WAYS TO ACHIEVE ETTR
There are several ways to achieve ETTR, which are described here. Not all are available for every camera, but at least one of the methods will apply to just about any camera.
It is important to understand that ETTR is always enacted at base ISO. The reasons for this are explained in detail in the Technical Underpinnings addendum found on page 2 of this article. Essentially, it is only at base ISO that the camera's exposure indictors can properly indicate ETTR. Of course, it is not always possible to achieve ETTR, and when this is the case, ISO may beneficially be added to the equation when shooting jpeg or when shooting raw with certain cameras. This situation is dealt with in the final section, but it is not the situation that pertains otherwise in this article. So, I repeat, ISO is not relevant to establishing ETTR, and for this purpose, ISO should be set to its base value.**
Any metering mode can be used in setting ETTR (with the one exception described below that requires spot metering). I personally prefer using center-weighted metering since it provides information that is less sensitive to small movements in the camera than either evaluative or spot metering and is more consistent while various camera adjustments are being made.
Your actual meter readings are irrelevant to ETTR. Instead you will be using the camera's histograms and/or blinkies to establish exposure – either live-view (LV) histogram/blinikes or post-shot histograms/blinkies or both.
Using Live-View Histogram or Blinkies
If your camera has live-view histograms or blinkies, these indicators provide the simplest means for setting ETTR – or at least an excellent first approximation. Let us first suppose you are using A-priority (generalizations to other modes are straightforward). Set the camera to display the LV histogram. Set the f-ratio according to your DOF requirements, and then frame the scene. While keeping the camera on the scene, adjust EC (Exposure Compensation) up or down as needed to place the top tail of your histogram just up against the right-hand limit. I'm assuming now that the resulting shutter-speed is acceptable (I'll deal with the case where it is not in the final section). Shoot. You have now accomplished ETTR. Don't worry about the brightness of the resulting image; you've got a proper exposure and you can fix the brightness later in processing or post-processing.
If your camera has blinkies (colored areas, usually red or white, that indicate over-exposure), you may find them to be a better means for determining ETTR than histograms. Sometimes there are inconsequential highlights or specular highlights that you might be perfectly willing to sacrifice (blow) to obtain a slightly higher overall exposure. ETTR, viewed most generally, assumes that you expose to place the brightest values you wish to retain just below the point of sensor saturation. The histogram, of course, cannot tell what parts of the image are blown, but the blinkies can. The blinkies tend also to be a more sensitive, as well as a more selective, indicator of over-exposure and, hence, are often better for determining ETTR.
The process for using blinkies is similar to that for using histograms described above, except now the camera is set to display the LV highlight blinkies. While viewing the scene, you adjust EC till the highlight blinkies just begin or just end. If there are highlights you wish purposefully to over-expose, simply adjust EC till the blinkies in the brightest regions you wish to retain just begin or end, ignoring their presence elsewhere.
If you're using S-priority, the process is essentially the same, except first set the desired shutter-speed. Then use up or down EC as needed to adjust the f-ratio so that either the histogram kisses its right-hand edge or the pertinent blinkies just begin or end.
With M mode, you adjust either or both f-ratio or shutter-speed (making sure to keep them within whatever bounds you need for DOF or camera-shake/motion-blur) to place the histogram at its right-hand edge or to just acquire the pertinent blinkies.
One refinement is in order: The live-view histograms and blinkies are often not as accurate as the post-shot histograms and blinkies. The live-view versions are based on a quickie jpeg made on the fly, whereas the post-shot histograms and blinkies are based on a more processed jpeg. So, after taking the shot, it is worthwhile checking either the post-shot histogram or blinkies to see if they are telling a consistent story. If they indicate clipping (either a histogram that is piled-up on the right-hand edge rather than just kissing it or the presence of undesirable blinkies), then – if it remains possible – retake the shot with 1/3 to 2/3 less EC and recheck the post-shot indicator. Iterate as needed – usually one retake is enough, although with the right camera settings and a bit of experience, you will find that you hit ETTR with the first shot more often than not.
Another advantage of reassessing ETTR with the post-shot indicators is that many cameras display color histograms after the shot, whereas LV may only display the luminance histogram. You are therefore able to assess whether any of the color channels has individually gone beyond ETTR and retake the shot as may be necessary. See, however, below on the effects of WB on the post-shot color histograms.
When using the LV blinkies as your indicator, whether to stop when the pertinent blinkies first display or just before depends on the camera and its settings (dealt with below). The post-shot blinkies will help you to determine which is right for you. If the post-shot blinkies frequently indicate you've gone too far with the LV blinkies, make the obvious operational adjustment.
Using Post-Shot Histogram or Blinkies.
If your camera lacks LV histograms and blinkies but has them as post-shot indicators (all serious cameras do as well as many others), then an iterative procedure can home in on ETTR quite quickly, provided, of course, you have the opportunity to take the shot more than once. The post-shot histograms tend to be best for this use.
Again assume A-priority. Set the desired f-ratio, compose the scene, and dial EC to achieve what appears to be a good exposure, say, based on the VF or LCD image. Take a shot (at base ISO, of course). Check the post-shot histograms to gain an idea of how much EC is needed (up or down) to bring the exposure to ETTR. A little experience may be needed to be able to judge how much adjustment is needed, but it is quickly learned. Make the EC adjustment and retake the shot. Re-check the post-shot histogram and do further adjustments as needed until you get a shot that is properly ETTR. Typically at most one retake is needed, rarely more than two.
When setting the initial exposure, it tends to be better to err on the side of too little than too much since it is easier to assess the needed adjustment from the size of a gap on the right-hand side of the histogram than from a pile-up on the edge.
The modifications for this procedure when using S-priortiy or M-mode are straightforward and completely analogous to those described in the preceding section.
If you wish purposefully to blow some highlights, you will want to consult the post-shot blinkies as well to determine just what is and what is not being over-exposed.
As a practical hint, when using this method it is wise to note just where the initial focus point is situated in the scene so that you can reframe quickly and accurately on any retakes.
Spot metering the brightest part.
While the preceding method is effective if you are able to retake the shot one or more times, it is less systematic than may be desirable. There is another (non-LV) method for ETTR that has a far better chance of getting it right the first time. This process makes use of the spot-metering mode to lock the exposure on the brightest part of the scene. When using this method, then, it is good to have your camera's exposure-lock function assigned to a convenient function button and the spot-metering area set as narrow as possible.
If there are highlights that you purposely wish to over-expose, interpret the phrase "the brightest part of the scene" to mean the brightest part you wish to retain.
Choose either A- or S-priorty and set the desired f-ratio or shutter speed. Set the metering mode to spot. Set EC initially to zero. Frame the scene and note two things: first, the focus point (so you can return to it) and, second, the part of the scene that is brightest. Move the camera so the spot metering area is in the brightest spot and lock the exposure. Now dial in 2-1/2 to 3 EV of EC. Recompose the scene (returning to the original focus point), and take the shot.
In M-mode simply spot-meter the brightest part (adjust f-ratio and shutter speed to zero the meter) and then add 3 EV of exposure though any combination of f-ratio and shutter speed. Recompose the scene and shoot.
This process will either give you ETTR or something very close. The reason it works is that, when EC is at zero, the camera's spot metering will attempt to make whatever is in the spot-metering area "middle gray." But, since you've metered the brightest part of the scene, the brightest part now becomes locked at middle gray (and the histogram will top-out in the middle of its range). Clearly, were you to take the picture with this exposure, it would be very dark. But the camera has upwards of 3 EV of highlight headroom; that is, the sensor can accommodate luminance that exceeds middle gray by upwards of 3 EV. Thus, by dialing in 3 EV of EC, middle gray (the brightest part of the scene) will be boosted to the top of the sensor's range and, as a result, the histogram will top-out at its right-hand edge. Voila! ETTR.
The amount of highlight headroom varies from camera to camera, and you will have to experiment with your camera to determine whether you have 2-2/3, 3, or even more EV of headroom. You can determine this amount quite accurately by using software like RawDigger, which allows you to see the actual raw histograms for your raw files. In M-mode at base ISO, take a raw shot of an evenly lit, solid-gray wall or other such target (fill the frame) metered in any mode at zero. The resulting histogram will be thin and roughly centered; you've just acquired middle gray. Now take a second shot of the same target that is radically over-exposed, say by 4 EV. Use RawDigger to examine the histograms of the raw file of the over-exposed shot first. You will see a big pile up on the right with an abrupt end at the saturation point. Note the value of this saturation point; it will be listed as the maximal value given for each channel. You might find it useful to set EV0 to this value. Next, examine the raw file of the middle-gray shot in RawDigger. Your highlight headroom is the number of EV between the top edge of the leading color channel (almost certainly the green) and the maximal value you determined above. My E-M5 has about 3-1/3 EV headroom, for example.
Even if your camera has, say, 3 EV of highlight headroom, you may not be able to dial in that much EC when using this procedure to obtain ETTR. Sometimes a lesser amount is required. This depends on how well the brightest part of the scene fills the spot-metering area. If the brightest part of the scene is large enough to completely fill the spot-metering area, and that brightest part is very uniform, then you will indeed be metering the brightest part, and the full 3 EV of EC will likely be appropriate. If, however, the brightest part of the scene does not fill the metering area, or if the portion of the scene in the metering area is not uniform – picture very bright, narrow margins to otherwise darker, backlit clouds or small spots of bright sunlight filtering through the leaves of a tree –, then something darker than the brightest part will be metered as middle gray. Any attempt now to raise the exposure by the full 3 EV will push this lower-valued middle gray to the upper limit – and the brightest part within it over the edge. The brightest part will therefore be blown. In a case like this you will want to dial in EC of, say, only 2-1/2 EV – or even less. With experience, you will be able to judge quickly just what value of EC is required.
BEST CAMERA SETTINGS FOR ETTR
It is clear that successful ETTR requires histograms or blinkies that tell a reasonably accurate tale about the camera's sensor saturation. Since both the histograms and the blinkies are based on a jpeg image, and not on the raw data, and since this jpeg can be affected by the camera's settings (WB, saturation, contrast, sharpness, picture mode), one would want to employ settings that result in histograms and blinkies that best reflect the underlying raw data and are fairly accurate indications of sensor saturation.
Cameras vary with regard to the effects of the settings on the histograms and blinkies, and it will be necessary to experiment with your camera to determine what works best. No general set of rules exists. I find, for example, that with my Olympus E-M5 I do best using the neutral picture mode and setting saturation, contrast, and sharpness to their lowest values of -2. I also keep my histogram highlight sensitivity at its maximal value of 255. Similar considerations apply to my Nikon D300. I am told, however, that the Panasonic GH2 histograms and blinkies are relatively insensitive to the camera's settings. I have this on good authority, but I do not know it from personal experience. A little bit of experimenting will tell the tale.
White balance also affects the histograms, sometimes resulting in a story that differs greatly from that of the underlying raw-data histograms and therefore potentially providing misleading information for ETTR. This situation is likely to be most problematic if either the post-shot blue or red channel is highest since those are the ones that are "boosted" by the WB multipliers. Histograms that most accurately reflect the underlying raw data will be obtained with the use of UniWB (unitary WB, a WB that provides unit multipliers for all color channels). Here are some citations about what UniWB is and how to make a UniWB for your camera. While UniWB is of value, it is not required, and you should not avoid ETTR just because you don't have UniWB.
Rather generally, RawDigger is a very useful resource for gaining meaningful information in relation to shooting ETTR. With RawDigger you can see the actual raw histograms and see exactly how closely you have achieved ETTR. It can therefore provide a gold mine of information about the effects of the various camera settings and the behavior of the histograms and blinkies. And it provides accurate feedback about the efficacy of your ETTR techniques. It is, therefore, highly recommended software for any raw shooter wanting to perfect using ETTR.
ETTR WITH DIFFERING SCENE DYNAMIC RANGE
It was mentioned at the outset that images shot ETTR will very often look too bright or too dark as shot. ETTR aims to expose the brightest significant parts of the scene at or very close to the sensor's saturation point, and, unlike conventional metering, does not attempt to expose to result in an image with an "appropriate" average brightness. That is, ETTR attempts to maximize the light capture so as to achieve the best possible image information (highest signal-to-noise) with the idea that a "proper" final image of appropriate brightness will be created from this information later on during processing or post-processing.
Whether the ETTR image will initially be overly bright or depressingly dark depends on the Dynamic Range (DR) of the scene relative to that of the camera, low or high, respectively – and these two situations have very different processing implications. We'll look at them in turn.
ETTR in low-DR scenes
The concept of ETTR, I believe, initially gained traction in the context of shooting scenes with DRs well within the camera's range – scenes for which the camera's metering (at base ISO) would suggest an exposure (f-ratio and shutter speed) that provides an image of appropriate average brightness but which nevertheless has highlights that are significantly below sensor saturation (i.e., having a gap of one or more EV at the top of the histogram). Under these circumstances, ETTR would say to increase exposure, say by lengthening shutter speed (if possible) or widening aperture (if possible) – but not by increasing ISO, which does not affect exposure – to push the highlights up to the right-hand edge of the histogram, closing the gap. This, of course, would result in an image that is too bright, but the idea is that the increased exposure results in better s/n and the brightness can be reduced to an appropriate level during the processing of a raw file or the post-processing of a jpeg.
Obvious low-DR situations occur when the scene DR well fits into the sensor's capabilities. Thus, in contrast to the classic examples of a black cat in the snow or a white cat in a coal mine (both of which are high scene-DR situations), one might consider instead a gray cat in the snow on an overcast day, or a foggy scene with no bright highlights or dark shadows. Here is an example of the latter, which was not taken ETTR but rather with traditional metering techniques, along with its ACR histogram.
|Low-DR scene: foggy day, with neutral processing (all sliders at zero).||ACR histograms showing gaps to either end.|
Note the narrowness of this histogram (low-DR scene) and the gaps at the ends, the one at the right resulting from the fact that this original image was not ETTR.
The RawDigger histograms depicting the actual raw data verify that this image was underexposed by slightly over 1 EV from ETTR. Sensor saturation would produce values at EV0 at the D300's base ISO 200.
Thus, while traditional exposure techniques resulted in an image of appropriate average brightness, this image could have benefited from the higher s/n that would have come from an additional exposure of over 1 EV to attain ETTR followed by a reduction in brightness during processing.
Not all low-DR scenes are as extreme as that depicted above. Here is an example that shows a low-DR scene that was actually shot ETTR. As you can see from the top frame, the ETTR exposure has resulted in an image that is too bright. That this was ETTR is verified from the RawDigger histograms that follow, and the image, while bright, is not over-exposed. But, after processing (which included cropping, WB adjustment, other tonal and color adjustments, and a reduction in brightness of over 1.5 EV), the image as seen in the bottom frame is just fine and has benefited from the better s/n in both highlights and shadows.
While this kind of low-DR ETTR is best done with raw files, some of its benefits can also be obtained with jpegs. But, since post-processing is going to be required anyway, you might just as well shoot raw and gain the full set of benefits that attend raw processing.
ETTR in high-DR scenes
The situation is rather different when using ETTR to shoot high-DR scenes, ones having important features involving both high- and low-luminance values that are separated by many EV. Here, on the left, is an example of the ACR histogram for a high-DR scene (that of the first frame in Example 1 below) along with a repeat of the ACR histogram for the low-DR foggy-day scene for comparison. It is that wide gap in the high-DR histogram between the highlights and the rest of the image, each pushed toward the extemes of the histogram range, that produces a very different ETTR situation.
|ACR histogram of a high-DR scene: dark room with bright window, Example 1 below.||ACR histogram of a low-DR scene: foggy-day image above.|
Fortunately, there is nothing new to learn about the way one shoots ETTR in a high-DR scene; it is exactly the same as for the low-DR scene. But the resulting image and its attendant processing are completely different. In shooting high-DR scenes ETTR, one still sets the exposure (at base ISO) to place the desired highlights just to the right-hand edge of the raw histogram (or till the pertinent blinkies just appear), but, since there are major portions of the image that have luminance values very much below the highlights, the resulting image will now typically be very dark -- sometimes very, very dark.
So, rather than reducing brightness during processing, as we did with the low-DR scene ETTR images, we would want to do the opposite: add brightness to the shadows and mid-tones while being careful to retain the highlights, which are already at the highest values. Expanding the brightness of the shadows and mid-tones while holding on to the highlights is done in the raw converter using a process known as exposure-compression, which is more properly called brightness-compression (since the raw processor cannot affect an image's exposure). Some raw converters, such as RPP, actually embody facility for this. In ACR it is enacted by first setting the so-called Exposure slider to produce an image of appropriate average brightness, then dialing back on (compressing) Highlights (often to -100) to prevent highlight clipping, and then further expanding the shadow brightness with the Shadows slider.
Here are three examples of such ETTR. For each of these shots, the exposure was set so that the important highlights were just on the margin of producing blinkies. In each case you see the original shot followed by its RawDigger raw histograms to verify the ETTR. Then comes the processed shot after brightness-compression and other processing. For all the original images, the basic EXIF data are available at the top of the corresponding RawDigger histograms.
Example 1: (Click any image to see in larger size)
Example 2: (Click any image to see larger size)
|Original shot, ETTR, just till blinkies begin.|
|RawDigger histograms verifying ETTR.|
|Processed file after brightness-compression and other adjustments.|
Example 3: (Click any image to see larger size)
Many high-DR scenes can be captured in the manner indicated by the examples above by using ETTR combined with brightness-compression. If the scene DR is so great that an acceptable final image cannot be obtained with these methods (too little detail, too much noise), then you have encountered a shot that is simply not meant to be with a single exposure. Other HDR techniques, typically multiple exposures, will be required – if possible.***
JPEG vs. RAW
While the processing associated with low-DR ETTR, described above, can be enacted with both raw files and some jpeg files, the high-DR technique for ETTR works most successfully only with raw files. Jpegs have many characteristics that differ from raw files in ways that render them far less able to undergo the adjustments involved with the brightness-compression required by high-DR ETTR. These differences include:
• jpegs are 8-bit files rather than 16-bit and are less robust to significant tonal and color adjustments.
• jpegs are non-linear, already having a gamma applied. Processing is best done with linear data.
• jpegs tend to have less highlight headroom to exploit during ETTR post-processing.
• jpegs have a WB applied and are susceptible to color distortions when undergoing significant WB and tonal alterations.
• jpegs have already suffered data losses during compression and are less tolerant to tonal and color adjustments. These losses can never be recovered.
• jpegs have already had a color space applied and have thereby lost gamut (particularly if sRGB is used for the camera's color space) that can never again be recovered.
WHAT IF ETTR IS NOT POSSIBLE?
While the raw shooter would typically benefit from employing ETTR if it is possible, shooting considerations may make it impossible. DoF considerations, for example, may dictate a high f-ratio, and camera-shake considerations may dictate a fairly fast shutter speed. Together, these settings may entail an exposure that falls short of ETTR. What to do?
If ETTR is not possible, it is still important to maximize exposure, i.e., to push the histogram as far to the right as possible at base ISO even if not to the right-hand edge. Let me emphasize that maximizing exposure means finding an acceptable combination of shutter speed and f-ratio that gives the largest possible exposure, and it does not involve the use of ISO, which is not part of exposure. For these purposes, ISO remains at its base value. This allows capturing the greatest possible signal with least relative noise and creates the best foundation for the final image. If this maximal exposure is not ETTR, then brightening can be added to achieve an image of the desired brightness. But first comes the setting of exposure, then comes the brightening. Now the question arises: where should this brightening be done: using in-camera ISO or during raw processing or both?
The answer to this question depends on the "ISO-nature" of the camera. With an ISO-invariant camera (one whose read noise does not change with the camera's ISO setting), one could do either (brighten in-camera or during raw processing), but there are advantages to shooting dark (letting your image remain unbrightened) at the base ISO and brightening during raw processing. This will typically result in a final image with better IQ and less chance of clipped highlights. With an ISO-variant camera (one whose read noise decreases with increased ISO), the benefit is in favor of brightening with added in-camera ISO, which will typically result in less read noise than shooting darker and pushing in raw processing. Some cameras are partly-ISO-invariant, becoming ISO-invariant only after reaching a given ISO level, say 800 or 1600. Here there are benefits from increasing ISO in-camera, if needed, up to this level and then effecting any further brightening, if required, during raw processing. More on this topic can be found in Note 5 on page 2 of Exposure vs. Brightening and you can learn more about the ISO-nature of your camera from sensorgen.info.
Various constraints can come into play in determining an optimal exposure. These include DoF considerations affecting the f-ratio, motion-blur and/or camera-shake considerations affecting shutter speed, any desired amount of blown highlights (often, but not always, none), any desired shadow noise (or lack of it). Sometimes these constraints are incompatible and cannot all be achieved at once. In this case the raw shooter must make compromises. The maximal exposure satisfying these compromises while avoiding unwanted over-exposure is the optimal exposure, even if it is not ETTR. A shot based on any lesser exposure is under-exposed.
* I want to thank farcus and GeorgianBay1939 for encouraging me to write this article. I am particularly indebted to bobn2 for giving an early draft a good read-through and making comments leading to useful clarifications. Naturally any remaining errors are my own responsibility.
** Some cameras have known problems at base ISO, such as the Canon 5D and 6D series with their banding issues at ISO 100. For these cameras, one may well be better off shooting ETTR with at least 1 EV of ISO. As with any shooting technique, one must know one's equipment, and so, should problems arise, one need only experiment a bit to devise the needed modifications to make things work properly.
*** Some care may be required when making large adjustments to shadows and mid-tones. Nonlinear transformations in moving from the camera space to the working space, and twisted converter profiles (such as Adobe's Standard Profiles), can cause color shifts during large luminosity lifts. In many cases this is not a practical problem, but, if it is, finding the right raw converter and using a proper camera profile become important. This is not an issue that is peculiar to ETTR.