Authors: N. Lu
Affilation: IBM, United States
Pages: 817 - 820
Keywords: statistical models, SPICE models, process variations, distance-dependent mismatch, tracking, across-chip variations
For analog circuit designers, both reducing device mismatch and having a good device mismatch model are very important. Pelgrom characterized device mismatch between two devices separated by a finite distance D as sqrt[a^2/(W*L) + b^2*D^2]. While the modeling of nearby mismatch (D = 0) in compact models is straightforward, the modeling of distance-dependent mismatch (D > 0) among several devices remains a challenge. Further, mismatch and intra-die/across-chip variations (ACV) are generally treated as two separate problems. In this paper, for the first time, we (i) link characterizations among nearby mismatch, finite-distance mismatch, ACV, and correlation into a unified concept and description and improve Pelgrom’s mismatch characterization, (ii) highlight the challenges in implementing Pelgrom’s distance-dependent mismatch characterization in compact models, and (iii) present a set of compact solutions for modeling mismatch and ACV. We use only two additional model instance parameters in our general compact solution. They are the physical location (x, y) of each device instance on a chip. When netlisting a device instance, none of other device’s coordinates are needed. We also apply our solution to model distance-dependent mismatch/ACV in polysilicon and diffused resistors.