Problems in newly created Dark Masters

[jkmorse]

Juan et al.,

Up until 2 years ago I had issues with the way MaximDL writes FITS files.  The problem was that it often truncated the image to only 32,000 bits instead of the normal 65,000; thereby trashing all of the higher end signal.  By working out the exact settings in MaximDL to avoid the problem, it went away. 

But lately it’s made a comeback of sorts.  RGB images are fine, so far, I am guessing because they have stars that fall into the upper range even before stretching.  But my narrowband images are a mess.  I can do a work around since the issue materializes when I calibrate my subs so if I skip calibration I can still use the files, but that is not ideal.

I have tracked the problem to some newly created master darks.  Something in the way PI is writing the master darks under 1.8.5 (at least since mid-January when I took the darksubs and made the masterdark) is reintroducing the problem.  Attached is a file including everything you should need to test the problem, including new and old master darks, other calibration frames, and new darksubs, as well as a set of 18 OIII subs:

https://www.dropbox.com/sh/q1bum458ba460tk/AABAlbPbOh03cRQU2xhWThJfa?dl=0

I am using the latest version of PI for Windows and an up to date Windows 10 computer.  Let me know if you need anything else.

Thanks,

Jim

[Juan Conejero]

Hi Jim,

Nothing has changed in PixInsight (publicly) for many months, so if there is a problem with the way these files are being generated, it shouldn't be our fault! On the other way, from a pure data loading/decoding perspective, there are no differences between raw broadband and narrowband frames: both are loaded and decoded as two-dimensional arrays of unsigned 16-bit integers.

I'll try to download and check out these files, but I'd guess the problem is in the original FITS files and their corresponding acquisition settings.

[jkmorse]

Thanks Juan.  Its just strange that my 2016 Dark works fine but the new 2018 dark is a bust.