.interact() tpf is blank

Problem description

When plotting sector 8 TESS data with ‘interact()’, pixel data does not plot accurately. Potentially due to the change in postage stamp size for sector 8?

Example

import lightkurve
tpf = lightkurve.search_targetpixelfile('L 98-59').download_all()
tpf[2].interact()
# earlier sectors (tpf[0] and tpf[1]) have no issues

Expected behavior

Environment:

• platform (e.g. Linux, OSX, Windows): OSX
• lightkurve version (e.g. 1.0b6): 1.0.0
• installation method (e.g. pip, conda, source): pip

[gully]

Good, glad that the screen stretch resolved the problem!

I suspect all TESS Sector 8 targets will exhibit the same screen stretch default appearance. What’s happening behind the scenes, is that lightkurve computes the 1% and 95% percentile across all finite pixel values in the TargetPixelFile.

vlo, lo, hi, vhi = np.nanpercentile(tpf.flux - pedestal, [0.2, 1, 95, 99.8]) # line 312 of interact.py

Sector 8 is apparently an exception, because 1% of the cadences have extremely low values, and these anomalously-low-values are not automatically masked with quality flags.

We could consider increasing the lo threshold default to, say, 5%. That threshold might be too high for other, more typical TPFs. We’ve found 1% and 95% to be decent threshold defaults for a wide range of Kepler and TESS data.

If you want to over-ride the hardcoded 1% for yourself, you could install the lightkurve developer version alter line 312 of the file interact.py, and relaunch your jupyter-notebook.

[emilygilbert]

Yes, stretching helps! Thanks, that’s a good tip!

I tested this on ~50 sector 8 targets, and they all have the same issue. I know sector 8 had some spacecraft anomalies, so I guess that’s not too surprising. I can send you an example notebook if that’ll help!