Highlight Overload and Dynamic Range.

[voxmagna]

I've started this thread to see what others think. I've had several electronic cameras over the years and I am currently on the D750. But I'm finding the problem of sensor dynamic range, does not seem to have vastly improved. A captured image with blown highlights, like bad focus, cannot be recovered in post.

I distinguish exposure and all the fancy options first, because they are not a magic bullet. The camera sensor has a (limited) dynamic range window to capture the brightest highlight to the darkest low light. The exposure options only move that 'window' with its fixed sensitivity limits up and down. You can get crushed highlights and normally exposed middle and low lights, or dark underexposed low lights and perfect undistorted (clipped) highlights.

But I don't have the same experience so much when I use a pro video camera, so what is going on? Now some could say I'm just a lousy photographer and don't know how to use the camera. Take a landscape with a bright blue sky with lots of cumulus cloud detail with plenty of scene contrast. I want the clouds unclipped not blown and the foreground bright and not underexposed. Even cheap handycam video cameras seem to cope.

I take my landscape example and rattle off some bracketed shots with the D750 picking the shot that is not clipped and with underexposed foreground. In post I can fudge the restricted dynamic range of the camera sensor by modifying the gamma curve. Video cameras have been using gamma and black lift for years to compensate for limited sensor range. But why can't my D750 and other still cameras do this?

I read about the D750 using multiple exposure points in matrix mode, but somebody tell me, is the camera merely selecting the best 'average' exposure for the scene or is it using that matrix of information to modify the gamma curve? I don't think so since the matrix is confined to the central area as I understand it and not the whole image.

The Nikons have scene setting options. I've always avoided them as being a bit gimmicky trying to choose my shooting settings when I know what they should be. But am I wrong? Do scene options play with the camera gamma curve and introduce the black lift and white drop needed for a scene with lots of contrast for landscape? Modifying the gamma curve to avoid clipping is a fudge for limits on sensor performance, but I find that better than having clipped highlights and flat white sky in landscapes when there were nice cloud formations.

I know about Extended Dynamic Range and I'm interested in trying it but I would have expected the camera to do a better job with single shots. I am still a little puzzled by the camera preview which is based on 8 bit JPEG processing when the camera can shoot 14 bit RAW. If I see highlight clipping on the preview screen, does that mean it is less likely to be present on the saved RAW image? I don't expect an LCD screen to have much dynamic range, but the clip indicator should just be working on image pixels and my guess is the Nikons still work on 256 bit brightness levels, even when saving only RAW images.

 

[pforsell]

The dynamic range of Nikon DSLRs has increased 100-fold (6 to 7 stops) since the beginning. I still have my D1H and D1X and the comparison to newer cameras is easy.

What hasn't much changed is the "zero-point" of the meter, which is about 2 to 3 stops below saturation. Hence, the increased dynamic range goes in the shadows and most shooters don't know how to take advantage of it.

But, the D750 is different, the calibration has changed a bit and the reference is now 9 % gray instead of the previous 12.7 % gray and the yet older (film era) 18 % gray. There is now one extra stop of highlight headroom.

Unfortunately this change in calibration (D750 middle gray is now 4 stops below saturation) has been too difficult for the general user base. dPreview is full of posts titled "why my D750 underexposes". LOL. Instead they should post "awesome highlight headroom increase in D750".

 

[voxmagna]

That is useful background history - Thanks. I'm not putting the exposure metering in the equation because I know the camera only meters selected points and the overall results (for highlights) can be a lot different. I'm bracketing several shots 1/3 stop, checking afterwards on preview what highlights are clipped. Then import the file into PC and seeing the foreground under exposed as on the lcd preview. If I put the same image on a plasma TV I get the same thing.

Only after modifying the gamma curve can I lighten foreground lowlights without clipping highlights. Do you know if the Nikon 'scene' modes do any curve bending? I don't have a grayscale test chart to try mine out. On the D750 I've dialed in -1/3 stop EV correction which helps, but that is just compensating for its metering decision and not changing the fundamental dynamic range.

I agree with what you say about under exposed results because when I look at pictures taken by other cameras, I can spot the clipped sky straight away. It is almost as if this is becoming the accepted norm. even on TV. I have looked at some shots where the lowlights appeared to have very little detail, but when lifted in post detail is there. That is re-assuring when you get back a shot that appears badly underexposed.

 

[Horoscope Fish]

Have you tried using the D750's "Hightlight Weighted" metering mode? In my experience it works so well, so much of the time it's my go-to metering mode. Sure, there are those times I find it's not getting the job done, but those times are relatively rare.
....

 

[voxmagna]

I tried it and it just seems to drag the exposure down automatically together with the mid and low lights, It seems to work like auto EV. It's fine to keep the high lights from blowing out completely but it's not changing the curve. Curiously when I bring these difficult images into post and apply their 'auto exposure' corrections, they leave or set the highlights on the 256 bit limit and bring up the bottom end. I expected the camera to do that in its auto modes but it doesn't. I think the exposure system is very limited to simply averaging over a small area and adding in a peak clip detector for the Highlight Weighted feature. It isn't sampling a large number of pixels in the frame and bending the tonal curve. I have no problem with what the exposure modes do in the small areas they measure, it is what happens to the overall scene.

I'm going to do some comparisons with RAW and jpeg images where there is a marginal highlight clip to see if and how much headroom I actually get with RAW compared to JPEG and whether the camera preview display for clipping represents RAW or jpeg. If I can get a grayscale test card I can also see if the scene modes are doing anything clever.

Of course in all of this, larger dynamic range on the camera is useless if output can't be printed or displayed. But newer monitors are improving and achieving wider contrast range than before.

 

[Don Kuykendall_RIP]

I have been working on this problem for a little while now. What I have found is that you have to expose for the sky and bring up the shadow detail in post. I try to stay in Center Weighted metering most of the time but occasionally I will have to move into Matrix mode to get readings. On those occasions where there is just to much difference in the sky and the landscape I will use split neutral density filters to balance out the difference. This works very good to get a good overall exposure with no blown highlights and good shadow detail.

 

[voxmagna]

Now that is what I too have been working towards but I'm new to Nikons and thought I must be doing something wrong. I use fixed aperture mode mostly. What I have been doing is using the spot or small group metering modes to get a reading for the sky then lower the camera for the foreground, make a judgement on exposure perhaps a shutter stop equivalent down, shoot and look at the post shot preview to check the burn out areas. If they are few I'll keep the shot and sort out the dark foreground by bending the tone curve in post.

On a previous Canon I knew that a few small clipped areas showing in the preview were not there in RAW, but I haven't tested the D750 yet. I can shoot the daylight landscapes at the lower 100-200 ISO so I'm not going to worry too much about bringing up noise in post and at the lower ISOs you have more usable dynamic range. I've read about graduated ND filters so I am pleased to hear from somebody trying them. These are all 'workarounds' which convince me that dynamic range or latitude is still a limitation of current camera sensor technology.

 

[Don Kuykendall_RIP]

What I have been doing is using the spot or small group metering modes to get a reading for the sky then lower the camera for the foreground, make a judgement on exposure perhaps a shutter stop equivalent down, shoot and look at the post shot preview to check the burn out areas. If they are few I'll keep the shot and sort out the dark foreground by bending the tone curve in post.

Spot will probably never work correctly on the sky. Center weighted normally is the best mode for sky and then use exposure compensation from there. This is the method I used on vacation shooting a lot of landscapes. I would look at the histogram and make sure that I had t least a sliver of room on the left. That way I knew that I would have enough detail in the shadows to get a good exposure in post. So far I have had good results doing it this way.

 

[voxmagna]

Thanks, I've got a 2 day vacation coming up so I'll play with center weighted. Incidentally, a Hoya 77mm Grad filter for my Nikon zoom lens is quite expensive! Have you tried the Nikon HDR mode with any success or afterwards in post and do the multiple shots have to be stable from a tripod to get the correct overlay?

OK I got some ideas to try out for landscapes. I can set up my D750 for multiple shots on the timer, bracketed 1 stop so I choose say 5 shots. In 32 bit post I can overlay them for one HDR image. In theory I can pixel stitch the images, so hand held bracketing might be doable on fixed aperture if I'm without tripod a sunny day and with a steady hand. Good job I have a fast PC to stitch 5 RAW images. I learned the camera internal HDR only works for JPEGS. Nikon didn't build in the processing power to merge RAW in the camera (shame).

 

[aroy]

Thanks, I've got a 2 day vacation coming up so I'll play with center weighted. Incidentally, a Hoya 77mm Grad filter for my Nikon zoom lens is quite expensive! Have you tried the Nikon HDR mode with any success or afterwards in post and do the multiple shots have to be stable from a tripod to get the correct overlay?

OK I got some ideas to try out for landscapes. I can set up my D750 for multiple shots on the timer, bracketed 1 stop so I choose say 5 shots. In 32 bit post I can overlay them for one HDR image. In theory I can pixel stitch the images, so hand held bracketing might be doable on fixed aperture if I'm without tripod a sunny day and with a steady hand. Good job I have a fast PC to stitch 5 RAW images. I learned the camera internal HDR only works for JPEGS. Nikon didn't build in the processing power to merge RAW in the camera (shame).

If I want the highlights (when the scene DR is high), I use spot meter to meter for highlights, then recover the shadows in post. If I feel that the DR is quite high, and the shadows will be noisy, then it is handheld bracketing (in D3300 it is totally manual). Processing the bracketed images in most HDR software is quite easily accomplished.

I have tried in camera HDR, it is quite good, but gives only jpeg. So does in camera panorama.

 

[pforsell]

Spot will probably never work correctly on the sky.

Could you elaborate this part? I feel like you left something out and I cannot guess what it is.

I have succesfully spot metered the sky since the 1980's. Using digital is not much different than using film, only with digital I must know my camera, with film I had to know my film.

What I do, is first I need to know my camera: how much highlight headroom there is above middle gray. Let's say my camera has 3 stops headroom. I spot meter from the brightest part of the sky that I don't want to blow, and then I dial in EC +3.0 stops or +2.7 stops or whatever depending on the camera. This is no different from the old Adams' zone system: I am essentially placing the brightest highlights that I want to preserve into zone IX or X. And I let the shadows fall wherever they may.

So, since you said that this will probably never work, what should be done differently? I have thousands of images that suggest that it could work.

 

[voxmagna]

I would guess he recommends not using spot in case you missed a brighter cloud and got sky as the peak highlight instead? On the other hand anybody using spot which I do, will spend time evaluating different parts of the scene. I think once you disconnect automatic exposure features like averaging and multi zone, you become a 'human' photographer making your own decisions in real manual mode.

If you have the time setting up a shot I don't think you win much from using all the preset exposure modes and it could go wrong for you. My point starting the thread is once you establish the highlight and lowlight exposure range equivalent to the scene dynamic range, the camera is limited as to how it will handle it anyway and there is no magic bullet setting. It is peak luminance values that clip and that's what you probably need to know, not an average exposure value over a large area. I have similar feelings about automated focusing zones to. The exception being dynamic tracking which I do consider useful. Light and shade by its nature is peaky so why fool yourself with averages and zones, because only one result goes to set exposure. The sensor pixel array is not mapped and by some 'magic' individual pixels or pixel groups are exposed differently. Sometimes, I wonder if the exposure and focus options mislead users as to how they work? However tonal curve bending in post is crudely similar to pixel group manipulation, but as far as I know that is not a Nikon camera feature, unless it is in the preset scene modes.

I just tried some HDR tests using 5 shots 1 stop apart and merged in post. More darkroom work, but it definitely gets the dynamic range from bright pretty cloudy skies and foreground. You have to choose a shutter at your chosen fixed aperture that can go up and down by the equivalent of +/-2 stops. It is quite revealing to preview each shot afterwards and conclude no one shot is exposed correctly over the high contrast scene, then create the HDR image file and see the scene as it should be. What surprised me knowing the limitation of 256 bit jpeg, is a 16 bit HDR image down sampled to 8 bit jpeg still looked very good.