search

ratt-ru / QuartiCal

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

Optionally solve for a term over all antennas. #123

Closed JSKenyon closed 3 years ago

JSKenyon commented 3 years ago

This PR adds a (gain).solve_per option which must be either "antenna" or "array". All solvers now understand this parameter and can produce an appropriate gain term (which is still fully compatible with the rest of the chain) .

JSKenyon commented 3 years ago

There are still some question marks as regards testing - if I do not zero the off-diagonal components of the model, I cannot get precise agreement between the solved and true gains. This is particularly bizarre as the residuals are at the 1e-15 level. This suggests that I may be failing to compare the gains correctly, specifically in the 2x2 case.

JSKenyon commented 3 years ago

There are still some question marks as regards testing - if I do not zero the off-diagonal components of the model, I cannot get precise agreement between the solved and true gains. This is particularly bizarre as the residuals are at the 1e-15 level. This suggests that I may be failing to compare the gains correctly, specifically in the 2x2 case.

I have fixed this and updated the PR.