[sdiy] Analysis of frequency variation in analogue synths

[Tom Wiltshire]

On 4 May 2007, at 15:23, Scott Gravenhorst wrote:

> I suggest that this question will not be answered by banter.

Hear hear!

> Assuming that it is worthy, what is necessary is not clever debate,  
> rather a set
> of well thought out experiments designed to eliminate error,  
> ambiguity and
> argument.  The experiments should be repeatable by others.
>
> I've seen some good discussion with respect to defining things we  
> can quantify,
> so let's get to work and make some numbers.

Yes, that's what I've been trying to do, which is why I've been a bit  
quiet during a discussion which I began. I'll post some results when  
I've got them.

> As for me, I can live without the answer.  My own quest is just to  
> make something
> new and/or different.

Please don't misunderstand me - my aim is not to try and slavishly  
copy analogue VCOs using digital technology in an attempt to capture  
some mythical 'vintage' sound. I'll leave that for the big boys  
(Roland, Yamaha,etc) and their virtual analogs. Rather, what  
interests me most is the differences between artificial sounds and  
natural sounds. Most synthesizers, much as I love 'em, sound like,..  
well, like SYNTHESIZERS. I'm trying to create genuinely original  
natural-sounding sounds - the sound of a rubber string plucked in a  
helium atmosphere, for example, or a stone dropped in a pond of mercury.

It seems to me that nature includes all sorts of random variations,  
but that the crucial thing about these variations is that each has  
its own character. 1/f noise got mentioned during the discussion  
quite a bit, and that's a good example. 1/f noise sounds more natural  
than white noise - it has the right character. What I'm looking for  
is distinctive types or styles of noise and variation that I can then  
(mis)apply to other parameters in my synth architecture to generate  
noises that are _different_, yet still somehow natural-sounding.
My investigations into VCOs are a part of this, since the old  
argument about VCOs vs DCOs boils down to "VCOs sound more natural  
than DCOs". Assuming this is true, what are the characteristics of  
the random variations in VCOs that give them their natural sound?

The sort of thing I'm talking about is demonstrated by two recent  
experiments with noise generation:

http://tomwiltshire.co.uk/sdiy/noisegen/Experiment1.php

This is a 32-bit shift register noise generator producing 8-bit  
values, which are then fed into a simple averaging filter. The 8  
different graphs show the effect of averaging different numbers of  
samples. You lose high frequencies from the noise, and reduce its  
amplitude. Over the medium term, and averaging over a decent number  
of samples, the net result is zero.


http://tomwiltshire.co.uk/sdiy/noisegen/Experiment2.php

This is a "random walk". It uses the same random number generator,  
but only uses one bit. The algorithm stores the current value and the  
direction, either up or down. Each time period it moves a single step  
in the given direction, either +1 or -1. Then it looks at the next  
bit from the random number generator, and if it's a one, it changes  
direction. This generates a random line with a different character to  
the averaged 8-bit noise of experiment one. Although in the long term  
this should also tend to give a net result of zero, in the medium  
term it shows considerable drift away from the centre, since it is  
unlimited. It could theorectically drift many thousands away from  
zero (and given long enough, it will!).

Which of these is most like the drift/noise/variation in synth X/Y/Zs  
VCOs/Filter/VCA? Which is most like the variation in a human singer's  
vocal chords? Which is like the variations of mud when a hippo sits  
down?

I hope I've convinced you all that this _is_ a worthy  topic! It's a  
hard question and I could use some good ideas!

Tom