F-tile Architecture and PMA and FEC Direct PHY IP User Guide

ID 683872
Date 6/26/2023
Public

A newer version of this document is available.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

Document Table of Contents

2.4.4. Datapath Clock Cadences

The read and write frequency of the PMA FIFO interface determines if you need a standard or custom cadence.

  • Standard cadence: Use if the read and write frequencies of the PMA FIFO interface are the same with 0 ppm frequency delta.
  • Custom cadence: Use if the read and write frequencies of the PMA FIFO interface have different frequencies or have the same frequency but with a frequency delta of greater than 0 ppm.
Figure 55. Standard Cadence and Custom Cadence

See PMA Data Rates for supported data rates.

Table 24.  Supported Datapath Clock Frequencies and Cadences by Datapath Clocking Mode
Datapath Clocking Mode Configuration Datapath Clock Frequency Cadence

PMA clocking mode

(maximum 906.25 MHz)

PMA Direct

Datapath clock frequency = PMA clock frequency

PMA clock frequency = line rate ÷ PMA width

Use the standard cadence on the TX and RX (data is valid at every clock edge). 15

System PLL clocking mode

(maximum 1 GHz)

PMA Direct

Use Case A: Chip-to-chip applications where F-tile and link partner share the same reference clock

Datapath clock frequency ≥ (system PLL output frequency)min where (system PLL output frequency)min = PMA clock frequency

If (system PLL output frequency = PMA clock frequency and ∆ppm = 0), use the standard cadence on the TX and RX (data is valid at every clock edge). Otherwise, use custom cadence. 16 , 17

Use Case B: Applications where F-tile and link partner do not share the same reference clock

Datapath clock frequency ≥ (system PLL output frequency)min where (system PLL output frequency)min = (maximum ppm 18 ÷ 1000000 + 1) × PMA clock frequency

System PLL clocking mode

(maximum 1 GHz)

Other configurations with FEC, PCS, and MAC

Datapath clock frequency ≥ (system PLL output frequency)min where (system PLL output frequency)min = PMA clock frequency

For example, for 10GbE-1, use ≥ 322.265625 MHz; for 25GbE-1, use ≥ 805.6640625 MHz; and, for 50GbE-1, use ≥ 830.078125 MHz.

If (system PLL output frequency = PMA clock frequency), use the standard cadence on the TX and RX (data is valid at every 32 of 33 or 34 clock cycles). Otherwise, use custom cadence. 19

One 25 Gbps PMA Direct PHY IP Port Using the PMA Clocking Mode Example

  • All blocks between the PMA interface and core FIFO interface run on the PMA clock.
  • On the transmitter, the PMA FIFO interface is clocked by the TX PMA clock on both sides.
  • On the receiver, the PMA FIFO interface is clocked by the RX recovered clock on both sides.
  • Use the standard cadence. Data on the TX and RX is valid at every clock edge of the PMA clock.
Figure 56. One 25 Gbps PMA Direct PHY IP Port Using the PMA Clocking Mode ExampleThis F-Tile Clocking Tool screenshot shows one 25 Gbps PMA Direct PHY IP port using the PMA clocking mode.

25 Gbps Ethernet Without FEC Port Using the Overclocked System PLL Clocking Mode Example

  • All blocks between the PMA interface and core FIFO interface run on the system PLL clock.
  • On the transmitter, the PMA FIFO interface performs a clock transfer from the system PLL domain to the TX PMA clock domain.
  • On the receiver, the PMA FIFO interface performs a clock transfer from the RX recovered clock domain to the system PLL domain. Refer to F-Tile Ethernet Intel® FPGA Hard IP User Guide for how to clock the core interface.
  • Because the system PLL clock frequency is faster than the PMA clock frequency, datapath clocking is overclocked. Therefore, you must use custom cadence.
Figure 57. 25 Gbps Ethernet Without FEC Port Using the Overclocked System PLL Clocking Mode ExampleThis F-Tile Clocking Tool screenshot shows one 25 Gbps Ethernet without FEC port using the overclocked system PLL clocking mode.
15 The TX PMA and TX digital blocks use a PMA clock derived from the local clock. The RX PMA and RX digital blocks run on a recovered clock (the link partner clock).
16 Use Case A: Standard cadence can be used only when the TX PMA reference clock, system PLL reference clock, and link partner TX reference clock are coming from same clock source (with a 0 ppm frequency delta). At 32 Gbps, only the standard cadence can be used because the system PLL reaches a maximum frequency of 1 GHz (it cannot tolerate any difference in the frequencies; the frequency delta must be 0 ppm).
17 Use Case B: The system PLL frequency must be overclocked to compensate for a frequency delta of greater than 0 ppm between the TX PMA reference clock, system PLL reference clock, and link partner TX reference clock. It does not support 32.0 Gbps because the system PLL clock must run at speeds greater than 1 GHz to incorporate a frequency delta of greater than 0 ppm.
18

maximum ppm = maximum ∆ppm ÷ 2

maximum ∆ppm = max(∆ppm between the link partner TX (the recovered clock on the local RX) and system PLL, ∆ppm between the system PLL and TX PMA)

19 The data path clock is already overclocked compared to the PMA clock by approximately 3% because of PCS and FEC overhead. Therefore, a frequency delta of greater than 0 ppm between the TX PMA reference clock, system PLL reference clock, and link partner TX reference clock is allowed.