Hi Edgar,
yes crosscorrelation would be possible, but I
doubt that there will be a large additional benefit.
One method would be image-based spectrogram cross
correlation. Some time ago I have tried that on earlier DX captures. When
the search was fed with a correctly sized template, it automatically produced a
correlation peak at the right freqiuency and time offset. However when the peak
was strong enough to be statistically significant, at least part of the pattern
was also clearly visible in the first place. The naked eye seems to be pretty
good at detecting simple patterns in the noise.
Such a correlation process is linear and thus
additive, in the sense that a random very bright voxel might contribute as
much energy to the correlation result as a weaker matching
structure. However regarding extra certainty, you would rather have an
"AND" (or multiplicative) connection of individual parts: If the signature
contains three dashes, and each of them alone has a 20% false positive
probability, the total false rate should be more like 0.2^3 =
0.8%.
Another option would be going back to the original
complex waveform and using phase-sensitive correlation. This would have a
significant advantage if the message were spread over a large number of
symbols, which can then be added coherently (like in Op deep search). But with
only three or five dashes, there's not really much to gain. Also
the phase-evolution has to be known a priori or modelled, which is a bit
more difficult for a DFCW signal (integer duration * shift, eg 240s *
0.075Hz = 18 ?), and over an unstable path.
From my point of view, the best contribution
to a "proof" would now be verifying the receiver's frequency calibration
offset. On yesterday's screenshots, the HGA22 line seemed to be just
half a pixel below the tick. Simply by zooming in on it you
could measure the offset much better, even though the line is spread
by some tens of milliHz due to the FSK modulation. Last year
I took some effort to measure HGA's idle frequency in detail against a
calibrated Rubidium source. The result was 135430.022
Hz, with only very little variation of a milliHz or so (see
attached).
Best 73,
Markus (DF6NM)
Sent: Saturday, November 02, 2013 4:02 AM
Subject: Re: LF: Tasmania?
Hi Markus,
Stefan,
Is it possible to search for the dashes using
cross-correlation?
A wave file of the signals received at Orford is
available.
Format is 16 bit, 22050 Hz.
Unfortunately I would
have to travel to Orford to retrieve them due to their size!
But that can
be done if it is worth it.
Regards, Edgar
Moonah,
Tasmania.
