[sdiy] Daft Idea, LINUX SYNTH

[Rainer Buchty]

>    3.  The software (code) part, at least, is pretty easy.  If fact,
> that is probably the reason why Linux is a success.  The one problem I
> see however is that a processor will need to be chosen for the code to
> run on.  Because of the differences between processors, and the level at
> which the code must be written, it is going to be very difficult to make
> the code both portable and efficient on all processors.  And, I would
> imagine that there will be 600 different choices of which micro would be
> the best to use.

My opinion on that: Forget about the processor discussion at first place.

Taking the modular approach boards will eventually come with their very
own processor doing dedicated stuff (e.g. real-time filter calculation,
fx processing etc.) which is completely independent from what the main
controlling unit has to do. So the initial discussion is less "what
processor should I take" but more like "how should this interconnection
bus look like" (where I'd favour the good old 68k approach with UDS/LDS
making it possible to have either 16 or 8bit bus width).

The modules can use whatever they want and be programmed in any way - all
they have to provide is some control and status registers to allow the
main CPU to parametrize the module. Here, of course, standards of
operation and intercommunication need to be defined (PIO vs. IRQ vs. DMA -
where my voting goes towards PIO since it's the "least common multiple"
making each module just a batch of registers somewhere in the memory map
of the main controller which under all circumstances will keep the
control over the entire system).

The main controller, however, needs to run a lean, efficient RTOS to cope
with MIDI messages, user interaction (controllers) and sound synthesis
rules in real time. That IMO can be done almost in C which makes porting
the stuff to the processor if choice is much easier than writing
everything entirely in assembly language.

As you can see, my vision of this OpenSynth is a most modular and very
multi-processing approach where the sound generation is distributed among
all attached modules, the main controller just takes care of proper
parametrization of these.

There are a lot of ideas and concepts floating around which - if they'd be
put together in one big project - could really give some sort of monster
synth. I'm quite sure, that RTOS is already written. And for sure there
are many interesting DIY projects out there, some of them being pure
analog stuff (like those being discussed here), others are hybrids,
possibly another bunch purely digital like IIR filters, effect modules
etc.

>    Now that I have gotten that out of my system...another posible idea
> is maybe not so much of a working system as a collection (archive) of
> contributions from list members that is all in one place for anybody to
> use.

That would indeed be a great start.

Rainer

-- 

Rainer Buchty, LRR, Technical University of Munich
Phone: +49 89 289-28401, Fax +49 89 289-28232, Room S3240