ratt-ru / QuartiCal

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

Add tied delay (unislope) option to delay solvers #155

Open o-smirnov opened 2 years ago

o-smirnov commented 2 years ago

We need an option to solve for a single delay term on both diagonals (with separate phase offsets though), a-la the "unislope" solver of CubiCal. Needed by @phetogoportia.

JSKenyon commented 2 years ago

Before I add this (which I am happy to do), I would be interested if the absorption you mentioned when discussing this affects QuartiCal (as it treats the phase-only problem slightly differently).

JSKenyon commented 1 year ago

As a reminder to myself but also a more general observation, we probably want the ability to make any diagonal gain term scalar i.e. the same in both diagonal correlations. This can probably be accomplished in the solvers by a summing J^H J and J^H r over the correlation axis prior to computing the update, in a similar fashion to the modifications made for per-array terms.