[sdiy] jitter in oscillators for music purposes

[Czech Martin]

The partials of a saw wave do not really fade out soon.
the 1/N law is tranlated into -20log10(N), the 100th will thus
have only -40db attenuation. As long as they stay where they are,
this doesn't matter. If they are wrapped arround by aliasing,
it will become very obvious.

If the tone sweeps up, the reflected partials will sweep down.
Very ugly.

The problem here is not to create a table based cyclical
waveform by band limited synthesis.

The problem is to compute a saw core PLUS noise,
in order to get the jitter. This is not ADDING noise
in a linear fashion, but this pesky phase + amplitude modulation
business with high bandwidth noise. After that is done precisely,
bandwidth will of course be reduced to play back the result.
That means that a (very simple) equation solver has to be set up,
with sufficient small time step, this step size can reach from
1us to 1ns, depending on the asctual error.
After that 1st band limiting and equidistant sampling.
After that downsampling to something reasonable.


m.c.

> -----Original Message-----
> From: Sean Costello [mailto:seancostello2003 at comcast.net]
> Sent: Donnerstag, 15. Juli 2004 21:32
> To: Antti Huovilainen; Czech Martin
> Cc: Colin f; synth-diy at dropmix.xs4all.nl
> Subject: Re: [sdiy] jitter in oscillators for music purposes
> 
> 
> Rumor has it that the Clavia synths just run their saws, 
> squares, etc., at
> the output sampling rate of 96 kHz, and that no oversampling 
> is used. The
> reason that this might work is that the harmonics of a 
> sawtooth that are
> above 48 KHz will probably have insignificant amplitude, 
> depending on the
> frequency of the sawtooth. The Nord Modular also has a 
> control rate running
> at 1/4 the sampling rate, which suggests that they are 
> working with vectors
> of samples, as opposed to generating audio on a 
> sample-by-sample basis.
> 
> If you are working with a computer, oversampling is probably 
> a necessity,
> assuming you want to generate native waveforms, as opposed to 
> using BLIT, or
> crossfading between single cycle samples. However, a hardware 
> DSP system can
> use a higher output sampling rate, which gives you more 
> headroom (in Hz) for
> the harmonics to die away to a reasonable amplitude. Or, you  
> could use
> oversampling, but with lower oversampling ratios and less 
> steep filters
> (i.e. lower order filters).
> 
> Sean Costello
> 
> ----- Original Message ----- 
> From: "Antti Huovilainen" <ajhuovil at cc.hut.fi>
> To: "Czech Martin" <Martin.Czech at Micronas.com>
> Cc: "Colin f" <colin at colinfraser.com>; <synth-diy at dropmix.xs4all.nl>
> Sent: Wednesday, July 14, 2004 5:45 AM
> Subject: RE: [sdiy] jitter in oscillators for music purposes
> 
> 
> > On Mon, 12 Jul 2004, Czech Martin wrote:
> >
> > > But the creation of saws via discrete *integration*
> > > will create some alias as well....
> >
> > Here using minblep / blit should help. Google for 
> icmc01-hardsync.pdf and
> > blit.pdf
> >
> > > I wonder if oversampling will help a lot...
> >
> > Oversampling helps around 12dB/octave (one times 6dB before 
> aliasing and
> > another 6dB after aliasing since the oversampled part of 
> alias is filtered
> > away). I'd suggest combination of minblep/blit and oversampling.
> >
> > Antti
> >
> > Give a man a fire, and he'll be warm that day,
> > Set him alight and he'll be warm for the rest of his life
> 
>