DC25_PAPER_TRACK13_FreeSignalIntegrityHowUnderstanding _Skytte.pdf

1、 Free signal integrity?How understanding anisotropic materials and tolerances could increase performance at 112/224Gbps and beyond Alfred P.Neves,Wild River Technology Kristoffer Skytte,Cadence John Phillips,Cadence Frank Zavosh,Cadence Eric Bogatin,University of Colorado eric.bogatincolorado.edu 2

2、Abstract The“need for speed”in AI systems is driven by their requirement to process large data sets,both during training and application.Channel design is constrained by the balance between acceptable loss budget and the power consumed in equalization and error correction.Reducing channel loss can e

3、nable lower power or longer unrepeated channel lengths.Historically,high-speed serial links focused on material selection to manage attenuation.In 224 Gbps PAM4 systems,however,second-order factors like impedance variations,crosstalk,and power loss into cavities significantly impact the loss budget.

4、While identically similar virtual channels can be designed in simulations,real fabricated boards differ due to manufacturing variations and limited understanding of material anisotropy.As channel bandwidths increase to 56 GHz and above,accurately defining material behavior in simulations becomes cru

5、cial.This paper analyzes second-order design features using precision measurements of test vehicles.We propose a metric-driven methodology based on AI/ML to determine relevant parameters for simulating anisotropic behavior that matches both time and frequency domain measurements.When second-order fa

6、ctors are accounted for,accurate measurement-simulation correlation is possible,enabling the design of digital twins that predict system performance.These digital twins,combined with AI/ML techniques,allow for design space exploration and sensitivity analysis to identify manufacturing tolerances nec