[time-nuts] Sound Card Spectrum Analyzer

[John Miles]

>
> If the noise is "known flat" it's a good way to check system
> bandwidth. Some means of checking response is indeed very necessary.
>
> Not switching the preamp is indeed a good thing. The sound card
> does not have the range of the 3561, so with the sound card the
> beat note absolutely require a switch. The switch adds a second
> calibration step at audio.

Maybe not -- the 3561A has nowhere near the dynamic range of the sound card
(13 bits versus 20+, or about 80 dB versus ~117 dB for a good sound card).

Normally you do beatnote calibration by injecting a signal at (level of
DUT - gain of LNA), or 40 dB down for the 11729C.  Say your maximum PN level
of interest is -50 dBc/Hz at 1 Hz.  If you adjust the level going into the
sound card to position the full, HPF'ed test signal near the sound card's
rails, you'll have a theoretical floor near -170 dBc/Hz.  Your signal
generator might be set for -60 dB relative to the DUT input amplitude in
this scenario, yielding a clean calibration spur without switching gains
anywhere.

The signal generator used to inject the beatnote for calibration doesn't
need to be anything too fancy, just something with a good attenuator.

> Many of the free applications that are out there will put a tone
> out of the card and track it back into the card. That should at
> least provide a tone to work with. I also should be something
> that can be fairly easily verified. The issue of mixer output
> impedance is still a little tricky without RF noise loading.

It does need to be an RF tone, though, to characterize the whole signal path
(and even then you'd really like to do it with broadband noise if it were
practical, which it's not).  If you haven't looked over HP note 11729B-1 (
http://www.ke5fx.com/gpib/5952-8286E.pdf ) it will give you a good refresher
on what you're up against.

-- john, KE5FX