This page archives Slack comments from the splinter session on coronagraphy of the Improving JWST Data Products Workshop (IJDPW).


Julien Girard

Here is a (sorry, long) summary for Coronagraphy, @Ell Bogat, Jonathan @Marshall Perrin, others, please feel free to add ideas or omissions.
We started to discuss it in the ERS-extended (let us know if you want to join) Slack space specific channel with Jarron Leisenring, Aarynn Carter, Jordan Stone, Jens Kammerer etc. (some of the spaceKLIP main developers who could not attend the workshop).Coronagraphy: status

  • NIRCam has 5 masks/occulters (3 rounds, 2 bars) seen by both SW and LW channels simultaneously (since cycle 2)
  • MIRI has 4 masks (3 FQPM at 10, 11 and 15µm and 1 large Lyot maks)
  • jwst Pipeline stream is Detector1 => Image2 => Coron3, in general does a good job (a good quicklook for the presence of point sources or eventual , less so for the bar masks than the round mask.
  • spaceKLIP community package (ERS/GTO/STScI/IDTs etc.) wraps around the official pipeline step 1 & 2, substitutes Coron3 and offers intertwined high level products for publications: raw and calibrated contrasts (corrected for algorithm throughput), MCMC-type (forward modeling) astrometry and photometry. Anyone is welcome on-board, there is a Slack channel and a github.


Detector1 items by priority

  • dark subtraction: currently much better to skip subarray darks for NIRCam (on-sky darks are time consuming and not high enough SNR), higher SNR darks are “coming” for MIRI (TBC).
  • 1/f noise: not a huge problem (see blink between ERS reduced image from Aug 2022 and the current spaceKLIP). When 2 rolls are combined, the 1/f is less visible). The approach recommended in the #topic_1_over_f is the one implemented in spaceKLIP, code here
  • outlier detection to be improved with best parameters from spaceKLIP (cannot be the same for all modes).
  • specific MIRI subarrays 390 Hz EMI noise (on going)


Image2 items by priority

  • Ongoing for NIRCam: Improve distortion correction (plate scale in the vicinity of the masks) and global alignment (correct WCS and True North)
  • Future for MIRI: Global alignment and plate scale in coordination with NIRCam parallel (TBD)


Coron3 items by priority

  • Mark the central object and mask positions on i2d “quick look” image (or in the header).
  • KLIP: probably change the default “troncation” to 10 components (instead of 50)
  • Frame alignment should be robust for all masks (recent findings with spaceKLIP for the bar masks) with current algorithm
  • smarter masking is necessary
  • better bad pixel / remaining outlier cleaning (inspired from spaceKLIP)
  • Fourier shifting method introduces strong odd/even ripples for any bad pixel/outlier that remains in the image. It’s especially an issue for bar masks and LW detectors (cosmetics is worse).
  • Mid-term: Make sure products are ok for all modes and combinations of mask/filter (especially with dual SW/LW for NIRCam)
  • Longer term:
  • Produce contrast products (in FITS extension? TBD)
  • Output specific astrometric systematics errors (depending on the quality of the guiding, the global alignment, TA metrics)


spaceKLIP (a good platform to discuss test, prototype ideas)

  • Currently spaceKLIP does not take advantage of the distortion solution (which is under revision/improvement with jhat)
  • For astrometry large separation (> 1") companions with NIRCam, the Coron3 i2d product is better/ a good alternative.
  • In the future spaceKLIP should take use distortion corrected image AND/OR use the solution (correct plate scale) in its analysis step


(edited)

GitHubGitHub
GitHub - kammerje/spaceKLIP: Pipeline for reducing JWST high-contrast imaging data. Published in Kammerer et al. 2022 and Carter et al. 2022.
Pipeline for reducing JWST high-contrast imaging data. Published in Kammerer et al. 2022 and Carter et al. 2022. - GitHub - kammerje/spaceKLIP: Pipeline for reducing JWST high-contrast imaging data...

Marshall Perrin

I didn’t get a chance to weigh in on this earlier, apologies (on west coast time zone this week).
Briefly I think the priority for pipeline work should be the improvements in Detector1.  A very useful incremental step would be if the MAST _rate/_rateints files were sufficiently usable, and this is not currently the case. The major neds for that are

  • Disabling the subarray dark
  • Much improved outlier/bad pixel detection. Coronagraphy requires this be nearly flawless, so we’ve put a lot of effort into it (including a lot of manual masking at times, which is not straightforwardly adapted into an automated solution. This is a tough nut to crack given the particularly stringent needs for this mode.
  • The use of edge pixels as pseudo ref-pix
  • Correlated noise removal (1/f for NIRCam, 390 Hz for MIRI).


I think focusing on the stage1 improvements only as the first priority would be an efficient prioritization to start with. Not saying the other items aren’t also important. But rather, every subsequent step depends on the detector1 outputs being sufficiently good, so getting the pipeline detector1 tuned up sufficiently for coronagraphy would be a necessary precondition before doing any enhancements and later stages.

Julien Girard

I agree, first things first (and I forgot to include the use of pseudo ref-pix for nircam subarrays) in the above summary.


  • No labels