Machine Learning Based Backchannel Mechanism for Expert-free High-speed Link Equalization Tuning

Event Time

Originally Aired - Wednesday, August 18 8:00 AM - 8:40 AM

Info Alert

Create or Log in to My Show Planner to see Videos and Resources.

Videos

Resources


{{video.title}}

Log in to your planner to join the zoom meeting!

{{chatHeaderContent}}

{{chatBodyContent}}

Resources

Info Alert

This Session Has Not Started Yet

Be sure to come back after the session starts to have access to session resources.

Event Location

Location: Meeting Room 211CD


Event Information

Title: Machine Learning Based Backchannel Mechanism for Expert-free High-speed Link Equalization Tuning

Event Type: DesignCon - Technical Session

Pass Type: All-Access Pass, 2-Day Pass

Theme: High-speed Communications,Data Centers


Description

While RX EQ adaptation provides best-effort performance tuning, experts are often required for TX swing/EQ settings in system performance optimization. In this work, we develop the machine learning based autonomous system which guides TX tuning based on RX EQ adaptation codes. Targeting the eye opening, the system first predicts the eye margin based on RX adaptation codes. If the criteria are not met, it will then predict the "direction" for each TX setting to help increasing the eye margin. Unlike PCIe link training, the proposed system does not depend on protocol pre-definitions, and is applicable to all applications in the industry.


Speakers


Tracks