EDLB
GarageYears,
if you know how the signal looks like you searching for, you get much more signal to noise than with only digital reducing the ambient noise. However the current frequency tolerance and ping time interval tolerance are too sloopy to use this technique effective. My point is, that this can be improved very easily without adding much cost, weigt and volume to the pinger. Think about it as you know the exact conversation you where looking for in your example beforehand.
That technique is the only reason that GPS works with relativ small receive antennas and low transmitt power in the satellites.
One way to think of this is with a visual analogy:
Imagine of a large number of vertical lines on a sheet of paper, of random width colors and spacing (noise).
Then add precisely spaced 0.1mm wide red lines every 10mm.
Take a second sheet of black paper and cut 0.1mm slots every 10mm.
Now slowly move this sheet over the top of the first sheet and plot the
spectrum (color) of the reflected light.
As the slots move over the precisily spaced red lines there will be a pronounced shift of the reflected spectrum from all colors (white/noise)
to red.
Assume a detector with very narrow optical filter that filters out all but the frequency (red) of interest and one can easily find (if they exist) the
red lines.
Once they are found other information can be deduced from where the slotted mask was at the peak.
This is a (very) simplified description of part of the correlation techniques used by GPS to extract signals from noise.
Note that if the lines are not precisily spaced the slots will need to be wider and the "gain" will be reduced.
Also if the color of the lines is variable/iunknown (doppler shift) the "red" filter will need to be wider so signal to noise will again be reduced.
This is one reason that "cold start" of GPS systems can be much longer than with a warm start where known approcimate location, time and up to current satelite data are avaialable.
In a warm start the system is able to predict where (in time/frequency) each satelites should be.
EDLBs point is that if precision oscilators were used the frequency and spacing of the pings would be tightly controlled so the slot and filter could be much narrower allowing much higher gain.
(As I recall there is also additional gain if the frequency and ping rate are correlated)
A bit more precsion could be obtained if the frequency/temperature plot of the oscilator was well known and optimised for likely deep sea temperatures.
In case of deepwater one value easily determined is the likley temperature of the pinger.
BTW: Scanning too wide a range of slot spacings and colors can "find"
signals in random noise unless a very long sample is available, that could be what happened with the re-analysis of the sub data.