Optimization in Electronic Design Automation

Course description

(I) Definition of terms and overview of optimization methods. Unconstrained methods: first and second order gradient-based algorithms, direct algorithms and metaheuristics. Constrained methods: parameter constrains and penalty functions. Definition of analog circuit sensitivity and its role in gradient-based algorithms. Optimization of circuit topology.

(II) Parametric optimization. Measurement definition and cost function formulation. Properties of cost functions. Nominal optimization. Effect of operating parameters and random manufacturing variations. Optimization across corners. Worst case analysis. Yield and yield optimization. A demo run of an optimization tool and result interpretation.

(III) A medium size circuit is run by each student on a personal computer. Parallelization of the optimization process. A large analog optimization case is run on a computer farm and the results are analyzed.

Course is carried out on study programme

Objectives and competences

Theoretical knowledge of optimization procedures in EDA. Gaining practical experience with a circuit optimization tool.

Learning and teaching methods

Individual consultation, directed seminar work, independent project work.

Intended learning outcomes

Upon successful completion of this course, the students should be able to:

  • determine the explicitly limited parameter space for a given design problem
  • formulate a respective cost function for a given design problem
  • add corners to the cost function of a given design problem
  • select an appropriate optimization method for a given design problem
  • independently run a simple design case using the PyOPUS tool box on a single computer
  • contribute to a group project running a large optimization case with PyOPUS on a computer farm

Reference nosilca

  1. BÜRMEN, Arpad, FAJFAR, Iztok. Mesh adaptive direct search with simplicial Hessian update. Computational optimization and applications. [Print ed.]. 2019, vol. 74, str. 645-667. [COBISS.SI-ID 12715348]
  2. ROJEC, Žiga, BÜRMEN, Arpad, FAJFAR, Iztok. Analog circuit topology synthesis by means of evolutionary computation. Engineering applications of artificial intelligence, ISSN 0952-1976. [Print ed.], Apr. 2019, vol. 80, str. 48-65, ilustr. https://www.sciencedirect.com/science/article/pii/S0952197619300119, doi: 10.1016/j.engappai.2019.01.012. [COBISS.SI-ID 12361044]
  3. FAJFAR, Iztok, PUHAN, Janez, BÜRMEN, Arpad. Evolving a Nelder-Mead algorithm for optimization with genetic programming. Evolutionary computation, 2017, vol. 25, no. 3. [COBISS.SI-ID 11276628]
  4. BÜRMEN, Arpad, OLENŠEK, Jernej, TUMA, Tadej. Mesh adaptive direct search with second directional derivative-based Hessian update. Comput Optim Appl, Springer 2015, DOI 10.1007/s10589-015-9753-5

BÜRMEN, Arpad, PUHAN, Janez, TUMA, Tadej. Grid Restrained Nelder-Mead Algorithm. Computational optimization and applications, 2006, [Online ed.], [17] str. [COBISS.SI-ID 5222996]

Study materials

TUMA, Tadej, BÜRMEN, Arpad Circuit Simulation with SPICE OPUS, Theory and Practice. Springer, 2009, Approx. 480 p. 158 illus., Hardcover ISBN: 978-0-8176-4866-4. [COBISS.SI-ID 7248980]

Bodi na tekočem

Univerza v Ljubljani, Fakulteta za elektrotehniko, Tržaška cesta 25, 1000 Ljubljana

E:  dekanat@fe.uni-lj.si T:  01 4768 411