TY - GEN
T1 - A Bayesian Framework for Optimizing Interconnects in High-Speed Channels
AU - Torun, Hakki M.
AU - Larbi, Mourad
AU - Swaminathan, Madhavan
N1 - Funding Information:
resulting ADD-GP model is shown to agree well with full-wave simulations, having a maximum of 2.9% CV-MSE on all elements of RLGC matrices. This ADD-GP model is then used in the optimization loop driven by TSBO to directly maximize eye opening rather than conventional approach of tuning the frequency response of the channel. Compared to conventional approach, direct eye optimization resulted in 42.4% and 46.1% increase in eye height and width along with 54.4% reduction in peak-to-peak jitter at a data of 16 Gbps. ACKNOWLEDGEMENT This research is funded by the DARPA CHIPS project under Award N00014-17-1-2950.
Publisher Copyright:
© 2018 IEEE.
PY - 2018/10/23
Y1 - 2018/10/23
N2 - Increasing demand in higher bandwidth chip-to chip communications have resulted in challenges related to modelling and optimization of their electrical performance due to CPU intensive simulations arising from multiscale structures. Conventional approaches use various approximations to either reduce the design complexity or reduce the simulation time, however, this can lead to inaccurate models and sub-optimal designs. In this paper, we address this problem by using machine learning based techniques and propose a Bayesian framework to model and optimize interconnects in high-speed channels in an accurate yet efficient fashion.
AB - Increasing demand in higher bandwidth chip-to chip communications have resulted in challenges related to modelling and optimization of their electrical performance due to CPU intensive simulations arising from multiscale structures. Conventional approaches use various approximations to either reduce the design complexity or reduce the simulation time, however, this can lead to inaccurate models and sub-optimal designs. In this paper, we address this problem by using machine learning based techniques and propose a Bayesian framework to model and optimize interconnects in high-speed channels in an accurate yet efficient fashion.
UR - http://www.scopus.com/inward/record.url?scp=85057073937&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85057073937&partnerID=8YFLogxK
U2 - 10.1109/NEMO.2018.8503097
DO - 10.1109/NEMO.2018.8503097
M3 - Conference contribution
AN - SCOPUS:85057073937
T3 - 2018 IEEE MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization, NEMO 2018
BT - 2018 IEEE MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization, NEMO 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 IEEE MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization, NEMO 2018
Y2 - 8 August 2018 through 10 August 2018
ER -