This is going to be a repo for code used to localize the origin of various signals acquired during search for ULB using TLP-25 and other tools. Mostly relevant for flight MH370 search, but involves fairly generic data

So the story is like that (writing from memory): Royal Navy sub has recorded a pinger-like sequence, then Ocean Shield with US Navy specialist TPL contractor was called in by Australian Navy and immediately confirmed the signal and also recorded another signal in the vicinity. And then silence fell. Also there was an airdropped buoy signal and an oil slick, of which no data were released (only 'not jet fuel'). The only tangible data released were 2 videos, of which I extracted sountrack. It so happens that videos are composed of 3 and 2 fragments, presumably separated in time over a few hours, both videos describe single source with 33kHz peak detected on 5-Apr-14. Second detection was made closer to Zenith slope on 8-Apr-14, presumably at 27kHz, but no data exist for that apart from ship position at the time. I eventually got hold of the survey geometry and also of later minisub sonar coverage from ATSB report. Nothing substantial - like actual waveform recordings - was released ever since.

Looking at ping timings on spectrogram, it was obvious that they shift A LOT but apparently can be fitted to an exponent-like curve against time, and this change must be unrelated to Doppler shift (huge velocities implied), so a characteristic of source itself. I hypothesized it is due to the very end of battery life in each of the sources recorded (but was too cheap to actually buy a pinger and test its signature immersed for a month).  Then the question of disappearance of the sources on a return pass makes sense.  Also the survey was mostly done with TPL way too low to capture the refraction cone efficiently- as sound from sea bottom does not radiate everywhere, it only is detectable in the direct cone given the deep sea and high frequency. The timings are then useless for location due to Doppler shift being so small vs power source drift. It is possible to use the reception fact (requiring certain SNR threshold) for amplitude-based site filtering. I built a ray tracer to check the amplitude loss from a sea bottom source to TPL given a realistic ocean acoustic profile. The question was - are there places on the sea floor which fit amplitude constraint and were not covered by the sonar survey. And really, there are such regions for both detections, one being in the deep NW corner and second one to the South of the search area on the Zenith slope. The source separation is however too wide though for black boxes mounted on aircraft next to each other.

The maps and code are here on the Github: https://github.com/kprostyakov/beacon_location/blob/master/out/cvr.png https://github.com/kprostyakov/beacon_location/blob/master/out/fdr.png https://github.com/kprostyakov/beacon_location/blob/master/out/map.png

The IG+Ulich+sk999 work has obvious merits, but it is ultimately flawed because of entrenched assumption of no pilot input after so-called FMT. The pilot was very active before, why would he suddenly stop doing things for hours?  The Inmarsat+aircraft perfomance model is not sufficient alone to define the latitude on arc with any confidence, period. It is educational to view the search from a social dynamics viewpoint, watching media attention and group discourse stabilizing on the Only True Theory, and then the theory fail  (reminding me of living through the collapse of communist state in the 1990s). This last paragraph I suppose is why this text was silently removed from MH370 subreddit by its kind 'pigdead' moderator. Oh drama.