A Low Energy Realizable Model for Linear Phase Filtering: A block processing technique
Translated title
En lavenergy realiserbar model til linear fase filtrering
Author
Bastholm, Troels
Term
4. term
Education
Publication year
2015
Submitted on
2015-04-13
Pages
79
Abstract
This thesis analyzes and designs a forward-backward filtering model that achieves linear phase (all frequencies are delayed equally, preserving the waveform). The model is realized with a block processing technique that handles data in chunks. The thesis provides a step-by-step analytical derivation and explains how to compute the model’s exact amplitude response. The model is evaluated against a Folded FIR filter used as a reference. Results show that the narrower the transition band, the greater the advantage of the forward-backward approach in the number of computations. Fixed-point implementations were developed under the same amplitude specification. Register-transfer level (RTL) designs of both models were created to estimate energy use, and the forward-backward model reduced energy consumption by about 25% compared with the reference.
Denne kandidatopgave analyserer og designer en frem-tilbage-filtreringsmodel, der opnår lineær fase (alle frekvenser forsinkes lige meget, så signalets form bevares). Modellen realiseres med blokbehandling, hvor data behandles i dele. Opgaven giver en trinvis analytisk udledning og forklarer, hvordan modellens præcise amplituderespons kan beregnes. Modellen evalueres mod et Folded FIR-filter, der bruges som reference. Resultaterne viser, at jo smallere overgangsbåndet er, desto større fordel har frem-tilbage-tilgangen i antal beregninger. Fastpunkt-implementeringer blev udviklet under samme amplitudespecifikation. Der blev desuden udviklet RTL-designs (register-transfer-niveau) af begge modeller for at estimere energiforbrug, og frem-tilbage-modellen reducerede energiforbruget med cirka 25% i forhold til referencen.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords