Private Island Networks Inc.

PCS / SERDES Architecture for Lattice ECP5 FPGA Open Source Network Processor

A high level overview of the usage and configuration of the ECP5UM DCU (PCS/SERDES) for Private Island Open Source Project

Overview

This article describes the usage and configuration of the Lattice ECP5UM SERDES/PCS block for the Private Island ® open source project. The ECP5UM provides two embedded SERDES/PCS channels per Dual-Channel Unit (DCU) and two DCU's per device.

As depicted in the figure below, Private Island utilizes up to four channels for processing via Gigabit PHYs. Two 1000BASE-T PHYs are provided on the Darsena development board with support provided for additional PHYs (e.g., 1000BASE-SX Fiber) via an expansion connector.

Figure 1. High Level Use of PCS/SERDES Blocks
High Level Use of PCS/SERDES Blocks

This article refers to Lattice's "ECP5 and ECP5-5G SERDES/PCS Usage Guide" TN1261, and it is suggested that the reader have it open while reading this article.

The ECP5UM DCU is a hard macrocell, and it provides a great deal of flexibility for various protocols. For Private Island, the first DCU is configured for 1000-BaseT operation via SGMII PHYs, and the settings are summarized in the table below.

Table 1: DCU Parameters for GbE1
DCU (PCS/SERDES) Parameter Setting
Protocol GbE
TX/RX serial data rates 1.25 Gbps
TxPLL/CDR PLL Multiplier 25x
REFCLK (DCU0 only)2 50 MHz
DCU / FPGA data interface width 8-bit
FPGA PCS Clock Rate 125 MHz
Receive CTC Enabled
Receive FIFO Enabled
Transmit FIFO Enabled
Reset Sequence Disabled2
1: EXTREF0 drives both DCU0 and DCU1
2: It is assumed that the DCU reset sequence is supervised by a micro controller.

DCU I/O Definition

The Verilog definition for the DCU module is provided in "ecp5um.v", which can be found within the Lattice Diamond installation. For Diamond on Windows 10, the path is <diamond install path>\<release>_x64\synpbase\lib\lucent. In some cases, the module I/O names do not match the signal names provided in the TN1261 documentation. Therefore, a mapping for many of the utilized signals is provided in the tables below.

Table 2: DCU Dual I/O
DCUA Definition signal name I/O Notes
D_FFC_DUAL_RST rst_dual_c I AUX channel
D_FFC_MACRO_RST serdes_rst_dual_c I AUX channel
D_FFC_MACROPDB serdes_pdb I AUX channel
D_FFC_TRST tx_serdes_rst_c I Only one for both TX channels
D_REFCLKI refclko I TX PLL reference clock from EXTREF
D_FFS_PLOL pll_lol O Don't use. Use soft logic generated by Clarity instead
Table 3: Channel I/O
DCUA Definition signal name I/O Notes
CHX_RX_REFCLK1 refclko I CDR reference clock from EXTREF
CHX_FF_RXI_CLK txi_clk I RX FIFO clock from FPGA
CHX_FF_TXI_CLK txi_clk2 I TX FIFO clock from FPGA
CHX_FF_EBRD_CLK txi_clk2 I CTC FIFO clock from FPGA
CHX_FF_TX_D_[0:7] txdata[0:7] I Transmit data bus
CHX_FF_TX_D_8 tx_k I Assert for special code-groups
CHX_FF_TX_D_10 xmit I Auto negotiation / idle insert
CHX_FF_TX_D_11 tx_disp_correct I Assert for DCU to correct disaparity
CHX_FFC_SIGNAL_DETECT signal_detect_c I Enable RX link state machine
CHX_FFC_LANE_TX_RST tx_pcs_rst_c I Reset TX channel PCS logic
CHX_FFC_LANE_RX_RST rx_pcs_rst_c I Reset RX channel PCS logic
CHX_FFC_RRST rx_serdes_rst_c I Reset digital logic in SERDES RX channel
CHX_FFC_TXPWDNB tx_pwrup_c I TX Channel power up
CHX_FFC_RXPWDNB rx_pwrup_c I RX Channel power up
CHX_FF_TX_PCLK tx_pclk O Transmit Primary Clock
CHX_FF_RX_D_[0:7] rxdata[0:7] O Receive data bus
CHX_FF_RX_D_8 rx_k O Asserted for special code-groups
CHX_FF_RX_D_9 rx_disp_err O RX disparity error
CHX_FF_RX_D_10 rx_cv_err O code violation
CHX_FFS_RLOS rx_los_low_s O Loss of signal (LO THRESHOLD RANGE) detection
CHX_FFS_LS_SYNC_STATUS lsm_status_s O Channel is synced with commas
CHX_FFS_RLOL rx_cdr_lol_s O CDR loss loss of lock
CHX_FFS_CC_UNDERRUN() ctc_urun_s O CTC status flag
CHX_FFS_CC_OVERRUN() ctc_orun_s O CTC status flag
CHX_FFS_SKP_ADDED ctc_ins_s O CTC status flag
CHX_FFS_SKP_DELETED ctc_del_s O CTC status flag
1: Where CHX is CH1 or CH2.
2: The Clarity generated files for sgmii<N>.v are manually modified to route txi_clk to ebrd_clk and rxi_clk.

SERDES Client Interface

This interface supports both an alternative method to static configuration of the DCU block as well as some dynamic configuration.

Table 4: SCI Dual I/O
DCUA Definition signal name I/O Notes
D_SCIWDATA[0:7] sci_wrdata[0:7] I
D_SCIADDR[0:5] sci_addr[0:5] I
D_SCIENAUX sci_en_dual I Not defined in TN1261
D_SCISELAUX sci_sel_dual I
D_SCIRD sci_rd I
D_SCIWSTN sci_wrn I
D_SCIRDATA[0:7] sci_rddata[0:7] O
D_SCIINT sci_int O
Table 5: SCI Channel I/O
DCUA Definition signal name I/O Notes
CHX_SCIEN sci_en I Not defined in TN1261
CHX_SCISEL sci_sel I

DCU Clocking

The figure below shows the clocking architecture for DCU0, channel 0, and it is similar to Figure 4 in TN1261. The output of the TxPLL is routed to the FPGA fabric as tx_pclk and used globally to clock the DCU FPGA Bridge circuits (RX FIFO, TX FIFO, and CTC FIFO) for both DCU0 and DCU1. In other words, the output of the DCU0 TXPLL is used globally to clock all Ethernet circuits in the FPGA soft logic and both DCU FPGA bridge circuits. Therefore, the other tx_pclk signals and all recovered RX CDR outputs are left floating.

Figure 2. FPGA / DCU Clocking
High Level Use of PCS/SERDES Blocks

SERDES/PCS Parameter

Provided below are some of the parameters that can be changed in the Clarity generated Verilog file.

RX Equalizer:

  • enable: DCU1_inst.CH0_REQ_EN = "0b1";
  • disable: DCU1_inst.CH0_REQ_EN = "0b0";
  • equalization setting: DCU1_inst.CH0_REQ_LVL_SET = "0b10";

AC/DC coupling:

  • DCU1_inst.CH0_RCV_DCC_EN = "0b0";
  • DCU1_inst.CH0_RCV_DCC_EN = "0b1";

Reference Clock

  • 125 MHz: DCU0_inst.D_REFCK_MODE = "0b001";
  • 50 MHz: DCU0_inst.D_REFCK_MODE = "0b100";

Terms

  • CTC: Clock Tolerance Compensation
  • DCU: Dual Channel Unit
  • FB: FPGA Bridge
  • FF: FPGA Fabric
  • FPGA: Field Programmable Gate Array
  • PCS: Physical Coding Sublayer
  • PHY: Physical Layer Device
  • SB: SERDES Bridge
  • SCI: SERDES Client Interface
  • SERDES: Serializer (SER) / Deserializer (DES)

References:

Didn't find an answer to your question? Post your issue below or in our new FORUM, and we'll try our best to help you find a solution.

And please note that we update our site daily with new content related to our open source approach to network security and system design. If you would like to be notified about these changes, then please join our mailing list.

Related articles on this site:

share
subscribe to mailing list:

Please help us improve this article by adding your comment or question:

your email address will be kept private
authenticate with a 3rd party for enhanced features, such as image upload
previous month
next month
Su
Mo
Tu
Wd
Th
Fr
Sa
loading