XilinxSpartan6FPGAserdes应用指南_PLL

需积分: 49 98 浏览量 2020-08-24 22:18:24 上传评论 1 收藏 877KB PDF 举报

Spartan-6 FPGA系列是Xilinx公司推出的一款FPGA（Field Programmable Gate Array）产品，其中集成的SerDes（Serializer-Deserializer）模块是实现高速数据传输的关键组件。本应用指南详细阐述了如何有效地利用这些SerDes块进行高速串行化和解串行化电路设计，特别适用于连接高速ADC（Analog-to-Digital Converter）和DAC（Digital-to-Analog Converter）等应用场景。 1. SerDes功能与应用： SerDes的主要作用是将并行数据转换为串行数据，并在接收端再将串行数据转换回并行，从而在保持高速数据传输的同时减少信号线的数量，提高系统性能。Spartan-6 FPGA中的ISERDES用于输入，OSERDES用于输出，两者都可以实现2:1到8:1的速率比，支持单数据率和双数据率I/O时钟。 2. 输入延迟块与相位检测器：在源同步系统中，数据和时钟需要边缘对齐。为了确保正确同步，设计通常需要输入延迟块和相位检测器来调整数据和时钟之间的相位关系。这些组件有助于在高速传输中实现精确的时序匹配，确保数据的准确接收和发送。 3. ISERDES与OSERDES配置：每个Spartan-6 FPGA的输入/输出块（IOB）包含一个4位的ISERDES和一个4位的OSERDES。通过级联相邻的主从ISERDES或OSERDES块，可以构建8位的串行/并行转换单元。级联使得更灵活的速率比成为可能，但也带来了一些限制，特别是在使用单端信号标准和特定模式时。 4. 级联ISERDES的注意事项：使用差分信号标准时，级联ISERDES不会存在问题，因为这些标准通常会使用两个IOB中的SerDes寄存器。然而，对于单端信号标准，当需要5:1或更高的SerDes比率或相位检测器模式时，不能让两个数据线路进入设备的相邻主从IOB。如果SerDes比率不超过4:1且不使用相位检测器模式，虽然可以避免这个问题，但可能会导致数据丢失。 5. 设计考虑：设计者在使用Spartan-6 FPGA的SerDes时，需要考虑信号完整性、眼图分析、抖动管理以及噪声抑制等因素，以确保在高速传输下的可靠性和稳定性。此外，还需要根据具体应用选择合适的串行协议，如PCIe、Gigabit Ethernet、USB等，这些协议往往提供了内置的错误检测和校正机制。 Xilinx Spartan-6 FPGA的SerDes功能提供了一种高效的方法来处理高速数据传输，但设计者需要充分理解其工作原理、限制条件和最佳实践，才能充分利用这些资源，实现高性能、低功耗的设计。本应用指南作为官方文档，旨在帮助工程师更好地理解和应用Spartan-6 FPGA的SerDes特性。

资源详情

资源评论

XAPP1064 (v1.2) November 19, 2013 www.xilinx.com 1

United States and other countries. All other trademarks are the property of their respective owners.

Summary Spartan®-6 devices contain input SerDes (ISERDES) and output SerDes (OSERDES) blocks.

These primitives simplify the design of serializing and deserializing circuits, while allowing

higher operational speeds. This application note discusses how to efficiently use these

primitives in conjunction with the input delay blocks and phase detector circuitry. These designs

are for use in source-synchronous systems where the data and clock are edge aligned.

Different techniques are required when the clock and data are center aligned.

ISERDES and

OSERDES

Guidelines

Each Spartan-6 FPGA input/output block (IOB) contains a 4-bit input SerDes and a 4-bit output

SerDes. The SerDes from two adjacent blocks (master and slave) can be cascaded to make an

8-bit block. This gives the possibility of SerDes ratios from 2:1 to 8:1 on both output and input

for both single and double data rate I/O clocks.

Cascading the ISERDES blocks is not an issue when a differential signaling standard is being

used because these standards use the two IOBs (master and slave) associated with the two

sets of SerDes registers. Thus, using two ISERDES effectively reduces design cost. However,

when using a single-ended signaling standard, some care needs to be taken when the design

requires either a SerDes ratio of five or more or the phase detector mode. Specifically, two data

lines cannot enter the device in adjacent master and slave IOBs when using cascaded SerDes.

This limitation is not necessary when the SerDes ratio is four or less and the phase detector

mode is not being used because the SerDes is not cascaded. However, by not using the phase

detector mode, data loss will occur during calibration and the application will need to account

for this loss.

Introduction to

Deserialization

and Data

Reception

A deserializer design and its associated clocking primitives are dependent on the format of the

incoming receive data stream. This data tends to fall into three categories.

Case 1

The data stream is a multiple of the rate of the incoming clock, and the clock signal is used as

a framing signal for the received data. Multiple changes in the state of the data lines occur

during one clock period. A widely used example is the 7:1 interface used in cameras and flat

panel TVs and monitors. Other ratios are obviously possible, and the Spartan-6 FPGA

ISERDES can support ratios of 2, 3, and 4:1, and also 5, 6, 7, and 8:1 when cascaded. In this

example, the received clock is multiplied in a PLL, and the resultant high-speed capture clock

is passed to the input logic through the BUFPLL primitive. The BUFPLL capture clock is

designed to always be used in single data rate (SDR) mode with respect to the input data. For

example, a 150 MHz input clock with accompanying 7:1 data requires the PLL and BUFPLL to

operate at 1050 MHz (equals 150 x 7). This high-speed capture clock is used to clock the

receive data into the input deserializers and is capable of driving one whole edge of a device.

Parallel data is then presented to the FPGA logic at the speed of the original incoming clock.

Figure 1 shows this 7:1 data formatting example.

Application Note: Spartan-6 FPGAs

XAPP1064 (v1.2) November 19, 2013

Source-Synchronous Serialization and

Deserialization (up to 1050 Mb/s)

Author: NIck Sawyer

Introduction to Deserialization and Data Reception

XAPP1064 (v1.2) November 19, 2013 www.xilinx.com 2

Case 2

The data stream is a multiply by two of the incoming clock, commonly called Double Data Rate

(DDR) reception. A DDR data stream is shown in Figure 2. Each transition of the clock

indicates a change in the state of a data line. There are two ways of receiving this kind of data.

The first is to use a PLL and a BUFPLL (see Case 1), where the PLL is being used to multiply

the incoming clock by two and the BUFPLL allows use of the whole edge of a device. The other

method is to use the BUFIO2 primitive, where two BUFIO2s are required to receive DDR data,

and the BUFIO2s are only able to drive the same half edge of a device where the clock input is

located. The deserialization factor (ratio) can be chosen by the designer (values of 2, 4, 6, and

8:1). The necessary divided clock for the parallel data is generated through one of the BUFIO2

primitives. Two BUFIO2s must be used to multiply the incoming DDR clock by two to generate

an SDR capture clock.

Case 3

The data stream is at the same rate as the receiver clock (SDR). A drawing of an SDR data

stream is shown in Figure 3. Each data bit changes every two clock transitions, normally on the

rising edge of the clock. There are two ways of receiving this kind of data stream. The received

clock is multiplied by one in a PLL and the BUFPLL is used to receive data on a whole edge of

a device, or a single BUFIO2 or PLL is used to drive the inputs in the half edge of the device

where the clock input is situated. The BUFIO2 is also used to divide down the received clock to

be used with the deserialized parallel data. Using SDR, ratios of 2, 3, 4, 5, 6, 7, and 8:1 are

possible.

X-Ref Target - Figure 1

Figure 1: Data Stream Using a Low-Speed Clock with a 7:1 SerDes Ratio

X-Ref Target - Figure 2

Figure 2: 8:1 Data Stream Using DDR

X-Ref Target - Figure 3

Figure 3: 8:1 Data Stream Using SDR

Receive 7 to 1 Data

Receive Clock

D0 D1 D2 D3 D4 D5 D6

x1064_01_092809

Receiver 8:1 Data

Received Clock

(Clock polarity is not

important for DDR)

D0 D1 D2 D3 D4 D5 D6 D7

x1064_02_100509

Receiver 8:1 Data

Received Clock

(Clock rising edge

indicates the start of a data bit)

D0 D1 D2 D3 D4 D5 D6 D7

x1064_03_100509

Introduction to Deserialization and Data Reception

XAPP1064 (v1.2) November 19, 2013 www.xilinx.com 3

Higher Deserialization Factors

Deserialization using factors greater than 8:1 is possible when receiving data. The PLL can be

used to generate a third clock, which is intermediate to the high-speed I/O capture clock and

the low-speed parallel data clock. Examples of designs using SerDes ratios of 10, 12, 14, and

16:1 are included in the Reference Design Files. Essentially, the input SerDes primitives are

still used in 5, 6, 7, and 8:1 modes, receiving data through a high-speed capture clock from the

PLL through the BUFPLL. The received parallel data is transferred to the FPGA logic in the

intermediate clock domain and then further transferred to the main clock domain using a 2:1

gearbox, also in the FPGA logic. A drawing of the mechanism is shown in Figure 4. In any of

these examples of higher deserialization factors where the PLL is used, the receiver clock can

be SDR, DDR, or a divided clock.

Data Reception Using PLL and BUFPLL

The topology for data reception using PLL and BUFPLL is uncomplicated. The receiver clock is

multiplied as required in the PLL to generate an internal single data-rate capture clock. The

incoming clock signal needs to pass through an input delay block (to balance datapath delays)

and a BUFIO2 to reach the PLL. In the 7:1 video example, the received pixel clock must be

multiplied by seven. The clock signal is routed from the PLL to a BUFPLL primitive to drive one

whole edge of the device. CLKOUT0 and CLKOUT1 are the only outputs of the PLL that are

capable of driving high-speed capture clocks to the BUFPLL. The BUFPLL also requires a

global clock signal equal to the original non-multiplied source clock, which can be driven from

any of the PLL outputs through a global buffer (BUFG), and the LOCKED signal from the PLL,

which is required for synchronization inside the BUFPLL.

X-Ref Target - Figure 4

Figure 4: Receiving Data at Higher SerDes Factors

serdes_1_to_n_clk_pll_s16_diff.v/vhd

Differential Clock IOB

Calibration and

Bitslip State

Machine

Master

ISERDES

Slave

ISERDES

Master

IDELAY

Slave

IDELAY

(0)

PLL

BITSLIP

BUFPLL

BUFG

BUFIO2

BUFIO2FB

BUFG

BITSLIP

LVDS Clock P

Parameters:

S = serdes_factor

PLLX = PLL multiplier

CLKIN_PERIOD = Input clock period

BS = TRUE/FALSE to use Bitslip (or not)

Parameters:

D = Number of data lines

S = serdes_factor

LVDS Clock N

x1064_04_100509

Feedback

x1 Clock x2 Clock

SerDes

Strobe

I/O Clock

Feedback

serdes_1_to_n_data_s16_diff.v/vhd

“D” Differential Data IOB

Calibration and

Phase Detection

State Machine

Master

ISERDES

Slave

ISERDES

Master

IDELAY

Slave

IDELAY

BITSLIP

D x S

LVDS Data P

LVDS Data N

Gearbox

DxS

Introduction to Deserialization and Data Reception

XAPP1064 (v1.2) November 19, 2013 www.xilinx.com 4

The three input signals to the BUFPLL allow the BUFPLL to distribute the high-speed receiver

clock to the input delay and SerDes primitives in the same edge of the device, along with the

required SerDes strobe signal (appropriately aligned) that allows safe transfer of low-speed

parallel data to the FPGA logic from the input SerDes.

When using the PLL for data reception, PLL deskew is required. The feedback clock signal is

routed from an I/O clock destination at the input SerDes primitive of the clock input pin back to

the PLL using a BUFIO2FB primitive. This mechanism forces the multiplied clock to be in the

same phase as the original received clock.

The mechanism for centering and capturing data reliably is based on the IODELAY2 primitive,

used in input delay mode only. This mechanism is discussed in the Delaying Input Data and

Clocks section. The block diagram of the receiver is shown in Figure 5.

DDR Data Reception Using Two BUFIO2s

The topology for DDR data reception using two BUFIO2s uses the incoming clock to directly

capture data without the use of a PLL. The incoming clock signal is fed through two delay

blocks which are calibrated at startup using the incoming clock as a training signal. In this way

any clock tree skew is effectively cancelled out. The state machine to do this has to use a free

running clock supplied from user logic. Any clock in the 50 MHz to 200 MHz range can be used.

Once calibrated, the finished signal is asserted, and this signal should be used as a reset signal

to the data receiver. In the case of a differential signal, as shown in Figure 6, the true and

complement signals are fed through master and slave input delays (both set to zero) and then

to a pair of BUFIO2 primitives. The first BUFIO2 accepts both true and complement input clocks

and uses these to generate the appropriate divided clock and SerDes strobe for the input

SerDes primitives. For example, if the receiver clock is 311 MHz (622 Mb/s data) and the

X-Ref Target - Figure 5

Figure 5: Data Reception Using PLL and BUFPLL

serdes_1_to_n_clk_pll_s8_diff.v/vhd

Differential Clock IOB

Calibration and

Bitslip State

Machine

Master

ISERDES

Slave

ISERDES

Master

IDELAY

Slave

IDELAY

(0)

PLL

BITSLIP

BUFPLL

BUFG

BUFIO2

BUFIO2FB

BITSLIP

LVDS Clock P

Parameters:

S = serdes_factor

PLLX = PLL multiplier

CLKIN_PERIOD = Input clock period

BS = TRUE/FALSE to use Bitslip (or not)

Parameters:

D = Number of data lines

S = serdes_factor

LVDS Clock N

x1064_05_100509

Feedback

x1 Clock

SerDes

Strobe

I/O Clock

Feedback

serdes_1_to_n_data_s8_diff.v/vhd

“D” Differential Data IOB

Calibration and

Phase Detection

State Machine

Master

ISERDES

(SDR Mode)

Slave

ISERDES

(SDR Mode)

Master

IDELAY

Slave

IDELAY

BITSLIP

D x S

LVDS Data P

LVDS Data N

Introduction to Deserialization and Data Reception

XAPP1064 (v1.2) November 19, 2013 www.xilinx.com 5

design requires an 8:1 SerDes reduction, the BUFIO2 being driven by true and complement

receiver clocks with its divide parameter set to eight actually divides the input clock by 4 to

77.75 MHz. The resultant I/O clock is routed to the input SerDes primitives along with the other

(inverted) I/O clock generated from the other BUFIO2. These two clocks are doubled in the

input SerDes to give a 622 MHz sampling clock for the 622 Mb/s data.

The BUFIO2s need to be located in the same half edge as the clock input, and when using

input delays, it is not possible to simultaneously use the alternate BUFIO2s in the other half

edge. Reception of data buses when using input delays is therefore limited to the half edge

where the clock input is located. See the Spartan-6 FPGA Clocking Resources User Guide

(UG382

) for more details on BUFIO2 placement and usage.

The mechanism for centering and capturing data reliably is based on the IODELAY2 primitive,

used in input delay mode only. This mechanism is discussed in the Delaying Input Data and

Clocks section. The block diagram of the receiver is shown in Figure 6.

SDR Data Reception Using BUFIO2

The topology for SDR data reception using BUFIO2 uses the incoming clock directly to capture

data. The clock signal is fed through a delay block (set to 0) to a BUFIO2. The BUFIO2 uses this

input clock to generate the appropriate divided clock and SerDes strobe for the input SerDes

primitives. For example, if the receiver clock is 525 MHz (525 Mb/s data) and the design

requires an 8:1 SerDes reduction, then the BUFIO2 with its divide parameter set to eight

divides the input clock by 8 to 65.625 MHz. The resultant I/O clock is routed to the input SerDes

primitives for data capture. The limiting factor in this case is the maximum clock frequency

allowed through the clock input pin.

X-Ref Target - Figure 6

Figure 6: DDR Data Reception Using Two BUFIO2s

serdes_1_to_n_clk_ddr_s8_diff.v/vhd

Differential Clock IOB

Master

IDELAY

Slave

IDELAY

LVDS Clock P

External Clock

Finished

Parameters:

S = serdes_factor

Parameters:

USE_PD = Phase detector

circuitry on or off

D = Number of data lines

S = serdes_factor

LVDS Clock N

x1064_06_110713

N I/O Clock

SerDes

Strobe

P I/O Clock

System

Clock

serdes_1_to_n_data_s8_ddr_diff.v/vhd

D*Differential Data IOB

Calibration and

Phase Detection

State Machine

Calibration

State Machine

Master

ISERDES

(DDR Mode)

Slave

ISERDES

(DDR Mode)

Master

IDELAY

(DDR)

Slave

IDELAY

(DDR)

User

BITSLIP

D x S

LVDS Data P

LVDS Data N

BUFG

BUFIO2_2CLK

BUFIO2

剩余25页未读，继续阅读

评论收藏

内容反馈

Xilinx Spartan6 FPGA serdes应用指南

评论0

最新资源

Xilinx Spartan6 FPGA serdes应用指南

评论0

最新资源

相关推荐

xilinx Spartan6系列FPGA产品规格书术datasheet配置说明技手册等资料合集.zip

Xilinx Spartan6

xilinx spartan-6

一步步教你如何进行Xilinx SerDes调试

Spartan-6 FPGA 数据手册

Spartan6 Serdes BITSLIP 功能详解

23_sdram_ov5640_vga_sobel_SPARTAN6_

spartan-6+OV5640实现实时运动目标检测.zip

RUN_sobel_边缘检测_run_OV5640_视频采集_

xilinx 7系列FPGA应用指南

Xilinx FPGA选型指南

Xilinx FPGA高级设计及应用

Xilinx FPGA系统设计入门级实用指南

Xilinx FPGA system generator在伺服电机控制中的应用.pdf

Spartan3 FPGA 用户使用指南

Spartan-6选型指南（Xilinx官方资料）

lvds_ctr_top_SPARTAN6_fpgalvds_lvds_verilogLVDS_lvdsverilog代

Spartan 3E用户指南

spartan6 clock

SerDes知识详解

理解SerDes

xilinx 7系列GTP serdes仿真文件

Xilinx Spartan-6封装库

Xilinx FPGA的时序分析指南

应用于Xilinx FPGA中的热插拔技术解析

Xilinx_FPGA_FFT_应用笔记.rar_FFT笔记_FPGA FFT_Xilinx fft_fft_算法笔记

Xilinx新一代FPGA设计套件Vivado应用指南（2-2）

SmartFusion2和Spartan6的比较