[sdiy] super cheap synthisizers

[Theo]

Phillip,
Some snippets
:
> By the way are there two cross-threads occurring?  I thought this started
> with someone wanting to implement a "single synth voice per micro"?
>
Yes your right,
Hope that someone (sorry deleted the initial post already) got
his answers from this tread.

Ill try to resume:
The answer is yes, but if you want more than one oscillator,
you probably need to add external hardware.
This would be either a DAC  for phase accumulator DSP oscillators or
extra counters/timers for a "juno" type DCO.(see below).
A highspeed uC like SX or the new Dallas DS89c420 is also a solution,
but these may not be cheap uC.

I just checked the specs of the Dallas part.
It DOES have the two 16bit timers I wanted, actually it got 3 of them :)
Still, there are DSP available with more power for less...

> "The goal is analogue oscillators with digital timing correction."
>
Ok sorry , not THE goal but MY goal and probably different from what the
person originally posting the question had in mind.

> Let's see if i get the premise.
>
> The sum of all analog voltages into a DCO is samples with enough precision
> that a correlation to an expected waveform frequency is achieved. Assuming
> the waveform frequency is measured by it's period. A simple calculation
> (this is 1V/8ve isn't it) is performed and a correction voltage is
applied.

Not quite.
- A linear ramp is fed a _un-precise_ control voltage
   a 5 bit DAC is good enough, the only function is to keep the amplitude
   more or less constant over the audio range.
- A downwards counter running at a HIGH frequency is loaded with
   a number representing the period time.
- The ramp is reset when the counter reaches zero.
- Voila, a numerical controlled frequency stable saw-tooth wave :-)
This method is _very_ lean on processor power and needs minimal external
parts.

Cheers,
Theo



From: phillip m gallo <philgallo at attglobal.net>

>"The goal is analogue oscillators with digital timing correction."
>
>Let's see if i get the premise.
>
>The sum of all analog voltages into a DCO is samples with enough precision
>that a correlation to an expected waveform frequency is achieved. Assuming
>the waveform frequency is measured by it's period. A simple calculation
>(this is 1V/8ve isn't it) is performed and a correction voltage is applied.
>
>
> "Its just a personal taste, but IMO running a interrupt at a normal sample
> rate and do phase accumulation with DAC output as someone proposed  is out
> of the question."
>
> Of course if you do the phase accumulator approach and feed this into a
> phase comparitor with your oscillator output as the other input term you
> have the std phase lock loop with the correction voltage applied to your
> linear osc.
>
> By the way are there two cross-threads occuring?  I thought this started
> with someone wanting to implement a "single synth voice per micro"?
>
> regards,
> p
>
>