Constrained Diffusion for Inverse Design of Fabricable Multi-Layer RF Passives from Partial S-Parameter Specifications

We present a constrained diffusion approach that generates fabricable passive com- ponents from partial S-parameter inputs spanning 1–20 GHz on an 8 ×8 mm board discretized at 64×64 pixels with sub-pixel grayscale metallization. The framework generates two-layer copper layouts with vias, with hard physical constraints on feed locations enforced throughout the denoising trajectory, multi-modal condi- tioning on partial S-parameter specifications, variable port locations, dielectric properties, and reference topology. Candidate designs are generated in seconds, with the constrained sampler achieving an 89% valid-board rate, as opposed to 8.5% unconstrained, and surrogate-predicted S-parameters matching targets with sub-decibel average weighted mean absolute error. We validate the approach with two fabricated designs on RO4003C: a manufacturable alternative to a hairpin filter whose coupling gaps violate fabrication rules, and a combline bandpass filter designed from scratch given only target S-parameters.