On Thu, 11 Apr 2019, DK7FC wrote:
That's great! Amazing! 976 km distance in the ULF range!
A decode at 15.63 dB carrier S/N.
Improved parameters:
sq5bpf@jitter:/stuff2/sq5bpf/vlf_stefan2$ vtcat -p < stefan_2970_1 | vtfilter -a
th=4 -h bp,f=2970,w=2300 > pliktmp7
sq5bpf@jitter:/stuff2/sq5bpf/vlf_stefan2$ vtblank -v -d0.0007 -a22 -t 70 <
pliktmp7 | vtmult -f 2970.1 | vtresample -r 240 | vtresample -r 10 |vtraw -oa
| ebnaut -dp16K21A -r10 -c2 -v -k16 -S6 -N7 -PS -L 50000
vtblank: selected channel: 1 = -:1
vtblank: channels: 1, sample_rate: 48000
vtblank: hfactor: 0.000e+00
vtblank: afactor: 22.000
initial reference phase 88.1 amplitude 8.289e-04
phase 0 0 0 0 0
vtblank: end of input
vtblank: dropsum 0 50041841, nfp 359999999
vtblank: dropfactor 0 1.390e-01
phase 1 180 180 180 180
found rank 6348 ber 4.2147e-01 Eb/N0 -2.0 M -2.316530794e-02 ph 1
180,180,180,180 [DK7FC/P]
carrier phase: -10.7 deg
carrier Eb/N0: -0.3 dB
carrier Es/N0: -15.26 dB
carrier S/N: 15.70 dB in 133.5 uHz, -23.04 dB in 1Hz, -57.02 dB in 2.5kHz
elapsed 98
phase 2 30 30 30 30
I'm still trying to find better decoding parameters. For now i'm trying
a bandpass filter with varying widths, and trying to find the best
noiseblanker parameters for each of these filters. A 2300Hz filter
seems best for now. I'm trying different bandwidths in 100Hz increments,
started at 1800Hz, currently i'm trying 2500Hz and 2600Hz, about 840
different vtblank parameters are tried per each bandwidth. Later i will
repeat the tests with different th parameter, preset notches in the
passband etc.
Unfortunately the search is done on two workstations with i7-7700 (8
cores) at 4GHz (this is due to intel turbo-boost, these are 3.6GHz cpus).
I will try to set up a proper htcondor cluster later with more machines.
Can you describe your RX antenna and location/position of it?
This is an e-field antenna: BF862 amplifier, the e-probe is a big can
(probably around 20cm height, 13cm diameter). The antenna is located on
the roof of a 4-story building in Warsaw. Care was taken to use the best
engineering practices (zip-ties, duct tape etc).
The antenna location is bad, but i've tried not to add more noise than it
already has by eliminating ground loops, isolating from noisy power
supplies etc.
Thank you for your work! I will try to use your parameters and then we see
the digital loss of the vorbis path.
It is a good set of data to 'play' with, because this distance/frequency/mode
has never been done by radio amateurs before.
I can make the raw data avaliable in case anyone is interested.
A first detection/decode DL > SP.
Eventually i'll do more experiments this weekend. Let's hope for a suitable
wx. The last days have not been really quiet ;-)
And BTW, when is your next transmission together with Marcin?
There is no time set yet. I need to finish my tx first.
Another thing: Sometimes i git better results when using vtfilter -h lp... -h
hp... instead of vtfilter -h bp... because the edges are a bit smoother. Also
it can improve the SNR to place the center frequency of the filter not to the
TX frequ ency but some 100 Hz up or down, depending on the noise
distribution.
I've tried the hp/lp combination but usually it wasn't better.
What works is to look at the spectrogram and notch out the noise peaks in
the filter passband.
PS: I remember you asked a few times if i can go lower than 5170 Hz. Now we
are there :-)
Yes! But don't stop until you get to DC :)
VY 73
Jacek / SQ5BPF
|