Documentation Navigator and Design Hubs...................................................................... 143
Please Read: Important Legal Notices................................................................................. 144
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 4
Chapter 1
SendFeedback
Transceiver and Tool Overview
Introduction to the UltraScale
Architecture
The Xilinx® UltraScale™ architecture is the rst ASIC-class architecture to enable mul-hundred
gigabit-per-second levels of system performance with smart processing, while ecientlyroung
and processing data on-chip. UltraScale architecture-based devices address a vast spectrum of
high-bandwidth, high-ulizaon system requirements by using industry-leading technical
innovaons, including next-generaonroung, ASIC-like clocking, 3D-on-3D ICs, mulprocessor
SoC (MPSoC) technologies, and new power reducon features. The devices share many building
blocks, providing scalability across process nodes and product families to leverage system-level
investment across plaorms.
Virtex® UltraScale+™ devices provide the highest performance and integraoncapabilies in a
FinFET node, including both the highest serial I/O and signal processing bandwidth, as well as
the highest on-chip memory density. As the industry's most capable FPGA family, the Virtex
UltraScale+ devices are ideal for applicaons including 1+ Tb/s networking and data center and
fully integrated radar/early-warning systems.
Virtex® UltraScale™ devices provide the greatest performance and integraon at 20 nm,
including serial I/O bandwidth and logic capacity. As the industry's only high-end FPGA at the
20 nm process node, this family is ideal for applicaons including 400G networking, large scale
ASIC prototyping, and emulaon.
Kintex® UltraScale+™ devices provide the best price/performance/wa balance in a FinFET
node, delivering the most cost-eecvesoluon for high-end capabilies, including transceiver
and memory interface line rates as well as 100G connecvity cores. Our newest mid-range family
is ideal for both packet processing and DSP-intensive funcons and is well suited for applicaons
including wireless MIMO technology, Nx100G networking, and data center.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 5
Chapter 1: Transceiver and Tool Overview
SendFeedback
Kintex® UltraScale™ devices provide the best price/performance/wa at 20 nm and include the
highest signal processing bandwidth in a mid-range device, next-generaon transceivers, and
low-cost packaging for an opmum blend of capability and cost-eecveness. The family is ideal
for packet processing in 100G networking and data centers applicaons as well as DSP-intensive
processing needed in next-generaon medical imaging, 8k4k video, and heterogeneous wireless
infrastructure.
Zynq® UltraScale+™ MPSoC devices provide 64-bit processor scalability while combining real-
me control with so and hard engines for graphics, video, waveform, and packet processing.
Integrang an ARM®-based system for advanced analycs and on-chip programmable logic for
task acceleraon creates unlimited possibilies for applicaons including 5G Wireless, next
generaon ADAS, and Industrial Internet-of-Things.
This user guide describes the UltraScale architecture GTM transceivers and is part of the
UltraScale architecture documentaon suite available at: www.xilinx.com/ultrascale.
Features
The GTM transceiver in the UltraScale+ FPGA is a high performance transceiver, supporng line
rates between 9.8 Gb/s and 58 Gb/s. Based on the available PLL divider conguraons in the
GTM transceivers, the following line rates are supported:
• PAM4 modulaon:
○58 Gb/s – 39.2 Gb/s
○29 Gb/s – 19.6 Gb/s
• NRZ modulaon:
○29 Gb/s – 19.6 Gb/s
○14.5 Gb/s – 9.8 Gb/s
The GTM transceiver is Xilinx’s rst PAM4 enabled transceiver that is highly congurable and
ghtly integrated with the programmable logic resources of the FPGA. The table below
summarizes the features by funconal group that support a wide variety of applicaons.
Table 1: GTM Transceiver Features
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 6
GroupFeature
KP4 Reed-Solomon forward error correction (RS-FEC) for up to 2 x 58 Gb/s or 1 x 116 electrical and optical links
PCS
PRBS generator and checker
Programmable FPGA logic interface
Chapter 1: Transceiver and Tool Overview
SendFeedback
Table 1: GTM Transceiver Features (cont'd)
GroupFeature
LC tank oscillator PLL (LCPLL) for best jitter performance
Flexible clocking with one PLL per Dual (two channels)
Programmable TX output
PMA
Notes:
1.A dual is a cluster or set of two GTM transceiver channels. One GTM_DUAL primitive, one differential reference clock
TX FIR filter with de-emphasis controls
Continuous time linear equalizer (CTLE)
Decision feedback equalization (DFE)
Feed forward equalization (FFE)
pin pair, and analog supply pins. There is no channel primitive.
The GTM transceiver supports NRZ and PAM4
modulaon as well as the following protocols:
• 100GE CAUI2
• 100GE CAUI4
• 200GE CCAUI4
• 400GE (CDAUI8)
• 50GE LAUI
• 50GE LAUI2
• Ethernet AN/LT (auto negoaon/link training)
• OTU4
• Interlaken at 53.125 Gb/s, 25.78125 Gb/s, and 12.5 Gb/s
• CPRI at 48 Gb/s, 24 Gb/s, 12 Gb/s, and 10.1 Gb/s
The rst-me user is recommended to read High-Speed Serial I/O Made Simple, which discusses
high-speed serial transceiver technology and its applicaons. The Xilinx Vivado® IP catalog
includes an UltraScale+ FPGAs GTM Transceivers Wizard to automacally congure GTM
transceivers to support conguraons for dierent protocols and perform custom conguraons.
The GTM transceiver oers a data rate range and features that allow physical layer support for
various protocols. The following gure illustrates the clustering of one GTM_DUAL primive.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 7
IBUFDS_GTM /
SendFeedback
OBUFDS_GTM
REFCLK
Distribution
Chapter 1: Transceiver and Tool Overview
Figure 1: GTM Dual Configuration
GTM_DUAL
GTM Channel 0 (CH0)
TX
RX
LCPLL
GTM Channel 1 (CH1)
TX
RX
X20210-061418
The GTM_DUAL primive contains one LCPLL and two GTM channels. Contrary to other
UltraScale+ device transceivers such as the GTH and GTY transceivers, the GTM transceiver
does not contain channel/common primives. All channel ports and aributes are within the
GTM_DUAL primive. The following gure illustrates the topology of a GTM channel.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 8
Driver
SendFeedback
TX
TX
Pre/
Pre2/
Post
Emp
PISO
Pre-Coder
Chapter 1: Transceiver and Tool Overview
Figure 2: GTM Channel Topology
Pattern
Generator
Gray
Encoder
Polarity
FIFO
FEC
TX
Interface
RX PMA
TX PCS
To RX Parallel
Data (Near-End
PCS Loopback)
Pre-
Coder
RX PCS
Gray
Decoder
Polarity
PRBS
Checker
FIFO
FEC
RX EQSIPOADC
DFE/
FFE
TX PMA
UltraScale+ FPGAs GTM Transceivers
Wizard
From RX Parallel
Data (Far-End
PCS Loopback)
RX
Interface
X20211-052918
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 9
The UltraScale+ FPGAs GTM Transceivers Wizard (hereinaer called the Wizard) is the preferred
tool to generate a wrapper to instanate the GTM_DUAL. The Wizard is located in the IP catalog
under the IO Interfaces category.
RECOMMENDED
: Download the most recent IP update before using the Wizard. Details on how to
use this Wizard can be found in the UltraScale+ FPGAs GTM Transceivers Wizard LogiCORE IP Product
Guide (PG315).
Chapter 1: Transceiver and Tool Overview
SendFeedback
Simulation
The simulaon environment and the test bench must fulll specic prerequisites before running
simulaon using the GTM_DUAL primives. For instrucons on how to set up the simulaon
environment for supported simulators depending on the used hardware descripon language
(HDL), see the latest version of the UltraScale+ GTM Transceivers Wizard LogiCORE IP ProductGuide (PG315) and Vivado Design Suite User Guide: Logic Simulaon (UG900).
The prerequisites for simulang a design with the GTM_DUAL primives are listed:
• A simulator with support for SecureIP models: SecureIP is an IP encrypon methodology.
SecureIP models are encrypted versions of the Verilog HDL used for implementaon of the
modeled block. To support SecureIP models, a simulator that complies with the encrypon
standards described in the Verilog language reference manual (LRM)—IEEE Standard for
Verilog Hardware Descripon Language (IEEE Std 1364-2005) is required.
• A mixed-language simulator for VHDL simulaon: SecureIP models use a Verilog standard. To
use them in a VHDL design, a mixed-language simulator is required. The simulator must be
able to simulate VHDL and Verilog simultaneously.
• An installed GTM transceiver SecureIP model
• The correct setup of the simulator for SecureIP use (inializaonle, environment variables).
• The correct simulator resoluon (Verilog).
Ports and Attributes
There are no simulaon-only ports on the GTM_DUAL primives. The GTM_DUAL primive has
aributes intended only for simulaon. The following table lists the simulaon-only aributes ofthe GTM_DUAL primive. The names of these aributes start with SIM_.
Table 2: GTM_DUAL Simulation-Only Attributes
AttributeTypeDescription
SIM_RESET_SPEEDUPStringIf the SIM_RESET_SPEEDUP attribute is set to TRUE (default), an
SIM_DEVICE
StringThis attribute selects the simulation version to match different
approximated reset sequence is used to speed up the reset time for
simulations, where faster reset times and faster simulation times are
desirable. If the SIM_RESET_SPEEDUP attribute is set to FALSE, the
model emulates hardware reset behavior in detail.
versions of silicon. The default for this attribute is ULTRASCALE_PLUS.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 10
Chapter 1: Transceiver and Tool Overview
SendFeedback
Implementation
It is a common pracce to dene the locaon of GTM transceiver Duals early in the design
process to ensure correct usage of clock resources and to facilitate signal integrity analysis during
board design. The implementaonow facilitates this pracce through the use of locaon
constraints in the XDC le.
The posion of each GTM transceiver Dual primive is specied by an XY coordinate system that
describes the column number and the relaveposion within that column. For a given device/
package combinaon, the transceiver with the coordinates X0Y0 is located at the lowest posion
of the lowest available bank.
There are two ways to create an XDC le for designs that ulize the GTM transceivers. The
preferred method is to use the UltraScale+ FPGAs GTM Transceivers Wizard. The Wizard
automacally generates XDC le templates that congure the transceivers and contain
placeholders for GTM transceiver placement informaon. The XDC les generated by the Wizard
can then be edited to customize operang parameters and placement informaon for the
applicaon.
The second approach is to create the XDC le manually. When using this approach, you must
enter both conguraonaributes that control transceiver operaon as well as the locaon
parameters. Care must be taken to ensure that all of the parameters needed to congure the
GTM transceiver are correctly entered. A GTM_DUAL primive must be instanated as shown in
the following gure.
Figure 3: One-Dual, Two-Channel Configuration (Reference Clock from the LCPLL)
GTM_DUAL
GTM Channel 0 (CH0)
TX
IBUFDS_GTM
LCPLL
GTM Channel 1 (CH1)
RX
TX
RX
X20212-061418
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 11
Each dual contains an LCPLL. Therefore, a reference clock can be connected directly to a
GTM_DUAL primive.
Shared Features
SendFeedback
Reference Clock Input/Output Structure
The reference clock structure in the GTM transceiver supports two modes of operaon: input
mode and output mode. In the input mode of operaon, your design provides a clock on the
dedicated reference clock I/O pins that are used to drive the LCPLL. In the output mode of
operaon, the recovered clocks (RXRECCLK0 and RXRECCLK1) from any of the two channels
within the same Dual can be routed to the dedicated reference clock I/O pins. This output clock
can then be used as the reference clock input at a dierentlocaon. The mode of operaon
cannot be changed during run me.
Chapter 2: Shared Features
Chapter 2
Input Mode
The reference clock input mode structure is illustrated in the following gure. The input is
terminated internally with 50Ω on each leg to MGTAVCC. The reference clock is instanated in
soware with the IBUFDS_GTM sowareprimive. The ports and aributes controlling the
reference clock input are ed to the IBUFDS_GTM soware primive.
Figure 4: Reference Clock Input Structure
MGTAVCC
GTREFCLKP
GTREFCLKN
CEB
I
IB
Nominal
50Ω
Nominal
50Ω
+
-
IBUFDS_GTM
MGTAVCC
/2
1'b0
Reserved
To GTREFCLK or
GTM_DUAL
2'b00
2'b01
2'b10
2'b11
To
HROW
O
ODIV2
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 12
REFCLK_HROW_CK_SEL
X20917-061418
Chapter 2: Shared Features
SendFeedback
Ports and Attributes
The following table denes the reference clock input ports in the IBUFDS_GTM soware
primive.
Table 3: Reference Clock Input Ports (IBUFDS_GTM)
PortDirClock DomainDescription
CEBInN/AThis is the active-Low asynchronous clock enable signal for the clock buffer.
IIn (pad)N/AThese are the reference clock input ports that get mapped to GTREFCLKP.
IBIn (pad)N/AThese are the reference clock input ports that get mapped to GTREFCLKN.
OOutN/AThis output drives the GTREFCLK signal in the GTM_DUAL software primitive.
ODIV2OutN/AThis output can be configured to output either the O signal or a divide-by-2
Setting this signal High powers down the clock buffer.
Refer to Reference Clock Selection and Distribution for more details.
version of the O signal. It can drive the BUFG_GT via the HROW routing. Refer
to Reference Clock Selection and Distribution for more details.
The following table denes the aributes in the IBUFDS_GTM soware primive that congure
the reference clock input.
REFCLK_EN_TX_PATH1-bitReserved. This attribute must always be set to 1'b0.
REFCLK_HROW_CK_SEL2-bitConfigures the ODIV2 output port:
2'b00: ODIV2 = O.
2'b01: ODIV2 = Divide-by-2 version of O.
2'b10: ODIV2 = 1'b0.
2'b11: ODIV2 = Reserved.
REFCLK_ICNTL_RX
2-bitReserved. Use the recommended value from the Wizard.
Output Mode
The reference clock output mode can be accessed via the OBUFDS_GTM soware primive. The
reference clock output mode structure for the OBUFDS_GTM primive is shown in the followinggure. The ports and aributes controlling the reference clock output are ed to the
OBUFDS_GTM sowareprimive.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 13
Chapter 2: Shared Features
SendFeedback
Figure 5: Reference Clock Output Use Model for OBUFDS_GTM
MGTAVCC
GTREFCLKP
GTREFCLKN
O
OB
OBUFDS_GTM
From RXRECCLK0/1
of GTM_DUAL
I
CEB
X20877-061418
Ports and Attributes
The following table denes the ports in the OBUFDS_GTM soware primive.
REFCLK_EN_TX_PATH1-bitReserved. This attribute must always be set to 1’b1.
REFCLK_ICNTL_TX5-bitReserved. Use the recommended value from the Wizard.
Reference Clock Selection and Distribution
The GTM transceivers in Virtex UltraScale+ FPGAs provide dierent reference clock input
opons. Clock selecon and availability is similar to the GTY transceivers in UltraScale+ devices,
but the reference clock selecon architecture supports only one LCPLL shared per Dual (two
GTM transceiver channels).
Chapter 2: Shared Features
SendFeedback
From an architecture perspecve, a Dual contains a grouping of two GTM channels inside one
GTM_DUAL primive, one dedicated external reference clock pin pair, and dedicated reference
clock roung. The reference clock for a GTM_DUAL primive must also be instanated. For duals
operang at line rates lower than 16.3725 Gb/s (NRZ) and 32.7 Gb/s (PAM4), the reference clock
for a Dual can also be sourced from the Dual above via GTNORTHREFCLK or from the Dual
below via GTSOUTHREFCLK. For devices that support stacked silicon interconnect (SSI)
technology, the reference clock sharing via the GTNORTHREFCLK and GTSOUTHREFCLK ports
is limited within its own super logic region (SLR). Duals operang at line rates above
16.3725 Gb/s (NRZ) and 32.7 Gb/s (PAM4) should not source a reference clock from another
Dual.
See the UltraScale device data sheets (see hp://www.xilinx.com/documentaon) for moreinformaon about SSI technology.
Reference clock features include:
• Clock roung for northbound and southbound clocks.
• Flexible clock inputs available for the LCPLL.
• Stac or dynamic selecon of the reference clock for the LCPLL.
The Dual architecture has two GTM transceivers, one dedicated reference clock pin pair, and
dedicated north and south reference clock roung. Each GTM dual has three clock pair inputs
available:
• One local reference clock pin pair, GTREFCLK.
• One reference clock pin pair for the Dual above, GTSOUTHREFCLK.
• One reference clock pin pair from the Dual below, GTNORTHREFCLK.
The gure below shows the detailed view of a reference clock mulplexer structure within a
single GTM_DUAL primive. The PLLREFCLKSEL port is required when mulple reference clock
sources are connected to this mulplexer. A single reference clock is most commonly used. In the
case of a single reference clock, connect the reference clock to the GTREFCLK ports and e the
PLLREFCLKSEL ports to 3’b001. The Xilinx soware tools handle the complexity of the
mulplexers and associated roung.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 15
Each Dual has one set of dedicated dierenal reference clock input pins (MGTREFCLK[P/N])
that can be connected to the external clock sources. In a single external reference clock use
model, an IBUFDS_GTM must be instanated to use the dedicated dierenal reference clock
source. The following gure shows a single external reference clock connected to the LCPLL
inside the Dual. The user design connects the IBUFDS_GTM output (O) to the GTREFCLK ports
of GTM_DUAL.
Figure 7: Single External Reference Clock in a Dual
IBUFDS_GTM
GTM_DUAL
MGTREFCLKP
MGTREFCLKN
I
O
IB
GTREFCLK
X20898-061418
Note: The IBUFDS_GTM diagram in the above gure is a simplicaon. The output port ODIV2 is le
oang, and the input port CEB is set to logic 0.
The following gure shows a single external reference clock with mulple Duals connected. The
user design connects the IBUFDS_GTM output (O) to the GTREFCLK ports of the GTM_DUAL
primives. In this case, the Xilinx implementaon tools make the necessary adjustments to the
north/south roung as well as the pin swapping necessary to route the reference clock from one
Dual to another when required.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 16
Chapter 2: Shared Features
SendFeedback
Figure 8: Single External Reference Clock with Multiple Duals
D(n+1)
GTM_DUAL
GTREFCLK
D(n)
GTM_DUAL
GTREFCLK
D(n-1)
GTM_DUAL
GTREFCLK
X20215-061418
MGTREFCLKP
MGTREFCLKN
IBUFDS_GTM
I
O
IB
Note: The IBUFDS_GTM diagram in the above gure is a simplicaon. The output port ODIV2 is le
oang, and the input port CEB is set to logic 0.
These rules must be observed when sharing a reference clock to ensure that jier margins for
high-speed designs are met:
• The number of Duals above the sourcing Dual must not exceed one.
• The number of Duals below the sourcing Dual must not exceed one.
• The total number of Duals sourced by an external clock pin pair (MGTREFCLKP/
MGTREFCLKN) must not exceed three Duals.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 17
The maximum number of Duals that can be sourced by a single clock pin pair is three (six
transceivers). Designs with more than three Duals require the use of mulple external clock pins
to ensure that the rules for controlling jier are followed. When mulple clock pins are used, an
external buer can be used to drive them from the same oscillator.
IMPORTANT
! Upon device conguraon, the clock output from the IBUFDS_GTM which takes inputs
from MGTREFCLKP and MGTREFCLKN can only be used as long as the GTPOWERGOOD signal has
already asserted High.
Ports and Attributes
The following table denes the clocking ports and aributes for the GTM_DUAL primive.
Table 7: GTM_DUAL Clocking Ports
SendFeedback
Chapter 2: Shared Features
Port
PLLREFCLKSEL[2:0]InAsyncInput to dynamically select the input reference clock to the
GTGREFCLK2PLLInClockReference clock generated by the internal interconnect logic.
GTREFCLKInClockExternal clock driven by IBUFDS_GTM for the LCPLL.
GTNORTHREFCLKInClockNorthbound clock from the Dual below.
GTSOUTHREFCLKInClockSouthbound clock from the Dual above.
PLLREFCLKLOSTOutAsyncA High on this signal indicates that the reference clock to the
This input is reserved for internal testing purposes only.
phase frequency detector of the LCPLL is lost.
LCPLL
Each Dual contains one LC-based PLL, referred to as LCPLL, and cannot be shared with
neighboring Duals. The internal clocking architecture of the GTM Dual is shown in the following
gure.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 18
REFCLK Distribution
SendFeedback
Chapter 2: Shared Features
Figure 9: Internal Dual Clocking Architecture
GTM_DUAL
GTM Channel 0 (CH0)
TX PMA
TX PCS
RX PMA
RX PCS
LCPLL
GTM Channel 1 (CH1)
TX PMA
TX PCS
RX PMA
RX PCS
X20900-061418
The LCPLL input clock selecon is described in Reference Clock Selecon and Distribuon. The
LCPLL outputs feed the TX and RX clock divider blocks, which control the generaon of serial
and parallel clocks used by the PMA and PCS blocks. The LCPLL is shared between the TX and
RX datapaths.
The gure below illustrates a conceptual view of the LCPLL architecture. The input clock can be
divided by a factor of M before it is fed into the phase frequency detector. The feedback divider
N determines the voltage-controlled oscillator (VCO) mulplicaonrao. For line rates below
28.1 Gb/s (NRZ) and 56.2 Gb/s (PAM4), a fraconal-N divider is supported where the eecverao is a combinaon of the N factor plus a fraconal part. The LCPLL output frequency depends
on the sengs of LCPLLCLKOUT_RATE. When LCPLLCLKOUT_RATE is set to HALF, the output
frequency is half of the VCO frequency. When it is set to FULL, the output frequency is the same
as the VCO frequency. A lock indicator block compares the frequencies of the reference clock
and VCO feedback clock to determine if a frequency lock has been achieved.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 19
PLL
f
PLLClkout
= f
PLLClkin
*
N + FractionalDivider
M*LCPLLCLKOUT _ RATE
f
LineRate
= f
PLLClkout
*Modulation
FractionalDivider = N
SDM
+ 0. < FractionalPart >
FractionalPart
=
SDMDAT A
2
SDMW IDTH
SendFeedback
CLKIN
/M
Figure 10: Internal Dual Clocking Architecture
Lock
Indicator
Phase
Frequency
Detector
Charge
Pump
Loop
Filter
VCO
/2
Chapter 2: Shared Features
PLL
LOCKED
PLL
CLKOUT
/N-Fractional
LCPLLCLKOUT_RATE
The LCPLL VCO operates within 9.8 GHz—14.5 GHz. The Xilinx soware tool chooses the
appropriate LCPLL seng based on applicaon requirements. Equaon 2-1 shows how to
determine the PLL output frequency (GHz).
Equaon 2-2 shows how to determine the line rate (Gb/s).
Equaon 2-3 and Equaon 2-4 show how to determine the fraconal divider presented in
Equaon 2-1.
PLLRESETInAsyncThis port is driven High and then deasserted to start the
PLLRESETBYPASSMODEInAsyncReserved. Tied Low.
PLLRESETDONEOutcfg_mclkStatus signal that indicates when the PLL reset sequence is
PLLRESETMASK[1:0]InAsyncBit 0 enables bit mask for PLL reset. Bit 1 enables bit mask
PLLRSVDIN[15:0]InReservedReserved. This port must be set to 0x0000.
PLLRSVDOUTOutAsyncReserved.
BGBYPASSBInAsyncReserved. This port must be set to 1’b1. Do not modify this
BGMONITORENBInAsyncReserved. This port must be set to 1’b1. Do not modify this
BGPDBInAsyncReserved. This port must be set to 1’b1. Do not modify this
BGRCALOVRD[4:0]InAsyncReserved. This port must be set to 5’b11111. Do not modify
BGRCALOVRDENBInAsyncReserved. This port must be set to 1’b1. Do not modify this
InAsyncAn active-High signal powers down the LCPLL.
Clock
Domain
Description
logic. This input is reserved for internal testing purposes.
LCPLL feedback divider to the phase frequency detector of
the LCPLL is lost
is PLLFBDIV + 2. Valid values are from 14–158. (Actual divider
values are 160–160.)
the LCPLL frequency is within a predetermined tolerance.
The transceiver and its clock outputs are not reliable until
this condition is met.
reference clock signals to the LCPLL. The input reference
clock to the LCPLL or any output clock generated from the
LCPLL must not be used to drive this clock. This clock is
required only when using the PLLFBCLKLOST and
PLLREFCLKLOST ports. It does not affect the LCPLL lock
detection, reset, and power-down functions.
phase frequency detector of the LCPLL is lost.
LCPLL reset.
complete.
for PLL SDM reset.
value.
value.
value.
this value.
value.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 21
Table 9: LCPLL Ports (cont'd)
SendFeedback
Chapter 2: Shared Features
PortDirection
RCALENInAsyncReserved. This port must be set to 1’b1. Do not modify this
SDMDATA[25:0]InAsyncInput to set the Fractional Divider. Bits [SDMWIDTH +
SDMTOGGLEInAsyncReserved. Set to 1’b0.
Clock
Domain
Description
value.
1:SDMWIDTH] are the integer part of the divider in two’s
complement. Bits [SDMWIDTH – 1:SDMWIDTH] set the
fractional part of the divider.
Table 10: LCPLL Attributes
AttributesTypeDescription
BIAS_CFG016-bitReserved. Use the recommended value from the Wizard.
BIAS_CFG116-bitReserved. Use the recommended value from the Wizard.
BIAS_CFG216-bitReserved. Use the recommended value from the Wizard.
BIAS_CFG316-bitReserved. Use the recommended value from the Wizard.
BIAS_CFG416-bitReserved. Use the recommended value from the Wizard.
A_CFG16-bitReserved. Use the recommended value from the Wizard.
CRS_CTRL_CFG016-bitReserved. Use the recommended value from the Wizard.
CRS_CTRL_CFG116-bitReserved. Use the recommended value from the Wizard.
DRPEN_CFG16-bitReserved. Use the recommended value from the Wizard.
PLL_CFG016-bitReserved. Use the recommended value from the Wizard.
Bit NameAddressDescription
LCPLLCLKOUT_RATE[8]Sets the LCPLLCLKOUT_RATE factor either to provide full
LCPLL VCO frequency, or half of LCPLL VCO frequency at the
output:
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 22
1'b0: Half rate.
1'b1: Full rate.
PLL_CFG1
Bit NameAddressDescription
LCPLL_REFCLK_DIV[4:0]LCPLL reference clock divider M settings.
PLL_CFG2
PLL_CFG316-bitReserved. Use the recommended value from the Wizard.
PLL_CFG416-bitReserved. Use the recommended value from the Wizard.
PLL_CFG516-bitReserved. Use the recommended value from the Wizard.
PLL_CFG616-bitReserved. Use the recommended value from the Wizard.
16-bitReserved. Use the recommended value from the Wizard.
5'b00000: Div by 2.
5'b00001: Div by 3.
5'b00010: Div by 4.
5'b10010: Div by 1.
Other values are not valid.
16-bitReserved. Use the recommended value from the Wizard.
Chapter 2: Shared Features
SendFeedback
Table 10: LCPLL Attributes (cont'd)
AttributesTypeDescription
SDM_CFG0[15:0]16-bitReserved. Use the recommended value from the Wizard.
Bit NameAddressDescription
SDM_WIDTHSEL[10:9]This attribute sets the denominator of the fractional part of
SDM_CFG216-bitReserved. Use the recommended value from the Wizard.
SDM_SEED_CFG016-bitReserved. Use the recommended value from the Wizard.
SDM_SEED_CFG116-bitReserved. Use the recommended value from the Wizard.
A_SDM_DATA_CFG016-bitReserved. Use the recommended value from the Wizard.
A_SDM_DATA_CFG116-bitReserved. Use the recommended value from the Wizard.
16-bitReserved. Use the recommended value from the Wizard.
Reset and Initialization
The GTM transceiver must be inializedaer device power-up and conguraon before it can be
used. The GTM transmier (TX) and receiver (RX) can be inialized independently and in parallel
as shown in the following gure. The GTM transceiver TX and RX inializaon comprises three
steps:
1. Inializing the associated PLL driving TX/RX
2. Inializing the TX and RX datapaths (PMA + PCS)
The TX and RX in the GTM transceiver receive the clock through the LCPLL in the transceiver's
own Dual. In the power-on inializaon sequence, the LCPLL used by the TX and RX must beinializedrst. The LCPLL used by the TX and RX is reset individually and its reset operaon
independent from TX and RX resets. The TX and RX datapaths must be inialized only aer the
associated LCPLL is locked.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 23
The GTM transceiver TX and RX use a state machine to control the inializaon process. They
are paroned into a few reset regions. The paron allows the reset state machine to control
the reset process in a sequence that the PMA can be reset rst and the PCS can be reset aer
the asseron of the TXUSERRDY or RXUSERRDY. It also allows the PMA, the PCS, and
funconal blocks inside them to be reset individually when needed during normal operaon.
The GTM transceiver oers two types of reset: inializaon and component.
• Inializaon Reset: This reset is used for complete GTM transceiver inializaon. It must be
used aer device power-up and conguraon. During normal operaon, when necessary,
GTTXRESET and GTRXRESET can also be used to reinialize the GTM transceiver TX and RX.
GTTXRESET is the inializaon reset port for the GTM transceiver TX. GTRXRESET is the
inializaon reset port for the GTM transceiver RX. During inializaon reset,
TXRESETMODE and RXRESETMODE should be set to sequenal mode. All TX PMA, TX PCS,
RX PMA, and RX PCS component resets should be enabled by seng all required component
bits of TXPMARESETMASK, TXPCSRESETMASK, RXPMARESETMASK, and
RXPCSRESETMASK to High.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 24
• Component Reset: This reset is used for special cases and specicsubsecon resets while the
GTM transceiver is in normal operaon. The component that is required to be reset is selected
by seng the associated bit within TXPMARESETMASK, TXPCSRESETMASK,
RXPMARESETMASK, or RXPCSRESETMASK to High. A TX component reset is triggered by
toggling the GTTXRESET port. An RX component reset is triggered by toggling the
GTRXRESET port. Separate component reset ports are available. For the TX, these are
TXCKALRESET, TXFECRESET, TXPCSRESET, and TXPMARESET. For the RX, these are
RXADAPTRESET, RXADCCLKGENRESET, RXBUFRESET, RXCDRFRRESET, RXCDRPHRESET,
RXDFERESET, RXDSPRESET, RXEYESCANRESET, RXFECRESET, RXPCSRESET,
RXPMARESET, and RXPRBSCSCNTRST.
Chapter 2: Shared Features
SendFeedback
All reset ports described in this seconiniate the internal reset state machine when driven
High. The internal reset state machines are held in the reset state unl these same reset ports are
driven Low. These resets are all asynchronous. The guideline for the pulse width of these
asynchronous resets is one period of the reference clock, unless otherwise noted.
Note: Do not use reset ports for the purpose of power down. For details on proper power-down usage,
refer to Power Down.
Resetting Multiple Lanes and Quads
Reseng mulple lanes in a Dual or mulple Duals aects the power supply regulaon circuit.
Reset Modes
The GTM transceiver TX and RX resets can operate in two dierent modes: sequenal mode and
single mode.
• Sequenal mode: The reset state machine starts with an inializaon or component reset
input driven High and proceeds through all states aer the requested reset states in the reset
state machine, as shown in Figure 13 for the GTM transceiver TX or Figure 18 for the GTM
transceiver RX unlcompleon. The compleon of sequenal mode reset ow is signaled
when (TX/RX)RESETDONE transions from Low to High.
• Single mode: The reset state machine only executes the requested component reset
independently for a predetermined me set by its aribute. It does not process any state aer
the requested state, as shown in Figure 13 for the GTM transceiver TX or Figure 18 for the
GTM transceiver RX. The requested reset can be any component reset to reset the PMA, the
PCS, or funconal blocks inside them. The compleon of a single mode reset is signaled when
(TX/RX)RESETDONE transions from Low to High.
The GTM transceiver inializaon reset must use sequenal mode. All component resets can be
operated in either sequenal mode or single mode. The GTM transceiver uses (TX/
RX)RESETMODE to select between sequenal reset mode and single reset mode. The following
table provides conguraon details that apply to both the GTM transceiver TX and GTM
transceiver RX. Reset modes have no impact on LCPLL reset. During normal operaon, the GTM
transceiver TX or GTM transceiver RX can be reset by applicaons in either sequenal mode or
single mode (GTM transceiver RX only), which provides exibility to reset a poron of the GTM
transceiver. When using either sequenal mode or single mode, (TX/RX)RESETMODE must be
set to the desired value of 50 ns before the asserons of any reset.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 25
Table 11:
GTM Transceiver Reset Modes Operation
Operation Mode(TX/RX)RESETMODE
Sequential Mode
Single Mode
2'b00
2'b11
Table 12: GTM Transceiver Reset Modes Ports
SendFeedback
PortDirectionClock DomainDescription
CH[0/1]_RXRESETMODE[1:0]InAsyncReset mode port for RX.
2'b00: Sequential mode
(recommended).
2'b01: Reserved.
2'b10: Reserved.
2'b11: Single mode.
Chapter 2: Shared Features
CH[0/1]_TXRESETMODE[1:0]
InAsyncReset mode port for TX.
2'b00: Sequential mode
(recommended).
2'b01: Reserved.
2'b10: Reserved.
2'b11: Single mode.
LCPLL Reset
The LCPLL must be reset before it can be used. Each GTM transceiver dual has dedicated reset
ports for its LCPLL. As shown in the gure, PLLRESET is an input that resets LCPLL. PLLLOCK is
an output that indicates the reset process is done. The guideline for this asynchronous PLLRESET
pulse width is one period of the reference clock. Aer a PLLRESET pulse, the internal reset
controller generates an internal LCPLL reset followed by an internal SDM reset. The me
required for LCPLL to lock is aected by a few factors, such as bandwidth seng and clock
frequency.
Figure 12: LCPLL Reset Timing Diagram
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 26
Table 13: LCPLL Reset Ports
PortDir
PLLRESETInAsyncActive-High signal that resets the LCPLL.
PLLRESETMASK[1:0]InAsyncReserved. Tied to 2’b11.
PLLRESETBYPASSMODEInAsyncReserved. Tied Low.
PLLLOCKOutAsyncActive-High LCPLL frequency lock signal indicates that the
Clock
Domain
Description
LCPLL frequency is within a predetermined tolerance. The
GTM transceiver and its clock outputs are not reliable until
this condition is met.
Chapter 2: Shared Features
SendFeedback
TX Initialization and Reset
The GTM transceiver TX uses a reset state machine to control the reset process. The GTM
transceiver TX is paroned into two reset regions, TX PMA and TX PCS. The paron allows TX
inializaon and reset to be operated only in sequenal mode, as shown in the gure below.
The inializing TX must use GTTXRESET in sequenal mode. Acvang the GTTXRESET input
can automacally trigger a full asynchronous TX reset. The reset state machine executes the
reset sequence, as shown in the gure below, covering the whole TX PMA and TX PCS. During
normal operaon, when needed, sequenal mode allows you to reset the TX from acvang
TXPMARESET and connue the reset state machine unl TXRESETDONE transions from Low
to High.
The TX reset state machine does not reset the PCS unl TXUSERRDY is detected High. Drive
TXUSERRDY High aer all clocks used by the applicaon including TXUSRCLK are shown as
stable.
Figure 13: GTM Transceiver TX Reset State Machine Sequence
GTTXRESET
High
Wait until
GTTXRESET from
High to Low
TXPMARESETMASK[0]
= 1?
No
TXPMARESETMASK[1]
= 1?
No
Wait for TXUSERRDY
= 1
Yes
Yes
TX CKCAL Reset
TX PMA Top Reset
TXPCSRESETMASK[0]
= 1?
No
TXPCSRESETMASK[1]
= 1?
No
TXRESETDONE High
Yes
Yes
TX FEC Reset
TX PCS Top Reset
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 27
X20905-060518
Ports and Attributes
SendFeedback
The table below lists ports required by the TX inializaon process.
Table 14: TX Initialization and Reset Ports
Chapter 2: Shared Features
PortDir
CH[0/1]_GTTXRESETInAsyncThis port is driven High and then deasserted to start a TX reset
CH[0/1]_TXUSERRDYInAsyncThis port is driven High by the user application when TXUSRCLK
CH[0/1]_TXPMARESETDONEOutAsyncThis active-High signal indicates TX PMA reset is complete.
CH[0/1]_TXRESETDONEOutTXUSRCLK This active-High signal indicates the GTM transceiver TX has
CH[0/1]_ TXCKALRESETInAsyncThis port is driven High and then deasserted to start a single
CH[0/1]_ TXPMARESETInAsyncThis port is driven High and then deasserted to start a single
CH[0/1]_ TXFECRESETInAsyncThis port is driven High and then deasserted to start a single
CH[0/1]_ TXPCSRESETInAsyncThis port is driven High and then deasserted to start a single
InAsyncReset mode port for TX:
InAsyncTX PMA reset mask selection:
Clock
Domain
Description
sequence. The components to be reset are determined by
TXPMARESETMASK and TXPCSRESETMASK. In sequential mode,
the resets are performed sequentially. In single mode, the
resets are performed simultaneously.
finished reset and is ready for use. This port is driven Low when
GTTXRESET goes High and is not driven High until the GTM
transceiver has completed all TX reset steps.
mode reset on TX CKCAL. The reset is not dependent on
TXRESETMODE or TXPMARESETMASK setting.
mode reset on TX PMA. The reset is not dependent on
TXRESETMODE or TXPMARESETMASK setting.
mode reset on TX FEC. The reset is not dependent on
TXRESETMODE or TXPCSRESETMASK setting.
mode reset on TX PCS. The reset is not dependent on
TXRESETMODE or TXPCSRESETMASK setting.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 28
Chapter 2: Shared Features
SendFeedback
The following table lists aributes required by GTM transceiver TX inializaon. In general cases,
the reset me required by the TX PMA or the TX PCS varies depending on line rate. The factors
aecng PMA reset me and PCS reset me are the user-congurableaributes
TX_PMA_RESET_TIME, TX_PCS_RESET_TIME, TX_CKCAL_RESET_TIME, and
TX_FEC_RESET_TIME.
Table 15: TX Initialization and Reset Attributes
AttributeTypeDescription
CH[0/1]_RST_TIME_CFG016-bitReserved.
Bit NameAddressDescription
TX_PCS_RESET_TIME[14:10]Represents the time duration to apply a TX PCS reset. Use
TX_PMA_RESET_TIME[9:5]Represents the time duration to apply a TX PMA reset. Use
TX_CKCAL_RESET_TIME[4:0]Represents the time duration to apply a TX CLKGEN reset.
CH[0/1]_RST_TIME_CFG116-bitReserved.
Bit NameAddressDescription
TX_FEC_RESET_TIME[4:0]Represents the time duration to apply a TX CKCAL reset.
CH[0/1]_RST_LP_CFG016-bitReserved.
Bit NameAddressDescription
TX_PCS_RESET_LOOP_ID[11:8]Reserved. Use the recommended value from the Wizard.
TX_PMA_RESET_LOOP_ID[7:4]Reserved. Use the recommended value from the Wizard.
TX_CKCAL_RESET_LOOP_ID[3:0]Reserved. Use the recommended value from the Wizard.
TX_FEC_RESET_LOOP_ID[15:12]Reserved. Use the recommended value from the Wizard.
CH[0/1]_RST_LP_ID_CFG116-bitReserved.
Bit NameAddressDescription
TX_PCS_LOOPER_END_ID[15:12]Reserved. Use the recommended value from the Wizard.
TX_PCS_LOOPER_START_ID[11:8]Reserved. Use the recommended value from the Wizard.
TX_PMA_LOOPER_END_ID[7:4]Reserved. Use the recommended value from the Wizard.
TX_PMA_LOOPER_START_ID[3:0]Reserved. Use the recommended value from the Wizard.
CH[0/1]_RST_LP_CFG41-bitReserved.
Bit NameAddressDescription
BYP_HDSHK_TX_PCS_RESET_LOOP[3]Reserved. Use the recommended value from the Wizard.
BYP_HDSHK_TX_CKCAL_RESET_LOOP[0]Reserved. Use the recommended value from the Wizard.
the recommended value from the Wizard. Must be a nonzero value when TXPCSRESETMASK[1] is High and
GTTXRESET initiates the reset process.
the recommended value from the Wizard. Must be a nonzero value when TXPMARESETMASK[1] is High and
GTTXRESET initiates the reset process.
Use the recommended value from the Wizard. Must be a
non-zero value when TXPMARESETMASK[0] is High and
GTTXRESET initiates the reset process.
Use the recommended value from the Wizard. Must be a
non-zero value when TXPMARESETMASK[0] is High and
GTTXRESET initiates the reset process.
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 29
Chapter 2: Shared Features
SendFeedback
GTM Transceiver TX Reset in Response to Completion
of Configuration
The TX reset sequence shown in TX Inializaon and Reset is not automacally started to follow
global GSR. It must meet these condions:
1. TXRESETMODE must be set to sequenal mode.
2. GTTXRESET must be used.
3. All TXPMARESETMASK and TXPCSRESETMASK bits should be set to High.
4. GTTXRESET cannot be driven Low unl the associated PLL is locked.
5. Ensure that GTPOWERGOOD is High before releasing PLLRESET and GTTXRESET.
If the reset mode is defaulted to single mode, then you must:
1. Wait another 300–500 ns.
2. Assert PLLRESET and GTTXRESET following the reset sequence described in the following
gure.
RECOMMENDED: Use the associated PLLLOCK from the PLL to release GTTXRESET from High to
Low as shown in the gure. The TX reset state machine waits when GTTXRESET is detected High and
starts the reset sequence unl GTTXRESET is released Low.
Figure 14: GTM Transmitter Initialization after Configuration
GTM Transceiver TX Reset in Response to GTTXRESET
Pulse in Full Sequential Reset
The GTM transceiver allows you to reset the enre TX completely at any me by sending
GTTXRESET an acve-High pulse. These condions must be met when using GTTXRESET:
UG581 (v1.0) January 4, 2019www.xilinx.com
Virtex UltraScale+ GTM Transceivers 30
1. TXRESETMODE must be set to sequenal mode.
2. All TXPMARESETMASK and TXPCSRESETMASK bits should be held High during the reset
sequence before TXRESETDONE is detected High.
Loading...
+ 115 hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.