Hi, I am conducting a little study on filters and currently finding a way to obtain their frequency response by feeding them a wideband signal and plotting their FFT.

I am new to the PD environment and currently the best I could do is generate N/2 oscillators with different frequencies (N being the sample size of the audio buffer). I am sure there is a smarter way of doing this (generating lists for the oscillator frequencies and/or sequencing them). Any suggestions?

I could write an external but a relatively simple subpatch or expression is better.

As an alternative I could feed white noise and average the FFT curves over time, but the result would not be as smooth.

Thank you!

lop_example.pd

@dgd There has obviously been a lot of discussion about filters for Pd, latency versus processing overhead, minimum phase etc.
@Katjav is your guru, and I can do no better than to point you here........... http://forum.pdpatchrepo.info/topic/5786/low-latency-fir-filters-minimum-phase-and-partitioned-convolution
Follow the links in that thread.........
David.

Thanks a lot @whale-av ! I definetly remember seeing Katja's page before. It is a wonderful resource. I will dig a little bit into it to see if I can come up with something more elegant for my frequency response graphs. My problem is not actually with understanding the filter behaviour itself or the trade-offs but more of implementing nice "visualization" or debugging tools in pd.

Thanks for the hints!

why don't you just take a fft of the impulse response? what are you looking for? (an impulse is as wideband as you can get btw, if you feed something like white noise in then you are basically feeding randomly-phased higher frequency components that alias into the fft i think, whereas an impulse is periodic with the period of the fft size though the impulse response might not be so windowing might be needed).

you can generate a 1-sample pulse by using vline~'s delay capability and calculating a sample delay

Hi, somebody pointed out to me that there are already the filter-graph-1 and filter-graph-2 abstractions for plotting frequency responses of filters They seem to work by doing a frequency sweep in a way similar to spectrum analysers,.

/doc/3.audio.examples/H13.butterworth.pd has a nice example of it.

You are right about the aliasing if the filter's stop band isn't attenuating enough. I was counting on the aliased signal energy being negligible. I am going to give the impulse response approach a try nevertheless and maybe compare it to the frequency sweep method.

Thanks for the answer!