A Deep Learning Approach for Volterra Kernel Extraction for Time Domain Simulation of Weakly Nonlinear Circuits

Volterra kernels are well known to be the multidimensional extension of the impulse response of a linear time invariant (LTI) system. It can be used to accurately model weakly nonlinear, specifically, polynomial nonlinearity systems. It has been used in the past for white-box model order reduction (MOR) to model frequency-domain performance metric quantities such as distortion in power amplifiers (PA). In this paper, we train a neural network from time-domain response of high-speed link buffers to extract multiple high-order kernels at once. Once the kernels are extracted, they can fully characterize the dynamics of the buffers of interest. Using the kernels, we demonstrate that time-domain response is straight-forward to obtain using super-, or multi-dimensional convolution. Previous work has used a shallow feed-forward neural network to train the system by using Gaussian noise as the identification signal. This is not convenient for the method to be compatible with existing computer-aided design tools. In this work, we directly use a pseudo random bit sequence (PRBS) to train the network. The proposed technique is more challenging because the PRBS has flat regions which have highly rich frequency spectrum and requires longer memory length, but allows the method to be compatible with existing simulation programs. We investigate different topologies including feed-forward neural network and recurrent neural network. Comparisons between training phase, inference phase, convergence are presented using different neural network topologies. The paper presents a numerical example using a 28Gbps data rate PAM4 transceiver to validate the proposed method against traditional simulation methods such as IBIS or SPICE level simulation for comparison in speed and accuracy. Using Volterra kernels promises a novel way to perform accurate nonlinear circuit simulation in the LTI system framework which is already well known and well developed. It can be conveniently incorporated into existing EDA frameworks.

[1]  Timo I. Laakso,et al.  Modelling of Nonlinear Power Amplifiers for Wireless Communications , 2004 .

[2]  C. S. Edrington,et al.  Modeling and simulation of shipboard nonlinear dynamic loads using Volterra kernels , 2013, 2013 IEEE Electric Ship Technologies Symposium (ESTS).

[3]  P. Vitaliy,et al.  Identification Accuracy of Nonlinear System based on Volterra Model in Frequency Domain , 2013 .

[4]  Kartikeya Mayaram,et al.  A modified-Volterra-series technique for improving the accuracy of quasi-static harmonic balance analysis in coupled device and circuit simulation , 2004, Proceedings of the IEEE 2004 Custom Integrated Circuits Conference (IEEE Cat. No.04CH37571).

[5]  William A. Gardner,et al.  Simplified methods for identifying the Volterra kernels of nonlinear systems , 1991, [1991] Proceedings of the 34th Midwest Symposium on Circuits and Systems.

[6]  M. J. Madero-Ayora,et al.  Volterra Behavioral Model for Wideband RF Amplifiers , 2007, IEEE Transactions on Microwave Theory and Techniques.

[7]  Peter R. Kinget,et al.  High-Frequency Distortion Analysis of Analog Integrated Circuits , 1999 .

[8]  S. Billings,et al.  Estimation of generalized frequency response functions for quadratically and cubically nonlinear systems , 2011 .

[9]  Stephen P. Boyd,et al.  Analytical Foundations of Volterra Series , 1984 .

[10]  Chunyu Xin,et al.  Radio frequency circuits for wireless receiver front-ends , 2005 .

[11]  Razvan Pascanu,et al.  On the difficulty of training recurrent neural networks , 2012, ICML.

[12]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[13]  Tianjian Lu,et al.  Transient Simulation for High-Speed Channels with Recurrent Neural Network , 2018, 2018 IEEE 27th Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS).

[14]  Tianjian Lu,et al.  Fast Transient Simulation of High-Speed Channels Using Recurrent Neural Network , 2019, ArXiv.

[15]  Graham F. Carey,et al.  Frequency Domain Kernel Estimation for 2nd-order Volterra Models Using Random Multi-tone Excitation , 2002, VLSI Design.

[16]  F. Hauske,et al.  Intrachannel Nonlinearity Compensation by Inverse Volterra Series Transfer Function , 2012, Journal of Lightwave Technology.

[17]  J. Schutt-Ainé,et al.  Volterra Kernel Extraction Through Monomial Power Series Feed Forward Neural Network for Behavior Modeling of High Speed I/O Buffer , 2019, 2019 Joint International Symposium on Electromagnetic Compatibility, Sapporo and Asia-Pacific International Symposium on Electromagnetic Compatibility (EMC Sapporo/APEMC).

[18]  T.J. Brazil,et al.  Behavioural modelling of RF power amplifiers using modified Volterra series in the time domain , 2004, High Frequency Postgraduate Student Colloquium, 2004.

[19]  Sheldon X.-D. Tan,et al.  Nonlinear Transient and Distortion Analysis via Frequency Domain Volterra Series , 2006 .

[20]  Weng Cho Chew,et al.  Volterra Series-Based Time-Domain Macromodeling of Nonlinear Circuits , 2016, IEEE Transactions on Components, Packaging and Manufacturing Technology.

[21]  M. Schetzen.  The Volterra and Wiener Theories of Nonlinear Systems , 1980 .

[22]  P. J. Lawrence.  Estimation of the Volterra functional series of a nonlinear system using frequency-response data , 1981 .

[23]  S. Billings,et al.  Volterra series truncation and reduction in the frequency domain for weakly nonlinear system , 2006 .

[24]  Zlatica Marinkovic,et al.  Volterra kernels extraction from neural networks for amplifier behavioral modeling , 2014, 2014 X International Symposium on Telecommunications (BIHTEL).

[25]  Xingjian Jing,et al.  The Generalized Frequency Response Functions and Output Spectrum of Nonlinear Systems , 2015 .

[26]  N. Wiener,et al.  Nonlinear Problems in Random Theory , 1964 .

[27]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[28]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[29]  E. Ngoya,et al.  A Two-Kernel Nonlinear Impulse Response Model for Handling Long Term Memory Effects in RF and Microwave Solid State Circuits. , 2006, 2006 IEEE MTT-S International Microwave Symposium Digest.

[30]  Thomas Hélie.  ON THE USE OF VOLTERRA SERIES FOR REAL-TIME SIMULATIONS OF WEAKLY NONLINEAR ANALOG AUDIO DEVICES: APPLICATION TO THE MOOG LADDER FILTER , 2006 .

[31]  G. Stegmayer.  Volterra series and neural networks to model an electronic device nonlinear behavior , 2004, 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541).

[32]  Jeffrey L. Elman,et al.  Finding Structure in Time , 1990, Cogn. Sci..

[33]  M. Youssef,et al.  Distortion analysis using signal flow graphs and Volterra series , 2003, 2003 46th Midwest Symposium on Circuits and Systems.

[34]  Koen Mouthaan,et al.  Evaluation of IBIS modelling techniques for signal integrity simulations without and with package parasitics , 2010, 2010 IEEE Electrical Design of Advanced Package & Systems Symposium.

[35]  Ying Chen,et al.  Volterra kernels extraction from frequency-domain data for weakly non-linear circuit time-domain simulation , 2017, 2017 IEEE Radio and Antenna Days of the Indian Ocean (RADIO).

[36]  Gary G. R. Green,et al.  Calculation of the Volterra kernels of non-linear dynamic systems using an artificial neural network , 1994, Biological Cybernetics.

[37]  N.B. Carvalho,et al.  A new Volterra series based orthogonal behavioral model for power amplifiers , 2005, 2005 Asia-Pacific Microwave Conference Proceedings.

[38]  W. Rugh.  Nonlinear System Theory: The Volterra / Wiener Approach , 1981 .

[39]  D. Marquardt.  An Algorithm for Least-Squares Estimation of Nonlinear Parameters , 1963 .

[40]  Georgina Stegmayer,et al.  Neural Networks and Volterra series for modeling new wireless communication devices , 2007, 2007 International Joint Conference on Neural Networks.