I'm trying to patch a four-band compressor in pd. I'm new to pd, to DSP and I have little theoretical background, so please be patient. :D
I'm trying to emulate the multiband compressor from Adobe Audition (http://images.digitalmedianet.com/2007/Week_25/0ra7bf9i/story/multiband%20compressor.jpg) because I'm developing a sound toy for Android and I'm exploring the effects I want to produce on the device on my pc.
This is what I came up with proceeding by intuition, for the most part, since I don't really know how to design such a filter. I've also used a couple of patches I found on this forum (and I thank you all). Yeah, it's probably very naive and simple... and a bit messy too. The crossover frequencies are fixed for the moment, but that's ok.
Now a couple of questions. Since I'm using a different number of filters for the four bands (1 or 2 filters per band), do I need to add some delay where I use just one (first and fourth band)? In other words, do lp and hp introduce delay? I'm using the filters from iemlib.
What do I need for the limiter part? Can somebody help me with that? I'd prefer to use vanilla objects since I'm using libpd, and externals are not always guaranteed to work with it.

Any other general thought/advice/improvement?

Thank you!

http://www.pdpatchrepo.info/hurleur/multiband_compressor.pd

There is a limiter built in vanilla that's part of the MTL library http://puredata.info/Members/mtl/index_html/ it's called limiteur~
There's also a more complex one that is part of pdLive http://code.google.com/p/pdlive/called FX_st-limiter

I believe casual filters will produce some delay. This series of books is really good if you do want to get into some o' the maths.

https://ccrma.stanford.edu/~jos/filters/Phase_Delay.html

I haven't explored these but if you're worried about phase:

https://ccrma.stanford.edu/~jos/filters/Filters_Preserving_Phase.html

or you could add the phase delay "lookahead". I don't know how many Multiband compressors worry about phase tho (besides Waves linear phase multiband compressor apparently).

With a 10nth order Chebyshev
http://en.wikipedia.org/wiki/Chebyshev_filter
the phase delay could be a bit intense mebe,

While causal filters to introduce delay, the delay will be frequency-dependent unless it is a linear-phase filter, which the iemlib ones are not.

A good filter to use for the crossover is the Linwitz-Riley filter, which uses two second-order Butterworth filters in series for the high-pass and low-pass parts. It's particularly nice because when the signals are added back together, the frequency response is flat and there are no phase issues. So if you're worried about issues with delay, this is a good approach to take.

like dis?

http://www.pdpatchrepo.info/hurleur/multiband_compressor.pd

Yep.

Thanks for your help, I'm using this patch and it's working fine, but now I'm facing a problem. I've just introduced in my libpd project the possibility to change the recording and pure data sampling rate and now at low values I'm getting weird squeaks/whistles when I'm processing the recording through this patch.
I've identified the object causing these squeaks and it is the lp2_butt~ filter.
When the cutoff frequency of the filter is higher than or lower but close to the sampling rate I get the squeaks. So, especially at 11025Hz and 8000Hz I'm getting distortion. I'm not much into the math behind DSP, but I know that for the Nyquist–Shannon sampling theorem maybe it doesn't make much sense to apply a low pass filter with a 7khz cut off frequency to a 8khz sample rate signal, but for simplicity I just didn't care about that and I thought it wouldn't have hurt anyway.
Now what's going on here? Do I need to apply a coefficient to the cutoff frequency depending on the sampling rate? I tried googling about that and I found something about "normalized cutoff frequency for butterworth filters", but I don't know if that's what I need.

Edit:
I found these links:
http://www.dsprelated.com/showmessage/97866/1.php
http://www.wescottdesign.com/articles/Sampling/sampling.pdf

and yeah, actually it looks like it's not a clever idea to use lp filters like that.
So, what's the best thing I could do in terms of sound similarity with the higher sampling rates processed signals? I could use the [samplerate~] object to get the sample rate and just bypass my filters at the loadbang if the cutoff frequency is higher than half the samplerate. Does it make sense?