[sdiy] Analogue Drift (was Re: HF VCOs and tracking problems)

[rburnett at richieburnett.co.uk]

Hi Ian, Tom and all,

Thanks for your comments about my "VCO wobble" experiment, I had no 
idea if anyone else was interested other than me and Tom! Some people 
can detect incredibly small variations in pitch at certain frequencies 
or in particular surroundings.  Any sort of reverberation or static 
reflection in a room seems to make tiny pitch fluctuations in the source 
sound much more noticeable.  I guess constructive/destructive 
interference in the room converts miniscule pitch variations into more 
noticeable amplitude fluctuations.

Another quick & easy way to see if an oscillator's frequency fluctuates 
is to do a long-time FFT on its output using Octave or MATLAB.  Any 
frequency variation is clearly visible as FM sidebands in the spectrum 
(sometimes called "phase noise spurs")

Here's three example spectra that I produced just now from the SH-09 
hidden under my desk at work...

http://www.richieburnett.co.uk/temp/graph1.tif

This shows the fundamental, 3rd harmonic and 5th harmonic of the top C 
square wave output at 4232.2Hz, 12696.8Hz and 21160.2Hz respectively 
(albeit 19 cents sharp!)  I've also labelled the 2nd harmonic at 
8464.4Hz, who's presence indicates the squarewave isn't exactly square.  
The SH-09 has two CMOS octave dividers for the sub-oscillator options, 
and although the MIX slider for this is firmly down, they still leak 
through to the output in the form of the lines at 1058, 2116.2 and 
3174.2Hz.  You can also see their harmonics interspersed between the 
main squarewave harmonics right across the spectrum. Also notice the 
cluster of sidebands close in to the 4232.2Hz fundamental...

http://www.richieburnett.co.uk/temp/graph2.tif

This shows the spectrum zoomed-in around the fundamental frequency of 
the oscillator.  You can see there are multiple pairs of sidebands 
centred around the fundamental at various offsets.  The first pair 
(4182.2Hz, 4282.2Hz) are +/-50Hz either side of the nominal oscillator 
pitch.  The second pair at +/-100Hz are much smaller, followed by 
+/-150Hz, +/-250Hz, +/-350Hz predominently.  These are the common odd 
harmonics present on the mains line here in the UK.  The low amplitude 
of the +/-100Hz pair suggests that 100Hz ripple in the full-wave 
rectified power supply is not a very significant contributor to pitch 
wobble. Also notice that there is still another cluster of sidebands 
even closer in to the fundamental frequency...

http://www.richieburnett.co.uk/temp/graph3.tif

This shows the spectrum zoomed-in even more in the vicinity of the 
fundamental at 4232.2Hz.  You can see there are sidebands at +/-1.6Hz 
and +/-3.2Hz.  These more or less appear regularly every 1.6Hz but roll 
off quickly.  1.6Hz is the frequency that the LFO of the synth was set 
to during this test, and shows that significant VCO modulation is taking 
place despite the MOD slider being firmly down.

x-axis scale is Hz, y-axis is dBs.  Anyone can do this test on their 
own oscillators using a PC sound-card and a maths package capable of 
calculating large FFTs in the 100's of thousands of points.

Regards,

-Richie,

PS. Well done biting the bullet and building a digital oscillator! :-)