Sent: Sunday, March 24, 2013 8:49 PM
Subject: Re: LF: DJ8WX 8.97 kHz
Ok Marcus
That would be an interesting 'test' , the s/n
levels seem to be a little varied , depending on
the reference, I know when ros was first coded ,
Jose used fldigi as the comparison ,
but later changed to align with the
path-sim propagation simulator , which then changed the
base line , Op is also aligned with the
same simulator http://www.moetronix.com/ae4jy/pathsim.htm
The levels may not be inside the Opera decode
range , I think -50 was the best over short range
tests , but those tests also showed that the
signal still left a trace on the qrss
plot and its possible , using you correlation
process , detection at lower levels may be possible
as well ?
One thing that did seem to
arise from the early pie fight's ( eating , not
3.142) was testing over a real path and
a static simulation did provide differing results ,
assumedly based on the 'decode profile' , a lot of
tests may be in conflict with design parameters ,
rather than providing an evaluation , say like comparing
slick and treaded tyres in the rain , especially
in free running data mode's , where , for
example , interlacing would be based on qsb profile
(hf) and on MF are found to fail due to much
longer qsb -
I see a follow up from Uwe , may try other
modes in a week or so , if so , may be
the Op monitors using mini-whips for 136
could take a look for his signal as
well for a day/night ? , if there is to be
a test , can post on the user group looking for
monitors, in case any one needs to re arrange things , tnx
.
73 -G.
Sent: Saturday, March 23, 2013 6:14 PM
Subject: Re: LF: DJ8WX 8.97 kHz
Hi Graham,
yes this is an interesting option.
The correlation detector looks for spectral
peaks in a long narrow FFT, which integrates across the total
duration of the Opera sequence (about 2000s for Op 32). For a reliable
detection, a phase-coherent central peak needs to be at about 15
dB SNR in the corresponding bin width (0.5 mHz). Taking a 100% duty cycle
carrier for reference, at 50% duty cycle we have 25% (-6 dB)
energy in the central peak, and another 25% in both AM sidebands together.
Thus for Op-32, the threshold is around
21 dB carrier-SNR in 0.5 mHz, scaling to -46 dB in 2.5 kHz (or -50 "dBOp" on the
Opera scale). The modulation energy will ideally also bring up a 15 dB
correlation peak, which is normally sufficient to ensure unambiguous
callsign identification. Extending the
duration from 32 minutes to ~ 4 hours would gain another 9
dB, arriving at a threshold of -55 dB in 2.5 kHz (-59
dBOp).
On my rather noisy VLF grabber, the carrier from DJ8WX reached
about 9 dB SNR today. Scaling the bandwidth from 1.5*0.48 mHz to 2.5
kHz, we get -56.5 dB SNR (or -60.5 dBOp) - just 1.5 dB bit
short of an Op-4H detection. Of course other online VLF receivers (namely
Paul Nicholson and OK2BVG) do have significantly better receive conditions
so Op-4H correlation should be quite feasible there.
For the standard Opera software decoder, Rik's
measurements had indicated a decoding threshold of -23 dB in 2.5 kHz for
two-minute Opera-2. This scales to -35 dB for Op-32 and -44 dB for
Op-4H (-39 and -48 dBOp respectively). Thus for Op-4H
you'd want to see a carrier SNR of 21.5 dB in a "qrss 6000"
grabber window, which seems a fair challenge at the distances we're interested
in.
For comparison, Stefan's famous LF reception at VK2DDI showed about 15 dB
SNR in an 11 mHz spectrogram, equivalent to -37 dB in 2.5 kHz (-41
dBOp). Thus an "easy target" for Op-32 or even Op-8 correlation, but still
slightly below the edge of the classic Op-32 decoder.
Best 73,
Markus (DF6NM)
Sent: Saturday, March 23, 2013 1:39 PM
Subject: Re: LF: DJ8WX 8.97 kHz
Marcu's / Uwe
This would make a good test for the
Op4H ( 4 hour) mode and the
correlation detector ?
To date there have been no 'dx' test , with
the addition of the correlation ,then
that should give three possible detections ?
OP4H decode , qrss plot and Correlation
( and 50% saving on power hihi)
OP4H has AFC as the shorter modes and is
designed for normal sound card use. with sound-card
audio or CW-comport Tx keying , rx audio is direct in
from sound card , in short range tests -50 db
s/n has been reached , but that is a data decode , would
the correlation , take the level below the qrss
plot ?
73-Graham
G0NBD
|