More Ensemble Stuff

[Sean Costello]

Hi all:

I've figured out a few more ensemble tricks (in Csound, that is):

- The modulation works best when it is exponential. I approximate this using an
inverted half sine wave (similar to an inverted full-wave rectified sine wave)
as the wavetable for the oscili units that are acting as the LFOs.

- Running the ensemble through 5 parallel bandpass filters, with specific center
frequencies, bandwidths, and amplitudes, can yield a very nice vocal choir
sound. However, the amount of pitch modulation for the ensemble that produces a
good vocal sound is MUCH higher than the amount of pitch modulation that is
useful for the string ensemble sound. When the delays are modulated for a good
string ensemble sound, the resulting choir sound is rather flat, and not that
vocal. Turning up the pitch modulation makes the string ensemble sound horribly
out of tune, but makes the vocal sound really come to life. I have a nice
simulation of a Russian mens choir running on my computer! My theory is that the
added pitch modulation helps to "fill out" the area under the formants.

- For vocal synthesis, using 3 independent LFOs might work better than the two
120-degree LFOs that most string ensembles work. The vibrato sound is nice with
the latter method, but there is a "phasiness" that doesn't sound particularly
vocal (most likely due to the slower of the two LFOs). However, it still works
well enough to work in a device that was switchable between string ensemble and
chorus. In addition, it should be fairly trivial to modify existing "ensemble"
units to produce the higher pitch modulation needed for vocal simulation. 

Anyway, hope this is interesting to someone out there. Constructing a "choir
organ" would seem to be fairly trivial, if you have a string ensemble - just
install a switch that allows for higher amounts of pitch modulation in the
ensemble, and follow the ensemble with four or five parallel bandpass filters
(which should only take 2 opamps each), that are mixed in differing amounts to a
single output. 

The difficulty with the above "choir organ" is that it would be limited to a
single formant sound. If you wanted male and female formants, they would be
forced to share the same pitch, and wouldn't sound that effective. The solution
would be to have different footages going through seperate ensemble/filtering
units - more expensive, but it would produce a very nice sound. Putting the
filtering before the ensemble might be a solution, but from my own experiments I
have found that the vocal formants virtually disappear when this is done. Static
formants, without pitch modulation of the input, don't sound particularly vocal,
or interesting for that matter. (Note: a unit like the Korg Lambda, with its
three oscillator/divider chains modulated by a 3-phase LFO, would be easy to
convert into a choir organ, as the different footages will already have the
proper modulation. Or does the Lambda have a choir sound already?) 

If you want any pointers to tables for the center
frequencies/bandwidth/amplitude for different formants, let me know.

Sean Costello