pixinsight 1.6 hanging after failed ImageCalibration (macosx 10.6)

[pfile]

hopefully this screenshot says it all. for some reason my master dark, flat and bias frames as provided by lightbuckets have a different geometry than the lights. this seems to have caused ImageCalibration to hang pixinsight.

i do not know if this is reproducable yet.

[pfile]

unfortunately this was not repeatable, but it's not the first time pixinsight has just hung up on me.

edit: well, it did not hang, but the calibrated frame is garbage - diagonal lines throughout indicating aliasing of the incompatible geometries?

perhaps there is a bounds-checking bug? it actually did not complain that the image geometries were incompatible.

[Juan Conejero]

perhaps there is a bounds-checking bug? it actually did not complain that the image geometries were incompatible.

Yes I can confirm this bug. We've seen it also as a result of incorrect overscan parameters. I still don't know where the problem is exactly but I'll try to fix it for version 1.6.1. Thanks for confirming it!

On a side note, in the screenshot you've included the dark frame optimization routine is giving you a dark scaling factor of 0.006, showing that there's no correlation (in terms of dark current) between your master flat frame and your master dark frame. In other terms, your master dark frame cannot improve your master flat frame.

[pfile]

regarding the master dark... yes. i should make another thread regarding this, but here is the problem: the data is from lightbuckets.com; the master flat, bias and dark are f32 fits files written out by maximDL (i think). when PI reads these files it rescales them from [0..1] and as a result i believe the levels in the calibration masters are way too high. for instance, if i subtract the bias frame from the light, the light is almost completely zeroed out. the raw lights are i16, and as such don't seem to be subject to scaling. i'm not sure how to proceed here.

[Juan Conejero]

Unfortunately there are softwares out there that insist in writing raw CCD data as floating point images. This causes compatibility problems because there is no (standard) way to specify the actual range to which a floating point FITS (also TIFF) image is referred to. We have discussed this before in this forum.

If the floating point images have been written by Maxim, then they are referred to the [0,65535] range. Do the following to import and convert your 32-bit floating point frames to unsigned 16-bit frames, as they should be:

- Open the Format Explorer window.

- Double click on the FITS format item (left column)

- On the FITS Format Preferences dialog, enter the following:

* Lower Range = 0 (as is by default)
* Upper range = 65535.0

- Click OK

- Open the SampleFormatConversion tool and select the 16-bit unsigned integer format.

- Open the ImageContainer tool (Process > ImageContainer) and select all the 32-bit floating point calibration FITS images.

- On ImageContainer, you may want to specify an output directory (recommended) and an output template. For example, this template:

&filename;_16&extension;

will write your files with a _16 suffix to let you know that they have been converted to 16-bit format.

- Drag the blue triangle from ImageContainer to the bottom bar of SampleFormatConversion. Wait until all your selected files are converted and written to disk.

Now you can proceed with 16-bit images in the normal way for calibration.

Remember that you must reset the FITS input range (FITS Format Preferences dialog) to the default [0,1] range to work with floating point images generated by PixInsight. PixInsight always writes floating point images in the [0,1] range, where 0 is black and 1 is white.

[pfile]

thanks. i know that something similar had been discussed before (in fact i had read the thread but not absorbed it). i spent some time last night looking for the thread and could not find it... so thanks for the refresher.

[pfile]

these master calibration files must be messed up somehow.

first of all they are 2 pixels taller and wider than the CCD. not sure where the extra columns/rows were added yet, or why this would happen.

second of all, if i tell the fits reader to always ask about how to scale the file, when i open the bias frame i see that the minimum value in the file is 1388.800048828125. the maximum value in the file is 2120.828613281250. if i map these to [0,65535] the bias frame is way too bright, as you might expect. i mapped it to [1388,65535], and at least when i calibrate the light the amp glow seems to be gone. there are dead columns on the CCD. is it safe to assume that these read out as 0? in that case i guess i should map to [1388.800048828125,65535.0].

i should just ask lightbuckets for their bias, dark and light subs and integrate them myself, because this is kind of silly.