• Stine Martine Gullaksen
  • Morten Kramer Danmark Nielsen
Low-energy filters are needed in our everyday life and even further optimisation of these are needed. This project explores the question on how energy dissipation can be minimised for digital FIR filters on software and hardware programmable platforms.
The answer to this is found by analysing, implementing and evaluating several methods for this purpose. Realisations of the methods were utilised, as to make a practical comparison between the optimisation methods. Two different software approaches was conducted: multirate filtering and Hamming distance optimisation. The multirate filter gave an optimisation, in relation to the reference point, of up to 20 % and the Hamming distance optimisation method gave up to 2 %. Three different multiplierless approaches were applied on the hardware programmable platform, in addition to two reference filters. Of these realisations the shift based filter gave the best performance. Next to this a parallel direct form filter, and third a multirate filter. The least energy efficient methods was a generic direct form filter and a filter utilising distributed arithmetic.
The evaluation of the realisation showed that several methods could be used as energy efficient solutions compared to a generic reference point. Based on the methods investigated in this project a set of guidelines are conducted. These are aimed at designers with the objective of designing low-energy filters.
Publication date31 May 2012
Number of pages178
ID: 63452691