[sdiy] Analysis of frequency variation in analogue synths

[Magnus Danielson]

From: Richard Wentk <richard at skydancer.com>
Subject: Re: [sdiy] Analysis of frequency variation in analogue synths
Date: Fri, 4 May 2007 01:03:15 +0100
Message-ID: <E5206455-B398-4AA9-A219-5373C3DAB133 at skydancer.com>

> 
> On 3 May 2007, at 15:37, Colin f wrote:
> 
> > My point is that there is an untested assumption here about the  
> > degree to
> > which it depends on the kind of modulation.
> > Before you go to the effort of carefully analysing all the possible
> > influences, it's worth testing empirically whether a fairly simple
> > combination of low and high speed cyclic variation plus a random  
> > component
> > modulating a digital oscillator's pitch and amplitude sounds any  
> > more or
> > less analogue than a real analogue VCO.
> > Reproducing detail that you can't hear is wasted effort.
> 
> Yes, but this assumes that a steady state unfiltered waveform is  
> likely to be audibly analogue.

The one-cycle waveform repeat vs. continous waveform experiment is interesting
in that it attempts to remove time-variant modulations except those at and
integer multiples of the waveform fundamental itself. Weither it is PM, AM,
FM, drift or any obscure combo or something else doesn't really care. There is
however a problem with the approach and that is that care needs to be taken to
ensure that the VCO is tuned to form an integer (or near integer) multiple of
samples. The splicing needs some cross-fading to make the wrapping less
audioble or else that will dominate and by itself take attention away from
whatever else there might be there that we wanted to check.

> Does that seem like a convincing assumption? Might it not be the way  
> as much about the dynamics of the sound - pitch being only a small  
> part - than the harmonic content?
> 
> > The failure of analogue modelling so far to fool the hard-core of  
> > analogue
> > fans is probably more economic than technical.
> 
> It depends on the quality of the models. If the models aren't  
> detailed enough, the sound won't be convincing.

A VCO uses a very simple model indeed, we charge a cap with a current and when
the voltage reaches a certain level we reset it (or change direction of the
current). Yeat, that simple model results in problems. If we analyze it with
modern cyclostationary view on things, we infact see a high variance over the
phase in respect of noise injection and how 1/f noise injects into the
oscillator. With some care one should be able to model it up that way. The
Leeson model taught us how 1/f noise turns into higher noise forms.

Cheers,
Magnus