Dear Andy, Wolf, Markus, LF Group,

I tried some further experiments with the same USB sound card and an additional computer (the one I am writing this mail on). The "real" offset in sound card clock rate is fairly stable (within a couple of ppm) at +48ppm. I reset data in the Spec Lab calibration table to 48000 samples/sec nominal and re-measured the sample rate error with the lap-top PC; this was -115ppm. I then transferred the USB sound card and calibration source to this desk top PC and left it running during the day, checking the sample rate error from time to time. The SR error showed small drifts consistent with the thermal drift of the sound card clock crystal, together with much larger step changes in sample rate, which occured when there was "activity" on the PC like opening files or applications, plugging in a USB stick, checking e-mail etc. Once the step had occured, the sample rate would stay very close to the new rate until the next step occured. It became clear there were discrete step sizes - I observed:
+48ppm (for a frustrating few minutes...)
-33.5ppm
-114.9ppm
-277.7ppm
-359ppm

If you subtract the +48ppm actual clock offset, you get offsets of:
0ppm
-81.5ppm
-162.9ppm
-325.7ppm
-407ppm

if you calculate what this is in samples/sec at 48k sample rate, it is close to: 0, -4, -8, -16, -20 samples/sec, so it looks like samples evaporate in groups of four... interesting, if not directly useful! I will be interested to try this with other PCs to see if the result is the same.

Cheers, Jim Moritz
73 de M0BMU