[time-nuts] Using cheap sound cards for measurements

[Magnus Danielson]

Lux, Jim (337C) wrote:
> The idea of using a standard stereo sound card interface to do Allan deviation measurements has been discussed on the list in the past (i.e. Beat the two signals down to some convenient audio frequency, digitize, and find zero crossings by curve fits to the sampled data).
> 
> Several have commented that one needs good isolation between the channels of the digitizer to get good results, and inexpensive interfaces (e.g. The one that comes on the motherboard) often don't have good isolation.
> 
> Here's a question.. Is that coupling determinstic and "calibrate-out-able"?  Seems that the factors leading to lack of isolation are things like layout, capacitive coupling, shared ground paths, and the like.  If the card is in a constant environment, those shouldn't be changing, so, in theory, one could somehow measure it, and apply the inverse transformation.

The coupling should be fairly consistent, so you should be able to 
measure it and compensate it out. Just recall that a phase shift may 
also be expected.

If one uses a good quality sound card, the crosstalk is fairly low, and 
propper cabling makes sure it stays that way.

Using different but nearby frequencies as respective input frequency 
will make crosstalk measurable but distinct. Then, swap frequencies for 
each input to ensure consistency over the frequency range where 
beatnotes is to be measured.

I forgot to bring my better sound-card from the studio when I was there, 
otherwise I would be doing experiments now... sigh. Should debug that 
5328A instead.

Cheers,
Magnus