Strange-looking master dark

[Diane Miller]

I recently did a set of 50 darks (Canon 7D2 DSLR, Ha mod, ISO 1600, 90 sec) and when I stretched it I saw this.  I haven't seen a result like this before.  The air temp was very close to freezing.  A set i did a month ago with similar conditions and parameters didn't show this banding.

Here is a screenshot.

[bulrichl]

Judging from the name of the image this is a masterdark, not an individual darkframe, correct?

Please have a look at all the individual darkframes:
1) Does one of them show this strong banding or is the banding visible in all darkframes?
2) If only one darkframe is affected: discard your masterdark and deselect the darkframe with banding when integrating for the new masterdark.

Could you please do the following in addition: select 'No image crop' in Format Explorer/DSLR_RAW and reload this darkframe. Show us the region at the left margin adjacent to the banding in 1:1 zoom. Don't forget to deselect 'No image crop' again after that.

Bernd

[Diane Miller]

Yes, it is the master dark, stretched with the default settings.  I did a different target last night, at 3.5 min exposures, and just ran that master dark with the same results -- I always delete the first and last frames to reduce the chance of a frame with some light leak or a shorter time as a battery died.  Same result.  I opened several of the raw frames and stretched and see the same lines, but a little more indistinctly.

I'm not finding the crop option mentioned, but here I'll  is a screenshot of the same stretched master dark, zooming in to the left side.  I see no artifacts there or on any edge.  Also here is a screenshot of my integration parameters. 

I'll check back to some older darks, although I usually delete the raw folders after getting the master.  Wondering if this started after the Ha mod??  Strange thing is, after the mod I redid the bias masters for the ISOs I use (1600 and rarely 3200) and I saw it in the 3200 master bias but not in the 1600 one.

Assuming it is "real,"  is it a problems?  I haven't seen any trace of it in the linear master or nonlinear processed files so far.

[Diane Miller]

My second screenshot, of the integration parameters didn't show.  Trying again:

[Diane Miller]

FWIW I have looked back at master biases before and after a recent Ha mod and they are the same before as after:  In both cases, at ISO 1600 I see no banding but at 3200 I see a somewhat similar pattern to this master dark.  No idea what that might mean, or if it would cause any problem with processing.  I'm currently shooting everything at ISO 1600 and that master bias is "clean."  (And I'm not using it for the master dark -- that is just an integration.  I'm currently relying on using darks shot at "the same temp" (ambient, anyway).

[bulrichl]

Yes, it is the master dark, stretched with the default settings.  I did a different target last night, at 3.5 min exposures, and just ran that master dark with the same results -- I always delete the first and last frames to reduce the chance of a frame with some light leak or a shorter time as a battery died.  Same result.  I opened several of the raw frames and stretched and see the same lines, but a little more indistinctly.

Is the banding reproducibly located at the very same y-coordinate range in each of the darkframes?
Is the banding located at the very same y-coordinate range in the 90 s and the 3.5 min masterdark?

I'm not finding the crop option mentioned, but here I'll  is a screenshot of the same stretched master dark, zooming in to the left side.  I see no artifacts there or on any edge.  Also here is a screenshot of my integration parameters. 

I'll check back to some older darks, although I usually delete the raw folders after getting the master.  Wondering if this started after the Ha mod??  Strange thing is, after the mod I redid the bias masters for the ISOs I use (1600 and rarely 3200) and I saw it in the 3200 master bias but not in the 1600 one.

Assuming it is "real,"  is it a problems?  I haven't seen any trace of it in the linear master or nonlinear processed files so far.

Take the same lightframe before and after calibration with the affected masterdark and apply a strong STF to both of them (edit STF parameters if necessary). Look for banding in the affected region. You'll see either

1) dark banding in the uncalibrated lightframe, no banding in the calibrated one or
2) no banding in the uncalibrated lightframe, bright banding in the calibrated lightframe.

As for 1): no problem, the banding is corrected by calibration with the masterdark.
As for 2): the masterdark would have to be corrected for the banding before using it in the calibration process.

Maybe this discussion is purely academic. If you are going for pretty pictures and don't see a trace of banding in the finished processed image, why even care about it? Or are you doing photometry?

[Diane Miller]

Good idea Bernd,

I see no banding in either light frame.  It seems to just be in the darks and biases.  In them, it is in the same place, although the edges are a bit fuzzy.  But if I open the dark set in Blink and step through them, it jumps around noticeably but in the same general area of the bottom half of the frame.  It is almost swamped by the noise and each "band" is very indistinct.  Maybe I should be using ISO 800 instead of 1600 -- will do some checks and compare with the Canon 1DXMk II body.

I'm not doing photometry so it apparently isn't an issue but I was curious what might be going on.

[Diane Miller]

I looked at some recent master darks from sessions with the Canon 5D Mk III and 1DX Mk II (when I needed a larger frame), and none show this issue so it is apparently a noise issue with the Canon 7D Mk II.  For daylight shooting, it does exhibit somewhat more noise than either of the other two bodies. 

I'll do a dark set at ISO 800 for comparison and report back.  If it is "clean" (as I suspect) then the question becomes, as I better off at ISO 800, or is the difference so trivial in the processed images that I'm better off at ISO 1600?

[Diane Miller]

It is still there but much less obvious on an integration of 40 frames at ISO 800.