Indeed... when I make the FFT bins substantially narrower than the
expected drift cycle amplitude, I do in fact see the drift cycles in
the reported "interpolated peak frequency".
However, now that the bins are so narrow, the window duration is
significant in relation to the cycle periods of interest (20 minutes
per window, cool down time is 15 minutes). And so the amplitude of
the drift is very small -- much smaller than might have been expected.
Again, I think this is due to convolving with the window period --
taking the average deviation over the period of the window, and that
would be much smaller than the peak deviation experienced from TCXO
effects.
What seems amazing here is that one can derive a valid soundcard
frequency correction by using the shorter period windows and using
reported interpolated peak frequencies. I guess that is still
reasonable since, when you do make the deviation visible with long
window periods, you want those deviations to be equal-ripple about
the zero deviation line. And using an averaged deviation works just
as well for all averaging periods. There is a trade-off between
accuracy and one's human patience...
Dr. David McClain
Chief Technical Officer
Refined Audiometrics Laboratory
4391 N. Camino Ferreo
Tucson, AZ 85750
email: d...@refined-audiometrics.com
phone: 1.520.390.3995
web: http://refined-audiometrics.com
On Oct 12, 2010, at 22:52, David McClain wrote:
Yes, indeed. I demodulate in AM mode, specifically to remove any
sensitivity to the LO wandering with ambient temperature.
And no I wasn't seeing any variation bigger than 4 ppb with a 0.1
ppm TCXO wander. That's what the quandary was all about.
I think I have answered the question... You cannot get around the
uncertainty principle, which states that your precision in
resolving frequencies is limited by the inverse of your resolution
in time. Attempting some hair-brained "interpolation" across a peak
in the FFT is just a mathematical game without any meaning.
A *proper* interpolation in frequency space is performed by zero-
padding the time record. When you do that, you introduce many inter-
bin sidelobes. But more to the point, when the FFT bin-size is the
same width as the expected drift amplitude, you get a broad,
convolved bin content from the duration of the window, and
attempting to say, on the basis of adjacent bin amplitudes, that
you know where the frequency of *the peak* is to any better than
the bin-width is just nonsense.
So SpectrumLab, while offering a fancy interpolated peak frequency,
must be interpreted with caution. What it reports can be wildly off-
base.
Dr. David McClain
Chief Technical Officer
Refined Audiometrics Laboratory
4391 N. Camino Ferreo
Tucson, AZ 85750
email: d...@refined-audiometrics.com
phone: 1.520.390.3995
web: http://refined-audiometrics.com
On Oct 12, 2010, at 19:47, Mark Spencer wrote:
Just curious, if you have set your radio to AM mode to remove any
variations due
to wonder in the radio LO, how would any minor deviations in the
TCXO of the
radio affect the measurement ? If you are looking at the
difference between a
10.000000 mhz carrier and a 10.000100 mhz sub carrier wouldn't the
difference
always appear to be 100 hz in AM mode even if your TCXO drifted
slightly ?
You might be seeing the affects of varrying doppler shift between
WWV and WWVH
(although I'd expect the variation to be larger.)
Sorry it's been a long day for me and I may be missing something
obvious (:
All the best Mark Spencer
----- Original Message ----
From: David McClain <d...@refined-audiometrics.com>
To: time-nuts@febo.com
Sent: Tue, October 12, 2010 4:48:56 PM
Subject: [time-nuts] Question about SoundCard stability?
Hi,
I have a Flex-3000 receiver, running freely on its internal TCXO
(0.1 ppm). I
have been recording the reported deviations in the measurements of
the 100 Hz
sizeband of 10 MHz WWV all day long. I do this in AM detection
mode, to remove
any variations due to the wander in the radio LO. Recording from
the 48 kHz
audio stream and into SpectrumLab for analysis.
After taking out the measured frequency error in the
"soundcard" (which I
believe is the Flex Radio internal CODEC), of 17.5 mHz, I'm seeing
frequency
deviations of 0.2 mHz RMS, and +/- 1 mHz p-p, with no measurable
long term
drift. The FFT uses a bin size of 11.44 mHz. SpectrumLab
interpolates to the
peak assuming high SNR and a Hann window, with no nearby
interference.
This implies that my "soundcard" is giving me a stability of 0.004
ppm, or
around 10^-9. How can this be? I already know that the TCXO
wanders about by as
much as +/- 1 Hz at 10 MHz due to temperature variations in the
room. (10^-7).
If that TCXO were used to derive the soundcard (CODEC) clock then
I should be
seeing variations of 25x larger. And I can assure you the CODEC is
*not* being
driven by an OCXO or GPSDO of any kind.
Anyone?
Dr. David McClain, N7AIG
Chief Technical Officer
Refined Audiometrics Laboratory
4391 N. Camino Ferreo
Tucson, AZ 85750
email: d...@refined-audiometrics.com
phone: 1.520.390.3995
web: http://refined-audiometrics.com
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/
listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/
listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/
time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
