Aug 6 (Tue) @ 11:00am: ”Millimeter Wave Power Amplifiers: Embedding Techniques and Rapid Optimization,” Everett O'Malley, ECE PhD Defense

Abstract

Communication systems in millimeter-wave bands promise unprecedented access to high data rates in wireless links. Power amplifiers account for a large fraction of transmitter power dissipation. Consequently, increasing the efficiency of power amplifiers is critical for improving the efficiency of transmitters. The design of power amplifiers, and circuit design in general, consists of an exploration through the permutations of circuit topologies. This work offers two different methods for searching the myriad possible designs.

First, we describe an automation approach that we apply to designing D-band power amplifiers. We generated circuit models from EM-simulation and loadpull. The optimization then proceeds from these circuit models, without any further need to employ time consuming harmonic balance or electromagnetic simulations. As a result, the per iteration time cost of our method is much faster than in similar works, which used Bayesian optimization techniques to reduce the number of iterations used. We use the improved optimization efficiency to investigate 49 matching-network candidates for each stage of four stage amplifiers. We applied the optimization approach for three different amplifiers in two process-technologies. In all cases the optimization time was less than seven seconds. In measurement, the amplifiers achieve peak power added efficiency (PAE) between 13.6% and 15.5% with less than a 3% shift in peak PAE frequency from the 140 GHz design target. Measured output powers ranged between 14.5 and 14.9 dBm.

Second, we describe methods for designing gain-boosted amplifiers that simultaneously satisfy both gain matching and power matching conditions. Gain-boosted amplifiers are popular in literature at high fractions of fmax, but the usual design equations assume gain matching, which may not present the core device with a load impedance that results in good power performance. We identify all possible port-parameter embedding networks that result in the canonical unilateral form, parameterized in terms of the load impedance presented to the core amplifier. The parameterization of unilateralizing embedding networks is then used to find the embedding networks that simultaneously meet the gain matching and noise matching conditions, thus achieving minimum noise measure. A power amplifier design example is given at 2.5 GHz. We describe an embedding-network synthesis approach based on identifying the directional coupler that must be equivalent to both the target 4-port embedding network and to the candidate design structure.

Bio

Everett O'Malley received the B.S. degree in electrical engineering and physics from Northeastern University, Boston, MA, USA, in 2018 and the M.S. degree in electrical engineering from the University of California, Santa Barbara, Santa Barbara, CA, USA, in 2020. Since 2018, he has been a Graduate Student Researcher with the University of California, Santa Barbara. His research interests include the use of feedback in mmwave amplifiers to improve noise and power  performance, and also machine learning approaches for mmwave integrated circuit design automation and optimization.

Hosted by: Professor James F. Buckwalter

Submitted by: Everett O'Malley <omalley@ucsb.edu>