posted 04 April 2002 04:06            

mbmodulator.kym

The effect is much like a modulated filter cutoff fed by the amplitude follower on a sidechain, but unlike a modulated filter, this sound reinforces the original spectrum all the way up to Nyquist as per your settings. When the amplitude follower puts out a strong signal, the spectrum is very bright (brighter than a typical modulated filter would be), and when the amplitude follower is a weak signal, the output is darker, as per your settings.

The VCS has faders for the crossover frequencies, the amplitude multiplier on the amp follower and its slew rate. Then there are 3 banks of sliders corresponding to each of the 6 passbands. These are offset, scale, and delay. The amplitude follower is fed to 6 independent ScaleAndOffset Sounds to affect the passbands by different amounts, and then all of the modulated passbands are recombined after going through independent delays.

Setting a non-zero offset allows some level of passband sound through all the time. A low value works well here for pads, in the lowest frequency passband. That gives a steady output modulated by brightness changes according to the scale settings on the remaining passbands.

Setting a dispersive range of delays adds interesting comb filtering to the sound, but not in a constant manner, since each passband can have a different amount of delay applied. Hence, each passband will have predominantly different comb filtering. I find that a range of values from 0 to 0.005 (5 ms) make the image wider. This sound processes stereo input and the result is very interesting with these dispersive delay settings. There are a number of presets included for examples.

This sound makes use of an interesting quirk of 1-pole filters. All of the filters in the filterbank are 1-pole lowpass and highpass filters. Bandpass filters are cascaded low-pass and high-pass 1-pole filters.

If you add together the outputs of a 1-pole lowpass, plus half the output of adjacent bandpass filters, and then the full output of the final high-pass filter, the result is very nearly the original sound (plus or minus a small fraction of a dB). This only works for 1-pole filters, because phase cancellation is serendipitous in this case. Try that with higher order filters and you get serious phase cancellation at low and high frequencies.

Also, 1-pole filters have very shallow rolloffs of 6 dB/octave. So any sound fed into one filter actually gets through with much of its original character. That allows one, for example, to set the offset on the lowest passband to 0.1, set its scale to 0, and then set all the other passbands to offset 0 and scale 1. Then you can continue to hear a recognizable original sound at a low background level (e.g., for pads) which brightens according to the amplitude follower output on the sidechain sound.

Try this with a percussive sound like a vibraphone, or xylophone, while having speech on the side chain. Play the input in rhythm with the sidechain and you get very interesting spectral variations on each beat. The results sound typically brighter than using a modulated filter, but they are similar.