Massively multi-topology sizing of analog integrated circuits

Created by W.Langdon from gp-bibliography.bib Revision:1.4549

  author =       "Pieter Palmers and Trent McConnaghy and 
                 Michiel Steyaert and Georges Gielen",
  title =        "Massively multi-topology sizing of analog integrated
  booktitle =    "Design, Automation Test in Europe Conference
                 Exhibition, 2009. DATE '09",
  year =         "2009",
  pages =        "706--711",
  address =      "Nice",
  month =        "20-24 " # apr,
  keywords =     "genetic algorithms, genetic programming, EHW, MOGA,
                 Pareto, SPICE, analogue integrated circuits, integrated
                 circuit design, network topology, operational
                 amplifiers, SPICE accuracy, analog integrated circuit,
                 multitopology sizing, optimisation-based constraint
                 satisfaction, symmetrical transconductance amplifier,
                 Analog circuits, Analog integrated circuits, Circuit
                 topology, Constraint optimization, Knowledge based
                 systems, Libraries, MOS devices, Mirrors, SPICE,
  isbn13 =       "978-1-4244-3781-8",
  URL =          "",
  DOI =          "doi:10.1109/DATE.2009.5090756",
  size =         "6 pages",
  abstract =     "This paper demonstrates a system that performs
                 multi-objective sizing across 100,000 analog circuit
                 topologies simultaneously, with SPICE accuracy. It
                 builds on a previous system, MOJITO, which searches
                 through 3500 topologies defined by a
                 hierarchically-organised set of 30 analog blocks. This
                 paper improves MOJITO's results quality via three key
                 extensions. First, it enlarges the block library to
                 enable symmetrical transconductance amplifiers and
                 more. Second, it improves initial topology diversity
                 via optimisation-based constraint satisfaction. Third,
                 it maintains topology diversity during search via a
                 novel multi-objective selection mechanism, dubbed
                 TAPAS. MO-JITO+TAPAS is demonstrated on a problem with
                 6 objectives, returning a tradeoff holding 17438
                 nondominated designs. The tradeoff is comprised of 152
                 unique topologies that include the newly-introduced
                 topologies. 59 designs across 12 topologies designs
                 outperform an expert-designed reference circuit.",
  notes =        "2nd author should be Trent McConaghy. Also known as

Genetic Programming entries for Pieter Palmers Trent McConaghy Michiel Steyaert Georges G E Gielen