[sdiy] oscillator jitter / phase noise

[Fredrik Carlqvist]

Hi all!

I've been thinking along these lines too. My new project has a saw/square
DCO that doesn't sound too fantastic. This is somewhat improved by the
filters, but a certain "phatigue" is missing. So I was thinking about
introducing jitter. 

I am no fourier expert (although I should be, I'm a mathematician...), but
the jitter should introduce not only its own low-freq noise sidebands, but
also, and this is what I think is the key, widen the bands of overtones. 

With a quartz controlled oscillator, the peaks in the spectrum are very
narrow. In the human ear, there are N sensory cells for each logarithmic
frequency interval, i. e. every half note step has a certain number of
hearing cells. When widening the peaks, more sensory cells will be
stimulated and - here's my two cents - fatten the sound. 

Seems like Magnus is the real fourier expert on this forum, he can maybe
check the validity of my reasoning.

Slow waveform modulation (as in the wave not looking exactly the same for
each period) would introduce low-band noise without the fattening if the
timing is kept exact. This is not what we want.

Timing irregularities (jitter) would widen the spectral lines of the wave
and probably also introduce low-band noise. This could be kept to a minimum
if the jitter is kept in a short small cycle. Perhaps the wave shape could
be distorted to cancel the low-band noise exactly.

I have been thinking about both cyclic and random jitter. I think the cyclic
will be just as efficient, and it is a lot cheaper in software, but I
haven't tested it yet.

Fredrik C