Re: Fo frequency - Movie of Wendover Coi

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Jim Lux" <jimlux@xxxxxxxxxxxxx>


----- Original Message -----
From: "Tesla list" <tesla@xxxxxxxxxx>
To: <tesla@xxxxxxxxxx>
Sent: Monday, April 18, 2005 6:56 AM
Subject: RE: Fo frequency - Movie of Wendover Coi


> Original poster: "Mccauley, Daniel H" <daniel.h.mccauley@xxxxxxxx>
>
>
> Doesn't anyone else see the problem here????
>
> The sound embedded in this video was probably sampled at 10 or 22kHz.  I
> don't see how you would
> be able to extract any useful frequency information above this (or even
> half of this)  What
> would Nyquist say about that?

Harry Nyquist wouldn't say much today (I think he's dead, or, possibly lving
in the same Venezuelan tepui with Tesla, Marconi, and Elvis)...

But... the question is not the sampling rate, but the filter before the
sampler.   You can sample a signal at substantially lower frequencies, and
the signal just aliases down into the fs/2 region. This is used fairly
commonly these days as "offset IF conversion", where you pick an IF
frequency that is an odd multiple of a quarter of the sampling rate. Check
out the applications notes and literature for a venerable fast A/D like the
AD9038 for more details.  I've built several RF systems that do this.

So, perhaps you sampled at 22kHz, and there's this huge 30-50 kHz signal out
there... it might just alias down and look like it's at a few kHz.
Providing a) it makes it through (or around) the usual low-pass
"anti-aliasing" (or roofing) filter in front of the a/d; and b) that the
aperture time on the A/D is small enough to actually sample it (rather than
acting as a single pole low pass).

Even if the A/D has a huge aperture time and is acting like an integrator,
you can still recover the signal at a higher frequency.  The integrator has
a 6dB/octave roll off, which isn't too severe.  The key would be that you'd
be integrating several cycles, but the relative phase of the integration
window slides over the phase of the high frequency signal.  In reality, most
A/D chips do something in between an ideal integrator and an ideal
instantaneous sampler.


In a video recording situation, they also rely on the fact that there isn't
much acoustic power out there beyond the 22kHz or so (half of the 44.1 kHz),
and that the microphone amplifier probably has another pole or two of roll
off.  Not many things emit narrow band tones at those frequencies (which
would manifest as a whistle or tone when aliased down).  Broadband emitters
(which are probably more common) would just show up as another noise source,
in among the circuit noise, wind, and other enviornmental noises.

So, hypothetically, say it was digitized at 44.1 kHz. Say there was a 2 pole
filter in front of the A/D, with the cutoff at 20 kHz.  Say we're looking
for a 65 kHz signal. 65 is about an octave and a half above 20, so the
filter is down 9 dB/pole or 18 dB.  There's probably another pole in the A/D
itself, but it's going to be higher.. say the break is at 40 kHz.. so, maybe
3 dB?   The microphone (although with the RF fields, it might have just
coupled directly into the amp) probably has another 20 dB of roll off...
All told, you're looking at 40-45 dB of attenuation.

Even if you have a crummy digitizer, you might still see this.. If it were
reasonably stable, and the audio compression algorithms didn't compress it
out (if it's "noise like" they do, if "tone like", the audio compression
actually enhances it.)

If you have a hot stuff 16 bit digitizer... it'd be easy to see.


There's also the ever present possibility of some intermod product being in
band.




>
> Dan
>
>
>  >By taking 4mS chunks of the sound track and letting the FFT
>  >grind through
>  >the numbers looking for higher frequencies, I "might" see the coils Fo
>  >frequency too!!:
>  >
>  >http://hot-streamer.com/temp/WendoverSound-03.gif
>  >
>  >Way down in the noise, there is a consistent 35.8kHz signal
>  >and a 69.8kHz
>  >signal popping up during the streamers.  That could be due to
>  >many other
>  >things and it is kind of a stretch to imagine that those
>  >signal's would
>  >survive through the years and media it was stored on.
>  >
>  >But, we were thinking that the coil was running at about 40kHz and the
>  >35kHz line does sort of agree with that...
>  >
>  >Not sure I would bet a lot of money on this analysis, but it
>  >is interesting!
>  >
>  >Cheers,
>  >
>  >         Terry
>
>