Efficient Optimization Using Experimental Queries

In the process of hardware optimization, physical queries requiring laboratory experiments are often necessary. This is similar to optimization using software where queries are made to a computer model. In both the laboratory optimization and optimization using computer models, queries come at a cost: laboratory time or computer time. Finding efficient searches using a small number of queries on average is therefore motivated. In this paper, techniques used in computer search are shown to be transparently applicable to certain instances of hardware optimization. The hardware example presented is a load-pull peaksearch algorithm for power amplifier load-impedance optimization. The successful search shown in this paper allows high-resolution measurement of the maximum power with a significant reduction in the number of measured reflection-coefficient states. The use of computationally intelligent procedures for reducing time costs in design optimization using hardware has significant potential applications in a number of iterative experimental procedures performed in the laboratory.