Hey guys!

I want to make my own, but I'm having trouble figuring out how to set up the filters. What aspects are there to take into consideration to make sure the split signal is as close to the original as possible?

If anyone has a simple example patch I would really appreciate that.

@schitz It is a complex subject and will probably become a long thread once it gets going.
https://en.wikipedia.org/wiki/Audio_crossover
Any crossover points between bands will have a cut-off slope which means that 2 speakers will be producing a small band of frequencies at the same time.
That means that phase correlation between the bands is important, but I remember that the steeper the slope the harder that is to achieve.
And of course as flat a total response as possible.

That might not be your intended purpose but it is relevant.
And for filters generally @katjav is a great read........ all of her site is littered with fft gems.......
https://www.katjaas.nl/fourierfilter/fourierfilter.html

David.

@schitz
The difference of a modified signal to it's dry one is the opposite:
LP-Original=-HP
Original-BP=Notch
Dry-Notch=BP
....
see this post:
https://forum.pdpatchrepo.info/topic/13409/how-can-i-create-a-switchable-filter/4

If I'm not mistaken, using a Butterwoth filter instead, makes a quick and dirty digital Linkwitz-Riley-Filter https://en.wikipedia.org/wiki/Linkwitz–Riley_filter

[years later EDIT: Also see this patch: https://forum.pdpatchrepo.info/topic/7006/multiband-compressor/5 ]

@schitz said:

close to the original as possible

@whale-av said:

phase correlation

That's the difficult part:
I would love to have a chat and see examples of linear-phase
and minimal-phase filters in Pd!

https://www.katjaas.nl/fourierfilter/fourierfilter.html

anyone ported this to Pd yet?

2 speakers

would make the topic infinitely complicated, as different positions and responses of the speakers come into play.

@lacuna said:

anyone ported this to Pd yet?

@solipp's pp.fft-split~ in Audiolab might be close enough for @schitz's purpose

@schitz

this answer might be a bit terrifying, but i am afraid that what you had to do is to not change anything in the splitted signals.

(or in other words: only if you do that what happens in an ideal 3-way speaker)

otherwise any correctly set up bandsplitter - based on a butterworth or FFT - will lead to the usual "noticeable" filter artefacts.

for things like multiband compression and such you might want to try using phaselinear filters instead. it is not like it would be required, but it is an alternative.