Experimental Log: Roger Bland, November 26, 2001
Tests of Matched Filtering for RAFOS Sources
ifremer on RAFOS
Yesterday I found what seems to
be an acceptable form for the replica to use with RAFOS signals from
the Webb sources in the Eastern Pacific. I will carry out some tests
to show that the correlation process works like I think that it does.
I have set up software to calculate a "replica" for a RAFOS signal
and correlate it with the signal received on the PSM VLA. The signal
is exactly 80 seconds long, and consists of a sine wave starting at
frequency f1 and sweeping up by an amount W over the 80 seconds. [f1 = 259.375
Hz and W =1.54 Hz are probably about right.]
This signal is calculated for the series of time values which
correspond to the recorded time series from the VLA. The calculated
signal is set to zero outside of the 80-second period starting at
time t1, a parameter which can be varied from the
console as the program runs. t1 is also the starting
time for the frequency sweep.
The phase of the signal at time t1 is set to
2 p f1 t1 + f1,
where f1 is a parameter which can
also be adjusted from the console. (In the later use of this replica
the phase at time t1 will probably be set to zero.)
To compare this replica with observation I have proceeded as
If the replica were exactly correct (right frequency, phase and
starting time) and the data were without noise, the result would be
just the square of the replica, thus a sine-squared function with
41500 peaks. Shifting a little off the correct value of t1
the result would be
- Select a strong RAFOS signal. A good one is the 4:30Z broadcast
from SS1 on day 245 of 2001 (September 2). The data series is
available as a series of times, in "Julian Day," and a long-integer
signal array representing the averaging of the signals from the four
hydrophones. (This represents a sort of beam forming.) The data is
read into memory (from file s1245b.dat - see Nov. 25).
- Running the program p11125a.pro), choose a set of parameters (f1, W, t1, and f1) and calculate the RAFOS signal.
- Multiply the signal and replica time series, point by point, to
get a cross-correlated waveform.
- Average the cc waveform over naver points (naver = 100
for the moment).
- Plot the cc waveform, and print the sum of its values over the