Serial Lite III Streaming Intel® FPGA IP User Guide

ID 683330
Date 11/01/2021

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

6.2.2. Advanced Clocking Mode Structure For Intel® Arria® 10, Stratix® V, and Arria® V Devices

Advanced clocking mode provides users the option to have different user interface clock frequency than the PHY core clock. Hence, there is PPM differences in this clocking mode. You must specify the user clock frequency through the Serial Lite III Streaming FPGA IP core parameter editor. Based on the user clock frequency value, the Intel® Quartus® Prime software automatically calculates the lane rate and core clock.

The parameter editor provides guidance in selecting a source user clock frequency that meets the transceiver data rate constraints. For more information about the lane rate calculation, refer to the “Transmission Overheads and Lane Rate Calculations” section.

The core operates at higher clock rates in Advanced Clocking Mode. Therefore, when operating in this mode, it may be difficult to close timing at higher data rates (for example, 12 to 15 G) or higher number of lanes. You can implement the following qsf assignment when seeing timing recovery violations from sink coreclkin reset synchronizer to the sink transfer paths:
set_instance_assignment -name GLOBAL_SIGNAL OFF -to 

The Serial Lite III Streaming Intel® Arria® 10 FPGA IP core uses the transmit serial clock bus (tx_serial_clk) and the tx_pll_locked signal to connect the external transmit PLL to the Transceiver Native PHY IP for Intel® Arria® 10 devices.

Note: For sink core, the IP core does not provide any adaptation module. Users should consider a design that can manage the PPM differences between interface_clock and user interface clock.
Figure 22. Advanced Clocking Mode Structure for Intel® Arria® 10 Devices
Figure 23. Advanced Clocking Mode Structure for Stratix V or Arria V Devices
Table 38.   Intel® Arria® 10, Stratix® V, and Arria® V Clocks in Advanced Clocking Mode
Clock Name Description
user_clock User-defined. This clock is determined by the required throughput of the user application. For example, if the user interface is 384-bits wide (6 lanes × 64 bit/lane) and the required throughput is 120 Gbps, the user_clock frequency is 312.5 MHz. This clock is an input to the IP core and you should toggle this at the specified frequency.
tx_serial_clk ( Intel® Arria® 10 devices) This clock should toggle at one-half the data rate of the transceiver lane. When you enter the user_clock frequency in the IP parameter editor, the per lane data rate is calculated. Use that value and divided it by two to determine the tx_serial_clk. You are required to instantiate the TX PLL. An example of the TX PLL (ATX PLL) is generated with the IP core and is configured with the required reference clock and tx_serial_clk frequencies.
interface_clock ( Stratix® V and Arria® V devices) This is an internal clock and it is not exposed to the user. The frequency of this clock is derived from the transceiver data rate. It is lane data rate divided by 40.
interface_clock This is an internal clock and it is not exposed to the user. The frequency of this clock is derived from the transceiver data rate. The frequency is lane data rate divided by 64.
xcvr_pll_ref_clk This reference clock is used by the CDR unit in the transceiver. It serves as a reference for the CDR to be able to recover the clock from the serial line. The frequency of this clock must match the frequency you select in the IP parameter editor. It should also match the frequency of the tx_pll_ref_clk reference clock for the TX PLL at the Source variant.
interface_clock This clock is derived from the transceiver data. It is lane data rate divided by 64. It is an output of the IP core and should be used to clock the RX user application.

Example of Implementing Specific User Interface Clock Frequency

An application requires the Serial Lite III Streaming IP core to sustain a frequency of 240 Gbps at the user interface.

user_clock (frequency) × number_of_lanes × 64 bits/lane = 240 Gbps

The data rate for Intel® Arria® 10 GX transceivers is limited to 17.4 Gbps. Therefore, 240 Gbps / 17.4 Gbps = 14 (rounding up)

Choosing 14 lanes gives:

user_clock (frequency) = 240 / (14 × 64) = 267.86 MHz

A value of 267.86 MHz is out of the supported range for the user_clock frequency. Therefore, you need to add one more lane.

user_clock (frequency) = 240 / (15 × 64) = 250 MHz

Choosing 250 MHz as the user_clock, the IP core provides the following values:

Transceiver data rate: 16.78 Gbps

interface_clk: 16.78 / 64 = 262.18 MHz

tx_serial_clock: 16.78 Gbps / 2 = 8390 MHz