EM-based yield optimization exploiting trust-region optimization and space mapping technology

Abstract

Design centering is a design problem which looks for nominal values of circuit parameters that maximize the probability of satisfying the design specification (yield function). Direct yield optimization of electromagnetic (EM)‐based circuits is obstructed by the high expense of EM simulations required in the yield estimation process. Also, the absence of any gradient information represents an obstacle against the optimization process. In this article, a new approach for design centering and yield optimization of EM‐based circuits is introduced. In the proposed approach, the generalized space mapping (SM) technique is incorporated with a derivative‐free trust region optimization method (NEWUOA). Moreover, a variance reduction sampling technique is implemented in the yield estimation process. Two techniques suitable for the microwave circuit design centering process are introduced. The first technique exploits the surrogate developed using any circuit optimizer, for example, minimax optimizer, in the yield maximization process. While the second technique iteratively constructs and updates an SM surrogate during the yield optimization process. Our novel approach is illustrated by practical examples showing its efficiency. One of the examples is entirely designed within the sonnet em environment. © 2014 Wiley Periodicals, Inc. Int J RF and Microwave CAE 25:474–484, 2015.

Statistical Design Centering Optimization of 1d Photonic Crystal Filters

“…In this paper, we propose an optimization approach, belonging to the class of statistical design centering optimization approaches [18][19][20][21][22], in which the objective is to maximize the yield function explicitly. In this approach, the design centering problem is formulated as an unconstrained yield optimization problem.…”

Space mapping surrogate‐based microwave circuit design centering using a new statistical technique

“…Generally, there are two main approaches for design centering. The first is a statistical approach which explicitly optimizes the estimated yield values using any statistical analysis approach in a straightforward way regardless the size of the problem or its convexity [2,11,24,[29][30][31][32][33]. The accuracy of the statistical estimation of the yield values does not depend on the number of parameters and performance features, but on the sample size [34].…”

“…To overcome this, computationally cheap space mapping (SM) surrogate models can be employed instead of the computationally expensive high-fidelity fine model [3][4][5][6][7][8]. Several techniques for SM surrogate-based microwave circuit design centering have been proposed in the literature [2,[9][10][11][12][13]. The main idea of this approach of design centering is that SM surrogates are used to create a sequence of iteratively updated feasible region approximations.…”

“…The solution of this optimization problem leads to the required design center. It is to be noticed that the work presented in Hassan et al [11] makes use of a SM surrogate, generated by minimax optimizer, in the yield maximization process. This technique may be of great benefits to whom interested in modeling and constructing surrogates as it can be easily integrated with any SM surrogate-based framework.The work presented in this article introduces a new statistical SM surrogate-based design centering technique for microwave circuit's applications.…”

Statistical design centering of RF cavity linear accelerator via non-derivative trust region optimization

“…This approach has fast convergence properties for convex and small dimensional problems. The second is a statistical approach which explicitly optimizes the yield function in a straightforward way regardless the size of the problem or its convexity [5]- [6].…”