[sdiy] Voltage references in VCO

[René Schmitz]

Hi all,

elmacaco wrote:
> That sounds interesting,
> 
> It would be cool if you could use it to modulate osc tracking as well, for
> authenticity ;)
> 
> But that's assuming it is the drift in the oscillators that is what is
> pleasing.

I think its not that simple. But drift is part of it. Here's how I see 
it: The thing is, what is drift actually? It might be thermally, this is 
what we are mostly focussed on, but there are surely other factors. Like 
1/f noise.

In essence a voltage reference, an opamp, comparator or any other 
component has its own noise, some of that with so low frequencies that 
we can measure it on a timescale that aproaches minutes or even days or 
weeks. (This is then what you'll find as long term stability on 
datasheets.) This doesn't have to do with temperature as such, since 
such measurements would be made in a thermostated environment. The lower 
the noise, the better the reference. (Neglecting thermal influcence for 
now.)

If my interpretation is right, then for a controlled modulation of an 
ideal VCO (if that were possible) you would need a source of randomness 
that has a zero DC effect, and a spectrum which falls off sharply below 
some fractions of Hz. So all influences that you might interpret as 
detuning, cancel out on average.
Say a rise in pitch of 10 cents that would average out in weeks wouldn't 
be acceptable, but say 1 cents over timescales of 10s of seconds would 
perhaps.

Of course temperature fluctuations have a similar spectrum on similar 
timescales. But you have another variable than only time, and you can 
measure that. As Ian has said, you can cancel the net effect of thermal 
influences. For example with some sensor and appropriate scaling circuitry.

> I don't know what it is, but I find older oscillators much more interesting
> straight tone wise then newer ones.   Moog, older Roland (system 100,
> SH-09, -2,-5, & -7) pleasing in a similar way yet the rolands don't have
> much drift in my experience.

I think part of that has to do with the modern components. Low offset 
drift OPs mustnot have much 1/f noise, so the "interesting" range of 
"drift" is also reduced. With noisier components you buy the 
"interesting" and the "annoying" in one package. The tradeoff to me 
seems to be to have enough to be interesting and little enough not to be 
annoyed.

Cheers,
  René

-- 
uzs159 at uni-bonn.de
http://www.uni-bonn.de/~uzs159