Hi Markus, Jim, VLF,
Wow, i'm excited. Now there are results from the 2 character messages.
The early message starts at 17:50 UTC, the late message at 21 UTC.
During the early message, the complete propagation path is not yet
totally in darkness. The D-layer 'comes up' partly so we expect a small
negative frequency offset or constantly dropping phase. This offset
seems to be about -10 uHz.
During the late message, the path is already in daylight on the RX
site, which has the advantage that the QRN from the east is dropping on
the omnidirectional E field RX antenna, which partly compensates the
dropping signal levels.
The 28th August was a noisy day and was not used in the stack. So the
29th and the 30th are used. Both messages, M1 and M2 were transmitted
on these days. When knowing the message i can already tweak the
expected frequency offsets and after just 2 days i can decode M2 !!!!
carrier S/N: 10.72 dB in 108.5 uHz, after just 2 stacked transmissions.
It appears together will some false decodes but is already the
strongest decode, so the Windows decoder would show the message after a
complete run.
>From the phase observations of JNX/HWU in VK7 or NWC here in DL we can
say that the phase changes are public known, at least that there are
negative and positive phase changes to be expected. Thus the
experienced RX station attempting to decode the message would run a set
of frequency offsets hoping for a decode.
So i think i can tell the frequency offset. For M2 it is +75 uHz, i.e.
a much steeper phase change. And this is why the second SNR peak in my
Excel plot did not appear to be so high! When considering the phase,
then the second peak will be even better (certainly depending on the
QRN coming from the west).
Now, guess what my idea is!?!?! Right! What about running a 5 hour long
transmission, applying +10 uHz in the first half and then -75 uHz in
the second half (easily done with SpecLab) ? Just for the fun. I could
modify the script which generates the data for the Excel plot and apply
these frequency offsets to see what the best SNR and start time will
be...
But first i'll continue transmitting as long as the message will be
shown undoubly and without other false decodes. It must fullfill the
rules for a valid decode (only based on informations available to those
who do not know the message).
73, Stefan
Am 29.08.2018 14:28, schrieb Markus Vester:
Hi
Stefan,
only the first half hour of the transmission was used to obtain the
decode, and that was already displaying 5.9 dB carrier EbN0 in
Nuernberg. So all I know is it was more than enough ;-) The carrier
plot in Amberg showed around 26 dB in 1.5 x 424 uHz.
Best 73,
Markus
-----Ursprüngliche Mitteilung-----
Von: DK7FC <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Mi, 29. Aug 2018 14:14
Betreff: Re: VLF: New experiments on 17.47 kHz planned...
PS: Ah,
you decoded the message. That's fine because i didn't know yet
if all worked well. What was the carrier S/N?
73, Stefan
Am 29.08.2018 12:03, schrieb Markus Vester:
Hi
Stefan
and Edgar,
I am still deeply impressed by your groundbreaking work. With close to
12 dB SNR after three nights, it looks like you could chat at a rate of
one character per day.
The Excel plots are a nice way to observe the carrier with different
integration times and optimize time and bandwidth. It looks like there
is another weaker maximum after 21 UT, shortly after sunrise in Hobart
(now 20:41 UT).
For
a
more intuitive interpretation, I would suggest to plot the middle of
the integration time rather than the start time on the x axis, i.e.
offset each curve to the right by half the duration. Then all curves
should more or less peak at the time of a propagation maximum. There
may also be a minor offset in the SNR scale because the noise power is
slightly underestimated due to rolloff of SpecLab's decimation filter,
I think the carrier SNR is really about 0.8 dB lower than indicated on
the y axis.
The NWC phase plot at DL0AO shows a negative phase slope around -45°
per hour, so it may be worth to also plot SNR with a small negative
frequency offset (~ -35 uHz), possibly gaining a bit more at long
integration times.
After decoding your message, I have set up the DL0AO 17.47 kHz instance
to unspread the PSK, so it may be used as a local control of transmit
phase and sequence timing.
Good luck!
Markus
-----Ursprüngliche Mitteilung-----
Von: DK7FC <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Mo, 27. Aug 2018 15:12
Betreff: Re: VLF: New experiments on 17.47 kHz planned...
Hi VLF,
After transmitting carriers for the last 5 days, there are 3 good days
to stack, the others were noisy.
The resulting SNR via start time plot for different carrier length is
attached. With some uncertainty we can say that these days the ideal
start time for the path from DL to VK7 is 17:50 UTC and the message
duration should be between 2.5...3 hours.
This evening i'm going to start an attempt for a 2 character EbNaut
message. With some luck, just 2 days, i.e. 2 stacks will be enough to
get a clear decode.
f = 17470.1000 Hz
Start time: 27.Aug.2018 17:50:00 UTC (daily)
Symbol period: 12 s
Characters: 2
CRC bits: 23
Coding 16K21A
Duration: 02:56 [hh:mm]
Antenna current: 1.2 A
73, Stefan
Am 24.08.2018 14:02, schrieb DK7FC:
Hi Edgar,
Thanks for the spectrogram showing JXN nicely! Here i'm watching NWC
but the SNR is much lower, probably due to the strong QRN we still have
in late August...
I'm going to calculate the weighting factors for the stacking based on
the noise amplitude from 18...21 UTC. And obviously it is not necessary
to transmit until 0 UTC, so i will stop an hour earlier from now on.
Your SpecLab settings are fine as they are. There is no advantage from
a shorter and wider FFT. The observed spectrum width currently allows
to display more than 2000 FFT bins centered on the TX frequency,
allowing to display enough noise arround the peak to demonstrate that
the peak is most unlikely a by-chance peak...
Thanks for last nights txt file. We can produce a first stack from the
two days. Last night is weighted by -1.3 dB. The result is attached. We
can already see a peak of 11.3 dB SNR, not unrealistic but still not
significant enough. The spectrum peak produced from the suggestion of
the Excel plot (Start 17:50 UTC, duration 2 hours) is attached as well.
It shows a higher SNR, which depends on the noise floor calculation (0
dB line). I need to work on this but anyway the peak looks good
relative to the noise already.
Another carrier will follow tonite.
73, Stefan
Am 23.08.2018 23:40, schrieb Edgar:
Hi
Stefan,
Based on the past two days signal from JXN, would it improve the chance
of detecting the carrier if the "FFT window time (length)" matched the
time when the signal is at maximum?
The FFT length is now 8.847 hours. The period when JXN signal is white
is about 1.5 hours.
The FFT bin width would be wider, and a larger percentage of bins would
have a stronger signal.
|
|
