F-Tile Ethernet Intel® FPGA Hard IP User Guide

Download PDF
ID 683023
Date 6/05/2024
Public

A newer version of this document is available.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

Document Table of Contents
Document Table of Contents x
1. Overview 2. Getting Started 3. F-Tile Ethernet Intel® FPGA Hard IP Parameters 4. Functional Description 5. Clocks 6. Resets 7. Interface Overview 8. Configuration Registers 9. Supported Modules and IPs 10. Supported Tools 11. F-Tile Ethernet Intel® FPGA Hard IP User Guide Archives 12. Document Revision History for the F-Tile Ethernet Intel® FPGA Hard IP User Guide
1. Overview x
1.1. Ethernet System in F-Tile Overview 1.2. F-Tile Ethernet Intel® FPGA Hard IP Overview
1.2. F-Tile Ethernet Intel® FPGA Hard IP Overview x
1.2.1. Device Family Support 1.2.2. Device Speed Grade Support 1.2.3. Resource Utilization 1.2.4. Round-trip Latency 1.2.5. Release Information
2. Getting Started x
2.1. Installing and Licensing F-Tile Ethernet Intel® FPGA Hard IP Cores 2.2. Specifying the IP Core Parameters and Options 2.3. Reference and System PLL Clock for your IP Design 2.4. Generated File Structure 2.5. Generating Tile Files 2.6. IP Core Testbenches
2.4. Generated File Structure x
2.4.1. Generating IP-XACT File
4. Functional Description x
4.1. Datapath Description 4.2. F-Tile Ethernet Intel® FPGA Hard IP MAC 4.3. PCS, OTN, and FlexE Modes 4.4. Precision Time Protocol 4.5. Auto-Negotiation and Link Training
4.2. F-Tile Ethernet Intel® FPGA Hard IP MAC x
4.2.1. MAC TX Datapath 4.2.2. MAC RX Datapath 4.2.3. Congestion and Flow Control Using PAUSE or Priority Flow Control (PFC) 4.2.4. Link Fault Signaling 4.2.5. Order of Ethernet Transmission
4.2.1. MAC TX Datapath x
4.2.1.1. TX Preamble, Start, and SFD Insertion 4.2.1.2. Source Address Insertion 4.2.1.3. Length/Type Field Processing 4.2.1.4. Frame Padding 4.2.1.5. Frame Check Sequence (CRC-32) Insertion 4.2.1.6. Inter-Packet Gap Generation and Insertion 4.2.1.7. TX Packing Logic
4.2.2. MAC RX Datapath x
4.2.2.1. RX Preamble Processing 4.2.2.2. RX Strict SFD Checking 4.2.2.3. RX FCS Checking 4.2.2.4. RX Malformed Packet Handling 4.2.2.5. Removing PAD Bytes and FCS Bytes from RX Frames 4.2.2.6. RX Undersized Frames, Oversized Frames, and Frames with Length Errors 4.2.2.7. Inter-Packet Gap
4.2.3. Congestion and Flow Control Using PAUSE or Priority Flow Control (PFC) x
4.2.3.1. Conditions Triggering XOFF Frame Transmission 4.2.3.2. Conditions Triggering XON Frame Transmission 4.2.3.3. Pause Control and Generation Interface 4.2.3.4. Pause Control Frame Filtering
4.3. PCS, OTN, and FlexE Modes x
4.3.1. PCS Mode 4.3.2. OTN Mode 4.3.3. FlexE Mode
4.4. Precision Time Protocol x
4.4.1. Features 4.4.2. PTP Timestamp Accuracy 4.4.3. PTP Client Flow 4.4.4. RX Virtual Lane Offset Calculation for No FEC Variants 4.4.5. Virtual Lane Order and Offset Values 4.4.6. UI Adjustment 4.4.7. Reference Time Interval 4.4.8. Minimum and Maximum Reference Time (TAM) Interval for UI Measurement (Hardware) 4.4.9. UI Value and PMA Delay 4.4.10. Routing Delay Adjustment for Advanced Timestamp Accuracy Mode 4.4.11. Routing Delay Adjustment for Basic Timestamp Accuracy Mode
4.4.3. PTP Client Flow x
4.4.3.1. PTP TX Client Flow 4.4.3.2. PTP RX Client Flow
4.4.6. UI Adjustment x
4.4.6.1. TX UI Adjustment 4.4.6.2. RX UI Adjustment
5. Clocks x
5.1. Clock Connections in Single Instance Operation 5.2. Clock Connections in Multiple Instance Operation 5.3. Clock Connections in MAC Asynchronous FIFO Operation 5.4. Clock Connections in PTP-Based Synchronous and Asynchronous Operation 5.5. Clock Connections in Synchronous Ethernet Operation 5.6. Custom Cadence
6. Resets x
6.1. Reset Signals 6.2. Reset Sequence
7. Interface Overview x
7.1. Status Interface 7.2. TX MAC Avalon ST Client Interface 7.3. RX MAC Avalon ST Aligned Client Interface 7.4. TX MAC Segmented Client Interface 7.5. RX MAC Segmented Client Interface 7.6. MAC Flow Control Interface 7.7. PCS Mode TX Interface 7.8. PCS Mode RX Interface 7.9. FlexE and OTN Mode TX Interface 7.10. FlexE and OTN Mode RX Interface 7.11. Custom Rate Interface 7.12. 32-bit Soft CWBIN Counters 7.13. Reconfiguration Interfaces 7.14. Precision Time Protocol Interface
7.2. TX MAC Avalon ST Client Interface x
7.2.1. TX MAC Avalon ST Client Interface with Disabled Preamble Passthrough 7.2.2. TX MAC Avalon ST Client Interface with Enabled Preamble Passthrough 7.2.3. Using MAC Avalon ST skip_crc Signal to Control Source Address, PAD, and CRC Insertion 7.2.4. Using MAC Avalon ST i_tx_error Signal to Mark Packets Invalid
7.3. RX MAC Avalon ST Aligned Client Interface x
7.3.1. RX MAC Avalon ST Client Interface with Disabled Preamble Passthrough 7.3.2. RX MAC Avalon ST Client Interface with Enabled Passthrough and RX CRC Forwarding 7.3.3. RX MAC Adapter Limitations 7.3.4. RX MAC Avalon ST Client Interface Status
7.4. TX MAC Segmented Client Interface x
7.4.1. TX MAC Segmented Client Interface with Disabled Preamble Passthrough 7.4.2. TX MAC Segmented Client Interface with Enabled Preamble Passthrough 7.4.3. Using MAC Segmented skip_crc Signal to Control Source Address, PAD, and CRC Insertion 7.4.4. Using MAC Segmented i_tx_mac_error to Mark Packets Invalid
7.5. RX MAC Segmented Client Interface x
7.5.1. RX MAC Segmented Client Interface with Disabled Preamble Passthrough 7.5.2. RX MAC Segmented Client Interface with Enabled Passthrough and RX CRC Forwarding 7.5.3. RX MAC Segmented Client Interface Status and Errors
7.13. Reconfiguration Interfaces x
7.13.1. Ethernet Reconfiguration Interfaces 7.13.2. Transceiver Reconfiguration Interfaces
7.14. Precision Time Protocol Interface x
7.14.1. Time-of-Day Interface 7.14.2. TX 2-Step Timestamp Interface 7.14.3. TX 1-Step Timestamp Interface 7.14.4. RX Timestamp Interface 7.14.5. PTP Status Interface 7.14.6. PTP Tile Interface
7.14.3. TX 1-Step Timestamp Interface x
7.14.3.1. PTP Field Offset for 1-Step Operation
8. Configuration Registers x
8.1. Ethernet Avalon® Memory-Mapped Interface Address Space 8.2. Transceiver Avalon® Memory-Mapped Interface Address Space
8.1. Ethernet Avalon® Memory-Mapped Interface Address Space x
8.1.1. Ethernet Hard IP Core CSRs 8.1.2. FEC and Transceiver Control and Status Registers
9. Supported Modules and IPs x
9.1. F-Tile Auto-Negotiation and Link Training For Ethernet Intel® FPGA IP 9.2. PTP Tile Adapter
9.1. F-Tile Auto-Negotiation and Link Training For Ethernet Intel® FPGA IP x
9.1.1. Overview 9.1.2. Release Information 9.1.3. Functional Description 9.1.4. Parameters 9.1.5. Clocks and Resets 9.1.6. Registers
9.2. PTP Tile Adapter x
9.2.1. Overview 9.2.2. Clocks, Reset, and Interface Ports 9.2.3. PTP Asymmetry Delay Reconfiguration Interface 9.2.4. PTP Peer-to-Peer MeanPathDelay Reconfiguration Interface
10. Supported Tools x
10.1. F-Tile Channel Placement Tool 10.2. Ethernet Toolkit Overview
10.2. Ethernet Toolkit Overview x
10.2.1. Features
1. Overview
2. Getting Started
3. F-Tile Ethernet Intel® FPGA Hard IP Parameters
4. Functional Description
5. Clocks
6. Resets
7. Interface Overview
8. Configuration Registers
9. Supported Modules and IPs
10. Supported Tools
11. F-Tile Ethernet Intel® FPGA Hard IP User Guide Archives
12. Document Revision History for the F-Tile Ethernet Intel® FPGA Hard IP User Guide

7.9. FlexE and OTN Mode TX Interface

The F-Tile Ethernet Intel® FPGA Hard IP TX client interface in FlexE and OTN variations employs the PCS66 interface protocol.

The FlexE and OTN variations allow the application to write 66b blocks to the TX PCS, bypassing the TX MAC.

  • In FlexE mode and 200GE/400GE OTN modes, the TX encoder in the PCS is also bypassed.
  • In 10GE/25GE/40GE/50GE/100GE OTN mode, both the TX encoder and the scrambler are bypassed.

The client acts as a source and the TX PCS acts as a sink in the transmit direction.

Table 53.  Signals of the PCS66 TX InterfaceAll interface signals are clocked by the TX clock. The signal names are standard Avalon® streaming interface signals.

Signal Name

Width

Description

i_tx_pcs66_d[1055:0]

i_tx_pcs66_d[527:0]

i_tx_pcs66_d[263:0]

i_tx_pcs66_d[131:0]

i_tx_pcs66_d[65:0]

1056 bits (400GE)

528 bits (200GE)

264 bits (100GE)

132 bits (40GE/50GE)

66 bits(10GE/25GE)

TX PCS 66b data for 1 block.

  • In FlexE mode, the data presented is scrambled.
  • In OTN mode, the data goes directly to the RS-FEC or PMA.

i_tx_pcs66_valid

1 bit When asserted, indicates that the TX PCS 66b data is valid.

Must be asserted when the TX PCS 66b ready signal is asserted.

o_tx_pcs66_ready

1 bit

TX PCS 66b ready signal.

When asserted, indicates the PCS is ready to receive new data.

i_tx_pcs66_am

1 bit for each channel Alignment marker insertion bit.

In FlexE mode, asserting this signal causes the PCS to allow gaps for the alignment markers in place of the data presented on the TX PCS data signal. The application marks the block as an alignment marker and the scrambler does not process the data.

In OTN mode, RS-FEC must be aware of the alignment marker location.

In OTN mode without FEC, the i_tx_pcs66_am signal is optional. You can tie it low. The Transcode Decoder inserts and removes the alignment markers for Ethernet modes with FEC (200GE/400GE) using 400GE PCS. For100GE, the PCS inserts the alignment marker and remaps it using the FEC Transcode Decoder. In this mode FEC is simulated as a sublayer between 100GE PCS and PMA. Therefore, the 100GE alignment markers exist outside of the FEC.

Note: When the Deterministic Latency Interface or Deterministic Latency Measurement parameter is turned on, sync pulses should be provided for 25GE and 10GE variants. When Deterministic Latency Measurement parameter is turned on, the internal module handles this generation.
Figure 48. Transmitting Data Using the PCS66 TX InterfaceThe figure shows how to write the 66b blocks directly to the TX PCS in FlexE and OTN mode using the PCS66 TX Interface.

TX data is written as 66b blocks. The blocks are expected to be 66b encoded, with the sync header bits in the rightmost bit positions (bits 1 and 0).

  • In FlexE mode, the PCS scrambles and stripes the blocks for transmission.
  • In OTN mode, the PCS only stripes the blocks for transmission. The PCS66 symbols must be scrambled if they are less than 100GE before being transmitted to the FEC TX. The FEC RX does not descramble for symmetry. The PCS66 symbols for 200GE and 400GE are not scrambled; instead, the scramble is applied on transcoded 257b blocks in FEC TX. In FEC RX, 257b blocks are always descrambled since symmetry exists.

i_tx_pcs66_valid should conform to these conditions:

  • Assert the valid signal only when the ready signal is asserted, and deassert only when the ready signal is deasserted.
  • The two signals can be spaced by a fixed latency between 1 and 6 cycles.
  • When the valid signal deasserts, i_tx_pcs66_d must be paused.

The block order for the PCS66 mode TX interface is the same as the TX PCS interface. Blocks are transmitted from LSB to MSB; the first block to be transmitted from the interface is i_tx_pcs66_d[65:0].

The bit order for the PCS66 mode TX interface is the same as the TX PCS interface. Bits are transmitted from least to most significant; the first bit to be transmitted from the interface is i_tx_pcs66_d[0].

Figure 49. Inserting Alignment Markers

When the PCS66 TX interface is used for FlexE mode, the timing of alignment marker insertion can be controlled from the fabric. The same operations can be performed on single lane variants, with slight variance:

  • For 40GE/50GE/100GE/200GE/400GE channels and 25GE channel with RS-FEC, the signal causes the alignment markers to be inserted.
  • For 10GE/25GE channels with no RS-FEC or Firecode, the signal causes the cycle to be treated as invalid for PCS processing (no changes to scramble).

In FlexE mode, the timing of alignment marker insertion is very rigid. Alignment markers cannot be delayed without disrupting the Ethernet link. Use valid cycles to count the alignment markers. When i_tx_pcs66_valid is low, the alignment marker counters and input must freeze.

OTN streams are expected to include their own alignment markers. In OTN mode with FEC, you must assert i_tx_pcs66_am to indicate the position of the alignment markers. In OTN mode without FEC, i_tx_pcs66_am is optional and you can tie the signal low.

In FEC modes, the TX datapath does not come completely out of reset until at least 2 alignment marker periods passed. You must start driving i_tx_pcs66_am at the proper interval before o_tx_lanes_stable goes high. You may drive the signal as soon as o_tx_pll_locked has gone high and o_tx_pcs66_ready starts toggling. When the external custom rate interface is enabled, you must start driving i_custom_cadence. For more information, refer to the Custom Rate Interface.

Related Information
Level Two Title