Over the last couple of weeks I have been developing an analysis of possible tracks for MH370. After adjustment for a very small cumulative frequency drift term during the flight (2Hz), the final results predict search areas very close to those published by the authorities.
I have written a paper on the analysis, posted here.
https://www.dropbox.com/s/ez4sxffxyl...14_issue_1.pdf
I have tried to be thorough but at the expense of being long. This was a lot of work – my wife thinks I am barmy!
The main progress on the system analysis that allows this level of analysis was on the nature of the mysterious D3 component. I have worked from the Inmarsat note with the data logs that stated it is related to the AFC system, plus some documents from Miteq that describe the Inmarsat AFC system. The conclusion is that D3 is related to the total Doppler shift on the pilot frequency used in the AFC system at Perth for the C-band downlink. This pilot frequency is generated (at L-band) at the TT+C ground station for Inmarsat 3F1 which is at Fucino, Italy (the so called primary ground station in the Miteq document). Because this is in the Northern hemisphere and Perth the Southern, the Doppler on the combined link is less than the C-band link to Perth alone. The combined Doppler can be calculated precisely. Details are in the paper.
The set of possible tracks used for the analysis is generated with two parameters, a) a speed for the leg from the turn-back point to the 18:29UT ping-ring, b) a different but constant speed for the rest of the legs. This is a parameterisation of the flight path, the speeds of the first part of the flight (before 19:41UT) can be adjusted without changing the conclusions, but the BFO and BTO data do not in themselves require that complication. This is again discussed in the document. I have not used any of the possible radar contacts with MH370 as a constraint on the model.
In general I have been trying to replicate and understand the analysis techniques that I believe the investigating authorities have used. The published work has always emphasised constant speeds, presumably on the basis of some analysis of the possible aircraft modes. I am not trying to second guess what the authorities have done.
I have used the chi-squared statistical test to judge whether models of the track can be excluded on the basis of the data. This is a quantitative test, rather than just picking the best fit. I don’t know what statistical tests the official analysis has used.
Comments are welcome of course.