OK, for the biquad, consider that you'll have to recalculate all of the coefficients as the LFO value changes the Fc (or Q). And because you have so many coefficients, it's difficult to avoid calculations (and those divisions are a lot more expensive than multiplies, btw) nd do table lookups, because small errors between them stand to make a larger error in the filter. And a lot of opportunity for clicks--a lot of values that have to be smoothed if you're making a big Fc or Q change. (To get an idea of what's going on with the coefficients, you can write a little program to change FC linearly in steps from say 20 Hz to 20 kHz, and plot the changes of the coefficients.)
Now look at the digital state variable, or the Moog-like 4x one-pole sections. You have a single value that changes for a change in Fc. Not only that, but it's a lot closer to being a straight line--you may be able to hook the LFO directly to it in some cases, with a little scaling and offset. Or, more likely, and not much more work, run it through an interpolated table. And if you need to smooth it, to avoid pops when changing settings, you only have one value to smooth. (and thanks--but there are a lot of nice guys here ;-) On Jan 2, 2011, at 8:55 PM, Alan Wolfe wrote: > "Biquads are used heavily in digital audio, particularly for fixed > filters, and ones that are changed manual/rarely" > > So if i had an LFO (like up to 4hz sine, square, or saw wave) changing > the Q value, or the Fc, would a biquad not really be ideal for that > situation? > > Why don't they work well when parameters are changed more often? > > Thanks!! and btw thanks for the JS source code of the calculations, > you are such a nice guy! > > On Sun, Jan 2, 2011 at 8:12 PM, Scott Nordlund <gs...@hotmail.com> wrote: >> >>> One simple and poplar filter is the digital state-variable filter, which >>> has most of the popular synth filter forms available simultaneously: >>> >>> http://www.earlevel.com/main/2003/03/02/the-digital-state-variable-filter >>> >>> This is the digital realization of a popular analog synth filter (Oberheim >>> Expander, etc.), and the digital version is popular too. >> >> Not to nitpick, but that's not quite right. Another standard synth filter >> (probably the most common, especially in 80's polysynths) is a 4 pole >> lowpass cascaded integrator filter. The transfer function is similar to the >> Moog style filter in the Stilson paper, though the circuit is quite >> different. Example circuit: >> http://www.experimentalistsanonymous.com/diy/Schematics/Filters%20Wahs%20and%20VCFs/Polyfusion%20VCF.jpg >> >> The Xpander's multimode stuff is done by mixing the outputs of the filter >> stages. >> >> Details here: >> http://www.musicsynthesizer.com/Circuitry/Multi-Function%20VCF.htm >> >> This may also be relevant: >> http://www.experimentalistsanonymous.com/diy/Schematics/Filters%20Wahs%20and%20VCFs/SSM2040%20Filter%20analysis.pdf >> >> I bring this up not because I'm picky, but because it's a neat but >> under-explored possibility. The only commercial implementations I know of >> are the Oberheim Matrix 12/Xpander, Doepfer A-107 (36 filter types) and Siel >> Mono (though this only offered highpass/lowpass/bandpass). >> >> -- >> 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 -- 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
