Re: [music-dsp] Approaches to multiple band EQ
In addition to what Robert says about the lack of generality in the numerator of the Chamberlin SVF (it's an "all pole" filter), you can't set the Fc very high in the audio band. It's not a good choice for this application. On Jan 11, 2011, at 4:32 PM, robert bristow-johnson wrote: > > On Jan 11, 2011, at 7:01 PM, Tom Wiltshire wrote: > >> I'd approach this from a analogue-thinking angle and design a tunable >> parametric EQ stage and then parallel a load of them up, like Robert >> suggested. > > that's not exactly what i meant to suggest. what goes in parallel are not > simply these tunable parametric EQs. what's in parallel are these biquads > that come from a partial fraction expansion of these parallel EQs in *series*. > >> For the EQ, I'd start by looking at a digital 12dB/oct SVF design, like the >> Chamberlin filter. > > well, they'll all be 2nd-order filters, whether SV or biquad is another > issue. the SVF as depicted by Hal is not as general (in the numerator of the > transfer function) as a 4-coefficient biquad. one reason i was thinking of > the parallel filters being biquad is that, especially for the Direct 1 form > (DF1), the result of each biquad is a double-wide word and, being in > parallel, all of the "bands" can be added up before a final quantization is > made for the graphic EQ output. > >> This allows you to tweak resonance independently of frequency, so you could >> adjust it depending on how wide the bands are. The per-band calculation >> wouldn't be too bad, so you could easily do multiple bands. > > okay, if you're suggesting a bank of BPFs (not parametric EQs, how would you > reduce the gain at a particular frequency?) in parallel, you have to worry > about getting a flat response when all of the sliders are set to zero. > >> And don't worry about the "nonlinear phase responses" > > i might agree with that. it's unavoidable with basic IIRs. > > -- > > r b-j r...@audioimagination.com > > "Imagination is more important than knowledge." > > > > > -- > dupswapdrop -- the music-dsp mailing list and website: > subscription info, FAQ, source code archive, list archive, book reviews, dsp > links > http://music.columbia.edu/cmc/music-dsp > http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Approaches to multiple band EQ
On Jan 11, 2011, at 7:01 PM, Tom Wiltshire wrote: I'd approach this from a analogue-thinking angle and design a tunable parametric EQ stage and then parallel a load of them up, like Robert suggested. that's not exactly what i meant to suggest. what goes in parallel are not simply these tunable parametric EQs. what's in parallel are these biquads that come from a partial fraction expansion of these parallel EQs in *series*. For the EQ, I'd start by looking at a digital 12dB/oct SVF design, like the Chamberlin filter. well, they'll all be 2nd-order filters, whether SV or biquad is another issue. the SVF as depicted by Hal is not as general (in the numerator of the transfer function) as a 4-coefficient biquad. one reason i was thinking of the parallel filters being biquad is that, especially for the Direct 1 form (DF1), the result of each biquad is a double-wide word and, being in parallel, all of the "bands" can be added up before a final quantization is made for the graphic EQ output. This allows you to tweak resonance independently of frequency, so you could adjust it depending on how wide the bands are. The per- band calculation wouldn't be too bad, so you could easily do multiple bands. okay, if you're suggesting a bank of BPFs (not parametric EQs, how would you reduce the gain at a particular frequency?) in parallel, you have to worry about getting a flat response when all of the sliders are set to zero. And don't worry about the "nonlinear phase responses" i might agree with that. it's unavoidable with basic IIRs. -- r b-j r...@audioimagination.com "Imagination is more important than knowledge." -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Approaches to multiple band EQ
I'd approach this from a analogue-thinking angle and design a tunable parametric EQ stage and then parallel a load of them up, like Robert suggested. For the EQ, I'd start by looking at a digital 12dB/oct SVF design, like the Chamberlin filter. This allows you to tweak resonance independently of frequency, so you could adjust it depending on how wide the bands are. The per-band calculation wouldn't be too bad, so you could easily do multiple bands. And don't worry about the "nonlinear phase responses" - have you thought about what a analogue EQ does to the phase? Any digital implementation will be no worse, and will have less noise to boot. "Does it sound good?" is the question to ask, unless you've got some particular reason to need to protect the phase? HTH, Tom On 11 Jan 2011, at 18:23, Thomas Young wrote: > Hi all > > I need to develop a real-time multiple band EQ DSP effect, but I am unsure > about how to approach it. > > My preferred approach would be to FFT-> Modify Spectrum-> IFFT, however I > think that will end up being too slow (or at least using up far more > processing power than I would like.) The only other approach I can think of > is a number of IIR band stop filters in series, would this be practical? I am > concerned that there would be some negative interaction between the filters, > or some unpredictable results due to different (non linear) phase responses > of the filters. It's important that the DSP introduces minimal distortion and > is acoustically transparent when 'flat'. > > Information about any other common approaches to multiple band EQ's would be > helpful too. > > Thanks > > Thomas Young > > Core Technology Programmer > Rebellion Developments LTD > > -- > dupswapdrop -- the music-dsp mailing list and website: > subscription info, FAQ, source code archive, list archive, book reviews, dsp > links > http://music.columbia.edu/cmc/music-dsp > http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Approaches to multiple band EQ
On Jan 11, 2011, at 1:23 PM, Thomas Young wrote: I need to develop a real-time multiple band EQ DSP effect, but I am unsure about how to approach it. do you mean a graphic EQ? My preferred approach would be to FFT-> Modify Spectrum-> IFFT, if you do that, better look up the concepts called "overlap-save" or "overlap-add". it's doable, but a lot of coding and there is throughput delay. whatever is your frequency response in the "Modify Spectrum" step, you need to turn that into a real *finite* impulse response that is significantly shorter than the length of the FFT. however I think that will end up being too slow (or at least using up far more processing power than I would like.) The only other approach I can think of is a number of IIR band stop filters in series, would this be practical? i think what you mean is a bunch of parametric EQs (sometimes called peaking EQ or "presence filter") in series. these are essentially a BPF in parallel with a "wire". they would be spaced equally in log frequency (perhaps 3 per octave) and have equal and fixed bandwidth in octaves. for a graphic EQ, the boost/cut gain for each band would be adjustable. I am concerned that there would be some negative interaction between the filters, or some unpredictable results due to different (non linear) phase responses of the filters. the group-delay (and, i'm sure, the phase-delay) response of the peaking EQs are, unfortunately, dependent on both the tuned frequency and the boost/cut gain. that means that, as the user changes the sliders for each band to different non-zero (dB) settings, the delay through each band will be different. this is *definitely* not phase linear, unless all sliders are set to 0 dB. being that they are in series, i do not know how one would put in a compensation delay that would affect one band and not the others. It's important that the DSP introduces minimal distortion and is acoustically transparent when 'flat'. this would be perfectly transparent, right down to the LSB, if all bands are set to 0 dB. but if all the knobs are set to something else, there is a lot of quantization error from each biquad section in series that can sorta build up. make sure your word size is wide enough. i mentioned this before in some context (either here or comp.dsp). i think there should be a straight-forward way (using Heaviside partial fraction expansion) to turn this set of 31 peaking EQs (each a biquad with 5 coefficients) in series into a set of 31 biquads (but these need only 4 coefs each) in parallel with each other and *one* feedforward wire. there would be much less quantization error, but i am not sure how the "coefficient-cooking" math would be done in- between the 31 sliders (and the fixed frequency and BW spec) and the 4*31+1 coefficients. in my opinion, that would be the best way to do a graphic EQ, particularly with a fixed-point DSP (because the output of each biquad can be left in double precision and added to the outputs of all other biquads as double-wide words before quantizing back to single width). but, until someone pays me to design their graphic EQ to do that, i'll leave that math (to do the partial fraction expansion automagically from the slider settings) to some ernest grad student or someone else to do. Information about any other common approaches to multiple band EQ's would be helpful too. well, it would be good to get the semantic straight. i *think* you mean a graphic equalizer, but i don't know for sure. -- r b-j r...@audioimagination.com "Imagination is more important than knowledge." -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
Re: [music-dsp] Approaches to multiple band EQ
Take a look at the the peak/shelf filters here: http://www.musicdsp.org/showone.php?id=197 Best Regards, Thomas Am 11.01.2011 19:23, schrieb Thomas Young: Hi all I need to develop a real-time multiple band EQ DSP effect, but I am unsure about how to approach it. My preferred approach would be to FFT-> Modify Spectrum-> IFFT, however I think that will end up being too slow (or at least using up far more processing power than I would like.) The only other approach I can think of is a number of IIR band stop filters in series, would this be practical? I am concerned that there would be some negative interaction between the filters, or some unpredictable results due to different (non linear) phase responses of the filters. It's important that the DSP introduces minimal distortion and is acoustically transparent when 'flat'. Information about any other common approaches to multiple band EQ's would be helpful too. Thanks Thomas Young Core Technology Programmer Rebellion Developments LTD -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
[music-dsp] Approaches to multiple band EQ
Hi all I need to develop a real-time multiple band EQ DSP effect, but I am unsure about how to approach it. My preferred approach would be to FFT-> Modify Spectrum-> IFFT, however I think that will end up being too slow (or at least using up far more processing power than I would like.) The only other approach I can think of is a number of IIR band stop filters in series, would this be practical? I am concerned that there would be some negative interaction between the filters, or some unpredictable results due to different (non linear) phase responses of the filters. It's important that the DSP introduces minimal distortion and is acoustically transparent when 'flat'. Information about any other common approaches to multiple band EQ's would be helpful too. Thanks Thomas Young Core Technology Programmer Rebellion Developments LTD -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
