[time-nuts] Re: Accuracy of a sound card

Alberto di Bene dibene at usa.net
Tue Aug 23 12:28:38 EDT 2005

Ulrich Bangert wrote:

>Because it is a differential measurement between the two channels (which
>are sampled exactly the same time due to the adc hardware) the first
>idea was, that the effects introduced by the not so good sample clock of
>a ordinary sound card may cancel out completely. 
Beware ! The above is true only for high end cards. I learnt it the hard
way. I am the author of SDRadio, a software that implements the concepts
of a software defined radio. The sound card is fed with the quadrature I
and Q components from a QSD (quadrature sampling detector) and the
rejection of the unwanted side band is done by the program using the
phasing method. This requires a precise phase difference of 90 degrees
between the two components. On my and many other PCs the program worked
ok, but I received a few reports saying that the unwanted band rejection
was terrible. After some investigations it turned out that between the
left and the right channel of the sound card there was a fixed time
difference ( a few tens of microseconds, if I recall correctly).
This forced me to implement a fractional unit delay routine, as the
usual phase correction techniques didn't work in this case, being of
course the phase error depending on the frequency.

I don't know the reason, but my suspect is that there is just one ADC,
multiplexed between the left and the right channels, hence the time skew
between them.

73  Alberto  I2PHD

More information about the time-nuts mailing list