A Hybrid Analog-Digital Parameter Optimizer for ASTRAC II

Abstract

A new automatic multiparameter optimizer for iter ative differential analyzers employs sequential ran dom parameter perturbation. The nominal parameter point changes whenever the random perturbations improve the system performance measure. Binary counters operate simple digital-to-analog converters to implement parameter storage, multiplication, and step-size changes. All-digital logic yields different types of random perturbations, viz., simple random walk, random walk with reflecting or absorbing bar riers, and various types of correlation over successive perturbations.