[sdiy] Moog wander

[Magnus Danielson]

From: René Schmitz <uzs159 at uni-bonn.de>
Subject: Re: [sdiy] Moog wander
Date: Mon, 17 Apr 2006 00:19:57 +0200
Message-ID: <4442C30D.8000802 at uni-bonn.de>

> Hi Magnus and all,

Hi René,

> Magnus Danielson wrote:
>  > However, it is not bogus when you start to
>  > think about what we can do about the state of things.
> 
> Thats where I was aiming at. :-)

Fine! :-D

As I see it, a good understanding of the sources of modulation (instabilities
if you want) and their action gives us several choices, one is to design stable
oscillators (our main target have been temperature compensation IMHO) but it
may very well be to make "interestingly" unstable oscillators.

There are these physcoacoustical sensations that is refered to as "cold" and
"warm". They keep creating a heated (warm?) debate. I keep wondering what kind
of measurable characteristics is actually correlating to either of them.

> > I'm expecting to see AM, PM and FM in there.
> 
> Sure.
> 
> > Whenever I look over longer time intervals (such as 1024 frequency measures
> > back-to-back with 100 ms inbetween) I clearly see some random walking about
> > from the Mini (VCO 1 at about 167 Hz) and also a raise in the noise at the
> > bottom of the frequency scale when looking at the FFT of the frequency
> > measures (thus FFTing the frequency modulation). I have a longer run going
> > right now for tau = 1s.
> 
> Thats going to be very interesting!

Indeed.

When I came down I did indeed see a clearer flicker or random walk in
frequency. Sure enought, I had a nice slope all over the FFT spectrum. I need
to work some to make a good estimate of the slope thought, but work is under
way. This was a nice measure of VCO 1 of my MiniMoog running at about 167 Hz
with 1024 back-to-back frequency measures with a second apart, totalling in
1024 second measure. The Mini had a few hours of heating and has the new
oscillator board with 726 if you wonder.

The slope of the noise drops about 6 db as it goes from 48 mHz to 132 mHz.
Mind you that this measuing frequency, so the noise is FM noise and needs a
correction for phase slope. Doing a few more measures gives away something like
3-4 dB/Oct of slope (frequency does not come out log but lin). If it is
3 dB/Oct it would be flicker frequency modulation and have f^-3 as PSD
characteristics.

Normal averaging doesn't beat that, so at least Allan deviation is needed.

I think I have a plan to help out on the trigger issue. If it wheren't for the
big block of metal in the middle of my lab-bench...

Cheers,
Magnus