[sdiy] Analogue VCO drift/jitter analysis results

[rburnett at richieburnett.co.uk]

Hi all,

I've uploaded some raw data (and a nice graph!) for the VCO stability 
testing I did on my old Roland SH-09 analogue monosynth:

http://www.richieburnett.co.uk/temp/vcodrift.zip

The Excel file "vco drift.xls" contains a graph showing how the phase 
of the SH-09's 1kHz sawtooth VCO wandered relative to an accurate 
digital reference oscillator over a period of 5 seconds.  The prominent 
"shark's fin" shape repeats roughly every second and is due to residual 
modulation from the synth's on-board LFO which was running at approx 
1Hz.  The smaller ripples repeat every 20ms and are due to frequency 
modulation from the 50Hz mains supply here in the UK.  You can change 
the axes on the graph to look in more detail at specific features.  (The 
phase axis is in radians, so you need to divide this number by 2pi to 
get it in complete cycles, and then multiply by 360 if you want an 
answer in degrees.)

For a measure of the instantaneous pitch error we need to look at the 
gradient of this phase graph.  If you take the large rising edge of the 
first "sharks fin" on the left the phase slews through 0.3 radians in 
about half a second.  Thats roughly 1/20th of a cycle in half a second, 
or a tenth of a cycle in one second.  So the pitch error during the 
rising edge of that sharks fin is about 0.1Hz, not much for a 1kHz tone. 
You can do the same gradient calculation to find the amount of 
frequency variation due to the smaller 50Hz ripples too.

The file "saw_1kHz.wav" is the original 1kHz sawtooth VCO straight out 
of the synth recorded at 48kHz 16-bit.  You can use this to do your own 
analysis if you want.  For instance, windowing this data and performing 
a big FFT will show the 1kHz sawtooth's harmonics, plus all of the 
modulation sidebands at 1Hz spacings due to the LFO, plus all of the 
modulation sidebands at 50Hz spacings due to the mains.  (I've posted 
these phase-jitter sideband plots here before like: 
http://www.richieburnett.co.uk/temp/graph1.tif   ...graph2.tif  
...graph3.tif etc.)

The file "complex_ampl.wav" contains the real and imaginary parts of 
the complex amplitude of the oscillator in left and right channels of a 
stereo wave file.  This is just one of the steps towards accurately 
finding out how the phase of the oscillator wobbles about!  (If you have 
Goldwave you play this file with X-Y visual display enabled and watch 
the phase of the oscillator wander about as a tiny dot in realtime.  
Playing at normal speed you can easily see the periodic 
clockwise/anti-clockwise rotation of the dot due to the LFO modulation, 
and slowing the playback down to 5% speed reveals the much faster 50Hz 
modulation!)

If you're interested in analogue VCO fatness, drift, jitter, whatever, 
please take a look at this data and let me know what you think.  This 
topic comes up over and over again but nobody ever seems to share any 
investigative results.

Also, I realise that this test wasn't performed on a vintage Minimoog.  
If someone wants to provide one, or more likely a bare oscillator 
recording i'll gladly run the same analysis.

-Richie Burnett,