CAD for statistical analysis and design of microwave circuits

Abstract

Statistical circuit design is essentially a method of design for reliability and high yield. For all but the smallest circuits, this is a CAD problem. Within this problem there are three subproblems: design centering, tolerance assignment, and variability reduction. This paper concentrates on design centering. Yield optimization is inherently a difficult problem because yield cannot, in general, be calculated exactly; only estimates of the yield are available. Furthermore, yield gradients are not available when some yield estimation techniques are used. This paper presents a review of useful definitions using parameter space concepts. Yield estimation techniques and some of their properties are also presented. The two distinct yield optimization strategies commercially available today, statistical and deterministic, are each discussed in detail. Examples of each are given.

Interpolation modeling of S‐parameter data bases for use in Monte Carlo simulations

“…Here we examine the effect of the PDF error on Monte Carlo yield estimation [16]. The error in matching estimated PDF to real PDF creates error in the calculated value of the yield; we will call this "yield error."…”

Reduction of microwave amplifier gain variance

“…and denominator terms of Eqs. 3 To measure the effect of the embedding network, the following scatter metrics were used to measure the variance of the S-parameters of the HEMT and the embedded HEMT:…”

Yield sensitivity study of A1GaAs/GaAs high electron mobility transistor

“…These parameters are used in the device model program to calculate device performances. The choice of parameter statistics is an important part of a study such as this [14]. It is likely that L and Z are correlated, which is not accounted for in this study.…”