[sdiy] VCLFO and VCADSR details available online

[Tom Wiltshire]

On 19 Dec 2006, at 06:08, Mike wrote:

>>  I see a missing gap
>> in today's market for some kind of low end polysynth.
>>
>
> I was thinking the same, and how with a little circuit board 
> re-juggling
> labolida's 'TinySynth' might make a good voice card to experiment with.
> Initially I was just thinking of 8 Tiny's in a rack frame, a bit like 
> an old
> TX816, but with front panel controls, later some control software and
> another micro per channel to handle the 'knob's.
>
> Silly idea, but actually affordable for an experiment.

Definitely agree with you both. I'd thought that perhaps rather than 
using big powerful expensive processors (that need big powerful 
expensive development hardware) it might make more sense to use many 
smaller cheaper processors. My VCLFO and VCADSR designs were part of 
this philosophy. The idea was to follow the basic analogue polysynth 
design of the 1980's, with a control panel which is scanned by 
(another) processor and patch memory which then just output a heap of 
CVs to control the voices. Whilst all this goes on, another processor 
scans the keyboard, deals with voice assignment, and sends gate signals 
to the voices.

I've hit two problems (or difficulties anyway). One is that the 
separation of the panel scanning and keyboard scanning makes a complete 
MIDI specification quite difficult. The keyboard processor can deal 
with Note On/Note Off and perhaps pitch bend and such like, but it 
doesn't have the information it needs to send Control Changes when 
someone tweaks a panel control. This means you either need to make 
these two units talk to each other, or you need to do both jobs with 
one processor. And then you're to using one of those big powerful 
expensive processors that I was trying to avoid...

The second problem is that dealing with CVs for modulations is actually 
quite complicated. Even a fairly simple synth could have 4 or 5 
modulation sources and 10 destinations - pick your figures according to 
taste, but please bear with me for the purposes of example.
Assuming that we want a pretty complete modulation matrix where we can 
send any source to any destination, we need a CV mixer for each 
destination, and a VCA for each possibility - 40 or 50 VCAs in my 
example. This isn't practical, and it's never been done this way, as 
far as I know. The Prophet 5, for example, uses a VCA for each _source_ 
(so only 4 or 5 VCAs in my example) and then uses switches to choose 
which destinations to send the signal to. This still requires a CMOS 
switch for each point in the matrix for a programmable system. 40 or 50 
CMOS switches isn't impossible (10-12 quad packages) but requires a 
considerable amount of board space. This becomes an issue when this 
circuit has to be repeated for each voice.

So, in summary, CVs rapidly become very complicated to route. Keeping 
modulations in the digital domain is certainly one way around this 
problem, but again, the easiest way to do this is to have one big 
powerful processor that deals with it all. I've wondered if one 
'modulation processor' per voice might be a half-way house. It'd need 
to be fairly fast, but might not need to be hugely powerful. This would 
generate 2 or 3 envelopes and an LFO or 2 for each voice. All 
modulation routing would be done digitally and the chip would just 
output a few summed voltages to go directly to filter cutoff or 
wherever.

So, there you go! My thoughts about building a hybrid polysynth without 
selling your granny!

Tom


++++ Electric Druid ++++
Web Design & Development
http://www.electricdruid.net