I usually take the picture with neutral style and deactivating the D-lighting. I check in the histogram that there is not clipping.
Sometimes, after taking the picture I apply an style to check more or less what would be the final appearance.
Note that the point of the scene referred workflow is that the pipeline does not clip (as long as there’s no clipping in the raw and highlight reconstruction is turned off), not until you get to filmic. When editing filmic’s curve parameters, you can set your white point as a value relative to your image’s mid grey.
2 Likes
I use a Nikon D810 and D610 and I have made my own curves for both cameras. I dunno if they are “correct” but they make the outcome look the way I imagined the photo.
To get there I started with the Nikon or Nikon alternative curve (I don’t remember) and “softened” the curve slightly. Softer tones allow for more options. In comparison it’s way easier to add contrast afterwards than start with a “steep” contrast and try to find in-betweens.
This means that I get a good base for further development - I don’t get a satisfying picture when I open it for the first time, but I see the potential rather than a proposed result.
I’m not sure the DT histogram is considered the raw data either. You can select a color space profile for it in the histogram profile that will change what you see so the values are not truly raw but processed by the current pipeline…
medium rare perhaps?
3 Likes
Using the raw (unaltered, maybe even not yet demosaic’ed) data in the histogram would not even make sense, IMHO:
And whenever I do something to the image (e.g. add 1EV of exposure or do something with the colors) , I expect the histogram to change, which it does.
Therefore it can’t be for the raw data, right?
I was trying to dissect this statement …which I took to mean that the OP felt the DT histogram was perhaps raw data if the thumbnail selection was changed…I could be wrong…I was citing one example to demonstrate that it was not…yours would be another…A raw histogram would be nice to have
It’s unfortunate our camera don’t give us the raw histogram, that would settle the matter for those who want to do ETTR and get perfect exposure every time. Unless there is a workaround, I don’t know about, your best approach would be to measure in spot mode some good reference and figure out how much compensation to add to each shot. Aurélien as a video on this I think. By the way, a gray card won’t work well, unfortunately. Someone suggested to use Teflon tape wrapped around something you can carry easily. To find the required compensation, with the spot measure on the Teflon, start from +2.5 EV to + 3.5 EV in 1/3 increment. You’ll need something to view the raw histogram to find which compensation to use for your camera. Another approach would be to systematically bracket your exposure on the plus side, say + 1/2, + 1 and possibly more, to make sure you got one properly expose shot.
If you think you can get around compensating after the shot, well I have bad news for you, even if sometimes it’s more or less fine to do this. If you have captured a more or less medium to high dynamic range, you will amplify the noise as well as the exposure by compensating after. It’s always better to get the proper exposure in camera, and it’s a shame that we have to go through all sort of workaround to get perfect exposure every time.
I’ve tried that approach of permanently setting e.g. +1 EV exposure compensation in my Fujifilm X-T1, which - as other Fujis - is known to underexpose heavily.
It worked beautifully, especially with scene-referred, when no compensation in exposure module was required - great signal to noise ratio, super clean images, but… only for low contrast scenes or those without bright lights.
Unfortunately, it looks like many things come into play when a camera measures light. Just when I was happy with nailing needed “overexposure” for my sensor during tripod test scene shots, I got then great number of blown own highlights in real-life pictures with changing frame and light.
So I gave up being smarter than those who designed the sensor and its algorithms 
You are right, these guys are smart, and they provide you with different measure mode. You can get your camera to expose correctly your raw captures only when using the spot mode with a previously measured compensation, and always using the same target. I almost always shoot at ISO 100, so you may need to verify your camera needs the same compensation at every ISO.
Indeed it needs different values. Although ISO dial and specs say the native ISO of X-T1 ranges from 200-6400, I’ve discovered - thanks to darktable (!) - that anything set beyong ISO 1600 is actually shot like 1600 and brightened by camera’s firmware.
Camera’s metering is complicated and - at least in matrix mode - behaves very differently e.g. when there is a saturated red flower in the frame, so fixing the exposure compensation often failed.
On the other hand, when shooting in controlled condition, deliberately overexposing can get excellent results, for example the ability to not set any exposure boost usually required for scene-referred workflow.
X-T3 and other xtrans4 cameras(I think the X-T4 has another bump at 1600) work the same way, except with dual gain ISO onwards from 640. This means that if you’re shooting at 640, you should just put it on 800 since you’ll get a little extra dynamic range(at the cost of noise) with the new amplifier, after that they are iso invariant.
2 Likes
Yes, so I gave up and let the camera underexpose - this avoids clipping highlights and the penalty for raising the exposure in darktable is barely noticeable with my sensor.
This was quite an experience, because before I was using Canon 1000d and 6d - the 6d performed brilliantly in terms of noise at highest ISOs, but it was crucial to let it a decent amount of light reach the sensor, so I often even overexposed a bit.
Then I moved the same habits to Fuji and tried to fill the sensor with photons as much as possible 
The “read noise” and “shadow improvement” charts are even more telling (really the same thing looked from different sides), basically saying you can win back a stop in shadows if you have dual gain (lower noise actually), but of course you still have to sacrifice some highlights (the “extra dynamic range” you mention comes from the improved shadows):
https://www.photonstophotos.net/Charts/RN_e.htm#FujiFilm%20X-T3_14
https://www.photonstophotos.net/Charts/PDR_Shadow.htm#FujiFilm%20X-T3,FujiFilm%20X-T3(ES),FujiFilm%20X-T3(MS)
Anyhow, doesn’t help the OP since the D810 does not have a dual gain sensor (like the newer D850 and D780).
1 Like
Pascal. I have D810, only shoot raw, and use DT3.8. I stopped using Nikon NX-D software very quickly. My main work is landscape-type photography, with some portraiture, all on a tripod, all Live View.
I always try for ETTR to get the most light capture on the sensor. I have setup my camera so the replay histogram gives a very close indication of blown highlights on the resulting raw. I also have ‘highlight indicator’ on which show as zebra stripes in the replay image.
To create the set-up. I took a set of bracketed shots of an outdoor scene with stable lighting and pulled the raw files into DT. I switched all the modules off and the raw under/over expose indicator on (set at 100%) and chose the shot that was the closest to overexposed as my base, and noted the camera exposure used. On camera I viewed that image’s replay histogram and checked where the histogram was in relation to its right hand edge.
NB: the picture style that the camera is set at is very important because the camera uses it to process the raw and produce the jpeg from which it derives the histogram.
Changing the style (or a style’s settings) changes the replay histogram BUT NOT the raw.
I changed the style to Neutral and took another shot at the same exposure, and it’s replay histogram was further to the left. So I adjusted the Neutral style in camera adding sharpness and contrast to push the histo to the right. I repeated this shoot-adjust-style process until the replay histo was just touching the rhs and there were only a few very minor zebra stripes on the playback image. I rechecked the last shot taken in DT and it was indeed ETTR without blowing. I did these tests in highlight metering mode.
In the field, the dynamic range of the subject and the importance of highlights in the subject determine how much exposure compensation I need. For instance, I usually don’t care if a few spectral highlights are blown, but I may not want the water spray in the central sunlit cascade to blow. I have found that I can place exposure pretty accurately after one test shot. Bear in mind Nikon’s highlight metering averages what the camera thinks are the highlights so isn’t ‘correct’, which means I often have to take more than one test shot.
For landscape I use highlight metering, and have to add from between +1/3EV (brightly lit subjects) to +2EV (flat subjects). The setup works with integrated metering as well, but then I may be reducing exposure as often as increasing, which I find not as simple as with highlight metering.
Admittedly it’s a bit cumbersome for viewfinder shooting. I occassionally go out with minimal kit and a monopod to shoot general-interest subjects and use the iewfinder. I then tend to rely on my judgement rather that ETTR and often shots are somewhat under exposed so I generally bracket.
In DT I have created a set of scene-referred styles (eg landscapes D810, portraits D810) which I apply to a set of images in Lighttable and which almost 100% of the time produce images that require minimal intervention. The quality of the images from DT are MUCH better than anything I achieved with Nikon software (and improve with every release).
Having said that what the Nikon software does do well is to quickly produce a 16-bit colour rendition of a raw in the in-camera picture style, which matches the in-camera jpg, but with more quality. You can also change the style settings in NX-D and download them into the camera to further improve image quality. I know a press photographer who only ever uses the manufacturer’s software because it quickly and consistently produces what he needs. Usefull for large numbers of shots with ‘basic’ output requirements, but not what I use DT for.
A raw in-camera histogram would be so usefull!
Good luck.
2 Likes
I would like first to thanks the all people who replied to my concern.
I spent sometimes for finding the DT’s histogram definition and I was happy of finding it. It’s based on the lightroom views, if I’m correct. I guess there are some difference between the displayed images and the RAW information’s before any modification. I assume such differences are small.
One important point I completely missed with digital photos is the ETTR recommendation. I wasn’t aware of this important information and because I’m coming from film slides background, I did exactly the opposite: exposure to the left (i. e. for avoiding full white sky without information). I looked at Aurelien’s explanations regarding the basics of digital exposure and I understand the reason for ETTR, unfortunately too late for my son’s wedding.
Thank you a lot Dusenberg. I’ll try to use your recommendations for future shots.
Take care.
By default the lighttable view is the embedded jpg file. As soon as you enter darkroom basic raw edits are applied and the lightroom preview will be updated with those when you switch back to it. You can start with a raw preview if you like but that is not the default and as soon as you enter the darkroom to edit the basic modules will be applied. The histogram in DT is better thought of as reflecting the values of a processed pipeline whatever that is at the current step you are at which should be what lighttable view shows but only after you edit it in darkroom and go back…you can see this as you watch it as you move through your history…click on different stages…on original it will have some of the basic raw modules applied so clicking up through these will show no real change
But as you go up and hit say color calibration in the history stack of the basic applied modules you will notice a change…
And then further at the end of the default processing…
This history stack is what the histogram displays…
They are not smart. They just know under which assumptions their own algos work, while you are left with only you best guess and black-box reverse engineering. Putting the tool in a black box to exclude the customer from operating his own tool at a lower level is not being smart, it’s called being an asshole.
If camera guys were smart, raw histogram would have been standard since 2008.
3 Likes
Same. I have the D810 too, and with a bit of trial and error, I figured that the flat look (intended for video) with about -2 of contrast and +2 of brightness gives a JPEG rendition that matches roughly the RAW. Meaning the camera histogram becomes somewhat meaningful.
1 Like
I know this isn’t helpful for most (especially on a Nikon topic) but if you have a Canon camera that is supported by Magic Lantern you can see a RAW histogram or waveform as well as zebras based on the RAW.
1 Like