Home/ Questions/Q 860557
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-15T08:50:06+00:00

Implementing a low pass FIR filter, when should one use FFT and IFFT instead

  • 0

Implementing a low pass FIR filter, when should one use FFT and IFFT instead of time-domain convolution?

The goal is to achieve the lowest CPU time required for real-time calculations. As I know, FFT has about O(n log n) complexity, but convolution in the time domain is of O(n²) complexity. To implement a low pass filter in the frequency domain, one should use FFT, then multiply each value with filtering coefficients (which are translated into frequency domain), then make IFFT.

So, the question is when it is justified to use frequency-based (FFT+IFFT) filtering instead of using direct convolution based FIR filter? Say, if one have 32 fixed-point coefficients, should FFT+IFFT be used or not? How about 128 coefficients? And so on…

Trying to optimize an existed source code (convolution-based FIR filter), I am totally confused, either I should to use FFT or just optimize it to use SSE or not.

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    Convolution is actually O(m*n) where m is the width of the finite impulse response, and N the sample window.

    So the tipping point of m where it is useful to change to FFT+IFFT is related to log(N) of the sample window.

    In realtime operation the fact that FFT is batch oriented might be more important than the relative amount of clock cycles, as it may not be acceptable to wait 1024 sample points before filtering, if the application is in a regulation loop for example.

    Now a lot of development has been done is this area and plenty of code is available, so trying a couple of solutions and benchmarking is key here.

    • 0