Problem with images calibration

[Julien]

Hi,

I try to process an image with PixInsight for the first time. I followed this tutorial to generate the Master Bias, Master Dark and Master Flat : http://pixinsight.com/tutorials/master-frames/en.html .

Now I try to calibrate my images files. Seems that there is something wrong with my master flat. After calibration my images are darker in the center and lighter in the corners (it's a little like an inverted flat).

I tried to reprocess from the beginning but I get exactly the same result.

I also tried to process with another software (MaximDL) and I don't have this problem with it.


I put some of my files online :

- a Bias file : http://www.elendil.cc/divers/pixinsight/BIAS.fit
- a Dark file : http://www.elendil.cc/divers/pixinsight/DARK.fit
- a Flat file : http://www.elendil.cc/divers/pixinsight/FLAT.fit
- an image file : http://www.elendil.cc/divers/pixinsight/NGC_1499.fit

- the Master Bias : http://www.elendil.cc/divers/pixinsight/MasterBIAS.fit
- the Master Dark : http://www.elendil.cc/divers/pixinsight/MasterDARK.fit
- the Master Flat : http://www.elendil.cc/divers/pixinsight/MasterFLAT.fit

- an image file calibrated with the masters files : http://www.elendil.cc/divers/pixinsight/NGC_1499_c.fit


Can someone tell me if I have done something wrong ?


Best regards

Julien

[oldwexi]

Hi Julien!

Your single flat, bias and light in my opinion are not correct saved, they start from 32768 to 65535
The should start  at 0 !! This usually happens when the fits file is saved in signed format.
The histogramm of the single flat is therefore to far to the right.
Producing a Masterflat and Masterbias with this data does not achieve what you want.

Your single Bias also starts from 32768 to 65535.
Load your dark,flat, bias and light,  check the histogramms and you will see the bias is much much
brighter than the dark.
Its the same with the light it starts at 32768 too.

What to do?
Use PixelMath and subtract from the Bias 0.5, do the the same with the light and the flat,
and check the histograms afterward.

This is what i did when Maxim delivered mixed signed and unsigned Data.

Aloha
Gerald

[Julien]

Hello

Can you tell me where I can find these informations in PI ? In MaximDL the FITS Header tells that the data are unsigned. And whan I use the Information dialog box I have values around 250 for the bias, a little bit higher for the darks and between 35000 and 45000 for the flats.


Julien

[Juan Conejero]

Hi Julien,

The problem is with your bias and light frames. They are 16-bit signed integer images storing physical data. By default, PixInsight interprets signed integer images as representing logical data in the range from black=-32768 to white=+32767. This is done to ensure usage of the whole 16-bit numeric range to store pixel values.

These files (your bias and light frames), however, are only using the positive half of the 16-bit range, from black=0 to white=+32767, to store raw CCD data. The negative half from -1 to -32768 is simply not used, which means that the effective bit depth of these images is 15 bits. In our opinion, this is a conceptual error because raw CCD data are unsigned 16-bit integers in the range from 0 to 65535 (and in the case of a light frame, you are probably losing data by storing them with just 15 bits).

To load these files correctly in PixInsight, write the following in the input hints field of ImageIntegration (Format Hints section) when you create your master bias frame:

signed-is-physical

The same applies to the ImageCalibration tool, when you calibrate your light frames. With the above format hint, the FITS support module will know how to interpret these signed integer images.

I strongly recommend you change the necessary settings in your image acquisition software in order to generate exclusively unsigned 16-bit integer images to store your raw data. This is the only correct way to store them.

[Julien]

signed-is-physical

Ok, that works.

The problem is with your bias and light frames. They are 16-bit signed integer images storing physical data.

[...]

I strongly recommend you change the necessary settings in your image acquisition software in order to generate exclusively unsigned 16-bit integer images to store your raw data. This is the only correct way to store them.

In MaximDL I have an option "Save Data in Unsigned Format" which is enabled.

When I look at my files' FITS Header I notice that the BZERO values are not the same. for the Bias and light frame I have 0. But for the darks and flat I have 32768. So, If I have correctly understood, I should also have 32768 for my bias and light frames ?
What is strange is that if I have the "Save Data in Unsigned Format" I activated the value 0 is recorded. But by disabling the option, I have the value 32768. I should therefore disable this option to have unsigned values???


Best regards

Julien