DSLR. Help! What am I doing wrong

[ChrisV]

While my asi071 is being upgraded I've gone back to using my dslr (550D modded). I'm using APT to collect the data - ISO400 120sec subs - and saving as raw CR2 files. I've used APT with the 071 before but not my dslr.

When I looks at the FITS files (in fitswork) my dark and bias frames have negative values - down to -300 ADUs. The lights all have positive values range from ~300 up to 63,100. I didn't think I could get negative values?

I've put it into PI and everything comes out between 0-1 (good). But when I calibrate the lights and start to process they all end up being clipped at the black end (bad). The shots look OK, but as of the clipping I can't stretch and reduce the background.

What the heck am I doing wrong???? Have I got some sort of offset problem?
Hope this description makes sense.

[glend]

Chris, when you used the Canon in the past, whar a did you use for capture? How are you shooting the Darks and Bias frames and what does the histogram look like? Negative values seem impossible for Darks and Bias, but they will display a histogram slammed up against the left hand (0) margin. You mention FITS but why? If your capturing Darks and Bias with the DSLR they should be RAW files. Can you pull them up in BYEOS and check them?

[ChrisV]

Sorry I meant when I look at the raw CR2 files (not FITS files) there are negative values (in fitswork).

I used to get DSLR images with EOS Utilities. Have just tried that again and I see the same thing. Attached is a screenshot of a bias shot with EOS Utilities, and viewed in fitswork with histogram to show negative values - and they clip at about -300.

When I put these files into PI I think they are getting clipped as there's a lot of pixels with values of zero. I think same is happening with DSS.

[glend]

Ok that is showing you a typical standard deviation type bell-curve, with zero at the mid-point. That is not the typical default histogram display used in BYEOS, or SGP, etc. Perhaps rhe type of display is selectable.

[rcheshire]

Chris. There is probably nothing wrong with your bias or dark frames. With DSLR data, clipping of the light frames is sometimes a consequence of subtracting bias from darks flats and lights and then subtracting dark from lights (dividing by the master flat). That is, the traditional CCD work flow.

Depending on the image set, it's better to bias subtract your flats only, then subtract your master dark from your lights, like so.

Signal = (light - dark) / (flat - bias)

This issue is not new in PI when using DSLR data, occasionally.

[ChrisV]

I get what you are saying. But this is the raw data from the camera that had negative values. I just don't get it

[rcheshire]

Odd. Can you Dropbox a bias or dark sub frame?

[ChrisV]

Thanks Rowland. Here's a bias frame that shows negative values when opened in fitswork
https://drive.google.com/open?id=0Bz6m5_jzLbUvNTAtaVA0YnhsNj A

[rcheshire]

Chris. I ran the bias through PI and ImageMagick.

The bias looks normal and there are no reported negative values.

Here's the PI output - and IM with histogram - very much the same. All channels report +ve min and max values...

I converted your file with dcraw using the same syntax as PI.

I may have missed something, but it looks OK to me.

[ChrisV]

Thanks Rowland. Must be fitswork (or me). I've now looked at PI and it's all positive there .