Image integration issue with flats

[WesChilton]

Hello,

I have a set of flats that I shot last week that will not integrate properly... either with or without calibration, the roughly 20k ADU flats integrate to around 6,400 ADU... a massive loss!

A previous set taken a few nights earlier with identical equipment, software and integration settings, integrates correctly. I have had a few other imagers look at the data as well and have the same issue. Strangely none of us can find any issues with the data either through individual inspection or in the FITS headers.

The data is here... I have included the raw flats, plus the bias and superbias FITS.   
https://www.dropbox.com/sh/i2p0ej4int8du7t/AAB1vZaWIGCVKffCy5n6U-WTa?dl=0

I am running the latest version 1.08.04:1195 (X64) on a Windows 10 Dell laptop. The flats were acquired via The Sky X Pro from a QSI 6120 camera (Astrodon, SII filter) and an Alnitak Flip Flat

Thanks!
Wes Chilton

[WesChilton]

I have downloaded PixInsight 1.8.5 and have tried to correctly integrate these same SII flats with the new ImageIntegration tool, using multiplicative and multiplicative with scaling.... the results were the same, instead of a 20K ADU flat, I ended up with a flat that was roughly 6,400 ADU... so if this is a bug, the bug remains.

[pfile]

in imageintegration, the 1st image in the file list is the 'reference' image. if it is weak or otherwise compromised somehow, this can happen. can you check that first image and see what the statisitics are like?

rob

[WesChilton]

So I've run every kind of analysis I can think of on the data and there aren't any differences that are significant enough to cause a 14K ADU loss during integration...

However I did experiment with choosing different reference images... most of the files made no difference, and still combined to around 6,400... but I did manage to find a few out of the group that would combine to average a 20K ADU master integration. But the problem is that I can find no reason why these files would be special... there is no particular ADU threshold they reach, no noise level, nothing.

[pfile]

are the flats all exactly of the same exposure length? or some longer and some shorter? how were the flat subs calibrated?

rob

[WesChilton]

They are all 47 seconds, they were calibrated with a master bias (I also tried a superbias) with ImageCalibration. I did not use a flat-dark.

I also have a previous and following set of data with identical settings, hardware and calibration that integrate without issues.

[pfile]

when you look at these flat frames with the statistics tool, do you have it set to reject outlier pixels (meaning "unclipped" is checked)? i wonder if there are some hot pixels in some of the flats that are not corrected by the bias, and are 'invisible' to the statistics process because it's excluding outliers. when those frames happen to be the reference, the mean values are messed up by the hot pixels and that leads to bad scaling.

maybe unlikely since all the flats should be the same, but the thought crossed my mind.

rob

[msmythers]

rob

Not being a fits person but reading everything that comes across this forum could this be a pedestal issue where the collection software is writing the fits in a range that has not been corrected for PI? I have zero clue about this other then that issue seems to show up every so often.


Mike

[WesChilton]

Hey you guys are welcome to download and look at the data... I linked it in my first posts. The more eyes the better!

Mike, I thought along those same lines, that there might be a pedestal in one or more of the flats, but there definitely isn't one listed in the header, nor can I detect one through any analysis I and a few others (David Ault being one) have done.

Rob, actually, it does seem that there are a few hot pixels (this is a very clean Sony camera, BTW) but those same hot pixels are in all of my flat data, and for some reason this data set is the only one that has the problem. But you might be on to something with that.