Codecov Report

Merging #509 (f2a123a) into develop (1789c8c) will increase coverage by 0.07%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop     #509      +/-   ##
===========================================
+ Coverage    65.33%   65.41%   +0.07%     
===========================================
  Files           12       12              
  Lines         4922     4933      +11     
===========================================
+ Hits          3216     3227      +11     
  Misses        1706     1706              

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1789c8c...f2a123a. Read the comment docs.

@JarronL, not directly related to this PR, but at some point maybe you could help enlighten us on how the wide filter positions are calculated? I don't understand how the wavelengths are weighted for the broadband. It seems to me the W filters are all systematically shifted towards the wider edge of the mask compared to the M filters, and that's not obvious to me why that should be the case. In particular, I'd expect broadband filter averages to be tilted toward shorter wavelengths, because there's more stellar photons at those wavelengths.

In any case we should use the values in the SIAF; it's just for my own understanding I'd like to better follow the logic for how these are derived.

Hmm... this is more a John Stansberry questions, but I'll give it a shot...

According to the NIRCam Coronagraphic Ops document (NIRCam_CoronagraphOps_V4.pdf), the wide band filters are treated a little differently compared to the medium band filters. Medium band filter locations take into account targets with stellar spectral types. But, the wide band "equations must be modified to account for the longer wavelengths they admit relative to those admitted by the F210M and F460M filters. For this purpose, the reference wavelength of interest is defined by the long half-transmission point, lambda_1/2, of the F210M and F460M filters, which are 2.196 and 4.747 um respectively."

For instance, the long-half-power point for F444W is 4.992, so the location was calculated as -3.16 * (4.992 - 4.747) = -0.7742. This is from 2016, though, so things have obviously changed somewhat given the slight differences relative to the pysiaf values.

In short, I guess they wanted to be sure that the longer wavelength contribution from the filters were fully occulted even if that meant the mask position is oversized for the shorter wavelength portions of the filter.

Thanks, I added all of your suggested changes.

The use of a different (longer) effective wavelength for calculating the offsets for the W filters came from John Krist, based on modeling he did.
It would actually be useful to use the current version of WebbPSF to model the contrast as a function of displacement along the bar. John was using a pretty notional OPD map, I think.

Also, looking back at my script for calculating the offsets, the were based on pixel scales of 32 and 65 mas. My guess is that the difference due to that probably can't be measured, but we may as well use the right values. Even better, use the scale appropriate to the location of the bar masks and taking into account the new distortion model for the coronagraph FOV.

Thank you @kammerje! Much appreciated. I'll go ahead and merge this now.

@JAStansberry and @JarronL, we could indeed do some contrast calculations for different MASKLWB offsets as you propose. (I'm about to finally update the OPD maps for as-built latest & greatest, too, so these would be much better than earlier notional OPDs). That's probably better done as a task by Jarron or someone else on the NIRCam IDT rather than any of us in the webbpsf or telescope teams - I'd be happy to advise on how to do the modeling though. It sounds like the choices John Krist made may have been motivated by getting as deep as possible contrast, at a cost of inner working angle. Which may well be the right call but it's something to clarify and convey to users.

Tagging @juliengirard too - FYI see the conversation on this PR about MASKLWB offsets.

Sounds good. @mperrin, if you could make a new PR and tag me as the responsible, I can look into it ::waves hand vaguely:: sometime in the near future. Would like to understand how the most recent filter-dependent bar mask reference positions as defined in SIAF were calculated, though, which probably requires some combination of @mgennaro, @JAStansberry, @juliengirard.

Done - see #511, @JarronL