Hi,
I'm currently trying to learn by doing and so I went out today to play with ETTR and dual ISO. I have tried dual ISO a few days ago and it worked pretty nicely although I didn't see huge advantages. Well, I'm no experienced digital photographer (yet) :). With ETTR, however, I found some problems. After pushing all CR2s through cr2hdr as usual, I opened the DNGs in darktable and found all shots that were done by the ETTR module in autosnap mode way too dark. Some had a significant magenta or green tint.
I was shooting in "P", with ISO originally set to 100. Dual ISO was set to +4. Now I'm wondering if it was some operator's mistake or if there's an issue with the ETTR module on the 6D. As far as I understand, image done with ETTR shoudl appear too bright, i.e. overexposed, rather than too dark.
While shooting I noticed that ETTR was often doing multiple shots and it fairly often said "ETTR giving up" on the preview screen.
-Stefan
AutoETTR does not work in P, it works in Av,Tv, and M, but really you should only use it M (IMO), otherwise you have two different metering systems fighting each other (Canon metering vs. AutoETTR).
AAH! Will test, thx! :)
Tried in Av and M and now it looks like the second shot (autosnap) is a regular CR2 and no dual ISO...
Edit:
Not only the second shot - it looks like AETTR switches dual ISO off?
Edit2:
Confirmed! I just checked the ML settings after my last test shoots and dual ISO is set to 100/100 = effectively off.
ETTR thinks you didn't need dual ISO for the particular scene and you probably didn't. Try on a scene with a large dynamic range (for example: indoors with some bright exterior windows)
Or disable dual ISO linking and just use a manual dual ISO setting. Or adjust SNR limts.
Quote from: dmilligan on July 29, 2014, 05:26:33 PM
ETTR thinks you didn't need dual ISO for the particular scene and you probably didn't. Try on a scene with a large dynamic range (for example: indoors with some bright exterior windows)
The algorithm is that intelligent? So, basically, with AETTR, I don't need to worry about dual ISO - it will set the values as needed?
I'll experiment more!
I didn't quite get how ETTR works in Live View mode. Does it continuously analyze the picture or does it wait for a shot being taken? I.e., does it work for shooting video?
Quote from: sgofferj on July 29, 2014, 07:57:25 PM
The algorithm is that intelligent?
It measures the SNR of the shadows and mid-tones to determine this. If they are below the threshold set in the menu, ETTR will use dual ISO (if you enable it) to reach those thresholds. You can think of it like this: the lower ISO is determined via the histogram (increase the exposure until just before there is clipping), and the higher ISO is determined by noise analysis (increase the exposure until we reach the SNR thresholds).
Tons of testing today in the kitchen in the afternoon - medium bright inside, very bright outside.
But I couldn't get consistent results. I originally set dual ISO to +4 again and again tried with Av but also in M mode. Out of about 60 shots I got a single image which was dual ISO and a little bit overexposed so I could do something with it in darktable. Most of the pictures were heavily underexposed. I also tried setting dual ISO to -4 but with that I didn't get a single usable picture.
Maybe I'm lacking some understanding of the underlying principle here...
Is there any tutorial on dual ISO which describes the use in practice rather than just explaining the technical details? What I did was setting it to +4 and main ISO to e.g. 100 and then aiming at the bright places in the picture (the windows) or - in manual - trying to get the bright places exposed right. As I understood the technical basics of dual ISO, the camera would create an image file with ISO 100 and ISO 1600. As I adjusted my parameters for the windows, I would expect that in the ISO 100 lines, the windows would be exposed right while the room is too dark and in ISO 1600, the room would be exposed right while the windows are blown out.
Together with AETTR, I would expect about the same, just that everything is just a tad brighter (overexposed).
I also tried without AETTR and the results were a bit better. However, with dual ISO set to -4 and the camera's main ISO set to 1600, every other picture was not dual ISO - it was totally flat, no lines at all, with the room correctly exposed and the windows blown out.
You said you are on 6D, right?
Quote from: a1ex on July 30, 2014, 08:22:19 PM
With these results, ETTR on this camera is also not much better than random exposure.
So, wait until somebody will fix it (I'm going to break the nightly builds until then), and if it still won't work after that, prepare to do a proper bug report (for example, choose a static test scene and upload the first 4-5 test images until ETTR settles).
Right now, the exposure decision on 6D (and 600D, and maybe others) is completely wrong because of bad RAW offsets.
The offset used in the nightly builds for 6d seems,
Active area: 50, 84, 3708, 5554 (crop size 5470,3658).
They should be:
The offset seen in exif info from normal Canon CR2 is:
Sensor Left Border = 84
Sensor Top Border = 50
Sensor Right Border = 5555
Sensor Bottom Border = 3697
Which results in an image 5472 width and 3648 Height
Wrong.
what is wrong ?
Quote from: a1ex on July 30, 2014, 09:32:50 PM
You said you are on 6D, right?
So, wait until somebody will fix it (I'm going to break the nightly builds until then), and if it still won't work after that, prepare to do a proper bug report (for example, choose a static test scene and upload the first 4-5 test images until ETTR settles).
Right now, the exposure decision on 6D (and 600D, and maybe others) is completely wrong because of bad RAW offsets.
AAAAAAAHHHH!!!
Ok, I'm fairly experienced in software testing and bug reporting but not in photography. But if somebody gives me precise instructions on what they need, I'll be happy to do whatever testing necessary to help.
Is it possible that using negative EV offsets for dual ISO leading to single ISO pictures every 2nd shot is also a bug?
The exposure decision for dual ISO is based on the shadow noise level. Since that one is measured incorrectly, the exposure decision is also wrong.
No, I meant, dual ISO without AETTR. Setting it to e.g. +4EV and using camera ISO 100 works perfectly. Setting it to -4EV and using camera ISO 1600 produces a single-ISO picture every second shot.
There is an option to do a single-ISO picture every second shot. Are you sure it's not checked?
It could be a bug too, but no other camera does this afaik.
Quote from: a1ex on July 31, 2014, 12:02:06 AM
There is an option to do a single-ISO picture every second shot. Are you sure it's not checked?
Pretty much, because it works perfectly with +4 and I change just that value, nothing else. I'll test more tomorrow.
Whops, I just checked just to be sure and it *was* activated. But that would mean that the +4 doesn't work properly then? Well, I'll test tomorrow...