Master calibration....16bit/32bit? .fits or .xisf ?

[sharpie78]

As the title suggests I have a quick question regarding master calibration file type....if anyone can help that'd be great  ....

I am capturing using a 12bit camera which outputs 16bit .fits files. When creating a Master Dark/Flat/DarkFlat should I save it as :
 
16bit or 32bit ?

.fits or .xisf ?

Obviously I would then be calibrating 16bit .fits Lights. My logical head says save the master calibration files as 16bit .fits......but that's just what my head says and it makes sense to me. It is what I have done up until now and been happy with the results, however when using BatchPreProcessing recently (probably for only the 4th or 5th time) I actually looked at the master calibration files and noticed they were 32bit .xisf files.
Am I missing something here ? Feel free to either offer an opinion or wade in with a full explanation 
Thanks in advance
Jack

[pfile]

for any integrated image you should use a 32bit sample format. xisf is superior to fits in a lot of ways and is the native filetype for pixinsight, so it makes sense to use xisf.

rob

[Juan Conejero]

Hi Jack,

The FITS format is deprecated in PixInsight. We continue supporting it exclusively for compatibility with existing data, and to share data with other applications. You should always use XISF, which is the only format able to provide all of the required metadata and features to work with PixInsight in optimal conditions.

[sharpie78]

Hi Juan and Rob,
Thanks for your replies. I will make the switch to using 32bit XISF files for calibration files then. 
While the results I have been getting with 16bit fits calibration masters have been good, when I noticed the BPP files were 32bit XISF files I thought it best to question my existing workflow.
Many Thanks
Jack