exposing to the right (again)

[Ray]

Secondly, although many people advise to turn the contrast/saturation, etc settings all the way down for a 'neutral' histogram, I disagree strongly. A histogram where there is '0' contrast and therefore retains the highlights is misleading unless you intend to print with '0' contrast!! Yes you have more room to play using the curves with the raw file, but the histogram is still misleading and you are causing yourself a lot more work when processing so as to add the contrast while holding back the highlights. I advise setting the jpgs to at least the lowest common denominator of your regular printing (admittedly far easier with people photography than landscape's), infact with many cameras you can set up multiple sets of paramaters for different occasions.


[a href=\"index.php?act=findpost&pid=61276\"][{POST_SNAPBACK}][/a]

I can't quite follow your reasoning here, Pom. We're talking about exposing to the right without blowing wanted detail in highlights, aren't we? At least that's what I've been talking about. The histogram is based on an in-camera jpeg conversion which is influenced by the contrast settings under the 'Picture Style' heading on the menu (on the 5D). if you are shooting RAW, these settings do not influence the RAW image in any way, but they do influence the appearance of the histogram and the flashing 'blown highlight' warning.

Here in Hanoi it's a dull day with consistent, dull lighting. I've just pressed the camera against the glass window of my hotel room and taken a series of shots of a scene of backyards, hanging washing and grey sky, carefully aligning the focus square on the same point in each shot to ensure the scenes are identical.

With contrast for all picture styles set at +4, I need an exposure compensation of minus 1 2/3rd stops to avoid the blinking highlight warning. With contrast set at  -4, I need an EC of minus 2/3rds stop, one whole stop less.  

I shall examine these RAW images later on my laptop, but right now I'm off to see a puppet show   .

[Ben Rubinstein]

Ray, you've just proven what i was saying. My point was that the histogram will only be accurate if you keep in mind your print or output. In your case the RAW files will be exactly the same but the histogram gave two pretty different interpretations. You have to choose what kind of RAW file you want.

Pushing the RAW file to the right to the utmost level, almost blowing every channel, may give you a good file if processed with zero contrast, but the moment you try to use the file, to manipulate it in any way, you're screwed.

The usual injuction to expose to the right with the addition that overexposure is better than underexposure as stated clearly in 'Real Camera RAW' simply does not work with people photography where the facial tones are extremely difficult to recover even with clever curve tweaks as they just don't look natural after a recovery of even 1/2 stop or less 'blown'.

If I set my histogram to show me a reading based on '0' contrast then I will be in serious trouble when I come to edit the RAW files when the files and especially facial tones are screwed up in that I cannot and do not print at '0' contrast.

I want good contrast in the highlights as well, not a dull/greyish look caused by lowering the contrast curve to bring back blown areas of skin.

[Ray]

Pom,
I still can't follow your reasoning. The RAW file is not influenced by the histogram in any way. The tonal qualities of your image, when shooting RAW, are dependent only upon the aperture, shutter speed and external lighting conditions of the scene you are shooting. (Excluding obvious factors such as general camera quality and compositional skill etc). The image on the camera's LCD screen might look washed out and dull, or punchy and contrasty but that's of no concern whatsoever in relation to the RAW file, the converted image or any subsequent prints you might make. But it would make a difference if you are shooting in jpeg mode.

As I understand, the purpose of adjusting the contrast of one of the picture styles is to make the histogram a more accurate guide for exposing to the right. Having now had a chance to examine the series of shots I took through the hotel window a couple of days ago, I can provide the following data that is certainly relevant to the lighting conditions at the time the shots were taken but might not be relevant to different lighting conditions.

Using the 'evaluative' metering mode, producing a 'correct' exposure of 1/50th and subsequent exposures of 1/60th, 1/80th, 1/100th and 1/125th using EC, I found that setting a Picture Style contrast to its minimum (-4) resulted in the histogram and flashing highlight warning being the most accurate for exposing to the right.

Curiously, the evaluative metering mode was spot on for these lighting conditions. There was no need to even look at the histogram. At the 'correct' exposure of 1/50th, there was a modest amount of highlight flashing in parts of the sky. At 1/60th the flashing was gone. Either 1/50th or 1/60th is okay for this scene. Whilst the histogram in ACR is clearly pushed against the right for both the 1/50th and 1/60th shots, an EC adjustment of around 1 stop is able to recover sufficient sky detail.

When I examine the shots taken with a Picture Style set at +4 contrast, at the same shutter speeds, the images are of course identical. However, in all of those shots except the one at 1/125th, the highlight warning was flashing. If I'd been using the histogram as a guide for exposing to the right, I'd have used either 1/100th or 1/125th exposure for this scene and in doing so I'd have underexposed shadows by one full stop more than necessary, thus needlessly introducing unwanted noise in the deepest shadows.

Dale's technique of exposing 3 stops greater than the spot meter reading for the brightest part of the image, might prove to be the most reliable and consistent method. However, the downside of this approach is the risk of losing the moment whilst searching for the brightest spot in the scene and then doing the mental arithmetic of dividing 350 by 2. (Although, I should add that I am quite capable of doing this without resorting to the use of a calculator, but I can't speak for others   )

[Ben Rubinstein]

Ray, your histogram even for shooting 'RAW only' depends on the jpg paramaters set at that time right?

I know the paramaters don't effect the RAW file but they do effect the preview image that pops up and the histogram is based on that not the RAW file.

Therefore if the paramaters are all set to '0' the histogram is assuming that for a given exposure there are no highlights blown, assuming '0' contrast! The exposure reading is therefore wrong if you are not going to be using '0' contrast during post processing as highlights will blow unless tamed, but that isn't the point, the histogram was misleading.

For example, and again this is more for my side of the line, wedding/portrait work. I have to overexpose slightly what the camera thinks is the right exposure because I know that the picture looks darker in print. Therefore if I want to lighten the photo to make a good print, I introduce more grain than if I'd shot that way in the first place. I have a feeling that the 5D underexposes on purpose more than my 1Ds did but that is another issue.

Exposure with digital is a new creature as far as I'm concerned. You have to set an EI value to your iso's and to your exposures in general, using your histogram which you have experience in interpreting with each camera and with the final print in mind when shooting. Yes technically my exposure might well differ from the 'true' but it's the one that gives me the print I'm looking for. I don't have experience in hand developing and printing B&W but I understand that it works in a similar way.

Your test at +4 prove my point, the histogram is misleading with such a setting. What I am arguing is that a 0 it can also be misleading and tests should be done to work out what is the most accurate for your style of shooting.

BTW I've just worked out that I've been talking about the '0' contrast of the 1Ds which really is '0'. On other cameras it may be the default 'middle' value and I can't remember which camera you use or which way the 5D does it. This may put this whole conversation into perspective!

**EDIT**

Just checked, if you are talking about picture styles and '-4' then that has been where the problem is. There are some who suggest setting all the parameters to a true '0' or -4 in your case so that the histogram reflects the all the data captured. This is what I've being advising against for the reasons stated above.

[jliechty]

This is probably a dumb question, but I thought that the point of "expose to the right" was to push the histogram as far to the right as possible during exposure without blowing anything out, and then to pull it back in post-processing. Obviously, once pulled back, will there not be sufficient room to play with contrast adjustments et. al.? I have never found this to be a problem in my own shooting, though I must admit that I only recently got a camera with RGB histograms and detailed adjustments for in-camera JPEGs, and as such I haven't acquired a wealth of experience with it yet. Thus, what am I not understanding about this exposure latitude phenomenon?

[Ben Rubinstein]

the trick is not to blow the highlights though and having a histogram giving you perhaps false information can land you into territory where you need careful and extensive curves adjustments and/or multiple masked layers to obtain good results. This is less of a problem for landscape shooters but facial tones do not tolerate overexposure, they start looking false and plasticky when pulled back and you get colour shifts. Personally when I get back from a wedding with hundreds of images, I want the best exposure for my needs, not to work over each image for hours. Yes, expose to the right, just not too far! added to that only your experience with your equipment and raw converter will tell you what 'too far' is.

[Ray]

Your test at +4 prove my point, the histogram is misleading with such a setting. What I am arguing is that a 0 it can also be misleading and tests should be done to work out what is the most accurate for your style of shooting.


Just checked, if you are talking about picture styles and '-4' then that has been where the problem is. There are some who suggest setting all the parameters to a true '0' or -4 in your case so that the histogram reflects the all the data captured. This is what I've being advising against for the reasons stated above.
[a href=\"index.php?act=findpost&pid=61576\"][{POST_SNAPBACK}][/a]

Pom,
I understood what you meant by zero contrast, ie. the minimum setting... which is -4 on the 5D.

I shoot mostly landscapes. I am not so obsessed with DR when shooting portraits, although I recall your complaining about noise in dark suits at weddings, using your 1Ds, and that's the sort of thing, I guess, which would result from not exposing fully to the right.

The -4 contrast setting seems to be right for landscapes with blue skies and white clouds. When I see small patches of 'highlight flashing' in the sky, I know the exposure is right and that most of the detail in the 'flashing' areas can be recovered using up to 1.5 stops of EC during the RAW conversion. If the whole sky is flashing, then I know the exposure is too great and there'll be a color shift in the blue sky when I use EC during conversion (a shift from blue to cyan which is not appealing.)

I'm glad I've discovered this. I now consider my histogram to be more accurately calibrated for my purposes. However, the image on the camera's LCD screen is often very flat and not particularly impressive when showing other people the results, but that's of little concern to me if such people are not clients.

However, there is still the problem of specral type highlights which one might not want to be specral. A -4 contrast setting would be very misleading here. In fact a +4 setting would be much safer.

[Ben Rubinstein]

The screen has never been good for more than checking composition and details, I suppose ambient light would mean that it would be impossible to calibrate for reliablity in any case.