Concurrent surrogate modeling and adaptive sampling in load-pull measurements

Surrogate modeling combined with adaptive sampling allows to greatly lessen the number of samples required to reach the desired model accuracy. However, this comes at cost of substantial computational load. The cost can be limited by less frequent model extraction. This paper describes a surrogate modeling with varying intervals between the model extractions. The intervals are aligned to the sampling time in two concurrent processes. The approach has been evaluated in load-pull measurements of the unmatched gain block AG303 from TriQuint. The accuracy of the models has been preserved, while significant reduction of modeling time has been obtained.