Before the delay calibration start it does a zero-smoothness solve which is used for a phasediff score and to set the actual smoothness constraint for the delay calibration.
I'm thinking that we should perhaps add these plots (and maybe the h5parm as well) as a diagnostic output to the delay-calibration workflow, as a zero-smoothness solve can be rather useful to spot issues in the data. These can be a useful starting point in cases where the delay calibration is sub-optimal, which still happens regularly, without then first having to manually run a solve again.
Before the delay calibration start it does a zero-smoothness solve which is used for a phasediff score and to set the actual smoothness constraint for the delay calibration.
I'm thinking that we should perhaps add these plots (and maybe the h5parm as well) as a diagnostic output to the
delay-calibrationworkflow, as a zero-smoothness solve can be rather useful to spot issues in the data. These can be a useful starting point in cases where the delay calibration is sub-optimal, which still happens regularly, without then first having to manually run a solve again.