I am struggling with Dual ISO and the color calibration module in darktable. With it turned on I get purple highlights.
I am looking forward to see your results.
This file is licensed Creative Commons, By-Attribution, Share-Alike.
Color calibration module:
20220218_MG_2435.DNG (39.7 MB)
20220218_MG_2435.DNG.xmp (23.0 KB)
As a bonus, normal ISO 200 if you want to compare.
20220218_MG_2434.CR2 (21.6 MB)
Interesting. Clever of the Magic Lantern folks! I got the impression from the article that this was only for 5D3 but obviously you can do it with your 6D. (I have a 6D but haven’t taken the plunge with ML).
I’m guessing dt isn’t set up for these raw files. With just the white balance module at the usual “camera reference” the image is very yellow, so perhaps the WB factors are simply wrong for a dual ISO image. I see the white point is 60,732 which is very high. I think normal 6D whitepoints don’t go beyond about 16k.
dt is showing much of the sky burnt out. Perhaps again this is because dt is not set up for dual ISO. Or did you simply overexpose? Presumably the idea is to get the sky ok at 100 ISO and then the 1600 ISO gets the shadow detail?
I got a vaguely decent WB at 2350K and .767 tint which are pretty weird values!
Will be interesting to see what the experts say!..
20220218_MG_2435-dual-ISO-V1-S-sRGB.xmp (44.2 KB)
Maybe some non standard stuff in that DNG that gets created??
@Peter Try changing your norm in filmic to no and make sure highlight reconstruction is LCH…I don’t see the extreme color that you have even though it is quite blown…I think you have pushed it a bit with all those local contrast modules and you have used color calibration with non D65 values in the wb module…
These are two simple versions i get using WB and CC workflows…
20220218_MG_2435.DNG.xmp (7.2 KB)
20220218_MG_2435_02.DNG.xmp (7.2 KB)
This seems to be caused by highlight reconstruction module.
In this case it worked better for me using only filmic for highlight reconstruction.
DT 3.8.1
20220218_MG_2435.DNG.xmp (9.2 KB)
I’ve found this example incredibly helpful with the filmic highlight reconstruction.
Thanks to you and to Peter for posting.
I had wanted to try that. My main PC died last week. Limping along on an old backup PC with Intel graphics…Filmic reconstruction gives me a garbled display…might be Opencl. Will have to try and rerun with it disabled…glad you found a better result…It does seem also to me that the filmic norm has more impact on the results produced when using color calibration vs using legacy wb
Not processing related but you may want to try to reduce the ISO gap; e.g., 100/800 or 200/1600.
I think that is a good suggestion…perhaps trying a range of iso pair options under the same conditions and find out what gives you the best dynamic range without blowing out the hightlights when you make that processed DNG from the dual iso??
Have other pictures with ISO 100/400 from the same trip. Often better with smaller gap.
I use raw histogram for ISO 100 and I check how much highlight I clip. Then I activate the second ISO to bring up the shadows. Half resolution for clipped ISO 1600 highlights but that is interpolated from ISO 100.
With Magic Lantern it is also possible to increase the headroom for highlights a little bit without using Dual ISO CMOS/ADTG/Digic register investigation on ISO
Sounds like interesting options… I was going from the one you presented as it had that large blown area in the sky so it seemed that your dual iso had failed in this case or the subsequent processing of it to the DNG somehow did not perform properly from what you expected…
Hard to beat the sun. Turn off White balance module and color calibration module and check clipped highlights in the raw file.
Will see what I can do with the other Dual ISO pictures I have from this time.
If it is clipping, then you may want to couple Dual ISO with AETTR. The reason I suggested decreasing the ISO gap is because we do not want too much of the sky to have half resolution and interpolation-related issues which may throw off dt algorithms. Since it is the bright areas we are concerned about consider decreasing the upper ISO; e.g., to 100/400 or 100/800.
Last time I used AETTR it needed to use Live view to measure the light. I felt it was easier just looking at the raw based histogram. Wish every camera could have that.
If you are willing to take a more technical approach and a daredevil, you could try merging the ISO scan lines yourself and attempt to recover the detail. 
Feel free to show me what you mean. Hear is the original file.
_MG_2435.CR2 (24.4 MB)
It is more of a joke. I don’t have the time or energy for acrobatics.
Have you tried the raw zebras? I also forgot to mention that you could do this, 1600/100, to preview/read the effect of the brighter ISO.
I see.
I think that is on by default. I get this (I don’t mean the histogram part) if that is what you mean How to view RAW histograms after taking the image?
What I am missing right now is how to embed the recovery ISO. I find it in the Terminal when I run the script but not in the metadata for the DNG file.
Which part? I talked about several things. Sorry for the info dump. Certainly, dynamic range decreases with aperture, speed and now less commonly ISO changes, so you have to be careful about that. What we see on the spec sheet is almost always the maximum possible dynamic range. We are rarely at that point, unless we are in a controlled studio environment.
Dual ISO does a fairly good job of predicting what the combined dynamic range is. What it doesn’t do is tell you how/where to expose your image. That is why my last suggestion is to preview the upper ISO, so you can monitor what is happening with highlight exposure and adjust accordingly.
I am aware this discussion is more suited for the capture and processing categories, but I hope it helps inform your raw development as well.
PS - there is a way to embed the original CR2 and also copy over the metadata. Read their docs and command information.
I normally take the ISO 100 picture to decide how much highlight I clip. Seems like less than 1% of green channel.
Then I activate dual ISO. In this case ISO 400. In this case the white table is clipping at ISO 400 as the raw zebras show.
Are you saying I should try something else than that to get better pictures?
