Expose to right, it is as simple as

[fdisilvestro]

I've been playing with an unannounced raw converter that does even better with the highlights, sorry, can't talk about it :~)

Raw Therapee does an amazing job at highlight recovery. It has four different choices: Luminance recovery, CIELab Blending, Color Propagation and Blend. It will also show the RAW histogram. Would like to hear about the RAW converter you are testing.

[Alan Goldhammer]

I liked Michael's write up and I also like the technical discussion on this thread as well.  I read Jeff Schewe's example last year some time when he posted it on another thread and it's quite illuminating.  There was also a discussion about Dynamic Range (DR) at that time and what the effective range of various DSLRs is.  We all have different equipment based on the type of work we do and our own personal budgets.  That being said there is a lot we can test ourselves to find out the limits of what we can and cannot do.  After the DR discussion I prepared the attached test shot (really cheap as you can see; mounted my Passport and a test shot to a piece of matboard and duck taped it to a brick wall outside my house; you can see the bricks on the left edge which provides a good color check for the shot).  I made a whole series of images based on the original meter reading and then going through various over and under exposures.  This gives a good range from which you can then see how LR/ACR works in terms of being able to recover detail and colors with associated noise.  I found this quite informative and it gave me a good feeling for how to implement ETTR in the field.  It took all of 30 minutes or so to prepare the chart and capture all the images (I also shot it early in the day when the sun was on the other side of the house so that there was even illumination on the chart).

[michael]

Does this mean that every image I take I should have my EV set at plus 1.0 ( or similar ) regardless of the brightness - or lack of it - in a scene and if it is looking pushed to the right in ACR  - without it being overexposed - and then simply reducing the exposure till the histogram has an even distribution of data?

Sort of.

You may want to increase exposure up to about 3 stops, or not at all. It all depends on the scene. The important point is to place as much of the data to the right of the histogram, but without clipping any important highlights. The problem (until manufacturers automate this) is that the histogram on your LCD is not based on the real raw data in its very large colour space. This means it's going to yell "clipping" when you still have extra headroom available.

For this reason it's worth doing some testing to see what your particular camera's characteristics are. Keep adding a third of a stop well into clipping (on the LCD histogram) and then look at the files in your raw converter and see when they really do clip. This will tell you how much extra headroom you have.

You can also set your camera to autobracket.

Michael

[stamper]

Sort of.

You may want to increase exposure up to about 3 stops, or not at all. It all depends on the scene. The important point is to place as much of the data to the right of the histogram, but without clipping any important highlights. The problem (until manufacturers automate this) is that the histogram on your LCD is not based on the real raw data in its very large colour space. This means it's going to yell "clipping" when you still have extra headroom available.

For this reason it's worth doing some testing to see what your particular camera's characteristics are. Keep adding a third of a stop well into clipping (on the LCD histogram) and then look at the files in your raw converter and see when they really do clip. This will tell you how much extra headroom you have.

You can also set your camera to autobracket.

Michael

Thanks for the feedback. My primary camera is a Nikon d700 and according to DxO I can raise my EV by as much as 3 stops without clipping. I generally use 2 stops to be on the safe side. I have been trying to sometimes expose for dark areas of an image and lowering the EV by 2 stops to prevent blowout of highlights. It was giving me an underexposed image. Now I think I know why after reading your article. Much food for thought. I have used the Uni WB in the past so I am aware of the limitations of the Jpeg histogram. The raising of EV for dark areas is something to try.

[hjulenissen]

Hi,

Agree with the conclusion, we should to expose to the right and the cameras should support this.

On the other hand I´d say that Michaels explanation of why we should expose to the right is lacking. The reason to expose to the right is mainly that we want maximize signal (which is the number of photons detected). There is no simpler explanation, and it also happens to be the correct one.

Best regards
Erik

When audio recording people went from analog to digital, they went through the same troubles. While analog gear tends to have a somewhat "soft" clipping characteristic (increasing gain will increase distortion but decrease noise), digital tends to have a brick-wall. The solution is to record as "hot" as one dears (a little clipping can be disasterous), and to use components that offers a large capture DR if needed.

I am not sure that I understand why ETTR is discussed so much. I try to get good exposures in camera, but will change according to taste in my editor. For controllable, slowly changing scenes, I think this is a non-issue (at least for me and my quality expectations). For action-scenes where the light is out of my control, I have more issues with finding good aperture and exposure time settings in isolation (getting suitable DOF and motion blur), than with gross exposure errors. Of course, when margins are really low (little light, much movement, large desired DOF, flash unavailable), missing by 1 stop can detract from the usability of the image.

-h

[Robert-Peter Westphal]

Hello,

I don't understand the complete discussion, because what Michael states is not new, but absolutely right !

It is common knowledge that the way a sensor reacts to light is complete different than a chemical film does. So it makes no sense to keep on measuring and interpreting light in the way it was invented for chemical film.
Furthermore, it is also common knowledge that a sensor captures color-depths of 12 to 16 bit. So it makes no sense to show exposure in a heavy reduced range of 8-bit by first creating a jpg.

Perhpas I'm a little simple-hearted, I don't own any Phd in physics, chemestry, or electrical engineering, so for me all said by Michael makes sense and equates my personal experiences.

Best wishes

Robert

[dreed]

It always amazes / annoys / amuses me when I publish an article that has been painstaking peer-reviewed by some of the brightest minds in the industry – people who design sensors and write raw software (in this instances) and then "experts" whose credentials are unknown tell me (us) why the information in the article is wrong.

This was the case with my original ETTR article. I think I trust Thomas Knoll's knowledge of digital imaging (the original author of Photoshop and Camera Raw) over some online commentator.

Similarly in this instances. I'm sure that there are areas to quibble over (both scientists and artists love to quibble). But when it comes to arcania and subtleties in complex topics I'll continue to trust known experts whose bona fides are well established.

Michael

If there's one thing that I always enjoy about articles you post it's that you're always involved in discussion about it afterwards. It would be nice if more of your contributors were involved in discussions about their articles - particularly in this forum. Sometimes there's not much to say or be said, but others, yes.

Thanks for being involved.

Darren

[NikoJorj]

Michael's article discusses solely automatic metering based on ETTR, but is there not also a manual approach of using the Zone system at arriving at an optimized exposure (one shot) at RAW (ETTR)?

Yes, but if Live View can reliably automate this, why not profit?
There are indeed situations where you have to shoot rather quickly, because of some photographic opportunity passing by (a beam of light on the right spot, the head of the giraffe balancing nicely with the iceberg forms, or you name it), and in those cases I'd really like my camera not to pretend it is loaded with negative film - which seems the hypothesis of my canon DRebel matrix metering, leading generally to burnt-beyond-any-recovery skies eg.

And yes indeed, raw histograms and flashing highlights based on those raw values are needed as XXIth century tools as well.
Adjusting a proper exposure looking at the jpeg file or its values may be a bit like correctly exposing the polaroid positive hoping it will suit the negative part as well.
Because as good the raw converter is, when the raw data is clipped there isn't any info anymore and all that can be done is guesswork (but yes, some converters may guess better than others).

Does this discussion apply to the electronic viewfinder cameras?

Yes, because the sensor works the same way, no matter the viewfinder.
What may be a catch is that you'll see the tonality of the jpeg file in the viewfinder, making it counterintuitive to apply the right setting : a too contrasty scene will appear too dark and a low contrast scene may appear too bright (but as a side note, I would think that low contrast scenes don't benefit as much from ETTR as high contrast ones, because SNR is not a very constraining factor for those).

[hjulenissen]

Furthermore, it is also common knowledge that a sensor captures color-depths of 12 to 16 bit.

Although the bits "are there", I think it is common knowledge that the least significant ones tend to be noise (random), and the minute steps at the upper part of the range is largely irrelevant (unless you are going to modify the tonal range a lot).

So it makes no sense to show exposure in a heavy reduced range of 8-bit by first creating a jpg.

Having a way to observe the exposure in-camera that was directly based on the raw file would be great. Now, if it was binned in 256 or 1024 bins probably would not matter much to me - I tend to only care about the head, tail and body of the beast, not the minute scales along its back.

[digitaldog]

The problem (until manufacturers automate this) is that the histogram on your LCD is not based on the real raw data in its very large colour space. This means it's going to yell "clipping" when you still have extra headroom available.

Exactly! The main problem is folks think exposing for a JPEG where the rendering is totally out of their control, and based on this incorrect clipping information applies to the raw data. They think if they see clipping on the LCD, it affects the raw data. Yup, you’ll clip that JPEG and that clipping info is useful if you shot only JPEG!

Expose using a meter (based on film with its H&D curve) and the feedback for JPEG on the camera, you’ll very likely be under exposing the linear raw data. Expose properly (idealized) for the raw based on what you intend to render (the intend here does have a bit of Adam’s thinking in the mix), you’ll very likely blow out the JPEG.

A photographer would never treat (expose and develop) ISO 400 transparency film the same as ISO 100 color neg film. Why should they treat exposure and development of JPEG and raw the same way, considering the vast differences in the data and processing? How do we get the manufacturers to give us a LCD feedback based on the actual data we want to capture?

[dreed]

Sort of.

You may want to increase exposure up to about 3 stops, or not at all. It all depends on the scene.

That last sentence really is the key.

My concern is that if ETTR became the default then it would make it harder for a lot of people to shoot "sharp" pictures because ETTR needs more light and that either means slower shutter, wider aperature or higher ISO - none of which are recognised as being better for pixel peeping with. Whilst *I* wouldn't mind an "ETTR mode" rather than a exposure compensation mode, I just can't see that selling cameras that random folks in the street use because it'll be harder for them to get good JPG pictures compared to the other manufacturers' cameras that don't.

The important point is to place as much of the data to the right of the histogram, but without clipping any important highlights. The problem (until manufacturers automate this) is that the histogram on your LCD is not based on the real raw data in its very large colour space. This means it's going to yell "clipping" when you still have extra headroom available.

Something that I've found useful in judging whether something is blown or not is to (temporarily) tune down the exposure in LR to -4. It then becomes quite obvious if there are colour channels that are "blown". Unfortunately I don't know of any way to do that "in camera" :-(

In retrospect, something that might be missing from that article was advice on how to better setup the camera to give more "accurate(?)" histograms - or would that be too camera specific? I think the advice I've used for Canon cameras is to turn the contrast right down. This seems to have given me a more accurate relationship with LR.

[digitaldog]

That last sentence really is the key.

My concern is that if ETTR became the default then it would make it harder for a lot of people to shoot "sharp" pictures because ETTR needs more light and that either means slower shutter, wider aperature or higher ISO - none of which are recognised as being better for pixel peeping with.

Well yes but that’s a lot like saying if you shoot film, and you only have ISO 100 film in the camera, your concern is that if ISO100 became the default then it would make it harder for a lot of people to shoot "sharp" pictures because that film stock needs more light and that either means slower shutter, wider aperature or higher ISO...

That you set your camera at F8@250th shot a JPEG and find that you’ve under exposed the raw just means that exposure is incorrect for raw, you need to run a test to find the actual ISO and exposure settings that produce the correct raw exposure.

In the old days, it was always interesting to run film ISO and exposure tests on color negs. It was a wake up call to find that exposing ISO 400 neg film at ISO 100 produced a vastly superior print (when of course you compensated for the exposure in the enlarger for that process).

[dreed]

Well yes but that’s a lot like saying if you shoot film, and you only have ISO 100 film in the camera, your concern is that if ISO100 became the default then it would make it harder for a lot of people to shoot "sharp" pictures because that film stock needs more light and that either means slower shutter, wider aperature or higher ISO...

And you're quite right about that. There were many times that I was on holiday and found myself with ISO 100 film in the camera and faced with 1/10th to get the shot (this was before lenses had IS) so I had to invent ways to "mount" the camera to keep it still or (more often than not), blurry picture. Sometimes I had ISO 200 loaded but rarely more than that. On one occasion I had two cameras - one that I kept loaded with ISO 800/1600 and the other with either 100 or 200. It was a trifle cumbersome. And that was just to get a 19th century film shot that wasn't blurry.

[Phinius]

One thing that should have been mentioned is that the histogram is generated from the JPEG data, and therefore the camera settings for JPEG need to be as flat as possible to have the histogram come as close to RAW as possible. Why don't manufacturers just base the histogram on RAW data as this would also preclude the false impression of clipping Michael mentions in his article?

Ron Johnson

[rmyers]

Has anyone used this or something like it.  It says you can shoot a series of exposures to determine the DR of your camera.  If you knew that, couldn't you exposure compensate more accurately?  Wouldn't it even calculate the exposure settings for you?



http://www.sekonic.com/Products/L-758DR/Overview.aspx

[John R Smith]

Something that nobody has mentioned -

Even more ridiculous is that an MF digital back like my Hassy one can only shoot in RAW, but I believe the histogram you see on the LCD is derived not from the RAW but from the JPEG preview. How dumb is that?

John

[madmanchan]

The default setting in ACR isn't linear. There is a contrast boost. If what you state is desirable then the settings should be zeroed out? Adobe should zero the settings as a starting point rather than a contrast boost. At lot of users don't know about this.

stamper, yes, there are contrast curves applied by default (*), but those curves are applied after the lightness-dependent portion of the color profile.  The Exposure control, on the other hand, is applied before the lightness-dependent portion of the color profile.  This means that if you've used ETTR in the camera or bracketed exposures in the camera, and want to normalize the results afterwards, you should use the Exposure slider because that will get the image values in the appropriate tonal range before the 3D color table is applied.

Eric

(*)  Zeroing out the curves as a starting point doesn't make sense for the vast majority of users.  Having zero curves would mean a linear scene-referred rendering, which looks ugly.  The majority of captured scenes need to be tone-mapped and color-mapped to look acceptable in reproduction (e.g., on a display), and that's what the curves accomplish.

[fdisilvestro]

In cases of extreme white balance correction, something similar to ETTR (color channel specific) can be achieved by using color compensating filters instead of white balance correction.

Typical examples are shots under incandescent lights, where the blue channel is usually 2-3 stops underexposed compared to the red channel.

[Alan Goldhammer]

Isn't the simplest solution to this conundrum two sets of firmware, one for the users who need to have the current jpg emulation since they lack the sophistication to delve deep into how ETTR works and maximizing the image and a second for the professional user who wants these options in the camera menu?  The only problem with camera companies doing this is that the professional base is always going to be significantly smaller than the casual user.  I cannot believe the amount of programming to achieve this would be that significant (but then I'm not a software engineer).  At least this approach would let both parties have their cake to eat.

[michael]

No need for separate firmware for pros. Just have it as a custom function, or selectable from a menu.

Michael