Data search via MAST no longer returns correct lightcurves.

When searching a light curve file using an identifier (e.g. ‘16CygB’, or a KIC number), the list of lighcurves returned are from a range of different targets, instead of the requested target.

import lightkurve
data = lk.search_lightcurvefile('16CygB').download_all()
>>> data
LightCurveFileCollection of 26 objects:
	KIC 100002399 (1 KeplerLightCurveFiles) Quarters: 6
	KIC 100002401 (1 KeplerLightCurveFiles) Quarters: 6
	KIC 100002403 (1 KeplerLightCurveFiles) Quarters: 6
	KIC 100002741 (3 KeplerLightCurveFiles) Quarters: 7,11,15
	KIC 100002743 (1 KeplerLightCurveFiles) Quarters: 7
	KIC 100002745 (1 KeplerLightCurveFiles) Quarters: 7
	KIC 100003083 (3 KeplerLightCurveFiles) Quarters: 8,12,16
	KIC 100003085 (1 KeplerLightCurveFiles) Quarters: 8
	KIC 100003087 (1 KeplerLightCurveFiles) Quarters: 8
	KIC 100003502 (1 KeplerLightCurveFiles) Quarters: 9
	KIC 100003505 (1 KeplerLightCurveFiles) Quarters: 9
	KIC 100003508 (1 KeplerLightCurveFiles) Quarters: 9
	KIC 100003551 (2 KeplerLightCurveFiles) Quarters: 10,14
	KIC 100003554 (1 KeplerLightCurveFiles) Quarters: 10
	KIC 100003557 (1 KeplerLightCurveFiles) Quarters: 10
	KIC 100004031 (2 KeplerLightCurveFiles) Quarters: 13,17
	KIC 12069424 (4 KeplerLightCurveFiles) Quarters: 0,1,2,5

For reference, the correct KIC number for 16 Cyg B is: 12069449, which doesn’t show up in the search at all.

I’m using lightkurve version 1.1.1.

Any ideas? @danhey has noticed this too and we think its a MAST issue, not lightkurve, but it’s important to flag it here as it will break any automated algorithm.

[barentsen]

A consequence of the different target ID’s is that stitch refuses to work:

…but the following hack works around the restriction:

[barentsen]

It looks like searching for 16 Cyg B data is confusing because the target was observed under a range of different target mask identifiers (e.g. 100002741 in Q11 & Q15, 100004031 in Q13 & Q17). This will be explained by the fact that the star is extremely bright (V~6). Kepler’s standard method of allocating masks was not designed for such bright objects, so I believe that custom masks were manually designed to observe this target (i.e. it is a very special case). The person who designed the masks and will know the full details is [e-mail] .

A side-effect of using custom masks is that they are observed under awkward target mask identifiers (1000xxxxx), rather than the standard KIC identifier. Moreover, the target mask identifiers will change between Quarters for custom masks.

@ojhall94 Does that make sense? Were you able to piece the data for 16 Cyg B together? Note that the object was observed in short cadence, but search_lightcurvefile will only return long cadence by default.