AI-Based Transformer Matching Network Synthesis Using a Data-Efficient Two-Step Method

Simone G, Tas D, Schneider F, Schilling R, Engelmann A, Maiwald T, Breun S, Hagelauer A, Franchi N (2026)

Publication Language: English

Publication Status: Accepted

Publication Type: Unpublished / Preprint

Future Publication Type: Conference contribution

Publication year: 2026

Event location: Los Angeles (Hollywood), California, USA

Abstract

This work presents an artificial intelligence-driven

approach for automated on-chip impedance-matching (IM) transformer

design, achieving layout generation within seconds. Unlike

direct prediction approaches, where machine learning models

directly estimate the optimal transformer geometry based on

input and output impedances, this method follows a two-step

process. This new approach significantly reduces the required

training data for the neural network while maintaining high

accuracy and efficiency. As a proof of concept, the synthesis of a

D-band transformer in IHP’s SG13G2 technology is presented

and optimized using the algorithm proposed in this paper.

The approach is versatile and applicable across technologies,

frequency ranges, and matching network structures, making it a

scalable solution for automated IM design.

Authors with CRIS profile

Involved external institutions

Fraunhofer-Einrichtung für Mikrosysteme und Festkörper-Technologien (EMFT) / Fraunhofer Research Institution for Microsystems and Solid State Technologies Germany (DE) Otto-von-Guericke-Universität Magdeburg Germany (DE)

How to cite

APA:

Simone, G., Tas, D., Schneider, F., Schilling, R., Engelmann, A., Maiwald, T.,... Franchi, N. (2026). AI-Based Transformer Matching Network Synthesis Using a Data-Efficient Two-Step Method. (Unpublished, Accepted).

MLA:

Simone, Gianluca, et al. AI-Based Transformer Matching Network Synthesis Using a Data-Efficient Two-Step Method. Unpublished, Accepted. 2026.

BibTeX: Download