[sdiy] Fatman module??

[Richard Wentk]

At 20:00 03/03/2003 -0500, harrybissell wrote:
>Hello Richard and all... (inline and snipped)
>
>Richard Wentk wrote:
>
> > 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
>
>I think that probably once you reach maybe 8 voices, an analog design is
>unlikely
>to make any sense.

I think this depends how big the cards are. For a player, you really do 
need 16 voices or more, because then you can be sure that notes won't ever 
be stolen. And you can also do useful things like doubling up patches for 
really really big sounds.

And of course there's always the multitimbral option.

I don't think getting 32 voices into something rack mounted is all that 
impossible. In fact with the right design I'm not convinced it would be all 
that much more expensive or complicated to build than an 8-voice system 
with heavily discrete voice cards.

Thinking about it, with a standardised digital interface there would be no 
reason not to make the thing stackable and modular, as long as the s/ware 
could cope. Build another rack, plug it into the bus, let the processor 
know you have another 8 or 12 voices or however many fit into a standard 
rack size, and you're away.

>I was thinking that you would have things like
>
>(8) - note CV
>(1) VCO1 offset
>(1) VCO2 offset
>(2) VCO mix
>(4) ADSR 1
>(4) ADSR 2
>(1) VCF Fc
>(1) VCF Q
>(1) VCF ENV AMT
>(1) VCA ENV AMT
>(1) LFO rate
>(1) LFO depth
>
>all analog... maybe more.  I'm thinking these are not multiplexed - direct DC
>voltages

But then you immediately run into the problem of connector quality, board 
and backplane sizes, and so on. Even for an 8-voice monotimbral synth you 
end up with huge, expensive connectors and complex board routing issues.

The big advantage of a muxed digital system is you can use cheap and nasty 
ribbon cable and it will continue to work reliably. If you have 8 data and 
8 address lines you can squeeze everything into a couple of cheap sockets. 
If the data is sent in frames and there's a processor on each voice card 
you can eliminate all the analogue hardware needed to create analogue LFOs 
and envelopes, add a *lot* of extra functionality, and do the whole shebang 
with minimal or even negative extra cost.

>I agree... but don't think you'd need 32 voices.  I'd even think that you 
>might
>
>want to have different voice cards available.  (designer flavors ???)

Okay, you could still do this. If you standardise:

1. The form factor
2. The digital demux and interface (I'm thinking a 16 way CV demux should 
be enough for each voice)
3. The outline voice architecture

you could fill in the rest of the details however you wanted. If you know 
the card has two VCOs, you could still pick and choose which VCOs to use 
and how to build them. Standardising the digital interface for each card 
means you would know which control voltages would be present, and what 
parameters they're supposed to control, and where the outputs go. After 
that, the rest is up to you.

>I'm not a fan of soft envelopes myself. I'd rather go discrete ....

Any particular reason? Done well I don't see what the disadvantages would 
be. Not only do you get better control and also useful options like 
retrigger modes, but you can also do things like control the shape of the 
envelope curves so you're not limited to a standard exponential.

And this is all on a couple of chips that can produce up to (say) four 
complex envelopes per voice, and a couple of MIDI syncable LFOs with lots 
of fun waveshapes.

Yet another advantage is that you can add more complex modulation options 
like various logical and arithmetical combinations of mod sources. (Check 
out the Waldforf synths, which have a lot of good ideas here.)

And finally the whole thing becomes more open ended. S/ware is much easier 
to extend with new features than h/ware. It would be easy to build a 
bare-bones V1 system and then add extra tweaks and twiddles as the firmware 
gets updated.

>I'd probably beware the CEM chips because of availability. They would reduce
>board space quite a bit

They are a bit of a mixed blessing. Does anyone have a Star Trek replicator 
to hand? :)

Richard