Summary
Near-field phasing currently supports only a focal point fixed at scalar cat_lon, cat_lat and cat_dist. I'd like to allow cat_times on near_field catalog entries so that the focus can move during an observation.
Motivation
We want to be able to represent a moving source, like a satellite whose path is known from telemetry, in our data. There's no way to do that in a single UVData object today: generate_phase_center_cat_entry rejects cat_times for anything but ephem, and all other types coerce the coordinates to scalars. You have to split the observation into per-integration objects, phase each one separately, and concatenate.
Proposed approach
Extend the existing near_field type rather than adding a new one. With cat_times absent, behavior is unchanged. With cat_times present, equal-length cat_lon, cat_lat and cat_dist arrays describe a moving focus interpolated onto the integration times, reusing the existing ephemeris interpolator. The near-field geometry itself (w = r1 - r2) doesn't change.
Plan
Two PRs, following CONTRIBUTING's one-feature-per-PR guidance:
- Make the private near-field correction selection-aware, adding
select_mask and update_vis to _apply_near_field_corrections(). No behavior change apart from one bug fix that falls out of it.
- The timed near-field feature on top.
Does extending near_field sound like the direction you'd want before I open the first one?
Summary
Near-field phasing currently supports only a focal point fixed at scalar
cat_lon,cat_latandcat_dist. I'd like to allowcat_timesonnear_fieldcatalog entries so that the focus can move during an observation.Motivation
We want to be able to represent a moving source, like a satellite whose path is known from telemetry, in our data. There's no way to do that in a single
UVDataobject today:generate_phase_center_cat_entryrejectscat_timesfor anything butephem, and all other types coerce the coordinates to scalars. You have to split the observation into per-integration objects, phase each one separately, and concatenate.Proposed approach
Extend the existing
near_fieldtype rather than adding a new one. Withcat_timesabsent, behavior is unchanged. Withcat_timespresent, equal-lengthcat_lon,cat_latandcat_distarrays describe a moving focus interpolated onto the integration times, reusing the existing ephemeris interpolator. The near-field geometry itself (w = r1 - r2) doesn't change.Plan
Two PRs, following CONTRIBUTING's one-feature-per-PR guidance:
select_maskandupdate_visto_apply_near_field_corrections(). No behavior change apart from one bug fix that falls out of it.Does extending
near_fieldsound like the direction you'd want before I open the first one?