The University of Southampton
University of Southampton Institutional Repository

Support-vector-machine based automatic performance modelling and optimisation for analogue and mixed-signal designs

Support-vector-machine based automatic performance modelling and optimisation for analogue and mixed-signal designs
Support-vector-machine based automatic performance modelling and optimisation for analogue and mixed-signal designs
The growing popularity of analogue and mixed-signal (AMS) ASIC and SoC designs for communication applications has led to an increasing requirement for high efficiency performance modelling and optimisation methodologies in AMS synthesis systems. Recently, the support vector machine (SVM) method has been introduced into this challenging field. This research has studied the application of SVMs to AMS performance modelling in terms of the computational cost and prediction accuracy. A novel, general performance modelling methodology which could be applied to an arbitrary AMS system has been developed and integrated into an AMS performance optimisation system.
The contributions of this research can be summarised as follows: firstly, a new performance modelling methodology based on automatic generation of knowledge databases for AMS performance modelling using SVM techniques has been developed. Two performance model construction methods have been implemented: a linearly graded method and a support vector regression method. They can provide a basis for efficient design space exploration. Both methods construct performance models for AMS designs in a fully automatic way. A simulator, a performance extractor and an SVM trainer have been developed and integrated into a practical demonstrator system.
Secondly, a knowledge-based AMS performance optimisation system has been developed for system-level and circuit-level designs. Knowledge data bases created using the proposed methodology are reusable. This has been verified by the application of two optimisation methods, a dedicated genetic optimisation algorithm and the standard pattern search technique.
Finally, the proposed performance model construction methodology and the underlying performance optimisation system have been validated using two complex case studies. The first example is a high-level model of a mixed-signal sigma-delta modulator which comprises most of the practical design nonlinearities and imperfections that are known to affect the performance. The proposed method is able to find designs with significantly superior key performance figures compared to those obtained by the standard sigma-delta modulator design procedure. The second example is a radio frequency (RF) range Colpitts filter for silicon implementations. The non-standard structure of this filter with a lossy spiral inductor makes this type of circuit difficult to handle by standard filter synthesis methods. The proposed performance optimisation method has found solutions which satisfy performance requirements in cases where a standard manual filter design procedure failed.
Ren, Xianqiang
c7f83e76-c48c-4690-86e3-f5fddb5e2198
Ren, Xianqiang
c7f83e76-c48c-4690-86e3-f5fddb5e2198
Kazmierski, Tomasz
a97d7958-40c3-413f-924d-84545216092a

Ren, Xianqiang (2008) Support-vector-machine based automatic performance modelling and optimisation for analogue and mixed-signal designs. University of Southampton, School of Electronics and Computer Science, Doctoral Thesis, 203pp.

Record type: Thesis (Doctoral)

Abstract

The growing popularity of analogue and mixed-signal (AMS) ASIC and SoC designs for communication applications has led to an increasing requirement for high efficiency performance modelling and optimisation methodologies in AMS synthesis systems. Recently, the support vector machine (SVM) method has been introduced into this challenging field. This research has studied the application of SVMs to AMS performance modelling in terms of the computational cost and prediction accuracy. A novel, general performance modelling methodology which could be applied to an arbitrary AMS system has been developed and integrated into an AMS performance optimisation system.
The contributions of this research can be summarised as follows: firstly, a new performance modelling methodology based on automatic generation of knowledge databases for AMS performance modelling using SVM techniques has been developed. Two performance model construction methods have been implemented: a linearly graded method and a support vector regression method. They can provide a basis for efficient design space exploration. Both methods construct performance models for AMS designs in a fully automatic way. A simulator, a performance extractor and an SVM trainer have been developed and integrated into a practical demonstrator system.
Secondly, a knowledge-based AMS performance optimisation system has been developed for system-level and circuit-level designs. Knowledge data bases created using the proposed methodology are reusable. This has been verified by the application of two optimisation methods, a dedicated genetic optimisation algorithm and the standard pattern search technique.
Finally, the proposed performance model construction methodology and the underlying performance optimisation system have been validated using two complex case studies. The first example is a high-level model of a mixed-signal sigma-delta modulator which comprises most of the practical design nonlinearities and imperfections that are known to affect the performance. The proposed method is able to find designs with significantly superior key performance figures compared to those obtained by the standard sigma-delta modulator design procedure. The second example is a radio frequency (RF) range Colpitts filter for silicon implementations. The non-standard structure of this filter with a lossy spiral inductor makes this type of circuit difficult to handle by standard filter synthesis methods. The proposed performance optimisation method has found solutions which satisfy performance requirements in cases where a standard manual filter design procedure failed.

Text
PhD_thesis_XRen.pdf - Other
Download (5MB)

More information

Published date: October 2008
Organisations: University of Southampton

Identifiers

Local EPrints ID: 64880
URI: http://eprints.soton.ac.uk/id/eprint/64880
PURE UUID: c6aaf49c-40d5-4a9f-8add-adbdb6164114

Catalogue record

Date deposited: 21 Jan 2009
Last modified: 13 Mar 2019 20:20

Export record

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×