Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm

Tuomas Haarnoja, Kari Tammi, Kai Zenger

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

3 Citations (Scopus)

Abstract

Filtered-x Least Mean Squares (FXLMS) algorithm is a well-known method for adapting feedforward FIR filters to compensate selected disturbances in the system output. However, the nonlinearities in the algorithm complicates its analysis. Nevertheless, it can be shown that if a periodic, synchronously sampled reference signal is considered, FXLMS algorithm can be analyzed by a LTP system to which the conventional linear analysis tools can be applied. Earlier, this method has been applicable only for systems, in which the model of the plant is unity. This paper extends the analysis to cover also systems with an arbitrary, linear plant model. The new approach relieves some of the other assumptions made in the earlier analysis. The effect of filter length, convergence coefficient, and model errors to the stability and to the rate of convergence of FXLMS are also studied by using the LTP representation
Original languageEnglish
Title of host publicationProceedings
Subtitle of host publication18th International Congress on Sound and Vibration 2011, ICSV 19
EditorsMalcolm J. Crocker, Marek Pawelczyk, Nickolay Ivanov
PublisherInternational Institute of Acoustics and Vibration IIAV
Pages203-210
Volume1
ISBN (Print)978-1-61839-259-6
Publication statusPublished - 2011
MoE publication typeA4 Article in a conference publication
Event18th International Congress on Sound & Vibration, ICSV18 - Rio de Janeiro, Brazil
Duration: 10 Jul 201114 Jul 2011

Conference

Conference18th International Congress on Sound & Vibration, ICSV18
Abbreviated titleICSV18
CountryBrazil
CityRio de Janeiro
Period10/07/1114/07/11

Fingerprint

FIR filters

Keywords

  • FxLMS
  • convergence
  • active vibration control

Cite this

Haarnoja, T., Tammi, K., & Zenger, K. (2011). Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm. In M. J. Crocker, M. Pawelczyk, & N. Ivanov (Eds.), Proceedings: 18th International Congress on Sound and Vibration 2011, ICSV 19 (Vol. 1, pp. 203-210). International Institute of Acoustics and Vibration IIAV.
Haarnoja, Tuomas ; Tammi, Kari ; Zenger, Kai. / Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm. Proceedings: 18th International Congress on Sound and Vibration 2011, ICSV 19. editor / Malcolm J. Crocker ; Marek Pawelczyk ; Nickolay Ivanov. Vol. 1 International Institute of Acoustics and Vibration IIAV, 2011. pp. 203-210
@inproceedings{b284bdaff9d6469399c1c1d9c34c1961,
title = "Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm",
abstract = "Filtered-x Least Mean Squares (FXLMS) algorithm is a well-known method for adapting feedforward FIR filters to compensate selected disturbances in the system output. However, the nonlinearities in the algorithm complicates its analysis. Nevertheless, it can be shown that if a periodic, synchronously sampled reference signal is considered, FXLMS algorithm can be analyzed by a LTP system to which the conventional linear analysis tools can be applied. Earlier, this method has been applicable only for systems, in which the model of the plant is unity. This paper extends the analysis to cover also systems with an arbitrary, linear plant model. The new approach relieves some of the other assumptions made in the earlier analysis. The effect of filter length, convergence coefficient, and model errors to the stability and to the rate of convergence of FXLMS are also studied by using the LTP representation",
keywords = "FxLMS, convergence, active vibration control",
author = "Tuomas Haarnoja and Kari Tammi and Kai Zenger",
note = "Project code: 26153",
year = "2011",
language = "English",
isbn = "978-1-61839-259-6",
volume = "1",
pages = "203--210",
editor = "Crocker, {Malcolm J. } and Pawelczyk, {Marek } and Ivanov, {Nickolay }",
booktitle = "Proceedings",
publisher = "International Institute of Acoustics and Vibration IIAV",
address = "United States",

}

Haarnoja, T, Tammi, K & Zenger, K 2011, Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm. in MJ Crocker, M Pawelczyk & N Ivanov (eds), Proceedings: 18th International Congress on Sound and Vibration 2011, ICSV 19. vol. 1, International Institute of Acoustics and Vibration IIAV, pp. 203-210, 18th International Congress on Sound & Vibration, ICSV18, Rio de Janeiro, Brazil, 10/07/11.

Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm. / Haarnoja, Tuomas; Tammi, Kari; Zenger, Kai.

Proceedings: 18th International Congress on Sound and Vibration 2011, ICSV 19. ed. / Malcolm J. Crocker; Marek Pawelczyk; Nickolay Ivanov. Vol. 1 International Institute of Acoustics and Vibration IIAV, 2011. p. 203-210.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm

AU - Haarnoja, Tuomas

AU - Tammi, Kari

AU - Zenger, Kai

N1 - Project code: 26153

PY - 2011

Y1 - 2011

N2 - Filtered-x Least Mean Squares (FXLMS) algorithm is a well-known method for adapting feedforward FIR filters to compensate selected disturbances in the system output. However, the nonlinearities in the algorithm complicates its analysis. Nevertheless, it can be shown that if a periodic, synchronously sampled reference signal is considered, FXLMS algorithm can be analyzed by a LTP system to which the conventional linear analysis tools can be applied. Earlier, this method has been applicable only for systems, in which the model of the plant is unity. This paper extends the analysis to cover also systems with an arbitrary, linear plant model. The new approach relieves some of the other assumptions made in the earlier analysis. The effect of filter length, convergence coefficient, and model errors to the stability and to the rate of convergence of FXLMS are also studied by using the LTP representation

AB - Filtered-x Least Mean Squares (FXLMS) algorithm is a well-known method for adapting feedforward FIR filters to compensate selected disturbances in the system output. However, the nonlinearities in the algorithm complicates its analysis. Nevertheless, it can be shown that if a periodic, synchronously sampled reference signal is considered, FXLMS algorithm can be analyzed by a LTP system to which the conventional linear analysis tools can be applied. Earlier, this method has been applicable only for systems, in which the model of the plant is unity. This paper extends the analysis to cover also systems with an arbitrary, linear plant model. The new approach relieves some of the other assumptions made in the earlier analysis. The effect of filter length, convergence coefficient, and model errors to the stability and to the rate of convergence of FXLMS are also studied by using the LTP representation

KW - FxLMS

KW - convergence

KW - active vibration control

UR - http://www.proceedings.com/13257.html

M3 - Conference article in proceedings

SN - 978-1-61839-259-6

VL - 1

SP - 203

EP - 210

BT - Proceedings

A2 - Crocker, Malcolm J.

A2 - Pawelczyk, Marek

A2 - Ivanov, Nickolay

PB - International Institute of Acoustics and Vibration IIAV

ER -

Haarnoja T, Tammi K, Zenger K. Linear time-periodic representation for periodic-reference, discrete-time filtered-x LMS algorithm. In Crocker MJ, Pawelczyk M, Ivanov N, editors, Proceedings: 18th International Congress on Sound and Vibration 2011, ICSV 19. Vol. 1. International Institute of Acoustics and Vibration IIAV. 2011. p. 203-210