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

All Digital VCXO Replacement for

Gigabit Transceiver Applications
XAPP589 (v2.3) April 29, 2015 (7 Series/Zynq-7000)
Authors: David Taylor, Matt Klein, and Vincent Vendramini

Summary
This application note delivers a system that is designed to replace external voltage-controlled
crystal oscillator (VCXO) circuits by utilizing functionality within each serial gigabit transceiver.

Note: In this application note, transceiver refers to these types of transceivers:

Device Family Transceiver Type

Artix®-7 FPGA GTP transceiver
Kintex®-7 FPGA GTX transceiver
Virtex®-7 FPGA GTX and GTH transceivers
Zynq®-7000 All Programmable SoC GTP and GTX transceiver

A common design requirement is to frequency or phase lock a transceiver output to an input

source (known as loop, recovered, or slave timing). Traditionally, an external clock cleaning
device or VCXO and PLL components are used to provide a high-quality clock reference for the
transceiver, since FPGA logic-based clocks are generally too noisy. While effective, external
clock components carry a power and cost penalty that is additive as each individual clock
channel is generated. When using many channels or in low-cost systems, the cost can be
significant. Additionally, adding many external clock sources provides more opportunity for
crosstalk and interference at the board level.

The system described in this application note provides a method to effectively replace these
external clock components using a combination of unique Xilinx transceiver features in
conjunction with a high-performance FPGA logic based digital PLL (DPLL). Each transceiver has
a phase interpolator (PI) circuit in the high-speed analog PLL output circuits that provides, on a
individual transceiver channel basis, the ability to phase and frequency modulate the transmit
clock operating the transceiver. Using a fully digital interface, the phase interpolator can be
phase and frequency controlled from the FPGA logic resources under control of a
high-resolution programmable DPLL. In conjunction with the FPGA logic DPLL, the phase
interpolator provides the ability to phase or frequency modulate the transceiver data output
directly locking to an input reference pulse or clock while providing an built-in clock cleaning
filter function. Unlike conventional solutions, high-quality system results because the clocking
components are contained within the transceiver.

The reference design circuit provides a fully integrated DPLL and transceiver phase interpolator
system which can be instantiated for each transceiver channel used. The transceiver can phase
or frequency lock to an input reference signal. The DPLL enables generation of a synchronous
transceiver data output with run-time configurable parameters (e.g., gain, cutoff frequency, and

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 1

System Applications

clock divider values) to enable you to set up the operation specifically for the end application.
This particularly highlights the flexibility of the reference input signal and DPLL cleaning
bandwidth.

The reference design circuit can lock an individual transceiver channel to up to ±160 ppm from
the reference oscillator and programmatically provide jitter cleaning bandwidths in the range
from 0.1 Hz to 1 KHz. In the 7 series FPGAs, the transceiver is capable of operating at up to
12.5 Gb/s. Typical applications for this circuit include video SD/HD, Sync E, IEEE1588, SDH,
SONET, and OTN.

You can download the reference design files for this application note from the Xilinx® website.
For detailed information about the design files, see Reference Design, page 28.

System Applications
A number of different applications need external VCXOs and PLLs or clock-cleaning
components on a per transceiver transmitter basis.

Some application examples include:

• OTN muxponders for trunk to tributary slaving of outputs.

• Broadcast equipment including switchers and routers using SD, HD, and 3G SDI video
outputs.
• Synchronous Ethernet.
• Recovered media clock generation (per IEEE specification 1588).

External components used to do this task are a high-impact expense for several reasons:

• Significant BOM cost, an estimated $10 to $15 per additional VCXO/PLL or clock cleaner.
• Significant power consumption (300 mW to 500 mW) per additional VCXO/PLL or clock
cleaner.
• Board space and PCB complexity, both due to additional board area required and careful
noise-reduction design layout requirements.

Figure 1 shows a general use case where inputs are received through any one of a number of
types of links carrying data. There could be one input link per output link or a group of input
links that data is striped across, which are de-multiplexed to form output links. While each input
link can share one reference clock, the challenge is that each output link needs its own
VCXO/PLL or clock cleaning system to provide a clean reference to the transceiver to serialize
outgoing data and produce the expected low-jitter output signal.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 2

System Applications

X-Ref Target - Figure 1

VCXO/PLL or
Clock Cleaner

CTRL or CLK
Control/CLK

Clean Clock
New CLK

Data SerDes Output #1

Demultiplexer, Switcher, Router, Channelizer

SerDes Output #2

SerDes Output #3

Inputs
SerDes Output #4

SerDes Output #5

$10 to $15 per VCXO

SerDes Output #N

Inputs can share one XTAL, but each unique output needs a VCXO X589_01_101013

Figure 1: Typical Design with Multiple VCXOs (One per Unique Output Rate)
By using transceivers the need for external VCXO/PLLs and clock cleaners can be eliminated.
The basics of the reference design method are:

• The 7 series FPGA transceiver has a transmit clock phase interpolator (TX PI) for the
transmit serial/deserializer bit clock.
• Each phase interpolator in each transmit serial/deserializer can be independently,
dynamically, and continuously changed in phase and hence, shifted in frequency.

There are a number of benefits to this implementation:

• Significant BOM cost reduction, an estimated $10 to $15 per additional VCXO/PLL.
• Significant power consumption savings (300 mW to 500 mW) per VCXO/PLL.
• Reduced board space and PCB complexity.
• Ability to have four unique differentiated transmission rates within a transceiver Quad.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 3

System Applications

An example block diagram of this new method is shown in Figure 2. The output VCXOs/PLLs or
clock cleaners are brought into the FPGA using the transceiver Quad’s phase-shifting
functionality.
X-Ref Target - Figure 2

Control Fixed
Digital PLL New CLK REFCLK(s)

DATA SerDes Output #1

Demulitplexer, Switcher, Router, Channelizer

DPLL

SerDes Output #2

DPLL

SerDes Output #3

Inputs DPLL

SerDes Output #4

DPLL

SerDes Output #5

DPLL

SerDes Output #N

Output VCXOs Removed

X589_02_041212

Figure 2: Phase-Shifting Solution Block Diagram

Solution Examples
This section describes solutions for broadcast switcher or router applications and OTN trunk to
tributary demultiplexer applications. Normally, an external VCXO/PLL or clock cleaner is
required for each unique output transmit serial/deserializer channel. This is very expensive
because each parts per million (PPM) channel variation requires an additional circuit, even if the
output base rate is the same (e.g., 1.485 Gb/s + 50 ppm and 1.485 Gb/s – 20 ppm).

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 4

System Applications

Example 1: Broadcast Switcher or Router

In this example (Figure 3), a broadcast switcher or router is receiving SD/HD/3G streams from
different cameras or unrelated sources. The video is processed inside of an FPGA, but the
outputs of FPGA must be exactly locked to some of the input channels. Individual HD-SDI and
3G-SDI inputs, for instance, while nominally running at 1.485 Gb/s and 2.97 Gb/s might not be
locked to each other. The inputs could be locked to their original sources, which could vary from
the nominal frequency by up to 150 ppm. Normally, when individual outputs of the FPGA are
meant to be locked to these PPM varied inputs, the design includes an external VCXO/PLL or
clock cleaner for each PPM varied output. By using the FPGA’s TX PI, the complex and expensive
components are instead designed into the transmit serial/deserializer functionality.
X-Ref Target - Figure 3

REFCLK
(1-2 Fixed XOs)
SD/HD/3G #1 RX SerDes Logic Resources TX SerDes
SD/HD/3G #1
RX SerDes FIFO
TX PI +
Digital PLL SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL
SD/HD/3G #2

SD/HD/3G #2
RX SerDes FIFO
TX PI +
Digital PLL SerDes
Routing and Switching Function

Rate Generator Phase Det.

LPF, FRQCTRL TX PI CTRL
SD/HD/3G #3
SD/HD/3G #3
RX SerDes FIFO
TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL
SD/HD/3G #4
SD/HD/3G #4
RX SerDes FIFO
TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

SD/HD/3G #N
SD/HD/3G #N
RX SerDes FIFO
TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

X589_03_041212

Figure 3: Broadcast Switcher or Router without External VCXOs/PLLs or Clock Cleaners on the Outputs

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 5

System Applications

Example 2: OTN Muxponder

In this example (Figure 4), within an OTN trunk there are multiple streams of data. Each stream
within the trunk has a unique rate. In a number of cases, an FPGA or other device receiving the
trunk will demultiplex it into its component streams. Each source stream, even at approximately
the same rate, has an oscillator or clock system that originally sourced the stream in the trunk.
The timing of the recovered stream must be preserved on a stream by stream basis. For
example, when there are multiple synchronous Ethernet streams that are each nominally at
1.25 Gb/s, but not from the same source, each stream (even if it is the same type) can vary by a
few PPM. The recovered outputs must each be exactly locked to their original source, which
requires a phase detector, low-pass filter, VCXO, and PLL external to the FPGA per unique output
channel. The Xilinx FPGA’s TX PI in the transmit serial/deserializer can take fixed nominal
oscillator rates as REFCLKs and effectively slave them within the transmit serial/deserializer to
the unique recovered rate with low jitter and without the need for external VCXOs/PLLs or clock
cleaners.
X-Ref Target - Figure 4

REFCLK
(Fixed XOs)
RX SerDes Logic Resources TX SerDes
FIFO Stream 1
TX PI +
Digital PLL SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

FIFO Stream 2
TX PI +
Digital PLL SerDes
Rate Generator Phase Det.
10G/40G/100G LPF, FRQCTRL TX PI CTRL
Trunk with
Multiple Links
and Streams RX SerDes FIFO Stream 3
Demultiplexer

TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

RX SerDes FIFO Stream 4

TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

FIFO Stream N
TX PI +
Digital PLL
SerDes
Rate Generator Phase Det.
LPF, FRQCTRL TX PI CTRL

X589_04_041212

Figure 4: OTN Trunk to Tributary Demultiplexor and Delivery without External VCXOs/PLLs or Clock Cleaners

In these examples, and for many other cases, the unique Xilinx transmit clock phase interpolator
functionality built into the transmit serial/deserializer and the FPGA-based phase detector,
digital PLL, low-pass filter, and controlled transmit serial/deserializer phase interpolator replace
expensive external VCXO/PLLs or clock cleaners.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 6

VCXO Replacement Theory

VCXO Replacement Theory

The 7 series FPGAs contain a key functional block in the transceiver transmitter that enables
functional replacement of a VCXO. This block, a phase interpolator, produces an output clock
with a fine phase shift relative to an input clock. The fine phase shift is produced in response to
a control word. The control word allows selection of a phase between 0° and 360°.

To create the equivalent of a VCXO with only a fixed frequency source, a phase is selected by the
phase interpolator and the selected phase selection value is continuously updated with a
linearly increasing or decreasing phase. This is equivalent to a positive or negative frequency
shift proportional to the rate of change of the controlling phase. See Equation 1 through
Equation 5.
dΦ ( t )
f = --------IN
-------- Equation 1
dt

Φ IN ( t ) =  fIN dt = f IN t Equation 2

Φ OUT ( t ) = Φ IN ( t ) + Φ CONTROL ( t ) Equation 3

dΦ OUT ( t ) dΦ IN ( t ) dΦ CONTROL ( t )
Differentiating -------------------- = ---------------- + -------------------------------- Equation 4
dt dt dt

dΦ OUT ( t ) dΦ CONTROL ( t )
f OUT = --------------------- , f OUT = f IN + -------------------------------- Equation 5
dt dt

From these equations, the output frequency is shifted by the rate of change of the control
phase Φ CONTROL with respect to time.

Figure 5 shows a functional block diagram of the phase interpolator, which includes the inputs
and outputs used in conjunction with the high-speed serial clock input from the SerDes
transmit PLL and other circuits that participate in the complete solution. This block first
produces a number of primary phases from the n-phase generation block. This block generates
x phases separated by 360°/x. In the case of the transmit serial clock phase interpolator, there
are eight primary phases (0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°). The phase-select block
in Figure 5 selects two adjacent phases based on the control logic.

The phase interpolation function is performed by the phase mixer, which functionally combines
k parts of Φ 1 and (1 – k) parts of Φ 2 (resulting in an interpolated phase output of
Φ 1k + Φ 2(1 – k)), where k is a fraction between 0 and 1.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 7

VCXO Replacement Theory

X-Ref Target - Figure 5

Phase Interpolator

Phase Mixer
1
Transmit x High Speed
n-Phase Phase k
REFCLK SerDes Generation Select 1–k SerDes
PLL 2 Clock

Control Logic

Where: φ2 – φ1 = 360°/x
Dynamic Phase Control xapp589_05_091913

Figure 5: Functional Block Diagram of a SerDes Transmit PLL Feeding a Phase Interpolator

Figure 6 shows two primary phases (Φ 1 and Φ 2) selected from the n-phase generator that are
interpolated by the phase mixer. The output is a clock with a phase that is in between Φ 1and Φ 2
with a resolution determined by the number of fractional steps allowed.
X-Ref Target - Figure 6

OUT
Phase Mixer
1
1 12
k
1–k
2

High Speed
SerDes Clock

xapp589_06_042912

Figure 6: Phase Mixer Producing an Interpolated Phase from Coarse Phases

The primary phase generation and interpolated phase creation are all performed in a clean
analog domain. The final result is a high-speed clock that can be any one of 128 phases of the
high-speed transmit clock used for serialization of the parallel transmit data. This produces a
very low jitter clock with very fine phase resolution.

A phase interpolator operates on the line rate and is present in all 7 series transceivers. It takes
in an input clock and produces an output with a fine phase shift. The phase interpolator in the
7 series FPGA transceiver transmitter has a phase control port that can be dynamically accessed
to enable fine frequency control. The phase control port update rate is dependent on the
transceiver and clock speed, with greater than 200 MHz being achievable in Virtex-7 devices.

The maximum achievable frequency offset is a combination of phase interpolator step size and
update rate. An Excel spreadsheet is provided with the design download to help you estimate
achievable performance with your own system parameters.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 8

PICXO DPLL

PICXO DPLL
The phase interpolator controlled crystal or Xtal oscillator (PICXO) parameters must be set
appropriately to generate a transceiver channel locked to a reference signal. The DPLL can be
analyzed using standard methods from a derivation of the transfer function outlined in this
section.

The PICXO DPLL circuit, for analysis purposes, is considered to be in three functional blocks:

1. Phase frequency detector

The phase-frequency detector measures the phase difference between the reference (R) and
the PICXO (V) clocks and produces an error output. As the DPLL is second order when
locked, this error output is driven to zero. It has a transfer function that is defined in units
of radian-1 and gain, GPD.

2. Second order loop filter

The second-order loop filter consists of proportional and integral paths with digital gains
defined by the terms G1 and G 2. The output represents the required tune value for the
oscillator.

3. Numerically controlled oscillator

The numerically controlled oscillator function is performed by the transmit phase

transceiver’s interpolator block, the phase accumulator, and the sigma-delta modulator. This
has units of radians/s and gain GPICXO.

These are configured in a standard DPLL configuration shown in Figure 7.

X-Ref Target - Figure 7

PFD Loop Filter PICXO

GPD G2 GPICXO
H1(z) H2(z)
Reference + + + Line
In Out
– +
G1
+
Z-1

Z-1

X589_07_032015

Figure 7: PICXO DPLL Digital Equivalent

The transfer of the reference input clock to the line output data is represented by the function
in Equation 6. This allows the clock cleaning and tracking of the all-digital VCXO replacement to
be exactly controlled by your application.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 9

PICXO DPLL

H1 ( z )H2 ( z )G PD
H ( z ) = ------------------------------------------- Equation 6
1 + H1 ( z )H2 ( z )G PD

with:

( g1 + g 2 )z – g2
H1 ( z ) = ---------------------------------- Equation 7
(z – 1)

z ( G PICXO )
H2 ( z ) = ---------------------- Equation 8
(z – 1)

The gain parameters g1 and g2 are defined as:

( G1 – 2 )
2
g1 = ---------------- Equation 9
2 28

( G2 + 1 )
2
g2 = ----------------- Equation 10
2 28

And G PD and GPICXO are defined as follows for GTX transceivers where the phase interpolation
is controlled through the DRP:

CLK ( Hz ) × bitrate ( bps ) × 0.4 × 10 – 9

G PD = ------------------------------------------------------------------------------- Equation 11
V × CE DSP × 2π

CE PI × ACC_STEP × 2π
G PICXO = -------------------------------------------------- Equation 12
64 × TXOUT_DIV × 2 21

Where:
CLK ( Hz )
CE PI = ------------------- Equation 13
wr TIME

CE PI
CE DSP = ----------------------------- Equation 14
CE_DSP_RATE

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 10

PICXO Measurements and Performance

The previous equations use these constants:

• For 7 series FPGAs and Zynq-7000 AP SoC GTX transceivers, PI res = 64 and wr TIME = 6
• For 7 series FPGAs and Zynq-7000 AP SoC GTH and GTP transceivers, PI res = 64 and
wrTIME = 2

and these variables:

• bitrate: The final serial interface line rate

• CLK: The PICXO and/or the DRP clock sourced from TXOUTCLK
• TXOUT_DIV attribute: The main TXPLL divider setting that sets the operating rate range for
the transceiver
• ACC_STEP: The TXPI step size of the system selected by the user

The Excel spreadsheet tool included in the PICXO design file package allows you to estimate the
PICXO response when setting the configurable parameters listed above. The PICXO DPLL allows
complete flexibility with settings. Therefore, you should understand the performance trade-offs
in PLL operation.

For optimum jitter and cleaning performance, it is recommended that the PICXO DPLL
bandwidth be less than 100 Hz. Higher tracking bandwidths can be achieved, however, with
some increase in jitter. It might be desirable to have a high bandwidth to acquire lock, then
switching subsequently to a lower cleaning bandwidth. This is known as fast acquisition for the
DPLL.

The DPLL architecture allows on-the-fly changes to the G1 and G2 values to support this while
not losing phase lock. On-the-fly changes can be supported in user logic by applying variable
G1 and G2 values. It might be appropriate to monitor the error output from the DPLL as one
method to ascertain a suitable point at which to switch gain values.

PICXO Measurements and Performance

This section includes sample measurements of an example PICXO design implemented on the
KC705 board.

Figure 8 and Figure 9 demonstrate the DPLL error and virtual voltage during the locking
process when a step change in frequency is applied. The error range is ±219, the virtual voltage
±220. In this case, the virtual voltage is settling ~140000, indicating the PICXO is generating a
positive offset relative to the local GTX transceiver reference fixed source (GTX REFCLK
frequency) of approximately +10 ppm. As the local GTX transceiver reference drifts in
frequency, the output remains locked to the incoming data. This allows retransmission with no
external VCXO and it performs jitter cleaning of the recovered signal. The time unit is CE DSP
clocks.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 11

PICXO Measurements and Performance

The plots in Figure 8 and Figure 9 show the Error and Volt outputs when a step frequency
change of ~9 ppm is applied to the PICXO.
X-Ref Target - Figure 8

X589_08_032015

Figure 8: Plot of PICXO DPLL Volt and Error During Step Change
X-Ref Target - Figure 9

X589_09_032015

Figure 9: Plot of PICXO DPLL Error During Step Change

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 12

PICXO Measurements and Performance

Figure 10 shows the transmit phase interpolator codes being written from the PICXO in the GTX
transceiver. This demonstrates the frequency offset generation in progress. Figure 10 also
reflects the direct phase shifting of the transmitter PLL at its operating frequency. The phase
rotation shown results in a continuous phase ramp at the line rate. As a positive frequency is
being generated, the phase is being continuously subtracted to generate shorter periods. The
time unit is CE PI clocks.
X-Ref Target - Figure 10

X589_10_032015

Figure 10: Plot of PICXO Transmit Phase Interpolator Control When Locked

When using the transmit phase interpolator in the GTX transceiver a general expectation is that
the transmitter jitter will increase between 0.01 and 0.03 UI pk-pk due to the phase stepping
and rotational nature of the modulator.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 13

PICXO Measurements and Performance

Figure 11 shows a Virtex-6 FPGA example waveform and Figure 12 shows a Kintex-7 FPGA
waveform being operated in a jitter cleaning mode at a 9.83 GBs rate. When operating in
systems where margins are reduced, Xilinx recommends performing some evaluations.
X-Ref Target - Figure 11

X589_11_032015

Figure 11: Virtex-6 FPGA GTX Transceiver Data Output at 2.488 GB Generating +20 ppm Offset

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 14

PICXO Measurements and Performance

X-Ref Target - Figure 12

X589_12_032015

Figure 12: Kintex-7 FPGA GTX Transceiver Data Output at 9.83 GB Operating as a PICXO-based Jitter
Cleaner at +32 ppm Offset
Figure 13, and Figure 14 demonstrate the transfer bandwidth of the PICXO. The loop filter
programmability is exercised to show how the transfer function can be adjusted for varying user
requirements for bandwidth and damping.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 15

PICXO Measurements and Performance

X-Ref Target - Figure 13

9.95 GBs Jitter Transfer—Variable Bandwidth

10

-10

-20
Gain (dB)

Gain G2/G1
8/0
-30
10/2
12/4
-40 14/6
16/8
18/10
-50

-60
1 10 100 1000 10000

Frequency (Hz)
X589_13_032015

Figure 13: Kintex-7 FPGA GTX Transceiver 10 GB Jitter Transfer Measurements—Variable

Bandwidth
X-Ref Target - Figure 14

9.95 GBs Jitter Transfer—Variable Damping

5

-1

-3
Gain G2/G1
Gain (dB)

16/15
-5
16/15
-7 16/13
16/12
-9 16/11
16/9
-11
16/8

-13

-15
1 10 100 1000

Frequency (Hz)
X589_14_032015

Figure 14: Kintex-7 FPGA GTX Transceiver 10 GB Jitter Transfer Measurements—Variable Damping

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 16

PICXO Measurements and Performance

For broadcast equipment, 270 Mb/s, 1.485 Mb/s, and 2.97 Mb/s are standard rates for SD-SDI,
HD-SDI, and 3G-SDI. It is a challenge to meet the broadcast jitter requirements of all 3G-SDI
formats. One of the 3G-SDI formats is 3G Level A. The examples in Figure 15 and Figure 16 are
measurements showing the system passing with margin for 3G Level A SDI (3G Level A
1920 x 1080p at 59.94 Hz) in both 10 Hz and 100 KHz jitter measurement bandwidths,
respectively. In this design, the VCXO and re-clocking function is incorporated completely in the
FPGA using the PICXO scheme.
X-Ref Target - Figure 15

X589_15_032015

Figure 15: ML605 with SDI FMC Board for a 3G Level A SDI Output of a Triple-Rate SDI Pass-Through
Design (at 10 Hz Jitter)

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 17

PICXO Measurements and Performance

X-Ref Target - Figure 16

X589_16_032015

Figure 16: ML605 with SDI FMC Board for a 3G Level A SDI Output of a Triple-Rate SDI Pass-Through
Design (at 100 KHz Jitter)

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 18

PICXO Architecture Overview

PICXO Architecture Overview

Instead of using an external VCXO, as described in VCXO Replacement Theory with its
challenges, a complete digital PLL and clock cleaner can be created using the phase interpolator
in the transmit serial/deserializer as a phase interpolator controlled crystal or Xtal oscillator
(PICXO). The PICXO macro operation for GTX transceivers is shown in the functional block
diagram of Figure 17.
X-Ref Target - Figure 17

User
Reference DRP GTX Transceiver
Clock/Pulse
OFFSET_PPM CE_DSP_RATE DRPDEN

HOLD OFFSET_EN DRPRDY

CEDSP
DRPDATAO
CEPI CTRL DRP
/R
Arbiter
and FIFO
CE CE DO
CE DRP
Sign ADD_SUB Data DRPDATAI
2nd Order
Phase/ Loop
Frequency DO DI
Filter
Detector Error Volt Sigma 8-bit DRPADDR
Delta Phase
/V Modulator Accumulator

G1 & G2 DRPCLKIN
Variable BW
TXPCSOUTCLK
~0.1–1000 Hz ACC_STEP

TXOUTCLK
BUFG/H/R ‘1’ ENPI Ports
X589_17_032015

Figure 17: PICXO Macro Functional Block Diagram for GTX Transceivers

For Kintex-7 FPGAs, the dynamic reconfiguration port (DRP) arbiter/FIFO and control blocks
manage the clock and DRP data interfaces between the GTX transceiver, the PICXO DPLL, and
the User DRP.

A typical use model for a DRP operation is that a user application can, if required, program GTX
transceiver DRP parameters prior to operation. The detailed operation is described in Designing
with PICXO, page 21.

The phase in the GTX transceiver is under direct DRP control from the PICXO circuit consisting
of phase accumulator, sigma delta modulator and loop filter, and phase detector components.

The phase accumulator tracks the current phase of the phase interpolator and increments or
decrements the phase based on input from the sigma-delta modulator block. Incrementing or
decrementing phases directly results in a negative or positive frequency offset.

The required fine frequency control is achieved by the sigma-delta modulation block driven by
a second order DPLL consisting of filter and phase detector with user-configurable loop
parameters and comparison frequencies for maximum flexibility.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 19

PICXO Architecture Overview

The PICXO operation is synchronous with the DRP clock. The maximum phase interpolator
update rate (DRP CLK/6) is the clock enable rate for the sigma-delta modulator and
accumulator CEPI, shown in Figure 17. The DPLL runs at a sub-rate CE DSP, the clock enable rate
for the phase/frequency detector and second-order loop filter (in Figure 17). This allows the
sigma-delta modulator to run with high resolution and allows usable DPLL coefficients for
low-frequency clock cleaning.

The reference design circuit uses one BUFG/BUFH/BUFR per line rate generated. When locked,
this clock is synchronous with the reference clock and can be used for other user downstream
logic.

The PICXO macro operation for GTH and GTP transceivers is shown in the functional block
diagram of Figure 18.
X-Ref Target - Figure 18

Reference OFFSET_PPM CE_DSP_RATE

GTH Transceiver
Clock/Pulse HOLD OFFSET_EN

CEDSP
CEPI CTRL
/R

CE
CE
ACC_DATA TXPIPPMSTEPSIZE
2nd Order
Phase/ Loop
Frequency Filter
Detector Error Volt Sigma
Delta
/V
Modulator
TXUSRCLK2

G1 & G2
Variable BW
~0.1–1000 Hz ACC_STEP BUFG/H/R
TXOUTCLK

‘1’ ENPI Ports

X589_18_032015

Figure 18: PICXO Macro Functional Block Diagram for GTH and GTP Transceivers
For Virtex-7 FPGAs with GTH transceivers and Artix -7 FPGAs with GTP transceivers, the principle
of the PICXO operation is the same as for GTX transceivers. However, the control of the GTH and
GTP TX PI is through dedicated ports.

The PICXO macro provides phase increment information directly to the GTH/GTP transceiver
TXPPMSTEPSIZE input ports. This bus applies phase increment or decrement information to the
internal GTH/GTP TXPI phase accumulator. This offers several advantages over the DRP access
method:

• The DRP does not need to be arbitrated between the user functions and the PICXO.
• The dedicated ports allow a faster update rate for the TXPI.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 20

Designing with PICXO

Designing with PICXO

Physical Interface
Table 1 through Table 5 show the port definitions.

Table 1: Clocks, Reset, and Interface to the Transceiver Ports

Signal Name Direction Description
RESET_I Input Synchronous reset. Active-High.
REF_CLK_I Input Reference clock, can be any clock (local, BUFG, pulse, etc.)
TXOUTCLKPCS_I Input Unused. Connect to 0, 1, or TXOUTCLK_I.
TXOUTCLK_I Input Connects to TXOUTCLK of the GTX serial transceiver via a BUFG/BUFH/BUFR.
DRPEN_O Output Connects to the DRPEN port of the GTX serial transceiver.
DRPWEN_O Output Connects to the DRPWE port of the GTX serial transceiver.
Connects to DRPDO of the GTX serial transceiver. Internally connects directly
DRPDO_I [15:0] Input
to DRPDATA_USER_O[15:0].
DRPDATA_O [15:0] Output Connects to the DRPDI port of the GTX serial transceiver.
DRPADDR_O [8:0] Output Connects to the DRPADDR port of the GTX serial transceiver.
DRPRDY_I Input Connects to the DRPRDY port of the GTX serial transceiver.

Table 2: DRP User Port (GTX Transceivers Only)

Signal Name Direction Description
DRP_USER_REQ_I Input Asserted to request DRP port access. Active-High signal.
DRP_USER_DONE_I Input Reset DRP arbiter.
DRPEN_USER_I Input Same functionality as DEN GTX serial transceiver port [Ref 1] .
DRPWEN_USER_I Input Same functionality as DWEN GTX serial transceiver port [Ref 1] .
DRPADDR_USER_I [8:0] Input Same functionality as DADDR GTX serial transceiver port [Ref 1] .
DRPDATA_USER_I [15:0] Input Same functionality as DI GTX serial transceiver port [Ref 1] .
This signal is a mirror of DRDY in the TXOUCLK domain. Indicates data present
DRPRDY_USER_O Output
on DRPDO is valid [Ref 1] .
DRPDATA_USER_O[15:0] Output Directly connected to DRPDO_I[15:0].
DRPBUSY_O Output Indicates that the DRP port is not available to the user. Active-High signal.

Table 3: TXPI Port (GTH and GTP Transceivers Only)

Signal Name Direction Description
ACC_DATA[4:0] Output Connects to TXPIPPMSTEPSIZE[4:0] of the GTH or GTP transceiver.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 21

Designing with PICXO

Table 4: Debug Ports

Signal Name Direction Description
ERROR_O [20:0] Output Output of phase detector. Signed number.
VOLT_O[21:0] Output Output of low-pass filter. Signed number.
DRPDATA_SHORT_O[7:0] Output Output of the accumulator. Unsigned number.
CE_PI_O Output Clock enable for accumulator.
CE_PI2_O Output Clock enable for low pass filter and DAC.
Reset phase detector counters, load phase detector error into the low-pass
CE_DSP_O Output
filter.
OVF_PD Output Overflow in the phase detector.
OVF_AB Output Saturation of the low-pass filter inputs.
OVF_INT Output Saturation of the low-pass filter integrator.
OVF_VOLT Output Saturation of the low-pass filter output.

Table 5: PICXO Loop Parameters

Signal Name Direction Description
G1[4:0] Input Filter linear path gain: Range 0 to x12h.
G2[4:0] Input Filter integrator path gain: Range 0 to x14h. (1)

R[15:0] Input Reference divider: Range 0 to 65535. Divides by R+2.

V[15:0] Input TXOUTCLK_I divider: Range 0 to 65535. Divides by V+2.
ACC_STEP[3:0] Input PICXO step size: Range 1 to 15 (0 = no step).
CE_DSP_RATE[15:0] Input DSP divider: Default 07FF. Controls CE_DSP rate.
VSIGCE_I Input Clock enable of the TXOUTCLK_I divider. Connects to 1 for normal operation.
VSIGCE_O Output Reserved: Floating.
RSIGCE_I Input Clock enable of Reference divider. Connects to 1 for normal operation.
C_I[7:0] Input Reserved: Connects to 0.
P_I[9:0] Input Reserved: Connects to 0.
N_I[9:0] Input Reserved: Connects to 0.
Direct frequency offset control. Signed number. OFFSET_PPM overwrites the
OFFSET_PPM[21:0] Input
output of the low-pass filter (VOLT_O) when OFFSET_EN is High.
Enables direct frequency offset control input. Active-High: Enables
OFFSET_EN Input
OFFSET_PPM input to overwrite output of low-pass filter (Volt).
Holds low-pass filter output value (Volt). Clock enable of Volt that stops Volt
HOLD Input
to the latest known ppm.
DON_I Input Potential jitter reduction. Active-High.

Notes:
1. Incorrect G1 and G2 values (either too high or low) can stop the DPLL from locking. Normally the target loop bandwidth should be in
the region of a few to several hundred Hz with G2 ≥ 1. It should also be understood G values are 2 N the set value.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 22

Designing with PICXO

Interface Operation
General Operation
The PICXO parameters (V, R, ACC_STEP, CE_DSP_RATE) can affect the PICXO lock if changed,
therefore they are considered pseudo-static inputs. The gains G1 and G2 can be changed
without loss of lock. All input and output signals to/from the PICXO are synchronous to
TXOUTCLK_I except REF_CLK_I and R. Figure 19 shows the timing dependency between
TXOUTCLK_I and the main debug outputs.
X-Ref Target - Figure 19

TXOUTCLK_I

CE_PI_O

CE_PI2_O

CE_DSP_O

ERROR_O E1 E2 E2 E3

VOLT_O V1 V2 V2 O1 V2 V3

OFFSET_EN

OFFSET_PPM O1

X589_19_032015

Figure 19: Timing Waveforms of Main Debug Outputs

Reset Considerations
The PICXO main reset RESET_I requires a minimum of eight TXOUTCLK_I cycles to reset the
PICXO correctly. When applied, RESET_I resets all blocks, including the phase detector, low-pass
filter, and DRP arbiter. When releasing RESET_I, the first phase detector output (ERROR_O) is
zero, and the first word written in the transceiver phase interpolator is zero.

The PICXO second reset DRP_USER_DONE_I resets the DRP arbiter only. In 7 series FPGAs, a
reset of the DRP arbiter is not necessary to restart DRP operation.

The transceiver TX PMA reset sequence must be completed before the PICXO reset is released
for operation.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 23

Designing with PICXO

7 Series FPGA Transceiver Clocking

The primary clocking scheme is detailed in Figure 20. The transceiver TXOUTCLK connects to a
BUFG that drives the PICXO inputs clocks TXOUTCLK_I and TXOUTCLKPCS_I, as well as the GTX
DRP clock DCLK.
X-Ref Target - Figure 20

GTX/GTH/GTP PICXO
BUFG/R/H
Transceiver

TXOUTCLK TXOUTCLK_I

GTREFCLK

DCLK/TXUSRCLK2 TXOUTCLKPCS_I

X589_20_032015

Figure 20: 7 Series FPGA Primary PICXO Clocking Scheme

Notes relevant to Figure 20:

1. For 7 series FPGA GTX transceivers, TXOUTCLK_I must be the same clock as DCLK.
2. For 7 series FPGA GTH/GTP transceivers, TXOUTCLK_I must be the same clock as
TXUSRCLK2.

A secondary clocking scheme is detailed in Figure 21. This clocking scheme can be used when
TXOUTCLK exceeds the GTX DRP clock specifications. In this case, the GTX DRP clock must be an
integer divisor of the TXOUTCLK frequency.
X-Ref Target - Figure 21

GTX/GTH/GTP PICXO
Transceiver BUFG/R/H
PLL
TXOUTCLK TXOUTCLK_I

GTREFCLK

DCLK/TXUSRCLK2 TXOUTCLKPCS_I

X589_21_032015

Figure 21: 7 Series FPGA Secondary PICXO Clocking Scheme

Notes relevant to Figure 21:

1. For 7 series FPGA GTX transceivers, TXOUTCLK_I must be the same clock as DCLK.
2. For 7 series FPGA GTH/GTP transceivers, TXOUTCLK_I must be the same clock as
TXUSRCLK2.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 24

Designing with PICXO

7 Series FPGA GTX DRP User Port Operation

The 7 series FPGA PICXO DRP user interface (Table 2) must be synchronous to TXOUTCLK_I. The
7 series FPGA GTX DRP interface (Table 1) is synchronous to TXOUTCLK_I. The DRP arbiter is
implemented with a finite state machine (FSM). The FSM reset is triggered by
DRP_USER_DONE_I or RESET_I. However, the user DRP port can still operate as normal when
DRP_USER_DONE_I or RESET_I are High.

To operate the 7 series FPGA PICXO DRP user port, the application asserts the DRP_USER_REQ_I
signal and waits for DRP_BUSY_O to transition Low. After DRP_BUSY_O is Low, the application
can operate the DRP USER port as per the GTX DRP specification [Ref 1]. The application must
keep DRP_USER_REQ_I asserted during a DRP transfer. After the application is finished with the
DRP accesses, it drives DRP_USER_REQ_I Low. The PICXO regains control of the DRP port and
DRPBUSY_O goes High (Figure 22).
X-Ref Target - Figure 22

TXOUTCLK_I
ce_pi and ce_dsp Faster Rate
CE_PI_O When drp_busy = 0

DRP_USER_REQ_I

DRPBUSY_O Resume PICXO DRP Access

DRPEN_USER_I

DRPWEN_USER_I

DRPADDR_USER_I[8:0]

DRPDATA_USER_I[15:0]

DRPRDY_USER_O

DEN

DWE

D[15:0]

DADDR[8:0]

DRDY

X589_22_032015

Figure 22: 7 Series FPGA GTX DRP User Port Operation

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 25

Designing with PICXO

HOLD Input Operation

The HOLD input is a clock enable to the low-pass filter integrator and output (VOLT_O). While
HOLD is High, the phase detector continues to operate as normal. When HOLD returns to Low,
the low-pass filter output is not synchronized anymore with the phase detector. Figure 23
illustrates this behavior.
X-Ref Target - Figure 23

TXOUTCLK_I

CE_PI_O

CE_PI2_O

CE_DSP_O

ERROR_O E1 E2 E2 E3

VOLT_O V1 V1 V2

HOLD

X589_23_032015

Figure 23: HOLD Input Operation

Direct Offset Control

OFFSET_PPM and OFFSET_EN allow direct control of the frequency offset. When OFFSET_EN is
High, the output of the low-pass filter (VOLT_O) takes the OFFSET_PPM value. During this time,
the phase detector and low-pass filter integrator operate normally. When OFFSET_EN returns to
Low, the output of the low-pass filter (VOLT_O) takes the current value calculated by the phase
detector and low-pass filter. Figure 24 illustrates this behavior.
X-Ref Target - Figure 24

TXOUTCLK_I

CE_PI_O

CE_PI2_O

CE_DSP_O

CE_PI_O* CE_DSP_RATE

ERROR_O E1 E2 E2 E3

VOLT_O V1 V2 V2 V3

OVF_AB
Minimum Six Cycles
OVF_INT After CE_DSP_O

X589_24_032015

Figure 24: Direct Offset Control

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 26

Implementation

Implementation
Vivado Tools Implementation
The PICXO design is delivered as a custom IP. To add the design to a project:

1. Unzip the file in a location.

2. Add the IP repository to the project: Tools > Project Options, select IP on the left pan, click
Add Repository, and select the XAPP589_picxo folder.
3. Select the IP catalog. The PICXO IP is under FPGA Features and Design > IO Interfaces.
4. Right-click XAPP589_picxo and select Customize IP.
5. Select the IP module name and the type of GT. Click OK.
6. The example design can be generated by selecting the IP source, doing a right-click, and
generate example design.

The GTP/GTX/GTH transceiver associated with the PICXO must be constrained to a specific
location. Period constraints are necessary on TXOUTCLK_I and REFCLK_I.

Mandatory Conditions and Limitations

7 Series FPGA GTX Transceiver
• Transmit buffer bypass is not supported.
• TXDLY_LCFG[2] and PCS_RSVD_ATTR[1] must be set to 1.
• The Kintex-7 FPGA ports TXPHALIGN, TXPHALIGNEN, and TXPHOVRDEN must be
connected to 1.
• TXPHDLYPD must be connected to 0.
• TXOUTCLKSEL must be set to TXOUTCLKPMA (010)

7 Series FPGA GTH/GTP Transceiver

• Transmit buffer bypass is not supported.
• GTH/GTP port TXPIPPMEN must be connected to 1.
• GTH/GTP port TXPIPPMSEL must be connected to 1.
• GTH/GTP attribute TXPI_SYNFREQ_PPM must be set to 001.
• TXOUTCLK and TXUSRCLK2 must be the same clock. TXOUTCLKSEL must be set to 010
or 001.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 27

Reference Design

Reference Design
The reference design files are based on the 7 series transceiver wrapper v2.1 [Ref 2]. The
designs target the KC705, AC701, ZC706, and VC709 development platforms. They loopback the
receive data to the transmitter. The PICXO instance locks the transmitter to the recovered clock
RXRECLK.

The output error_o of the phase/frequency detector can be captured when CE_DSP_O is High to
monitor the PICXO response. When locked, ERROR_O should oscillate around 0 (see Figure 9).

Simulation of the example design is not supported. The drp_arbiter source code is provided to
enable functional simulation of DRP user access.

You can download the reference design files for this application note at
www.xilinx.com/member/vcxoff/index.htm (see Download the VCXO Removal Reference Design
for 7 Series FPGAs).

Table 6 shows the reference design matrix.

Table 6: Reference Design Matrix

Parameter Description
General
Developer names David Taylor, Matt Klein, Vincent Vendramini
Kintex-7 XC7K325T FFG900-1
Virtex -7 XC7VX690T FFG1761-2
Target devices
Artix-7 XC7A200T FBG676C-2
Zynq-7000 XC7Z045 FFG900-2
Source code provided Yes
Source code format VHDL
Design uses code/IP from existing Xilinx application
note/reference designs, CORE Generator™ tool, or third Yes
party
Simulation
Functional simulation performed No
Timing simulation performed No
Test bench used for functional and timing simulations No
Test bench format N/A
Simulator software tools/version used N/A
SPICE/IBIS simulations N/A
Implementation
Synthesis software tools/version used Vivado® Design Suite 2015.1
Implementation software tools/version used Vivado Design Suite 2015.1
Static timing analysis performed Yes

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 28

References

Table 6: Reference Design Matrix (Cont’d)

Parameter Description
Hardware Verification
Hardware verified Yes
Hardware platform used for verification KC705, VC709, ZC706, and AC701

Table 7 shows the device utilization table for the reference design.

Table 7: Device Utilization and Performance for Reference Design (Vivado Design Suite 2015.1)
Artix-7 FPGAs Kintex-7 FPGAs
(One GTP Transceiver) (One GTX Transceiver) Virtex-7 FPGAs
Target Devices
(Four GTH Transceivers)
Full Design Full Design
Slice LUTs 4,350 2,993 12,203
Slice registers 6,054 3,966 15,696
Occupied slices (1) 2,077 1,447 5,533
Block RAM 14 14 56
BUFG/BUFHCE 6/0 4/0 10/3
GTP/GTX/GTH 1 1 4
MMCM 0 0 0

Notes:
1. The number of occupied slices can vary depending on packing results.

Table 8 shows the statistics and performance expectations for a standalone PICXO.

Table 8: Statistics and Performance Expectations for a Standalone PICXO (Vivado Design
Suite 2015.1)
7 Series FPGA 7 Series FPGA 7 Series FPGA
Target Devices
GTP Transceiver GTX Transceiver GTH Transceiver
LUTs 862 940 873
Registers 950 992 950
SRLs 17 17 17
Occupied slices 348 355 357
Maximum PICXO Speed grade dependent, Speed grade dependent, Speed grade dependent,
clock rate matches TXUSRCLK2 matches DRP port limit matches TXUSRCLK2
maximum frequency maximum frequency

References
1. 7 Series FPGAs Configuration User Guide (UG470). See the “Dynamic Reconfiguration Port”
chapter.
2. 7 Series FPGAs GTX/GTH Transceivers User Guide (UG476)

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 29

Revision History

Revision History
The following table shows the revision history for this document.

Date Version Revision

05/08/2012 1.0 Initial Xilinx release.
06/19/2012 1.1 Updated to include Verilog version. Minor changes to the VHDL code in the
Reference Design. Updated Table 5. Revised Block RAM in Table 7.
10/16/2012 2.0 Added support for Kintex-7 FPGAs throughout document, including adding
figures and revising tables. See new Reference Design information with
updated design file.
10/15/2013 2.1 Updated Equation 9 through Equation 13. Updated bulleted list of constants
after Equation 14. Removed Figure 8: “Estimated Transfer Function using
Function H(z) in Equation 6” and Table 1: “Kintex - 7 FPGA PICXO Sample
Parameters for a Nominal 50 Hz Cleaning Bandwidth.” Updated PICXO
Architecture Overview. Replaced DIVCNT_TC with CE_DSP_RATE throughout.
Renamed, relocated, and updated PICXO Architecture Overview. In Table 1,
replaced Kintex-7 with 7 series FPGAs, added DRPDO_I [15:0], and updated
DRPADDR_O [8:0]. In Table 2, updated title, added DRP_USER_DONE_I, and
updated DRPADDR_USER_I [8:0]. Added Table 3. In Table 4, replaced CKLEN_O
with CE_PI2_O and RSTCNT_O with CE_DSP_O. In Table 5, replaced
DIVCNT_TC[15:0] with CE_DSP_RATE[15:0] and updated descriptions of V[15:0],
ACC_STEP[3:0], VSIGCE_I, RSIGCE_I, OFFSET_PPM[21:0], and HOLD. Updated
Interface Operation, Implementation, Mandatory Conditions and Limitations,
and Reference Design. Updated Table 8 and moved to Reference Design.
09/12/2014 2.2 Updated for Vivado Design Suite 2014.1. Removed information about
Virtex-6 FPGAs and the ChipScope™ tool. Defined the transceiver types
covered in the application note on page 1. Removed application support for
3G-SDI from the Summary, page 1. The “Phase Resolution and Frequency Shift
Example” section was deleted. Added OVF_VOLT in Table 4 and DON_I in
Table 5. The Table 6 title changed from Reference Design Checklist to
Reference Design Matrix. Added support for AC701 and ZC706 to Reference
Design, page 28 and to the Table 6 Hardware Verification section. Virtex-6 LXT
XC6VLX240T FF1146-1 was removed from Table 6 target devices and Artix-7
XC7A200T FBG676C-2 and Zynq-7000 XC7Z045 FFG900-2 were added. XST, ISE
Design Suite 13.4 and 14.4 were removed from Implementation software tools
used. Removed ChipScope IP data from Table 7. The MUXFXes row was
removed from Table 8, and the Virtex-6 LXT and SXT column was replaced with
7 series FPGA GTP transceiver information. Values in Table 7 and Table 8 were
updated. Removed the “ISE Tools Implementation” section from
Implementation, page 27. Removed Virtex-6 documents from References,
page 29.
04/29/2015 2.3 Updated for Vivado Design Suite 2015.1. Updated constants for the equations
in PICXO DPLL. Updated Table 1, Table 2, Table 3, Table 5, Table 6, Table 7, and
Table 8. Added GTP to Table 7 and Table 8. Updated Figure 20 and Figure 21.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 30

Please Read: Important Legal Notices

Please Read: Important Legal Notices

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 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 http://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 http://www.xilinx.com/legal.htm#tos.
Automotive Applications Disclaimer
XILINX PRODUCTS ARE NOT DESIGNED OR INTENDED TO BE FAIL-SAFE, OR FOR USE IN ANY APPLICATION REQUIRING FAIL-SAFE
PERFORMANCE, SUCH AS APPLICATIONS RELATED TO: (I) THE DEPLOYMENT OF AIRBAGS, (II) CONTROL OF A VEHICLE, UNLESS THERE
IS A FAIL-SAFE OR REDUNDANCY FEATURE (WHICH DOES NOT INCLUDE USE OF SOFTWARE IN THE XILINX DEVICE TO IMPLEMENT THE
REDUNDANCY) AND A WARNING SIGNAL UPON FAILURE TO THE OPERATOR, OR (III) USES THAT COULD LEAD TO DEATH OR
PERSONAL INJURY. CUSTOMER ASSUMES THE SOLE RISK AND LIABILITY OF ANY USE OF XILINX PRODUCTS IN SUCH APPLICATIONS.
© Copyright 2012–2015 Xilinx, Inc. Xilinx, the Xilinx logo, Artix, ISE, Kintex, Spartan, Virtex, Vivado, 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.

XAPP589 (v2.3) April 29, 2015 www.xilinx.com 31