[sdiy] Polyphonic MIDI to CV
brianw at audiobanshee.com
Sun Jul 11 03:29:14 CEST 2021
On Jul 9, 2021, at 13:59, Tom Wiltshire <tom at electricdruid.net> wrote:
> On 9 Jul 2021, at 18:49, Rutger Vlek via Synth-diy <synth-diy at synth-diy.org> wrote:
>> Interesting project! I have been playing with the same idea for a while. It's quite a challenge to replace keyboards with buss bars typically used in divide down architectures. The range probably does not need to cover alle 128 midi notes, but at least the range of a stringer. A large array of quad VCAs (2164) could do the trick, but is rather costly. Via mcu you would be able to combine VCAs for multiple octaves on a single key (something normally done with multiple buss bars).
> I think we need to be looking at the electronic organ tech this string machines were derived from. There’s a ton of stuff there about how to do polyphonic enveloping/gating of signals for cheap. Mostly along the lines of the Roland swing-type VCA with transistors, CMOS switches, or logic gates.
I recall that the TR-808 uses these transistor VCA circuits. Are they called "swing-type" because the output has a DC offset? I think the CV for the VCA level ends up appearing as a DC offset, but that's reasonably solved by AC-coupling the output to the next stage of the audio mixer. I also recall that the TR-808 has simple Decay envelopes.
>> An alternative architecture I wondered about is a Bottom Octave generator and Saw wave multipliers to achieve the higher octaves. In theory it could give you saw and square waves simultaneously for all notes, but I don't know if it can be implemented in analog reliably. Has anyone ever tried it before?
> I've tried it using triangle waves, which can be pitched up an octave by using a precision fullwave rectifier. Any offset in the triangle shows up as a fundamental in the supposedly-doubled output. I found it was possible to make it work, but fiddly to adjust when I tried with four or five octaves. And that was a monophonic application!
I assume that the triangle offset shows up as a fundamental because the output alternates between two similar periods.
I believe that a comparator could be used to turn a saw into a square, which could then be inverted and added to the original saw to double the frequency. The precise threshold would probably create the same problem with a fundamental showing up in the output, because there would again be an alternating pattern between two similar periods - unless they were precisely the same.
> The great advantage of the standard flip-flop divider is that you get a genuine solid octave-down with no messing about, even if the input duty cycle isn’t exactly correct.
Indeed. Not only is the flip-flop divider immune to input duty cycle variations, but it's output has a perfect 50/50 duty cycle.
The disadvantage is that you have a square :-) that begs conversion into another shape.
More information about the Synth-diy