Dark Current Suppression and the PI image calibration process

[bulrichl]

I think I've found evidence that strongly supports this last possibility: I looked at the "Overscan" region (optical black area) of long exposed (ISO 800, 6 min) darkframes from my Canon EOS 600D. For this camera, the normal image region is width=5202 and height=3465 pixels. The Overscan area is 51 pixels above and 142 pixels left of the normal image region. I only inspected the left optical black area and discovered, that sometimes there are darkframes with a bright artifact in that area, which I take to be a cosmic. When this artifact is strong enough (this is a rare event), dark banding occurs over the whole width of the image (Overscan region + image region). In y-direction this band was a few pixels wider than the artifact. Here is a screendump of an example clipping (b/w CFA raw image). Left half of the image showing the overscan region, right half showing a small part of the image region:

This is just wonderful. Please could you share your data set? I should investigate this.

Please let me point out that this finding has been possible because you can read the overscan areas of DSLR images in PixInsight. Let's hope you discovered something important. :-)

Best regards,
Vicent.

I will gladly share my data but don't have a website of my own. On the album site of my email service provider only pictures in JPG or TIF format are allowed. So where should I upload the data? Do you want CR2 files (20 - 25 MBytes each) or XISF files (36 MBytes each)? Please send me a personal message to clarify.

Reading the Overscan area is made possible by dcraw which PI depends on. 

However, please don't expect to much.
The left Overscan area in the 600D is 142 pixels wide. The picture area has a width of 5202 pixels. So the relation of the corresponding areas is 142 / 5202 = 2.7 %. This implies that the probability of a cosmic-ray hit in the Overscan area is about 37 times lower than in the picture area. Only strong artifacts (high intensity and extension over a few pixels are required) give rise to this sort of dark banding. Furthermore the width of this banding is quite low: about 6 - 8 pixels more than the extension of the artifact in y-direction.

For these reasons I believe that only a small portion of the Canon Banding is caused by bright artifacts and in-camera preprocessing. There must be other causes as well.

Bernd

[bulrichl]

Perhaps a stupid question. Is there any reason to suppose that the in-camera dark current handling of a light image, may, in any way differ from a dark image of the same exposure time.

I don't think so. Canon consumer cameras are made for daylight photography. Who takes darkframes? Only the small community of astrophotographers. Nothing that Canon is supposed to care about.

Bernd

[bulrichl]

Perhaps a stupid question. Is there any reason to suppose that the in-camera dark current handling of a light image, may, in any way differ from a dark image of the same exposure time.

It may work as well as calibrating with a single dark.  It certainly won't work as well as calibrating with a master dark integrated from multiple darks, which will inject less noise.

What is the connection? I didn't intend to use Overscan data for calibration like it can be done with scientific CCD cameras. I was just interested to understand what happens during in-camera preprocessing.

Bernd

[sharkmelley]

I don't have the technical expertise to comment on Clark's description of dark current suppression technology but it is certainly true that the better CMOS sensors have much lower dark current than previous generations.  The Canon 7D mkII certainly seems to be much better than previous Canon cameras.  However Sony Exmor sensors have had this low level of dark current for a long time and so Canon is very late to the game.  Exmor sensors are found in Nikon, Sony and Pentax cameras, though this list may not be comprehensive.

I use the Sony A7S (at ambient temperatures from -5C to 20C) but I still find a combination of dithering, dark frame subtraction and sigma rejection is important to eliminate all trace of thermal pattern and so-called walking noise.

One annoying feature of many Exmor sensors is the "edge glow" which typically appears along the bottom edge of an image and in a couple of patches on the left hand edge.  This edge glow appears in bias frames (even cooled to -20C), dark frames and light frames.  There was a discussion here: https://www.dpreview.com/forums/thread/4091328 I have yet to see a satisfactory explanation of the effect but it is not dark current. In any case, the edge glow in the master bias will remove the edge glow in the master dark and so PixInsight dark frame optimisation works well - I use it all the time.

Mark

[whwang]

Some more fuel to the edge glow issue.  See this:


In the image, the top one is a long exposure dark taken with the Bulb mode, the bottom one is a short exposure bias taken with the M mode.  Both were taken with the same Pentax 645z camera.  The bias has edge glow and the dark doesn't.  Until I found this, my 645z images cannot be perfectly calibrated using PixInsight.  After I found this, I took all exposures in Bulb mode, including flat and bias, and the calibration problem is gone.

So, if this is the same edge glow you saw, then somehow Pentax figured out a way to turn it off in Bulb mode.  It could be simply ampglow, as I saw some indication that its strength changes with exposure time (when taken with different exposure times in M mode).

Cheers,
Wei-Hao