From: timkoelm@iconz.co.nz
Date: Tue, 27 Jan 1998 23:13:13 -0800
Subject: [jv1080] Taurus Patch Programming #8

+ List Help : finger: jv1080@emccta.com, web: http://www.emccta.com/~jv1080
+ List Admin: mvitek@emccta.com, http://www.emccta.com/~mvitek

This is a multi-part message in MIME format.

- --------------6EBA6FC62348
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Greetings JV/XP'ers!!

Welcome to this week's installment of my patch programming series.


Today, choir sounds. Some time ago, I programmed a choir sound using the internal waves,
and at the time it was the best approximation of a choir I could come up with. Recently,
as I went over some of my older patches, I was dissatisfied with my effort, and
consequently reprogrammed the patch using techniques I had learnt to make the patch as
effective as I could. I think the results are as good as one can manage using the
internal samples as the basis of a patch. So here, I am presenting two patches: "Clear
Light", my original effort at a choir sound, and "Clear Light 2", the revised and
reprogrammed version. We'll examine the making of "Clear Light 2", and use "Clear
Light" as comparison. Whether or not this patch is truly successful is up to the player,
but certainly there is a technique at work here which you can integrate into your own
patches when on programming adventures.

The internal samples didn't have the true "massed choir" effect I was looking for in
trying to make this patch - I believe the only samples like these are located in the
"Session" expansion board. Rather, the internal samples are more like the breathy, synth
type choirs to be found on instruments like the Fairlight CMI and Korg M1. I thought
about what synthesis resources were available to me to make a choir as "vocal" and
realistic as I could, given these waves. I found myself looking towards several things - 
the resonant multimode filters, LFOs, random pitch and tuning parameters and the EFX 
processor, all of which are at work here.

In picking the sample I was to use as a foundation, I looked for the sample which had
formants truest to the human voice. A formant is a fixed frequency/overtone or group of 
frequencies/overtones which gives a particular sound its character. The choir "oohs" 
were too artificial for my taste, as were most of the choir "aahs". In the end, I chose 
the "Voice Aahs C" waveform, which I copied and pasted to tone 2. I've created this 
patch using tones 1 and 2, and then copied those tones to tones 3 and 4.

A real choir is a very animated sound - each voice is like a Super JV tone in constant
oscillation. Reproducing this phenomenon was important, so I drew upon the Super JV's
pitch and LFO resources. First, on the pitch page, I randomly detuned each tone by a
maximum of 10 cents using the Random Pitch parameter. This gave me some degree of 
movement. Then using a random LFO (LFO1 in this case) I modulated pitch and filter 
cutoff by very small amounts. Using a triangle-wave LFO (LFO 2) I simulated the vibrato 
heard in a singers voice, by modulating pitch. I did this for each tone. Then for one 
pair of tones, using the pitch envelope, I created a downward pitch-swoop on key-press. 
For the other pair of tones I created an upward pitch swoop - again on key-press. I also 
enabled the Analog Feel parameter for the patch, to add further pitch variation. The 
resulting effects are subtle but important. For stereo depth I panned one group of tones 
hard-left, and the other group of tones hard-right. The resulting stereo depth is so 
great, no chorusing or delay is needed. 

The most important factor at work in this patch is "Tone Control". Selecting a waveform
which most accurately portrayed the overtones at work in a human voice was not enough.
Listening to "Clear Light 1" you'll notice the tone is light and breathy - not like a
real massed choir at all. Using the Super JV's filters, I was able to pinpoint harmonics
in the raw waveform which most represented human vocal formants, and emphasize those
harmonics accordingly...In other words, emphasize the "aah" sound in a choir. Here's how
I acheived this.


******************************************************************

WG ---- TVA ----|          |---- TVF ----| 
                |----------|             |
WG -------------|                        |---- TVF ---- TVA

******************************************************************

WG   -  Wave Generator
TVA  -  Time Variant Amplifier
TVF  -  Time Variant Filter

The waves of tone 1 and 2 (or 3 and 4) are mixed by the TVA of tone 1, and then are 
filtered by the TVF of tone 1 before being filtered by the TVF of tone 2. The master 
output is controlled by the TVA funtions of tone 2. 

First, for the two pairs of tones, I switched to Structure 2, which puts the filters in
series. Then for tone 1, I selected a Peaking Filter, and for tone 2 a Low Pass Filter.
A peaking filter adds a resonant peak to a given waveform which can be swept up and
down, if so desired - I've described this before, see my last post for more information. 
Using this filter, I was able to sweep the resonant peak up and down the frequency 
spectrum until I pinpointed frequences which most resembled vocal formants. Try adding 
more resonance to the Peaking filter, and sweep the center frequency using the alpha 
dial - see what other interesting frequencies you can pinpoint. Using the LPF after the 
Peaking filter, I was able to cut out most of the undesired high-frequency breath noise. 

Putting the filters in series opens up a whole world of sonic possibilties. For example, 
you could have a Band Pass Filter feeding a High Pass Filter, or two peaking filters in 
series for even more accurate tone control. Putting a High Pass Filter after a Low Pass 
Filter results in a variable-width Band Pass Filter. Using Structure 2 effectively 
allows you to create hybrid filter types and when combined with Ring Modulators and 
Boosters, the effects can be fairly mind-blowing.  

To further the "Tone Control" factor, I dialled up the Stereo EQ EFX program and piped 
the tones through it. I was able to pinpoint additional specific frequencies and 
emphasize them accordingly. The best way to locate specific frequencies when using any 
sort of equalizing device is to crank the gain of the band you want to adjust to its 
maximum, and then sweep the frequency until you find what you're after. That's the 
technique I've used here. Experiment with the frequencies I've specified in the Stereo 
EQ program, and see what variations you can come up with. 

And that's "Clear Light 2". You can now listen to "Clear Light 1" as a comparison, and 
decide how succesful the reprogrammed patch is. These techniques are useful for all 
sorts of sounds, not just choir sounds. They're good for creating "formanty" leads as 
well, and alive, moving pads. Some of you who out there who have the "Keyboards of the 
60's and 70's" and "Session" expansion boards may want to try these tips with the choir 
samples on them. I'm sure the results would far exceed what is capable with the internal 
waves. And if anyone comes up with something hot, let me know!


Best Regards,

Taurus3

******************************

P.S. The .mid file which comes
    with this post, will over-
    -write patches 1 to 2 in
    user memory. You have
    been warned.


