Automated Architecture-Agnostic Synthesis of Wideband Load-Modulated RF GaN Power Amplifiers

6G radios require GaN power amplifiers with more bandwidth and more efficiency at once, and efficiency is decided at back-off, where a modulated signal spends most of its time. Load-modulated amplifiers, such as Doherty, load-modulated balanced amplifiers (LMBA) and outphasing amplifiers, address that need. The networks that split the input and combine the outputs are still designed by hand, one architecture and one frequency at a time. This PhD will make that step automatic and architecture-agnostic. The candidate will:

• extend one synthesis theory (including a proven realizability criterion) from Doherty and LMBA to outphasing comprising 2 or more amplifiers and assuming a lossless splitting and combining networks;
• measure load-pull with its efficiency contours across frequency and drive level on imec GaN devices, and make the contour the unit in which every compromise is valued;
• make wideband network synthesis automatic: search topologies and element values together, with learned search where exhaustive search fails and with the constraint on buildable passive networks;
• design, fabricate and measure GaN demonstrators of at least two architectures from one data set, and compare measurement with the predictions.

Given a device and a specification, the method and its tool return an architecture, its networks and their predicted bandwidth. The realized hardware shows how far the prediction holds. The PhD-candidate leaves with a theorem-backed method, a toolchain validated on the bench, and published GaN prototypes: network theory, algorithm design and RF measurement in one profile.