ratt-ru / QuartiCal

CubiCal, but with greater power.
MIT License
8 stars 4 forks source link

Improvements to delay solver + sundry fixes #156

Closed JSKenyon closed 2 years ago

JSKenyon commented 2 years ago

@o-smirnov @landmanbester This branch adds delay estimates (currently assuming that the solution interval on delay is 0/infinite).

I now also have a generic mechanism for initialising terms, which should make life easier. This also paves the way for being able to load parameters from disk instead of only gains.

I have made a tentative change to the flagging for delays (although I will likely need to propagate it to all phase-only terms). The issue is that the delays often take a few iterations to settle down/start converging, likely due to the slight ambiguity between the delays and offsets. We may need to consider splitting these into separate terms eventually to avoid this sort of problem.

I am still working on a solution to the poor convergence properties we see when the amplitudes are wrong.

Edit: Also, you can now write out the model (summed over direction) to a column.