Showing posts with label Many-Objective Evolutionary Algorithm. Show all posts
Showing posts with label Many-Objective Evolutionary Algorithm. Show all posts

Oct 6, 2020

[paper] gm/ID-Based Sizing for Analog ICs

Tuotian Liao and Lihong Zhang
An LDE-Aware gm/ID-Based Hybrid Sizing Method for Analog Integrated Circuits
Analog Integrated Circuits. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 1–1. doi:10.1109/tcad.2020.3025068 

Abstract: Layout-dependent effects (LDEs) have become increasingly more important in the synthesis of analog integrated circuits. In this paper, a two-phase hybrid sizing method for high performance analog circuits is proposed. It consists of gm/ID-based device characterization, circuit modeling, sensitivity-based constraints for LDEs, and mixed-integer nonlinear programming in the first phase, and many-objective evolutionary algorithm (many OEA) based sizing in the second phase. In the first phase, accurate device characterization is handled with little modeling effort thanks to the gm/ID design methodology. Then the LDE parameters that are linked to the normalized DC current are further optimized with the aid of sensitivity analysis. Thus, a variety of electrical, geometrical, and LDE-related constraints can be conveniently integrated into modeling of the sizing problem. In the second phase, the many OEA-based sizing refiner can further optimize the LDE parameters by using more detailed layout information via our proposed model. A new floor plan variation scheme is also applied to improve computation efficiency and enhance optimization effectiveness. The experimental results demonstrate high efficacy of our proposed methodology in LDE-aware analog sizing optimization.
Fig: Module-level of the LDE-aware gm/ID EA two-phase synthesis flow

Thanks to the contribution of the EKV model [1], inversion coefficient (IC) can be used to indicate the biasing inversion level of a MOSFET. This helped Binkley et al. [2] change the design freedom from the conventional W, L, and ID to IC, L, and ID. Since IC is related to DC bias, device geometry, and device characteristics (e.g., gm/ID), it can reflect performance tradeoff (e.g., intrinsic gain vs. bandwidth) of a single MOSFET. In [3], bias information rather than gm/ID parameters was set as variables, while a small-scale LUT was built to find MOSFET aspect ratio (i.e., W/L) and eventually W.

Aknowlegement: This work was supported in part by the Natural Sciences and Engineering Research Council of Canada (NSERC), Canada Foundation for Innovation (CFI), Research and Development Corporation (RDC) of Newfoundland and Labrador, and Memorial University of Newfoundland.

References:
  1. C. Enz, F. Chicco, and A. Pezzotta, “Nanoscale MOSFET modeling: Part 1: The simplified EKV model for the design of low-power analog circuits,” IEEE Solid-State Circuits Mag., vol. 9, no. 3, pp. 26–35, 2017.
  2. D. M. Binkley, C. E. Hopper, S.D. Tucker, B.C. Moss, J. M. Rochelle, and, D. P. Foty, “A CAD methodology for optimizing transistor current and sizing in analog CMOS design,” IEEE Trans. Comput-Aided Design Integr. Circuits Syst., vol. 22 no. 2, pp. 225-237, 2003.
  3. C.-W. Lin, P.-D. Sue, Y.-T. Shyu, and S.-J. Chang, “A bias-driven approach for automated design of operational amplifiers,” in Proc. Int.