Re: sound card calibration?
OK, found it.
The discrepancy was from my uncareful attempt to measure the MCLK
reference oscillator. I was using a shortwave receiver and a crude
sig gen and WWV carrier as the reference. Turns out I found a
receiver spur. The sound card generates a number of spurs around 12.
By ear, the one that may be the biggest does correlate with the
frequency error measured with the WWV 500 and 600 Hz tones. I'd have
to get out an o'scope or use (another) sound card as a spectrum
analyzer to sort it out any better. Good enough for sorting out the
measurement discrepancy issue, though.
So, in conclusion, fldigi has a method built in that uses the WWV
audio content (the 1 second pulses). Or, you can capture WWV audio
and use spectral analysis to figure out the sound card sampling error
to whatever degree you'd like. If you need a recipe to do the
spectral analysis, let me know.
Best regards,
Drew
kb9fko
On 5/24/16, Drew Arnett <arnett.drew@gmail.com> wrote:
> OK, looked into it a bit more.
>
> The online documentation for fldigi for calibration does work. And
> partially answers the question of how good does it need to be?
> Answer: two different sound cards I have here are probably good
> enough. Slick, too the wwv transmit feature to share your calibration
> with your friends.
>
> So, in fldigi, using an AM receiver, first turn on the docked scope
> with view->waterfall->docked scope. Then set the opmode to wwv. [Not
> wwv freq scan. Not wwv freq analysis. Need docs to explain those;
> haven't chased down, yet.] Should start to get a skinny stripe in the
> scope. If it's not near the center, fool with the RX correction to
> move it there. Then right click once in the scope to zoom in so it
> looks like the pictures in the online docs.
>
> http://www.w1hkj.com/FldigiHelp-3.21/html/digiscope_display_wwv_mode.html
>
> Both my cards look fine per his notes.
>
> I repeated my measurements. 12 MHz xtal seems to be -2 ppm. Tones
> measured with wwv and spectral analysis are about -8 ppm, meaning the
> sampling frequency is +8ppm. If anyone can explain the discrepancy, I
> would like to know. However, 2 or 8 ppm seems to be fairly decent.
>
> I'd be curious to know the accuracy requirements for the sampling
> frequency, and it's dependency on mode and or software.
>
> I would also love to get my hands on an example of a soundcard with
> poor frequency analysis to repeat this set of experiments.
>
> Thanks for the interesting measurement problem.
>
> Best regards,
>
> Drew
> kb9fko
>
Reply to: