More trouble with DSLR Flats

[bmhjr]

I have been having a hard time getting flats to work correctly with my DSLR images.  This last batch I was very careful to make sure the Statistics showed a median value around 7500 with scale set to 14 bit.  However, when I ran BPP the Master Flat looks very underexposed with a median value of 0.065 on the Normalized Real[0,1] scale.  I would have thought it should be closer to 0.5.  Needless to say the calibrated lights are over corrected.

Does anyone know what I may be doing wrong?  I have added links to a flat sub and the master flat after calibration.

https://www.dropbox.com/s/ilg7i01bys0elbb/35MM_Tv1125s_1600iso_f4-5_20151223-19h47m44s113ms.CR2?dl=0

https://www.dropbox.com/s/3wss6wyo4ck1jcy/flat-BINNING_1.fit?dl=0

Thanks,
Bill

[bmhjr]

Here is a Flat subframe with No STF applied

[bmhjr]

Here is the Master Flat with No STF applied

[pfile]

what capture program are you using and does it have the ability to show histograms in linear mode? it may be easier to just make sure you are near 1/2 well depth at exposure time...

rob

[bmhjr]

I am using BYEOS.  I first took test frame and looked at back of camera histogram, not using BYEOS, and made sure it was in upper 1/3 near right side.  I then connected BYEOS and checked the histogram in the imaging capture and noted the same.  I then opened the CR2 file in PI and ran statistics to make sure it was near mid range of 14bit 0-16000 median value.

When the CR2 file is opened in PI, there is no stretch needed and the mean value is near mid point. For some reason the Master Flat is different. I just used standard settings in BPP

Should I be using the gray scale Pure Raw CR2 flat subframe to adjust levels instead of the RGB?

[msmythers]

Bill

You opened your CR2 image with the DSLR_Raw setting of de-Bayer RGB. That uses the cameras white balance setting along with debayering the image. The BPP master used the CR2 files as RAW not debayered which is the correct way of using them in BPP. If you change the DSLR_Raw setting to Pure-Raw and open your CR2 flat you will see the image statistics are much more like the master flat.


Mike

[bmhjr]

Got it.  So then do I adjust the exposure on my Flat subframes so that the Median value is close to 0.5 when measuring the RAW CR2 USING [0,1] SCALE?

[msmythers]

I'll let someone else answer that. I don't use a Canon camera and I don't worry as much with my own images since I only shoot with an old 12 bit Sony mirrorless camera. With my own images as long as I get a good quality flat that is not clipping the data the results seem fine but that's just me.


Mike

[bmhjr]

Thanks Mike

[pfile]

i am not sure about this 14-bit scaling mode you're using, but if you use the regular 16-bit scaling (the default) then a completely overexposed flat should be 0.25 everywhere. that means a "well exposed" flat should have a mean of about 0.125.

if you are talking about the 14-bit mode in the histogram tool, this is simply a resolution. it does not change how the histogram is scaled on the X axis. you can see what this control does by setting it to 8-bits and seeing the effect on the plotted histogram.

rob

[bmhjr]

i am not sure about this 14-bit scaling mode you're using, but if you use the regular 16-bit scaling (the default) then a completely overexposed flat should be 0.25 everywhere. that means a "well exposed" flat should have a mean of about 0.125.

if you are talking about the 14-bit mode in the histogram tool, this is simply a resolution. it does not change how the histogram is scaled on the X axis. you can see what this control does by setting it to 8-bits and seeing the effect on the plotted histogram.

rob

Thanks Rob, I will take some more flats and try for the 0.125 mean.

[hcostache]

Hi Bill,

in my experience, PI interprets DSLR raw files as being 16bit integer, no matter what the DAC resolution is. This (16bit) is the only setting that will give you the statistics in terms of the actual numbers coming out of the camera.

Regards,
Horia

[bmhjr]

Hi Bill,

in my experience, PI interprets DSLR raw files as being 16bit integer, no matter what the DAC resolution is. This (16bit) is the only setting that will give you the statistics in terms of the actual numbers coming out of the camera.

Regards,
Horia

Thank you Horia.  I have been working on this all day.  After several sets of flats at different exposures, I managed to get a set that should be exposed correctly.  With the DSLR format setting at Pure RAW, a flat CR2 file shows about 0.13 for a mean value with the Statistics scale set to Normal[0,1].  Ill see how that works out.

Bill

[bmhjr]

After taking more flats which showed a Mean value in Statistics of 0.13, I was able to get a pretty decent result.  The histogram looked over exposed in my capture software so I can't rely on that for judging the proper exposure.  I also used manual Calibration and Integration instead of the BPP process.  Don't know if that made any difference.

Thanks everyone for the tips and info.

Bill

[pfile]

yeah most capture software (including BYE) by default shows you a histogram with a stretch applied. i think this is to mimic the back of the camera histogram display, where the camera firmware has applied a stretch to the raw data. the reason BYE does this is that there's a rule of thumb for DSLRs which says to get the back-of-camera histogram "well detached" from the left side of the histogram display to ensure that your lights are sky-limited (rather than read noise-limited).

a properly exposed flat (~1/2 well depth) looks completely overexposed on the back-of-camera histogram.

rob