DesignCon is part of the Informa Markets Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Early Bird Registration Now Open till November 30th. Save Up to $300 Today!


DesignCon 2019 Presentation Viewer

Purchase procecdings

Welcome to the DesignCon Presentation Store. Here you can view and download conference and/or show floor theater presentations before, during, and after the event. If you’re looking for a presentation from a specific session that you’re unable to find here, note that it’s likely because the presenter has not provided permission for external use or has not yet shared their presentation with us. Please check back after the event for a more complete catalogue of available presentations.

If you’d like to do a bulk download of all conference presentations or technical papers at once, please click here for conference presentations or click here for full technical papers. For sessions not included in the main conference, click here for Chiphead Theater presentations or click here for sponsored session presentations.

Spec-driven CTLE Model Synthesis Through Reinforcement Learning

Zhaoyin Daniel Wu (Principal Engineer, Xilinx Inc)

Location: Ballroom C

Date: Wednesday, January 30

Time: 2:50pm - 3:30pm

Track: 15. Machine Learning for Microelectronics, Signaling & System Design, 08. Optimizing High-Speed Serial Design

Session Type: Technical Session

Vault Recording: TBD

Audience Level: All

To enable link level simulation for given CTLE specs, it is necessary to develop models to represent the behavior of these CTLE circuits under these specs. Traditionally, a lot of manual tuning and reiteration is needed to find the desirable pole/zero locations to build a CTLE transfer function to meet the spec. In this paper, given a set of frequency domain specs of CTLE, a reinforcement learning technique is used to automatically adjust the pole/zero location of the CTLE transfer function to meet the spec, which significantly reduced the manual work of finding the desired CTLE model.

Takeaway

CTLE modeling can be automated by Q-learning given the targeted specifications. (peaking, Nyquist, bandwidth, etc.)

Presentation Files

SLIDES_15_SpecDrivenCTLESynthesisThruRL_Wu.pdf
PAPER_15_SpecdrivenCTLEModelSynthesis_Liu.pdf