I’m sorry about this, but the previous histogram images are considering a base curve applied to the image.
Turning it off, the 0 EV histogram is like this:
and the +1 EV
Could you redo your raw files with spot metering (and the typical grey card shot considerations)?
Also, +2/3 EV and +1 EV have both “1/15” marked on the histogram… EXIF error?
Would the 0 EV be enough?
_MG_4250.CR2 (10.4 MB)
What would they be?
Don’t have a clue. On Digikam it shows the same way.
@gadolf, the camera metering works under the assumption that illumination of pixels in a typical normal photograph averages out to middle grey.
Therefore, if you take a photo of a grey card with camera suggested exposure, its histogram will be bunched in the middle. So your camera’s metering is working as expected.
If you take a photo of a white card, the output should be a grey card and the histogram will be the same!
1 Like
… but it doesn’t, right?
Take a look at the last histogram I showed (with base curve turned off). I shot that one without any exposure compensation (0 EV), with the camera suggested exposure. Shouldn’t the histogram be in the middle?
Every camera have certain tendencies. The manufacturer likes to give some signature so to speak. Your camera seem to be underexposing a stop. This does not mean malfunction. Once you know this, and if you don’t like it, then you can correct it manually.
Also, I think that you should apply the base curve to the photo for the histogram evaluation. Because the expected output is a jpeg for manufacturers. Thus the exposure is also computed with basic adjustments in mind.
If you say so, then, I’ll accept that.
In fact, my previous experience with digital cameras is almost non-existent. Before this one, there’s my powershot, and before that, a 4MP Nikon point and shoot.
The strange thing is that if I compensate, say, 2/3 stops, depending on the subject, the resulting image gets really overexposed. It’s like it’s not a constant deviation.
18% grey subject.
I really do not want to make things more complicated, but are you sure your grey subject is reflecting 18%???
@agriggio’s link covers that. Fortunately, you don’t need RawDigger to examine your raw files.
I wouldn’t do that if I were examining the raw file. And as I and then others have indicated, I wouldn’t process with the base curve most of the time, esp. if you are using filmic.
Other remarks
1. It is entirely possible that your metering is broken. It is not unusual but don’t assume it is so right away.
2. Metering is tied to focusing (and possibly WB). You can separate the two for composing if you prefer a feature like back button focus, etc.
3. Metering is tied to the histogram (or the live view), which is unfortunately the case for all cameras tied to the JPG preview, not the raw file.
4. Due to point 3, your metering would be off say about 1-1.5 EV if your goal is to do ETTR. We call that headroom. For less technical photographers, the headroom actually protects them from overexposure; it is a good thing unless you know what you are doing.
5. Use Manual mode. That way, you get full control over your settings. The are many settings to turn off if you want your live preview to reflect the raw output as much as possible.
6. I had other things to say but I forget… 
The problem is evaluative metering, you can’t really count on it as a base for personalized exposure. If you need to be precise measuring the light then you’ll have to resort to center-weighted or spot metering.
Not sure what you mean.
It is probably the cheapest card in the market. It’s only a cheap cardboard. But it’s sold as 18% grey, for photography.
Not sure why you’ve pointed to a 12% grey, although I’m aware of the 12 x 18% grey dispute (although I don’t understand it 
).
If you are really pedantic, not all materials are equal and useful in all situations. 
Thanks @afre for your comments.
Fyi, I turned off every possible setting that might be pushing the image from the neutral stance, although I’m not looking at the camera’s histogram. (But it’s good information knowing that the in camera histogram comes from the jpg)
As for point #1, that’s the one I’m trying to discard, although I agree with @shreedhar that even if the metering presents a bias I can compensate for that.
That’s what I thought it could mean, and probably it isn’t the best one. But even so, would it cause more than one stop shift, like it seems to be the case?
One thing: doens’t this dcpreview post explain what I need to understand but can’t?
A bias is normal but a broken meter isn’t. You need to verify.
You need to check your raw files. Each stop should represent a doubling or halving of the raw values. I am sure you know how to stop up or down with aperture, shutter speed and ISO…
It depends on how your camera meters; and how consistent the material is under various conditions, along its surface, and how reflective and opaque it is, etc., etc. The key is to know your material so that you know what to expect when viewing the raw file. If your values don’t match your expected value by a lot, then you know that something is amiss.
yes, I know, although I haven’t been using that since I had my SLR, back in the nineties…
EDIT: Rest sure that for the grey card pics I took, the camera was correctly set up. (In D5, no blinking shutter speed / aperture indicator.)
That was fun… I opened _MG4250.CR2 in rawproc, with rawdata=1, so no processing at all was applied. Your camera produces 12-bit raw data, so your maximum value is 4096. libraw says your camera’s min black is 128, but the dcraw adobe_coeff entry says 0, so we’ll just assume dcraw is correct and your camera has a 0-4096 data range.
The image appears to be a full-framed neutral tone, no borders, no other stuff to influence the histogram. Without white balance, the gray tone appears to be at about 518, with the camera white balance applied it goes to 737 . 18% gray in 12-bit would be 737, so I’d say your meter is right on.
I did all this with rawproc data, which is floating point 0.0-1.0, so I had to do some math to convert back to the unsigned integers delivered by the camera. I also found that a monochrome histogram of the raw data isn’t that useful, better would be RGB collected from the individual values of the mosaic pattern, so I’ll be programming that change shortly…
1 Like
Okay, I think I did this a bit wrongly. I used the max channel values to imply the gray tone, when I probably should have used the means. So, for the opened image without any processing, that mean is 421 (@gadolf, from the rawproc info dialog I averaged the rmean, gmean, and bmean, then multiplied it by 65535 to get the integer value). After white balance, that number is 550, so it’s still about 1/2 stop below the 737 value that is 18% gray for your 12-bit camera.
Others, pipe in if you see other considerations or corrections. A good exercise, has given me pause to think through my histogram mods a bit differently…
2 Likes