Hello All.

what would be the most elegant, cpu-cheap and phase-accurate way to build a 32band 'crossover-bank' where I can manipulate the volume of the bands and sum them back together afterwards?

bpbank~ has gotten me quite far, but I think it's not really a crossover, but a filter bank. With low quantity of bands it would not work for me as it does not cover the whole frequency spectrum then.

I also tried building a crossoverbank with 32 Linkwitz–Riley filters ( 2 hip.bw~ & 2 lop.bw~) but they don't sum back together accurately. (probably doing something wrong)

I am also aware that most multiband compressors solve the problem with a filterbank for measurement and a chain of shelf filters for manipulation, but I guess it's cpu-expensive and almost impossible for me to figure out the values for all 32 shelf filters across the frequency spectrum.

Context: I am patching a 32band processor that manipulates the loudness of the bands depending on certain criteria like 'loudness of a band compared to loudness of fullrange signal and a target-curve'. It will be what I call a 'spectral ballancer'.

Here is what works quite well with bpbank~ sb_demo.pd

Thanks!

@trummerschlunk Have you tried using FFT resynthesis? The bands won't have constant pitch width and the whole thing will introduce extra latency, but it might not matter depending on the sound you want. Take a look at I06.timbre.stamp in Pd help. Also, the min number of bands would be 256, so you'd need to combine bands to get down to 32.

@trummerschlunk I made a little test patch for the different options that you described, hope it's helpful in some way ... crossover-filter-test.pd

The version with shelving filters shouldn't be more cpu-expensive (it might even be cheaper), and the math for the gains is quite simple. Or am I missing something?

But I don't think any of these approaches is suitable for as much as 32 bands! Wouldn't you need much steeper filter slopes for that? Of course, you could use higher order Butterworth shelving filters, but that's gonna be really expensive! So maybe, as @jameslo suggested, you should go for frequency domain techniques in that case.

@jameslo @manuels thank you so much for your feedback and taking the time!

For the 32band spectral ballancer, I will dig into fft now.

It's very interesting to see the different ways in the testpatch. It helps for a future module in my patch, a 5band multiband expander/compressor which I will probably build with shelf filters.

Again, thanks a lot.