Anyone up for building a serious multiband limiter

[dscro]

I've wanted to build one for a while. I'm thinking of something in the vein of the Waves L16. I understand basically what they did with it and what must be going on "under the hood" and I could offer the limiting portion of the plug with variable release times.

I just haven't been able to get the crossover frequency thing right to separate the audio spectrum into parts and still have it sound good when recombined. I don't know if we could use the FFT stuff in SB or just the crossover or other filters. There must be some theory behind setting those crossovers right.

Anyone want to investigate that part of it?

[zmix]

dScro,

The real issue is indeed the filters. This is certainly the weakest part of Sonic Birth. They simply don't sound good enough for pro audio.. You would need to find a way of coding some filters on your own.

-CZ

[eesn]

is 16 bands still considered serious?
http://esem.name/?p=91

[dscro]

In fact I think 16 is silly. But more is always better right? I think we should go for 64

Actually I was thinking of maybe 4 or 5. zmix is right that the overall sound of the filters is the weak link. And you can't create new filters from scratch using math in SB because I think filtering is a recursive process of calculating coefficients then affecting the waveform, then calculating coefficients the . . . something like that. At least I think that's the case having seen sample C++ code on DSP sites in the past. Then again, maybe there is a way to do it with convolution in SB.

I don't have the coding experience to write new source for SB.

[dscro]

I take it back. We might be able to code filters from scratch in SB using sample delays and simple math. See http://www.users.qwest.net/%7Evolt42/ca ... ilters.pdf

maybe even make them linear phase

[makira]

are the filters that bad ? they are pretty much textbook...

[dscro]

No they aren't terrible at all, they just don't sound linear phase and I think they are strictly Butterworth as well. What I wonder is when you coded them did you use a recursive approach y(n) = Ax(n) + By(n-1) where x is input and y is output so that these are IIR filters or did you use a limited number of coefficients, non-recursive formula and essentially do FIR x(n) = Ax(n-1) + Bx(n-2) + . . .

For me the big issue is that I was never able to chop up the audio spectrum into say four parts with the crossovers and then put it back together without things sounding out of phase. To really cut bands sharply I stacked a few filters in series to increase the number or poles but that also increased the phase issues.

I was amazed that I was able to make a simple LPF from scratch in SB today using nothing but sample delays and constant multipliers. It's FIR style since I have to just go for a fixed number of samples and coefficients because I can't do recursive. The feedback loop doesn't go down to the sample level. I forget how many milliseconds it's limited to. So I can't say, take the last sample and feed it back to be averaged with the next sample.

It's truly interesting just what you can do from scratch using delays and a little math in SB.

[zmix]

Yes, the filters are that bad. They are way too steep for any sort of non-effect work. Textbook is -bad- too, by the way... I've managed to create a few shelving filters but I tend to av0id using the built in filters in the audio path.

I don't think that linear phase is a useful approach for 'good sounding' filters, either.

I wish this forum had some picture or file hosting....

-CZ

[dscro]

Yeah, I would agree that linear phase filters aren't necessary or even desirable for building a classic sound eq plug. But I always thought that if you wanted to chop up the audio spectrum into four or five slices, compress/limit each piece separately and then put them back together, you do want to try to keep the phase constant then. Maybe it's not necessary. By the way, I hate the fact that linear phase filters are called "linear phase." I think it's a confusing term. They should be called "no phase shift" filters.

zmix, I would love to see one of your custom coded shelf filters.