Normalizing FFT magnitudes

Reply to Normalizing FFT magnitudes on Fri, 12 Jul 2019 10:33:20 GMT

dom1817 — Fri, 12 Jul 2019 10:33:20 GMT

I've been trying to normalize the magnitude results of a FFT signal analysis in a range between 0 and 1.

What I'm doing now is to do list operations on the tables to find the maximum value of the current FFT results and divide all the values by it (see patch attached). This works nicely with Pd generated waves, but it doesn't look good with samples (it's very narrow on the lower frequencies and not spread at all, see picture below).

Here are my questions:

Is there a way to normalize the values on the signal domain, without these table operations?
If not, why does the normalized FFT looks so narrow?

Many thanks!

fft_normalization.pd

Reply to Normalizing FFT magnitudes on Fri, 12 Jul 2019 13:37:27 GMT

Jona — Fri, 12 Jul 2019 13:37:27 GMT

@dom1817 for me it works when I use [array max $0-magnitude] instead of [list-max]...
...or with the normalize message

Reply to Normalizing FFT magnitudes on Fri, 12 Jul 2019 22:05:50 GMT

dom1817 — Fri, 12 Jul 2019 22:05:50 GMT

Thank you, I somehow didn't notice the [array max] object in the [array] help file.

I did a little research on the Pd list tonight and actually using a [log~] object with appropriate scaling is a good solution. Still not perfect but much better than before, will upload an updated patch soon.