JP-2071: Added a boolean switch to turn off calculations for one group ramp #75
Conversation
…oup ramps is turned on.
…ion ramps with only one good groups.
… making changes to the change log.
…ata with group dimension of only 1 when one group ramps are suppressed.
Codecov Report
@@ Coverage Diff @@
## main #75 +/- ##
==========================================
+ Coverage 88.63% 88.70% +0.07%
==========================================
Files 16 16
Lines 2262 2285 +23
==========================================
+ Hits 2005 2027 +22
- Misses 257 258 +1
Continue to review full report at Codecov.
|
…prevent reuse of variable name.
… it more explicit that the ramp_data.groupdq array got updated.
| good_groups = np.zeros(intdq.shape, dtype=int) | ||
|
|
||
| # Compute the number of good groups in each ramp | ||
| good_groups[intdq == 0] = 1 |
There was a problem hiding this comment.
I'm not sure that relying on intdq being exactly zero to indicate a good group is entirely appropriate. It's possible that for some flag values to be present (non-zero) that are to be treated as information only (e.g. a warning), but does not render the data unusable. It should be only cases where a group has the DO_NOT_USE flag set that they should be considered completely bad. This assumes that any group that has either the SATURATED or JUMP flags set also has DO_NOT_USE set. We would need to check that.
There was a problem hiding this comment.
This will require discussion because this is how segments are computed at
stcal/src/stcal/ramp_fitting/ols_fit.py
Line 1703 in 47aec18
The assumption that if a flag is non-zero then the group is not used has always been the assumption for this code.
| ngood_groups = good_groups.sum(axis=0) | ||
|
|
||
| # For each pixel that has only one good group, mark those groups as | ||
| # DO_NOT_USE, making the ramp effectively zero good groups. |
There was a problem hiding this comment.
Is this DO_NOT_USE setting only being applied to some internal data array, for use solely within the slope calculations? We don't want those flags to get set in the original input data to the step (i.e. we don't want to make them permanent).
There was a problem hiding this comment.
The modification to the input DQ flags is not saved. In ramp_fit_step.py the process function runs ramp_fit, then the input ramp model is closed without updating or saving.
|
This PR does not correctly address the feature request. |
This PR resolves spacetelescope/jwst#6013
A boolean switch has been added to STCAL ramp fitting to suppress ramps that have only one good group. These ramps will be treated as having zero good groups. This is done by finding the one good group in the ramp, then setting the DQ flag for that ramp to DO_NOT_USE.
Two tests have been added to test this new feature.