I’ve been working hard the Z 6’s Highlight-Weighted Matrix mode, which is separate and distinct from their Matrix mode. It can definitely blow highlights in some cases, mostly from my observation when the scene includes light sources, but it does a good job of not blowing anything else. It does so at the expense of a stop or so of headroom, mostly, but the camera’s sensor in the depths of the shadow pit does quite well in allowing some curve to yank them into presentability.
My ‘ETTR’ technique of late is to bracket, if the scene will stay still for it. Otherwise, on-the-go I’m relying on this mode and post-process lifting, and I’m satisfied with the results. I don’t think I’d buy another camera that doesn’t have such a metering mode. Not that I wouldn’t enthusiastically embrace a proper ETTR mode…
The measuring method doesn’t matter: you just need to avoid clipping. It doesn’t help to measure just the brightest spot since the camera evaluates this as mid grey - then you need to know, how much you can increase exposure until there’s clipping in the raw.
Unless there’s a raw histogram or zebra indicating clipping based on sensor output its up to your experience…
To complicate things a little more, in my case I use ‘back-button focus’ exclusively, I just got hooked to it many years ago. Meaning “focus in the center, recompose, meter/adjust and shoot”. That’s one reason I’ve been using matrix metering forever - metering before composing did not sound reasonable, but I guess I’ll have to try it and see what happens.
I’ll also try @paperdigits recco, perhaps setting +2EV overexposure permanently during a few sessions. Taking a look at recent and old raw histograms going back a few years, they all seem to have 2-3 or more EVs of headroom.
I will also use bracketing harder while testing this, that’s a great idea. I have been using 3-frame / 0.5 or 0.7EV bracketing more for “fine tuning” exposure. I’ll try 5 or 7 if the scene allows. It’s free anyways
I wonder why Nikon & Canon do not add ETTR exposure mode. It seems like there is a demand for this
Probably because it’s a lot less useful for jpeg shooting, and it sounds like it’s rather hard to get correct (especially in the higher ISO settings).
And how many users really care about such a mode (as a proportion of the total userbase for those cameras)?
I’d highly recommend using the HDRMerge darktable plugin when bracketing - it will align and merge your bracketed RAW files into a DNG that is auto-imported into darktable. This way you can just process the resulting DNG as you normally would. As long as you have a scene that lends itself to bracketing, this is really powerful
I have a Sony A7 III. So what I tell is about this camera.
The default settings in the camera are for exposing an in-camera JPG (a devleoped photo). The same applies for the live histogram. However some cameras, like mine, allow you to change those settings and the histogram. I do that for landscape photographs, the reason is that I want to get out as much as possible from the dynamic range and expose that I don’t clip highlights.
I think some Nikon cameras have an ETTR exposure mode, as does my Sony. A couple of points though:
in some cases it may be preferable to blow some specular highlights in order to reduce noise in the shadows, otherwise you might end up with some images with low exposure. Exposure compenation with zebras can be useful here.
The default settings in exposure module and filmic assume the image was exposed using matrix metering, and adjusted fir ETTR using manual compensation dial. If the camera automatically exposes for ETTR, the exposure module no longer knows what the camera would have estimated to be middle grey, and it cannot correct for the ETTR adjustment. This means that there may be more variability in exposure between different shots, and you may need to adjust the exposure slider more often to set the middle grey point, whereas if you use matrix metering, then on many cases the amount of tweaking needed on the exposure slider will be less.
My workflow is a bit different. I use “zebra” or “highlight” indication showing me (before taking a photo) which part of the image is overexposed in jpg. After doing some trials I found that when jpg overexposed indication starts to appear I can still add 2/3 EV and recover the highlights in raw.
I think filmic doesn’t assume much about your photo. It assumes middle gray is at the correct ev. If it isn’t, raise or lower exposure before filmic.
I think in modern times people thinking about ETTR is a mistake. It was done with older digital cameras with a heavy noise floor. These days a shot can easily be recovered 3ev stops…(ahem) even Canons.
The exposure a camera picks is more and more influenced by ‘tricks’ in cameras. The jpg you see on the camera screen can have shadows boosted, or might actually be taken 1ev less than you told the camera to do but it autoboosts the exposure by 1ev (this way having 1ev more of highlight room).
This all is not bad, (not at all) but it means that predicting what your camera is doing is getting more complicated, and trying to do ETTR will cause some heart broken moments when you realize back home you’ve sensor-clipped some important parts.
This - I think - is what aurrelien meant with
“by experience, this is a dangerous game where highlights might be blown much sooner than you expect, so I personnaly prefer a bit of noise and to give some safety margin to highlights”.
ETTR is to maximize signal to noise, which is a problem cameras don’t have that often these days. Use it if you know what you are doing. Most people are NOT. And most articles online are not useful at all.
Remember that your camera has something of a base curve, default look applied to raw sensor data,making it brighter in most cases. This look differs per camera model (not just camera brand). So one person having to apply +2ev before filmic and another person having to apply 0.5ev makes sense.At least to me.
If anything I underexpose a bit by default. I rather brighten the image then having to mess with highlight saving tricks.
If your middle-grey is not at 18% of the medium peak luminance, then I have no idea what you call middle-grey, but it’s certainly not middle-grey.
But, to be honest, I don’t really know what you call highlights clipping too since that depends on the white-balance coeffs for your camera, hence it depends on the individual channel ADC gain × light spectrum presently recorded × WB coeffs used to normalize it × software gain applied to normalize the 12-14 bits signal - dark current threshold in the [0 ; 1] range.
Set matrix metering -1 EV bias and be done with it.
Camera sensors only know dark current and ADC saturation level. Whatever in-between is none of their concern.
So you are basically wasting a lot of energy trying to record the holy value of something highly unpredictable in order to make systematic exposure decisions based on processed JPEG histograms and Rawspeed heuristics.
Sounds like a nice way to not get the expected results while making the process a lot more complicated.
I simply reported what values my camera puts into raw files when I use its metered exposure to photograph an out-of-focus evenly-lit plain card. The values are about 380 on a scale of 0…16383, which is over 5 stops of headroom before sensor saturation.
This is in raw files, so before any white balancing etc. There is no need for WB to cause clipping. Some software (such as dcraw) can clip values in the WB process, and I regard that as a software flaw.
We might choose to process the image to place that value at 0.18 on a linear scale of 0…1, which is about 2.5 stops down from maximum. Nothing wrong with doing that, of course.
By definition of what? Practically no camera meters for 18% of sensor saturation, if only because highlights are not necessarily 100% diffuse reflectance. That’s basically the point of HDR:
One type of highlight is the specular reflection. The advantages of having more accurate specular reflections enabled by HDR include better surface material identification [2] as well as in depth perception, even with 2D imagery [3] [4].
Cameras using matrix metering assume the average luminance is more or less middle grey, aside from a couple which have highlight-weighted spot meter. So far, only the Fuji seem to have some HDR approach, by reducing exposure by 0.7 EV compared to other makers (which makes users scream a lot about “underexposed” RAW, you know, the infamous “I set the exposure right in camera and RAW is dark”). So, as far as I know, in 2021, camera still deal with diffuse white as the upper bound, ± safety factor for some.
For the matter at hand, they are. All we care is the saturation level and the dark current, aka the 2 bounds of the DR.
The definition of middle grey is not 18% of highlights.
The definition for saturation-based ISO speed accounts for half a stop of highlight headroom above 100% diffuse reflectance compared to the exposure index values that are computed from mid-tones (it’s 78 lx·s ÷ saturation exposure instead of 10 lx·s ÷ mid-tone exposure, so ~3 stops instead of the ~2.5 stops that there are between 18% and 100%). DxOMark then took that saturation-based definition and decided to apply it to raw data (and misleadingly call it “measured ISO” in their measurements).
That means that for a camera where the manufacturer ISO setting coincides with DxOMark’s “measured ISO”, what the manufacturer considers a mid-tone is 3 stops below saturation, so if it meters for 18%, it means that diffuse white is half a stop below saturation. Most cameras have “measured ISO” values even lower, meaning that they saturate even later.
If we look at the α7 III, for example, its “measured ISO” at ISO 100 is 75, which means that at that setting, the camera considers 0.1 lx·s a mid-tone, and saturates at 1.04 lx·s, so 3.4 stops above. Same story with the R6 (63 vs. 100), the E-M1 II (83 vs. 200), the S1 (62 vs. 100), or the Z6 (77 vs. 100).
