Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MOTIF-RF: Multi-template On-chip Transformer Synthesis Incorporating Frequency-domain Self-transfer Learning for RFIC Design Automation

Houbo He, Yizhou Xu, Lei Xia, Yaolong Hu, Fan Cai, Taiyun Chi

TL;DR

This work tackles the need for accurate, fast XFMR S-parameter surrogates to enable AI-assisted specs-to-GDS for RFICs. It systematically benchmarks four ML architectures (MLP, CNN, UNet, GT) on XFMR datasets and introduces a frequency-domain self-transfer learning scheme that exploits spectral continuity to boost prediction accuracy by about 30–50%. Building on these surrogates, the authors demonstrate a CMA-ES-based inverse-design workflow for impedance matching, validating fast convergence and reliable design outcomes across multiple XFMR topologies. The findings offer practical guidance for RFIC designers seeking AI-driven automation and provide a pathway toward integrated, end-to-end AI-assisted RFIC design flows.

Abstract

This paper presents a systematic study on developing multi-template machine learning (ML) surrogate models and applying them to the inverse design of transformers (XFMRs) in radio-frequency integrated circuits (RFICs). Our study starts with benchmarking four widely used ML architectures, including MLP-, CNN-, UNet-, and GT-based models, using the same datasets across different XFMR topologies. To improve modeling accuracy beyond these baselines, we then propose a new frequency-domain self-transfer learning technique that exploits correlations between adjacent frequency bands, leading to around 30%-50% accuracy improvement in the S-parameters prediction. Building on these models, we further develop an inverse design framework based on the covariance matrix adaptation evolutionary strategy (CMA-ES) algorithm. This framework is validated using multiple impedance-matching tasks, all demonstrating fast convergence and trustworthy performance. These results advance the goal of AI-assisted specs-to-GDS automation for RFICs and provide RFIC designers with actionable tools for integrating AI into their workflows.

MOTIF-RF: Multi-template On-chip Transformer Synthesis Incorporating Frequency-domain Self-transfer Learning for RFIC Design Automation

TL;DR

This work tackles the need for accurate, fast XFMR S-parameter surrogates to enable AI-assisted specs-to-GDS for RFICs. It systematically benchmarks four ML architectures (MLP, CNN, UNet, GT) on XFMR datasets and introduces a frequency-domain self-transfer learning scheme that exploits spectral continuity to boost prediction accuracy by about 30–50%. Building on these surrogates, the authors demonstrate a CMA-ES-based inverse-design workflow for impedance matching, validating fast convergence and reliable design outcomes across multiple XFMR topologies. The findings offer practical guidance for RFIC designers seeking AI-driven automation and provide a pathway toward integrated, end-to-end AI-assisted RFIC design flows.

Abstract

This paper presents a systematic study on developing multi-template machine learning (ML) surrogate models and applying them to the inverse design of transformers (XFMRs) in radio-frequency integrated circuits (RFICs). Our study starts with benchmarking four widely used ML architectures, including MLP-, CNN-, UNet-, and GT-based models, using the same datasets across different XFMR topologies. To improve modeling accuracy beyond these baselines, we then propose a new frequency-domain self-transfer learning technique that exploits correlations between adjacent frequency bands, leading to around 30%-50% accuracy improvement in the S-parameters prediction. Building on these models, we further develop an inverse design framework based on the covariance matrix adaptation evolutionary strategy (CMA-ES) algorithm. This framework is validated using multiple impedance-matching tasks, all demonstrating fast convergence and trustworthy performance. These results advance the goal of AI-assisted specs-to-GDS automation for RFICs and provide RFIC designers with actionable tools for integrating AI into their workflows.

Paper Structure

This paper contains 18 sections, 2 equations, 10 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. XFMR Fundamentals and S-parameters
  4. Dataset Generation
  5. Surrogate Modeling
  6. ML Surrogate Models
  7. Tabular-based Method Using Multi-layer Perceptron (MLP)
  8. Image-based Methods Using CNN and UNet
  9. Graph Transformer (GT)-based Method
  10. Proposed Frequency-domain Self-transfer Learning
  11. Model Evaluation
  12. Evaluation of Different Modeling Approaches
  13. Evaluation of Frequency-domain Self-transfer Learning
  14. Inverse Design
  15. Impedance-matching Task and Its Cost Function
  16. ...and 3 more sections

Figures (10)

  • Figure 1: “Specs to GDS” for AI-assisted design automation of XFMRs in RFICs, incorporating forward surrogate models and inverse design algorithms.
  • Figure 2: (a) EM model of a generic $M$:$N$ XFMR and (b) its geometric parameters.
  • Figure 3: (a) EM model of a generic 8-shaped-inductor-based XFMR. (b) EM model of a generic parallel-inductor-based XFMR.
  • Figure 4: (a) CNN-based and (b) UNet-based model architectures.
  • Figure 5: (a) XFMR segment embedding. (b) GT-based model architecture.
  • ...and 5 more figures