UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

AR# 21222

Virtex-II Pro X RocketIO - What is the lane-to-lane skew between the RocketIO, and how do I control/reduce it?

Description

General Description:

What is the lane-to-lane skew between the RocketIO, and how do I control/reduce it?

Solution

What is the lane-to-laneTX skew?

There are three components to the TX skew, as follows:

USRCLK Phase

USRCLK phase results in a one-word skew.

For an internal data path of 32 bits, it is 32 UI (Unit Interval).

For an internal data path of 40 bits, it is 40 UI (Unit Interval).

PISO Alignment

PISO clock alignment can have as much as a one-word skew as well, depending on the PLL alignment on power up.

For an internal data path of 32 bits, it is 32 UI (Unit Interval).

For an internal data path of 40 bits, it is 40 UI (Unit Interval).

BREFCLK Routing Skew

BREFCLK routing skew should generally be 2 UI at 2.5Gbps, and a maximum of 4 UI.

Consequently, the absolute worst case number for an internal data path width of 32 bits is 68 UI and for 40 bits it is 84 UI. In practice it is always less because the PISO and USRCLK skews tend to cancel each other out; this is pending characterization.

Is this skew guaranteed and stable?

These skews are deterministic after power-up; this has been tested and does not vary once the device is stable.

Are there ways to control or mitigate the skew?

For mitigating skew, you can eliminate the one-word skew due to the clock phase by using a closed loop approach that allows for encoded data. This method includes sending a training pattern and observing each lane at the receiver to determine the skew, or reporting back when each lane is aligned. The transmit side can adjust the fabric interface delay until the lanes are aligned. This solution can be a simple state machine that tries the various combinations of delay until each lane is aligned. The delay can be a simple two-stage pipeline delay on the fabric user interface with optional bypassing on each stage to adjust the delay.

AR# 21222
Date Created 09/04/2007
Last Updated 12/15/2012
Status Active
Type General Article