|
Hi Stefan and Paul,
thank you for sharing the most interesting NAA
phase plot.
Yes I understand that variations of
ionospheric height will Doppler-spread any spectral line especially
during the dawn and dusk periods, and ultimately the signal might be smeared out
so much that there is no more benefit in higher resolution. But I do not think that a 47 uHz FFT
is beyond the limit. Though the signal is integrated over about 6 hours,
due to the usual windowing the tails are less heavily weighted,
reducing the "effective" duration to about 4
hours.
A perfectly linear phase slope would simply place
the maximum in an adjacent bin. So the trace will wobble a bit, with a
lower frequency while the terminator moves across the path during the
evening, and higher in the morning period. Only the deviation from a linear fit will lead to a loss
of bin magnitude. The effect depends a bit
on the probability distribution of the phase deviation, uniform
distribution gives a sin(x)/x behaviour. If all phases within a 90° interval
(ie. +-45°) appear equally often, the degradation is only 0.9 dB. With
180° you'd still have -3.9 dB left. To kill the signal completely,
you'd have to spread the phase across the whole 360°.
To illustrate this and to coarsely estimate the
phase deviation, I have sketched slant coloured boxes across the slopes in
Paul's graph. The slopes have a duration of about 6 hours (morning,
+60 uHz) and 4 h (evening, -40 uHz), and the vertical width of the
boxes is on the order of 90°. (It looks as
if the sign of the phase was opposite to the usual convention, as the
sunlit path (12 to 18 UT) should be shorter, with phase advancing
towards more positive (up) rather than negative values).
Assuming that NAA is perfectly GPS synchronized,
the systematic shift of the daytime phase is astounding. A possible simple
explanation is that days are becoming longer during the spring season,
and midday solar elevation is increasing from day to day.
With a given absolute path length
variation, phase deviation should scale more or less proportional to
frequency, so the adverse effect on 9 kHz should be 2.5 times less than on 24
kHz. In essence, I expect only little phase
spreading loss across an intercontinental path at 9 kHz and 47 uHz
resolution.
On the other hand, a very much longer integration
across day and night (eg. 12 uHz) would not seem to make much sense, not
only due to the phase effects but also due to the
variations in signal level and background noise.
Best 73,
MArkus (DF6NM)
Sent: Thursday, March 24, 2011 7:01 PM
Subject: Re: LF: Wasilla Alaska VLF - offline
Hi Laurence, Markus,
VLF,
Hmmmm, about the 60000 windows, Paul Nicholson means.... (see a part
of our emails below)
Am 19.03.2011 09:41, schrieb Paul:
Hi
Stefan,
> Well, you see people are running slower and slower
windows,
> such as "DFCW-60000"
There is a limit set by the
variability of the path length.
For example, see
http://abelian.org/vlf/tmp/110319a.gif
This
shows the absolute phase of NAA at 24kHz over a great
circle distance of
4672km. The graph spans 5 and a bit days.
During the day the
phase advances because the D-layer height
falls from about 90km to
70km. The pattern repeats each day
with the night-time phase
(path length) pretty much the same
each night, but the daytime path
varies.
The path change between day and night is almost a complete
cycle at 24kHz, so we might expect 100 to 120 degrees at 9kHz.
The
value of a long coherent integration will be significantly
reduced if the
phase changes by more than 30 or 40 deg.
If the phase changes by more than
90 deg, a long integration
will be worse than a shorter
one.
Maybe "60000" is better anyway, as seen between DJ8WX or
OE5ODL and TF3HZ. Anyway i would start with a faster mode and go for a slower,
if nothing is visible. It's your dicision, Laurence :-)
73,
Stefan/DK7FC
Am 24.03.2011 11:40, schrieb Markus Vester:
Hi Laurence,
if I had only one to choose from, I'd definitely opt for the
slowest "60000" window at 47 uHz. Depending on antenna orientation, this would
definitely give Scott a chance, or perhaps one of us over
here. Would an onnidirectional E-field work as well for you as the RX
loops?
Running the second FFT for the Alphas nearly doubles the CPU
load, so you may want to skip that.
Thanks for the great work!
Best 73,
Markus
|
Paul_NAA_phasefit.jpg
Description: JPEG image
|
