
D  1

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 13

Reference Design

The MMCME2 output clocks are calculated in Equation 12.

F OUT = F IN •  ---------------- 
M
Equation 12
D • O 
Where:
O is the divider factor of the output counter of the MMCME2 clock output.
D is the value used in the MMCME2 attributes.
• The CLK0 output of the MMCME2 accepts a real number value as divider. This is
perfect for generating 310 MHz from 1250 MHz.
- The D value is 4.0322.
• The CLK1 and CLK2 outputs of the MMCME2 are used to generate the 625 MHz
clocks that get distributed to the ISERDESE2 via the BUFIO clock buffers.
- The D value for both clocks is 2.
• Clock outputs CLK3 and CLK4 are then used to generate the IntClkDiv clock of
312.5 MHz and the IntClkv clock of 625 MHz. These clock outputs must be
phase-shifted. Therefore, the attribute for the phase-shifting operation of the MMCM
must be turned on.
- The D value for CLK3 (312.5 MHz) is 4.
- The D value for CLK4 (625 MHz) is 2.
- The attribute to turn on for both is CLKOUTn_USE_FINE_PS.
• n = clock output (3 or 4).

Reference The reference design files are available for download at:
Design https://secure.xilinx.com/webreg/clickthrough.do?cid=184349
Table 2 shows the reference design checklist.

Table 2: Reference Design Checklist

Parameter Description
General
Developer Name Marc Defossez
Target Device Kintex-7 and Virtex-7 FPGAs
Recommended speed grades: -2 and -3,
and only high-performance I/O banks in
-1 devices. (The -1 speed grade needs
special care. View the source code for
recommendations.)
Source Code Provided Yes
Source Code Format VHDL
IP Used No
Simulation
Functional Simulation Performed Yes
Timing Simulation Performed No
Testbench Format VHDL
Simulator Software/Version ISIM_13.3 or later
SPICE/IBIS Simulations No

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 14

Reference Design

Table 2: Reference Design Checklist (Cont’d)

Parameter Description
Implementation
Synthesis Tool/Version XST 13.3 or later
Implementation Tool/Version ISE Design Suite 13.3 or later
Static Timing Analysis Performed Yes
Hardware Verification
Hardware Verified Yes
Hardware Platform Used for Verification KC705 board

Table 3 shows a summary of the reference design utilization. The reference design contains a
dual receiver. For demonstration and test purposes, a set of PRBS transmitters and a set of
PRBS reception blocks is added on the KC705 board. Only the design utilization of the dual
receiver is listed. The XC7K325T-2-FFG900 device is used for implementation.

Table 3: Reference Design Utilization

Component Percentage (%) Total Number Number Utilized
Flip-Flops 1 407,600 191
Slices: 203,800
As LUT 1 174
As memory 1 17
As SRL 1 17
As route-through 2
BUFG/BUFGCTRL 9 32 3
ISERDESE2 1 500 5
OSERDESE2 1
IDELAYE2 1 350 4
IDELAYCTRL 1 10 1
MMCME2 (1) 1 10 1

Notes:
1. One MMCME2 is needed per I/O bank. One I/O bank can fit up to 19 receiver channels.
One set of ISERDESE2/OSERDESE2 must be kept free for the clock phase adjustment
(CDC logic).

The design uses floorplanning constraints in the UCF file. This implementation technique
makes each receiver use the same amount of FPGA logic and allows easy multiplication of the
design in any 7 series FPGA. For more information about this, refer to the documentation inside
the reference design ZIP file.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 15

Receiver UI and Jitter Tolerance

Receiver UI and Given the DRU method used, two valid sampling points are required at all times. This means
Jitter Tolerance the starting point is 0.500 UI. The oversampling is based on evenly spaced sampling points.
Any error in that spacing causes the receiver jitter eye requirement to increase.
Receiver Jitter Eye Requirement = Eye Requirement of DRU + Sampling Phase Error
0.625 UI = (0.500 UI) + (0.125 UI)
Sampling phase error comprises all of the effects of taking a 125 MHz clock and multiplying it
up to 625 MHz, then feeding it to the two BUFIOs with a phase shift and the IODELAYE2’s
200 ps phase shift.
Sampling phase error includes:
• MMCME2_ADV jitter for the exact setting in the reference design
• MMCME2_ADV phase error between CLK0 and CLK90
• MMCME2_ADV DCD
• IODELAYE2 delay accuracy (ability to create a 200 ps phase shift)
• IODELAYE2 pattern-dependent jitter
• Any offset between the two paths of the master and slave ISERDESE2
Sampling phase error does not include:
• Any other frequency of the clock or setting of the MMCME2_ADV
• Signal integrity losses (ISI, board jitter, etc.)
• Internal device jitter
Process, voltage, and temperature characterization are performed to qualify the interface. The
allowable overall jitter tolerance is 0.375 UI.

Reference The design is highly hierarchical. This provides great flexibility and allows easy reuse of design
Design modules. The top structure of the design is shown in Figure 12.

Directory Setup X-Ref Target - Figure 12

X523_12_012012

Figure 12: Top-Level Setup of the Design

Figure 13 and Figure 14 show the directory structure of the Common and SgmiiReceiver
folders, respectively.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 16

Reference Design Directory Setup

X-Ref Target - Figure 13

X23_13_012012

Figure 13: Directory Structure of Common Library

X-Ref Target - Figure 14

X523_14_012012

Figure 14: SgmiiReceiver Design with Directory Structure of the Libraries

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 17

Conclusion

SgmiiReceiver is the only part of the design needed for custom implementation. This is also
the part discussed in this application note.

Conclusion Xilinx 7 series FPGAs can implement asynchronous communication using SelectIO interface
resources, allowing transceivers to be reserved for other uses. This implementation also helps
to reduce cost by making smaller FPGA selection possible.

Revision The following table shows the revision history for this document.
History
Date Version Description of Revisions
04/06/12 1.0 Initial Xilinx Release.
05/17/17 1.1 Added Zynq-7000 SoC to document title and Summary. Removed “GT”
before “transceivers” in Introduction and Conclusion. Updated
OSERDES and ISERDES connections in Figure 2. Updated OSERDES
connection in Figure 10. In Figure 11, updated BUFIO rate to 625 MHz.

Notice of The information disclosed to you hereunder (the “Materials”) is provided solely for the selection and use of
Xilinx products. To the maximum extent permitted by applicable law: (1) Materials are made available "AS
Disclaimer IS" and with all faults, Xilinx hereby DISCLAIMS ALL WARRANTIES AND CONDITIONS, EXPRESS,
IMPLIED, OR STATUTORY, INCLUDING BUT NOT LIMITED TO WARRANTIES OF
MERCHANTABILITY, NON-INFRINGEMENT, OR FITNESS FOR ANY PARTICULAR PURPOSE; and (2)
Xilinx shall not be liable (whether in contract or tort, including negligence, or under any other theory of
liability) for any loss or damage of any kind or nature related to, arising under, or in connection with, the
Materials (including your use of the Materials), including for any direct, indirect, special, incidental, or
consequential loss or damage (including loss of data, profits, goodwill, or any type of loss or damage
suffered as a result of any action brought by a third party) even if such damage or loss was reasonably
foreseeable or Xilinx had been advised of the possibility of the same. Xilinx assumes no obligation to
correct any errors contained in the Materials or to notify you of updates to the Materials or to product
specifications. You may not reproduce, modify, distribute, or publicly display the Materials without prior
written consent. Certain products are subject to the terms and conditions of Xilinx’s limited warranty,
please refer to Xilinx’s Terms of Sale which can be viewed at https://www.xilinx.com/legal.htm#tos; IP
cores may be subject to warranty and support terms contained in a license issued to you by Xilinx. Xilinx
products are not designed or intended to be fail-safe or for use in any application requiring fail-safe
performance; you assume sole risk and liability for use of Xilinx products in such critical applications,
please refer to Xilinx’s Terms of Sale which can be viewed at https://www.xilinx.com/legal.htm#tos.

XAPP523 (v1.1) May 17, 2017 www.xilinx.com 18