[sdiy] Fatman module??

[Richard Wentk]

At 10:52 03/03/2003 -0800, harrybissell at prodigy.net wrote:
>What I'd like to see is a common specification for a synth "backplane"
>
>This would be some form of card cage and connector spec that could be
>used to support the building of polyphonic or 'multiphonic' synths.
>
>I'd see having a MIDI/CV card, maybe a memory or programming card,
>and a number of voice cards. If we agreed on a standard board size
>and connection scheme, then any DIY member could design a voice card
>(or any other card) for the rack.

If I were going to do this I would:

1. Create LFOs and envelopes digitally using a single card to emulate a 
bunch of hardware
2. Multiplex this and other voice data over the bus so you could have up to 
(say) 32 channels without using a ton of cabling

Addressing would either be by way of a handful of digital address lines and 
a clock, or you could use two lines - a master clock and a counter reset - 
onto two lines and put demux control counters onto the boards.

I'd imagine you'd need these lines as a bare minimum, which would 
pre-summed digitally to include all the possible modulation sources (except 
poly mod):
VCO 1 pitch
VCO 2 pitch
VCF cutoff
VCF Q
VCA volume
(Output pan?)

The problem with using discrete lines for individual cards is that you're 
inevitably limited in what you can do. Using the muxed approach you can 
squeeze the information for 32 cards into a handful of lines, and the 
design is inherently multitimbral. Doing it with a discrete backplane I 
doubt you'd be able to get more than 8 voices together.

The voice cards themselves would have just a couple of VCOs , sync, polymod 
control (you *have* to have filter and VCO FM on an analogue synth :) ), 
perhaps a ring mod, the VCF and VCA, and a bunch of latches and S&Hs for 
the modulation inputs.

The hard part would be deciding how to handle the possible poly mod 
routings, and also mix control for the VCO waveform outs, the ring mod and 
noise, and so on. Some of these could be switched digitally instead of 
given continuous level controls.

Then again, given how cheap processors are today, it's not so very nuts to 
do *all* the distribution digitally, with a processor on each card making 
sense of the input information from an 8-bit data bus on the backplane. 
That would certainly be more reliable and easier to design than analogue 
muxing, and probably wouldn't be all that much more expensive. (If at all.)

So you'd have:

1. A master processor, including the envelope and LFO simulation as before, 
and also MIDI decoding and basic pitch calculations. Something moderately 
beefy (but not insanely so) would be needed here.
2. A distribution bus. 8-bit data and 8-bit address plus some control lines 
and a couple of audio out lines would probably be enough at a pinch. More 
would be nicer, but maybe not essential.
3. A voice card with two VCOs, one VCF, a VCA, a ring mod, noise, poly mod 
mixers/switchers, a processor, and a 16-way analogue demux to source the 
CVs. All of this should fit onto a Eurocard without too much of a stretch, 
especially if it's based around one of the CEM chips.

It's still not cheap, but by the time you're done you'd have a home-made 
Rhodes Chroma in a rack of manageable proportions that wouldn't be so 
insanely difficult to make.

Of course it's nothing like a Fatman any more. But if you want a poly one 
of those, I'm thinking the best way would be to make a rack that's 19" high 
and 8U wide and just stack them vertically for effect. :)

The other plus is you can go nuts with the LFO designs, and do a bunch of 
stuff analogue designs can't usually do. (MIDI sync and all that, weird 
waveshapes, and so on.) Likewise it's easy to do things like pitch tracking 
for the envelopes, which is a high-overhead kind of thing for straight 
analogue.

Richard