PHY Lite for Parallel Interfaces FPGA IP User Guide

ID 683716
Date 3/31/2025
Public
Document Table of Contents

2.2.2.2. Dynamic Reconfigurable Delays

The following table lists the delays that can only be reconfigured when the corresponding read/write path is not being used. For differential data, the output delay settings should be programmed for both pins in a differential pair. The input settings, however, should be programmed only for the even pin.

Table 17.  Dynamic Reconfigurable Delays
Configurable Settings Width Description Unit Granularity
TxDqDelay 11 Output delay for data and strobe 1 / 128 of VCO cycle per pin
RxDqsNDelayPi 7 Phase shift in negative edge of DQS 1 / 128 of VCO cycle per pin
RxDqsPDelayPi 7 Phase shift in positive edge of DQS 1 / 128 of VCO cycle per pin
RxRcvEnPiRank0 11 RcvEn delay 1 / 128 of VCO cycle per nibble
DqsSenseAmpDelay 5 DQS sense amplifier delay PHY clock cycle per nibble
DqSenseAmpDuration 4 DQ sense amplifier duration PHY clock cycle per nibble
DqSenseAmpDelay 5 DQ sense amplifier delay PHY clock cycle per nibble
DqOdtDuration 4 DQ ODT duration PHY clock cycle per nibble
DqOdtDelay 5 DQ ODT delay PHY clock cycle per nibble
DqsOdtDuration 4 DQS ODT duration PHY clock cycle per nibble
DqsOdtDelay 5 DQS ODT delay PHY clock cycle per nibble
DqsSenseAmpDuration 4 DQS sense amplifier duration PHY clock cycle per nibble
read_enable_offset 4 Delay before reading from the RX FIFO PHY clock cycle per lane
RxDataVrefL 9 IO reference voltage lower nibble 1 / 512 of VCCIO per nibble
RxDataVrefU 9 IO reference voltage upper nibble 1 / 512 of VCCIO per nibble
TrainReset 1 Reset the training to clear non-permanent states. Internal to the I/O lane and resets stateful logic on the read and write paths. per lane
RLTrainingMode 1 Enables read leveling training mode per lane
DataTrainFeedback_N0 12 Provides feedback for different training steps.

In RL Training mode it is simply a counter.

per nibble
D0DlyRange 2 Enable/Set delay line inside the RxDqD0 for matched Rx. 0: Disable, 1: 1 delay cell, 2: 2 delay cells, 3: 3 delay cells. 1/8 of VCO cycle. common for all data pins
read_enable_offset_msb 1 Extends the shift register from 16 to 32 when rxfifo_offset_mode is set to 1 per lane
rxfifo_offset_mode_1 1

0: do not bypass read_enable_offset shift register

1: bypass read_enable_offset shift register

per lane
rxfifo_offset_mode_0 1

0: reading from Rxfifo is delayed by rddata_en delay + read_enable_offset[3:0]

1: reading from Rxfifo is delayed by {read_enable_offset_msb, read_enable_offset[3:0]}
per lane

If a PHY Lite group occupies two I/O lanes (for example, using a x16 DQS tree), expect additional delay on DQS due to the longer DQS path. For x16 input data groups, use the D0DlyRange register to shift out the DQ data so that the left edge of the data eye can be observed during DQS sweeps. At the maximum supported double data rate of 2500 MHz, 100ps is 0.25 UI; thus, setting the D0DlyRange register to 1 (1 delay cell) is usually sufficient.

Use the D0DlyRange register to shift out the DQ data so that the left edge of the data eye can be observed during DQS sweeps. At the maximum supported double data rate of 2500 MHz, 100ps is 0.25 UI; thus, setting the D0DlyRange register to 1 (1 delay cell) is usually sufficient.