Expose to the right, ETTR - with E-P3

This blog is really about my experience with the Olympus E-P3 camera (and to some extent E-P2 before it). This is also about the advantage, that is obtained when shooting in RAW format, and always exposing as much as possible. The expression "as much as possible" does not mean here overexposing. What, then, is overexposure?

Gamma

Because of this issue of exposure I´m not going directly to the E-P3, but we need to discuss a few things. First a look at the the crucial difference between the RAW file and the film. Ignoring this difference leads into blurred concept of overexposure or correct exposure. The difference is gamma curve.

In practice, in almost all digital cameras, AE follows the old film era definition. It is based on the average reflectivity of the object. The camera may be set to expose by either 12% or 18% reflectivity. Film has a gamma curve, which is kind of a ski hill-shaped curve. On the top it is bright and curve is gentle, the middle is steep, even slide and down hill evens out into shadows. Most pictures appear to be good, when its central tones have a good contrast. Dark colors must not suddenly change into black and light shades must not go into white all of a sudden. Such sharp edges do not seem natural. Therefore, tonal changes are slowed down at each end with gentle arc. For film photographers it is most important to use the contrasty middle tonal range as completely as possible. According to this need for visual naturality different types of film require somewhat different exposure. Slide films need less exposure to prevent highlights from becoming white (to be overexposed), and negative films must not be exposed too thin and prints appearing too dark in shadow areas (would be underexposure).

For the majority of subjects, all goes well, if it is exposed according to the camera's meter. Only experience tells when, in what direction and how much it is necessary to deviate from the camera´s value.

RAW

RAW file is different from the film in that it does not have the same gamma curve, it is linear. From white to black, there goes a straight line with no curving at either ends. The requirement of naturalness (to the human eye) in the image, is carried out in the RAW conversion when RAW file is converted into pixel file and saved for example as TIFF or JPG format. Then the final image gets a gamma curve, and this curve is adjustable by the user in every good converter. However, this curve is not always visible as itself in most converters, only the effect in the image is visible.

Now, one could ask where's the difference. The subject is just exposed at the middle (or somewhere there) of that straight line, gamma curve is included in converter and all is well. Actually that is just what most photographers are doing - and at the same time they are wasting a significant part of tones from their images.

Tonal values in a RAW-file

In the case of RAW file we need to discuss bits in images. Bits tell us into how many tiny levels our direct line is divided. And it must be divided, otherwise we would not have a digital file. The overall impression is still a straight line, because there are so many levels. Typically, a modern digital camera has a 12-bit (per channel) converter, which converts the sensor's analog (linear) signal into digital (small steps in straight line). The best cameras can have a 14- or 16-bit converter. This does not change the issue at all, the same scale from white to black is only divided into larger number of smaller steps.

A 12-bit image consists of 2 ^12 = 4096 possile steps or tonal values from white to black. At this point, we unfortunately must say goodbye to the linearity of tonal distribution: tone values are not evenly distributed into bits. Starting from white towards black, half of the ALL tonal values are inside the first bit. For simplicity, we can speak in photographic terms, ie steps in aperture or shutter speed scale: One full step in opening the aperture (eg, F5.6 -> F4) or an equivalent increase in exposure time (eg, 1/60s -> 1/30s) increases the amount of light to double. Then the first step starting from pure white includes half of all possible tones in a RAW file! In a 12-bit digital camera, the first step starting from white has thus 2048 possible tonal values (per channel), the next step has 1024 possible tonal values, again the next step has 512 possible tonal values, and so on.

Now you may (hopefully) no longer think that it is indifferent to where on this scale of tonal distribution your RAW file is exposed, if you want the best possible tonality in your image!!!

Above we have two RAW images as they are opened in Lightroom 3. The upper one is exposed according to what camera meter (AE) says and the second is exposed to the right. The change in exposure is +2 steps. Below images we have their histograms in Lightroom with default settings, left camera AE and right my ETTR version.

 Next pair of images has the same images normalized. Mainly I clicked Auto plus changed slighty sliders. Again above is camera´s version and below ETTR version. Histograms also in the same order as above. The most obvious differencies are 1,8 steps more exposure in camera´s version and  more blacks for ETTR version. There was no need for any change in exposure for ETTR image. A quick glance reveals not much difference in images, but they have. The real clou is seen in the next pair of images.

The clou is the amount of detail and noise. These 100% crops are in the same order as above, and neither has any luminance noise reduction. ETTR is one of the most efficient ways of minimizing noise in shadows and elsewhere. Note also how two stops more exposure did not diminish detail in light ares, it only gives more detail as s/n ratio gets better.  

Histogram and exposure

The histogram reflects the distribution of tones in the image. The histogram has white on the right side and black on the left. Therefore, we can say that in the histogram the first step from the RIGHT has half of the possible tonal values of the picture.

What does a histogram in a digital camera show? It does not show the tones of a RAW image, but tones of a JPG preview calculated from a RAW image. These tonal values are dependent on the camera's contrast and saturation settings. The histogram might show over- or underexposure (or both), even though the sensor itself would have plenty of exposure latitude left. In addition, the histogram is affected by the selected color space. In particular, the sRGB color space is narrower than the sensor color space. When it is selected, the RGB histogram displays too often saturated colors that do not actually exist in the RAW file.

Worse still, the histogram is not even a good ETTR-exposure meter, because it reflects the image globally and not its individual tone areas. It does not directly tell what is the exact area in the image, which is now going blocked. ETTR-exposure requires the use of each pixel in the image area as a separate light meter and display.

Finally, the worst of all: a traditional reflex camera´s exposure metering (prior to taking a photo) is even more problematic in terms of shooting in RAW. It does not understand anything about the behavior of the sensor. They are two completely different things.

Expose To The Right, ETTR

As much as three quarters of possible tonal values of a RAW file are in the first two steps from the right. This should be a sufficient reason for every RAW photographer, to expose as to the right as possible, without burning the light tones of the image, of course. This method deviates from the traditional concept of "correct" exposure. ETTR RAW file can appeear be too light when opened into a converter. The converter is used to extend the tonal range into dark direction until the tonal range is as desired. In most cases, successful exposure means that the brightest tones of the image do not need to be adjusted at all. The resulting  image has richer tones and is even significantly cleaner regarding noise.

ETTR has been also criticized or denigrated in online blogs. The problem is not the idea itself, but the cameras. Only few of the cameras provide photographers with a good tools for reliable control of ETTR. Without control, there is danger of overexposing the sensor, and the image is beyond repair. ETTR does not always mean more generous exposure than the camera's exposure meter indicates. Exposure may be also less when the subject has lots of contrast, but even then the exposure goes for the lightest tones.

ETTR means that you should set the ”weakest" settings that can be found in your camera: lowest contrast and saturation as well as a widest color space. With Live View cameras exposure control with sensor feedback is possible before exposure. With reflex cameras exposure can be seen only after the exposure. The quality of information received depends on how well or poorly the camera settings in imitate RAW files. In addition, all the automatic adjustments (Highlight, Shadow ...) which shape the gamma curve must be switched off.

The second ETTR example is again exposed for the lightest area in image and that is the lightest details of relief. We have now almost a one step increase over camera´s meter. Histogram and sliders are here at Lightroom´s default. Lightroom´s warning colors show ares where at least one RGB channel is blocked at Lightroom´s default gamma (2.2). 

This before/after pair shows details from the lightest are in the image. On the left are default settings and on the right you can see how two tiny moves in sliders fix the toe and shoulder of gamma curve so that nothing is blocked. ETTR does not lead into more work at post, vice versa

Shooting JPGs

As a clarification: ETTR is not applicable when shooting JPGs. It is not possible to ETTR a JPG file and darken afterwards unless subject contrast is so limited that it can be included in the straight part of gamma curve. Even then, most likely, light tones show banding

ETTR for me is always observing the lightest non-specular tones in image. Specular highlights and reflections may and should go white to give the image the right tonality and contrast. In this example ETTR has led into one step "over exposure" compared to camera AE. In reality this exposure is correct and the engineer who designed my camera´s AE has it wrong. It is a matter of taste where is the limit or start of specularity. It can be tweaked with a tiny change in Lightroom, if needed. Otherwise the image is OK.

Olympus E-P3 and ETTR

So, using ETTR might lead into struggle with inferior equipment. I learned ETTR however during my profession as advertising photographer in the late 90's. The tools then were Leaf three shot digital backs and Leaf software. With them and studio flash equipment it was possible to set the exposure at 1/20-stop precision exactly where it was supposed to. Later PhaseOne CaptureOne allowed the same in other cameras. But doing so demands shooting tethered.

Beyond these devices only the Olympus VF-2 viewfinder gave me a new hope toward guaranteed ETTR shooting anywhere, anytime. With E-P2 I learned to observe the viewfinder image and see when exposure at highlights was as wanted. It demanded these camera settings:

Shooting Menu 1 > Picture Mode > 4 Muted > Contrast -2, Saturation, -2, gradation: Normal

Custom G > Live View Boost: Off, Color Space: Adobe RGB

Setup Menu > VF-2 Brightness: Experimenting until I learned the appropriate setting. After that Custom D > Info Settings > Histogram: Off

This was just a start, because human eye is really not a good instrument, it is affected by the overall brightness of the scene. I had yet to check the captured image in the VF-2.

The Olympus E-P3 finally fixed these deficiencies, because it has an adjustable highlight and shadow warning by colors in viewfinder image, that is, before the exposure. This control is set at:

Custom D > Histogram Settings > Highlight: 245

Auto white balance must be set to neutral:

Custom G > Auto WB, Keep Warm Color: Off

Now, I adjust the exposure until the red warning color just disappears from the critical tone area. In principle, this allows a 1/6-stop accuracy at best, but in practice, the reference tone is very close to the value of 96% when the image opens in Lightroom. Same area will be at RGB value 245 in the final image with very little need to adjust as seen above. I do not need to use histogram, I do not need to check captured images, at least not because of the exposure.

I do not know of any other camera body, which gives me such a perfect exposure control with the same precision and certainty. E-P3 is my ETTR King!

-p-

Comments

[ithinkthatsme]

Hi, i have seen the link you posted on the dpreview forum and followed it. Nice article really helpful. I should start experimenting with that. the noise difference between the two shots you showed is quite significant. keep on and thanks for sharing your thoughts.

[Pekka Potka]

Hi, thanks. Yes, the advantage in signal/noise ratio can be huge. Practically noise free ISO 1600 without NR is worth some experimenting... ;-)
-p-

[mie]

You mention about ETTR using the highlight red colour warning in E-P3. The mistake was the function included for highlight warning in E-P3 might only applies to the JPG file not the RAW files, IMHO. The camera was showing the theoretical images if you apply the exposure value with the JPG setting you have set earlier such as contrast, saturation and etc hence probably the ETTR might not be accurate even if you use the histogram feature.

[Pekka Potka]

Mie: You are, of course, right about the highlight warning showing the values of JPG preview image. But it is not a mistake, it´s the only way to do it, it´s all we have until something else is figured out by Olympus or other camera makers. With mentioned settings this warning has been moved as far right as possible in E-P3. For the wast majority of images it is then set quite perfectly for Lightroom conversion. There is of course some leeway left. By lessening contrast and moving exposure and brightness sliders to the left in Lightroom, you could expose still further right. I have not found that to be a benefial starting point in practice.
-p-

[pdh]

Thanks for a straightforward restatement of this technique.
I'd like it if you could clarify the settings you used on the E-P2 - in particular "Keep Warm Color: Off" - as I can't find this in the menus.
Is this actually a setting on the E-P3 rather than the E-P2?

[Pekka Potka]

pdh: My mistake. Thanks for asking for clarification. This strange feature came with E-P3. I have now corrected it in text also. Sorry for making you do my work...
These Olympus menus can be confusing every now and then. But on the other hand I really like the possibilities they give.
-p-

[d]

I may be missing something, but why don't you just take the photo at two stops lower ISO? What gain do you have, shooting at 1600 ISO, when the shutter speed is cancelled out by the +2EV?

[Pekka Potka]

d: I guess you missed the whole point. ISO or shutter speed is not important here. The benefit of ETTR can been seen at any ISO, any shutter speed.
-p-

[Pat Donnelly]

Very clearly explained!

Thanks!

[Tom Hearne]

Just to clarify: on the E-P3, are the settings noted for the E-P2 still needed? That is, do you still need to set Muted, Contrast -2, Saturation -2, etc?

Also, I use Aperture rather than Lightroom and apply a preset to most RAW imports that balances the histogram, boosts contrast, etc. Do you do something similar in Lightroom - or adjust all photos manually?

BTW, thanks for your many insightful reviews....

[Pekka Potka]

Tom: Yes, those E-P2 settings are still needed for E-P3 as basis. Many times I apply a "normalization" preset similar to yours to a set of images because it makes it easier to visually select images. Beyond that, I adjust images individually when working with the best images.
-p-

[Tom Hearne]

Hi Pekka,
I've done some testing to see how ETTR compared to 'normal' metering with my E-P3.

I set up a tripod and took a series of shots at iso200, 800, 1600, 3200 and 6400 - all at f2.5, allowing the shutter speed to vary. In each case using the parameters above. For each iso I took two shots: one using average metering, the second ETTR. For my scene the difference was 1EV.

The ETTR shot was consistently much better than the average metering shot, as expected. Comparing the ETTR shot at each iso to the average metered shot at the next lower iso (which means that the two shutter speeds were the same) yielded results that were very hard to distinguish. I generally preferred the lower iso shot. This surprised me. I expected the opposite: that the higher iso at ETTR exposure would have more information, less noise, etc.

What am I missing?

[Pekka Potka]

Tom:
1. A sensor has only one real ISO speed. All the other values are electronically amplified from a lesser signal.
2. The sensor in E-P3 is good only up to ISO 1600, IMO. Beyond that the quality deteriorates very fast.
3. When you use tripod, your ISO should be the lowest possible, which here looks like to be 200.

So, ETTR improves the quality when all the other factors are equal. And changing the other factors may negate the benefits.
-p-

[Assaf T]

Hi Pekka,
Thanks for the info!
I experimented with the ETTR method recently and got very nice results. This is something from 2 weeks ago:
https://fbcdn-sphotos-a.akamaihd.net/hphotos-ak-ash4/414405_10150597063686424_535606423_11114462_163723340_o.jpg

Just one question - why do you prefer using Highlight: 245 over 255?
As far as I understand, by reducing the value to 245, more red areas will appear although no clipping occurs for this data in the RAW/.
I guess I missed something, because it seems to me that this will make the highlights display mode even less reliable as a measure for RAW clipping than it is.
This is what I found in the manual, "areas above the upper limit of brightness for the image are shown in red". So if for 255 it shows certain areas in red (i.e. clipped for a jpeg output) while these areas might be OK in the RAW, reducing the value to 245 will make the false alarms even more frequent.

Could you please explain this point?

Thank you!
Assaf

[Pekka Potka]

Hi Assaf, nice image, great feel of space and atmosphere!

From technical point of view you are absolutely right. You should go as far right as ever possible without saturating the sensels (as they are read into RAW values).

But I´m a photographer and photography for me is a visual art. I think about the image, about the tonal values in FINAL image. I have a long career as an advertising photographer behind me and that taught me to always think about 245. As a rule of thumb: RGB tone 245 corresponds to the brightest possible tonal value in printing, above which everything will be just paper white. In high quality ink jet printing you can manage to go beyond that but not really in offset printing. That´s why I want to set my exposure to where I want the brightest tone in my image to be. Everything brighter, maybe specular highlights, reflections, light bulbs etc. will go where ever they want to, and they blink red in viewfinder (and LR3). Of course I can expose for some tones in reflections or whatever if I want to do, still the principle is that I set MY brightest tone.

I have found that coincidentally EP-3 245 (with all the other settings) corresponds quite nicely to LR3´s tone 96% (at LR default settings) which corresponds quite nicely to AdobeRGB 245. E-P3 245 with LR3 default settings also gives you some needed leeway in Lightroom, because many times situations just happen too fast to let you set exposure absolutely right. So, my aim is just to turn my visual idea into the best image with minimum fuss.

-p-

[Assaf T]

Thanks for the explanation!

[ZoranC]

If one can afford to drop shutter speed (one shot was at 1/50 ISO 1600 0EV, other at 1/15 ISO 1600 2EV) why not just drop ISO and have shot at 1/15 ISO 400 0EV? That should should have been also included in comparison.

[Pekka Potka]

ZoranC: Why? I guess everyone knows that ISO 400 gives better result than ISO 1600. Please, see my answer to Tom above. This pair of images is just an example to show the benefits of ETTR in SNR and detail at any ISO. Food for thought... Seems like I should have included a comparison pair at ISO 200 where you can´t go any lower (with E-P3). It is very easy to see how the noise in e.g. blue sky drops when you ETTR at base ISO and with any camera. Try and see it for yourself!
-p-

[ZoranC]

Pekka Potka: Re "why?", answer is: If goal is to get best image quality then one needs to show results of all paths leading to it and why one presented here would be best and in which scenarios. If example is a bad example then that bad example might mislead people into doing ETTR when they should have just stayed at 0EV and dropped ISO.

[Pekka Potka]

The goal in that example was not to get THE best image quality of that subject. If I were to show THE best path leading into THE best image quality of that subject, I would not have used E-P3, I would not have shot handheld and I would have used studio flashes. But no, I was not doing that. ZoranC, you are free to think what ever you like about the concept of ETTR and now your opinion has been heard. OK?
-p-

[Brian Mosley]

Pekka, thank you so much for putting this very clear tutorial together - I really appreciate it, and can confirm that it is a very practical way to improve exposure success.

I only hope that Olympus make flashing highlights available in every info view - it's a bit irritating having to cycle through 3 views to get back to the screen which displays flashing highlights when I also use the electronic levels to compose landscapes.

I'll continue to recommend that people read your article for excellent advice on practical ETTR.

Cheers

Brian

[Robin Lee]

Thanks Pekka. Great read.... :)