Statistical analysis and yield optimization in practical RF and microwave designs

Abstract

This presentation provides a historical perspective of the advances in statistical design methodologies for microwave design tools. We focus on some of the early efforts by John Bandler to turn his research advances into practical results. We describe the formulation of the yield optimization problem, in particular the powerful one-sided ℓ 1 objective function. We discuss modeling requirements to properly account for statistical variations in IC manufacturing. We also describe applications of yield sensitivities and design of experiments (DOE) as well as a recent technique for fast CMOS mismatch analysis.Index Terms -Design methodology, design for manufacture, design optimization, microwave theory and techniques, Monte Carlo methods, statistical analysis, yield estimation, CADCAM.

Rapid Yield Estimation and Optimization of Microwave Structures Exploiting Feature-Based Statistical Analysis

“…In practice, the accuracy of a yield estimate using will degrade as moves away from so yield optimization should preferably be implemented as an iterative process, namely, (8) where are approximations of obtained by optimizing , i.e., a yield estimation function set up similarly to , but centered around rather than and using corresponding perturbations. The cost of each iteration is EM simulations, where is the number of designable parameters.…”

Parallel Computational Approach to Gradient Based EM Optimization of Microwave Structures

“…Significant progress has been achieved in several areas such as EM simulation challenges with large mesh and complex geometries [2]- [4], [80]- [83], yield-driven design [5]- [8], in the circuit simulation area for capturing the dynamic behavior of nonlinear components and systems [9]- [15], space mapping [22]- [31], and adjoint sensitivity [15]- [17], [67], [90].…”

Advanced Parametric Modeling and Yield Optimization Approaches for Microwave Passive Components

“…Furthermore, uncertainties introduced in the manufacturing process pose in-herent randomness on physical dimensions and material properties of microwave components. Under this consideration, performing yield-driven EM optimization is an essential step in manufacturability-driven designs of microwave passive components [25], [26]. Many techniques have been developed for yield-driven EM yield optimization in the last two decades.…”

“…2 Uncertainties, introduced by the manufacturing process, pose inherent randomness on both geometrical dimensions and material properties of microwave components. Under this consideration, performing yield-driven optimization becomes an essential step in manufacturability-driven designs in a time-to-market development environment [25], [26]. The two decades after 1970 have witnessed the development of various yield optimization approaches, such as Monte Carlo-based approaches [25], [26], [80]- [82] and geometrical approaches [83]- [85].…”

“…The objective function U (x 0 ) is congregated from the simulated responses related to design specifications for all the circuit outcomes randomly generated around the nominal point. Increase of the yield can be achieved by minimizing U (x 0 ) since such minimization leads to a better center in the feasible region [26]. In this chapter, we use the formulation in [25] as a starting point and propose a new and different objective function to facilitate yield-driven EM optimization.…”