[time-nuts] Sound Card Spectrum Analyzer

Demian Martin demianm_1 at yahoo.com
Fri Feb 19 06:47:16 UTC 2010


While I have little experience looking at phase noise I have a lot with
using commercial sound cards in acoustic analysis. I use a system from this
guy : http://libinst.com/ called Praxis. The designer had to resolve the
same issues re: calibration of the chain. He uses an external box that
generates a square wave of known amplitude and frequency. The spectrum is
known so it's easy to check the system calibration. I would think its
possible to inject the square wave into the freq trim pin of the reference
oscillator for cal purposes. If the varactor diode isn't filtered you could
get response to 100 KHz.

To deal with the dynamic range issue you can equalize the preamp. If you
know the eq you can correct for it in the fft. The same problem exists in a
magnetic phono cartridge (remember those?) and the preamp has a 40 dB
difference in gain between 20 Hz and 20 KHz. This measurement is a little
different since the compensation is different, less gain at low frequencies
and more at high frequencies.

Some of the semi-pro soundcards sample at 192 KHz. You need to check the
response, since to get better s/n numbers some mfr's roll of the top early
and are way down at 90 KHz. Low frequency response is a different issue. The
mfr's don't publish schematics. Look suggests that DC response is possible.
The absolute DC accuracy won't be good since it's not an issue for
soundcards.

I have a lot of experience working with this card:
http://www.esi-audio.com/products/julia/ and can furnish measurements of it.
I also have resources for modifying the analog section to some degree. It
also has single ended and differential inputs to help reduce common mode
noise. Its well supported under Linux, important, since both the Windows and
the Mac sound system will muck around with your bits.
    -Demian







More information about the time-nuts mailing list