I am trying to get a general sense of how many multiplication operations are needed for a 4th-order Butterworth filter (IIR filter) with a rational sampling rate change. I am excluding any multiplication by 0's.

For example, for an upsampler of order 5 and downsampler of order 7 with a 99-tap FIR filter in between, the number of multiplication operations needed would be of order

$ \frac{99N}{7} $

Where $N$ is the sampling rate (per second). This is since an upsampler would effective have 4 0's for every 5 upsampled from the sampling rate of $N$.

I would like to know what the logic would be for the case of a 4th-order lowpass Butterworth filter. There are cascading designs, but there are also the canonical transpose and direct forms of implementing a Butterworth filter. What would be the difference in multiply operations (excluding multiply by 0) in each of these designs? Also, how can I calculate this? I have read somewhere essentially any IIR filter uses 6 multiplies per 2 orders of the IIR filter (4th order in this case, with $\frac{4}{2} = 2$, thus $6 * \frac{4}{2} = 12$ multiplies per second of input sample), but this is not the correct logic in my opinion.

What is the best way to think about this? I am pretty much self-teaching myself digital signal processing, so I am lost as to how to approach this.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.