Apologies if I am posting this in the wrong place.
In the darktable user manual, under “Getting started” exposure is discussed as well as ETTR, and then it is stated “As a general rule of thumb, it is safe to under-expose all images by 0.5 to 1 EV”. Surely this is contradictory, or am I missing something?
unless you don’t have a raw histogram on your camera it’s hard to do ETTR without risk for clipping. You can’t get back lost information. On the other side with modern cameras there’s enough room for boosting shadows.
So don’t do stuff just because you heard, but make your own experience how the metering on your camera is working. If found, that i don’t need to underexpose since my camera already does 
3 Likes
As Martin already stated: It depends.
I have a constellation here (text in German, but the images may speak for themselves):
https://www.bilddateien.de/blog/2021-06-06-raw-oder-jpeg.html#belichtungskorrektur
The first image of that blue tit is the in-camera-jpg accompanied by it’s histogramm.
You see it’s ETTR and the brightest parts of white parts of the face are really white.
This is with exposure correction of -1EV
When I import the same image, but the RAW into darktable, I get the second histogram and have room to improvement of the image.
This clearly shows that “in the field” it’s often hard to do things right in terms of ETTR …
1 Like
Related to the previous post, I usually choose a flat profile (natural, flat…) for the jpg picture that shows the camera.
In that way, the histogram “lies” less than if you apply a profile with some contrast.
1 Like
Just to get back to the original question, the formulation in the manual is indeed somewhat confusing.
Either some steps are missing in the explanation, or someone mixed up “over-exposure” (which is what you do with ETTR, relative to the camera metering) and “under-exposure” (used in the text).
1 Like
ETTR for me means giving your sensor as much light as possible, without clipping highlights. Sometimes that calls for over-exposing and sometimes for under-exposing. Everything depends on a particular scene.
3 Likes
Yes personally I’d rather have just said ETTR and let other resources teach that. The darktable manual is not a camera manual.
4 Likes
My Sony’s pre-exposure zebras equate to the post-exposure blinkies when these are set at 100+.
If I aim to have minimal zebras/blinkies in the highlights when taking the picture, then I will be able to subsequently increase the raw file’s exposure by as much as 1 stop in darktable.
I think that’s exactly what’s supposed to happen, as in both cases you select areas where clipping occurs, but still only in the jpeg version.
That’s my experience as well, in fact, I use a base style that has that 1EV extra exposure included.
It also fits with the basecurve for Sony. That curve flattens out about 1EV before white…
It also means that you could increase the camera exposure by up to 1EV, which sometimes can be handy.
Unfortunately, I usually struggle to have enough light, so I’m not really looking for longer exposures 
Ettr, means you expose as high as you can, without clipping a single pixel anywhere.
This often means you gotta expose lower then you might think.
And that you must not be freaked out by this, because (most) modern camera’s can expose like -3 or -4 (or even more) and still produce a very nice clean picture.
If you want to preserve highlights, they can’t be clipped, so expose as high as you dare.
I say ‘dare’ in this, because here comes the thing why I always say to forget about ETTR: i know of no camera that can give you a good true readout that shows you if highlights are being clipped while taking the shot .
There are no raw histograms as far as I know, and the way white balance works means it’s quite hard to know if one of your channels is clipping.
So, you have to expose as high as you can without clipping anything , but you don’t know if you clipped anything till you open the raw file. Unless you start to know your camera (technically) very well, it’s safer for most people to be cautious about your highlights and expose a bit lower , instead of trying to reach the limits of your sensor cells.
Of course, if you don’t care about highlights you can go a bit higher.
(Do not forget the basic ettr lesson : a white balanced jpeg preview doesn’t tell you anything about your raw sensor data ).
5 Likes
If I set my Sony’s zebras at 107+ they precisely coincide with darktable’s raw OE warnings but I much prefer having the headroom/insurance that 100+ provides.
My camera has a PDR of 11.43 at iso 100 according to https://www.photonstophotos.net (for reference the Nikon Z7II has 11.6. At iso 400 my camera is 0,64 EV more iso invariant than the Z7II)
I find that anything over 2 EV underexposure compared to final image results in noticeable deterioration of colours. Since my camera is relatively iso invariant I do wonder why anyone would underexpose more than that unless absolutely necessary. It is often necessary for me but it comes at a significant cost in IQ.
It doesn’t make sense to reason about underexposing or not in general. You need to know your camera - it’s all about avoiding clipping. If your camera‘s metering has a tendency to underexposed images it doesn’t make sense to underexpose if there’s no risk for clipping just because someone told you that’s appropriate, if it has a tendency to overexpose, then ETTR doesn’t mean simply overexpose a bit
2 Likes
This is what I’m trying to complicate.
There are plenty of cases where clipping is fine and the cost of fidelity to high to under expose. The question is, and has always been, how to expose for the scene and your creative intent. We have more leeway in the shadows now but the photographer still has to decide. I was myself overzealous about clipping a few years ago but realised that i lost iq avoiding clipping.
I do ETTR, but I have raw histogram thanks to Magic Lantern. Clipping 1% of one channel is ok for me.
2 Likes
As with almost everything in photography (and life) it’s all a compromise between competing desires. Decide what’s most important to the image you want to produce and compromise elsewhere to make it. All things being equal (if it won’t disturb other important considerations) ETTR is a good idea.
1 Like
I’ve messed with most of the ETTR metering pet-tricks, and found all wanting for support of my rather active style of shooting. I’m more interested in paying attention to what compositions present themselves to the frame, and want something automatic that’ll expose well most of the time. That came in my recent camera acquisition in the form of Highlight-Weighted Matrix metering. It compels an exposure that preserves highlights as determined from the JPEG rendition, so it typically leaves a stop or so of headroom, but not always - it’ll let really, really bright areas go, and I find that to be okay most of the time. Here’s one example :
The brightest parts of the sun-glint on the tender are blown, but what the hell, that’s what it looked like when I stared at the scene in-person. I just kicked in an instance of the librtprocess highlight recovery tool so the magenta would go away, but the shoulder of the filmic curve can also render them acceptable. Thing is, I didn’t have to stand there fiddling with the EV dial. Now, this sort of exposing makes shadow lifting more challenging, but dynamic range is what I bought the camera for.
Probably the most enabling realization of my journey into digital imaging has been that there’s no such thing as an exactly “correct” exposure. There’s room between sensor saturation and the noise floor, and all one really needs to do is place their image somewhere between the two…
13 Likes
With “PP-Off” standard setting lower limit 109+ zebras correspond nearly exactly to the raw exposure on my Sony A7III. So I can easily ETTR.
It’s slowly coming, but for example the a7 II doesn’t have those zebra options . It’s a sort of recent development (and still isn’t officially a good raw exposure indicator).
Magic lantern did add it, which is very nice. But still after market :).
5o be honest, when talking about -3 and -4 i was mostly talking about iso invariant.
But it’s tricky if your camera picks a higher iso , which means you already ‘used up’ some or the levels of under exposure you can do.
Color degradation I didn’t notice that much past -2ev (and my old sensor is far from how clean modern ones are ), but maybe I am easily pleased :).
And yes, i added the line in my post to make clear there you have to judge what is more important for you, highlights or shadows (or noise, or …). Expose accordingly. This means mastering the technical side of your camera some photographers don’t want to be bothered with.
I normally don’t watch it, but if I’m outside in bright sunlight i learned to start with dialing in more than -1ev and start from there :).