[sdiy] super cheap synthisizers

[Theo]

Ingo,

With a phase accu the output is frequency is independent from the sample
frequency.
The modulo counter is always increased with the same number
and used as a pointer to a sample look up table.
This way the frequency is rock solid but you wave form does NOT always start
with sample #1.

For saw the counter output it self will do.
Note that there will be amplitude jitter when doing a saw wave this way.
This only is a problem when your output frequency get "close" to the sample
frequency.
A high sample rate like Paul uses solves this problem.

The phase accu method is only good if you use a DAC.
But because the generated wave forms are always full code,
you don't need a high resolution DAC.
Think 8 bit R2R will be plenty.

Using a phase accu to reset a ramp like some pp proposed won't work.
This would cause frequency jitter AND give low frequency resolution.
Unlike the sample read out, the ramp has no fixed phase relation with the
accu.
Frequency jitter because the number of clocks in-between resets will differ.
Low resolution because this would be the same as using a "slow" timer.

Theo




From: Ingo Debus <debus at cityweb.de>
>
> harry wrote:
> >
> > I would NOT assume that you could make one pulse 'shorter' and one
'longer'...
> > if you did it evenly... it would sound like a pulse of 1/2 the frequency
you intended...
> > with a non 50% duty cycle...
> >
> > If it was a random 'short-long' it would sound like sh!t.  (well to be
fair... I am known
> > for saying things sound like sh!t that some folks like...).  This would
be 'jitter' in the
> > frequency and would not be a good feature.. OK to add it if you want
'that' sound..
> > but if you could not avoid it... really bad.
>
> Isn't that more or less what the phase accumulator oscillator does? As I
> understand it, it runs at a fixed sampling rate, regardless of the
> oscillator frequency. Given the values used are precise enough (someone
> here wrote 24 bit) it will try to keep the frequency as precise as
> possible; but since the output is only updated at the sampling frequency
> the period time is sometimes one sample longer, if the sampling rate is
> not exactly an integer multiple of the output frequency. Sort of jitter,
> correct? (Probably not random.)
>
> Is this the reason why Paul Maddox suggests such a high sampling rate, 1
> MHz? To keep the jitter low?
>
> Ingo
>
>