In part 1 of this discussion I talked about how determining exposure for a digital capture is substantially different than it was for film, yet digital camera makers utilize the same methodology as film cameras for determining exposure when capturing an image and converting it to a Jpeg file.  The entire premise of these two articles is based on a different method of determining a digital exposure, commonly referred to  “Expose to the Right” (ETTR), which I and many others feel is a better technique for determining exposure.  Using ETTR  allows the photographer to maximize signal to noise ratio as well as record subtle detail in shadows because we capture that information higher in the dynamic range of the sensor.

While there are those that debate whether this concept is valid or not, it really isn’t very debatable. Sensors really work this way, and you really can use ETTR to lower noise and increase detail in shadows.  Any debate should center on the practical application of ETTR, not whether ETTR itself is valid.  I’ve seen discussion that erring on the underexposed side is actually a better system, but as mentioned this highly compromises the data we are capturing, and while the results may be acceptable, capturing the image using ETTR would result in purer data spread across more levels with a far better signal/noise ratio.  As I mentioned, cheating on the underexposure side of things actually predates ETTR, and those that advocated it were simply trying to ensure highlight detail wasn’t blown, without fully realizing the consequence.  (I speak from experience ... been there and done that).

The challenge with ETTR is knowing when and how to use it.  There are workflows that make ETTR challenging, and sometimes while ETTR may achieve a better “technical” result, the difference in real world results is nearly imperceptible.  As one that has been practicing ETTR for quite some time I can offer some of my observations and how I apply the principle in my photography.

Just to make things clear, understand that ETTR should only be used in a RAW workflow.  The entire concept is based on the fact that the capture device is linear, so you can capture the RAW information anywhere within the cameras dynamic range and normalize the exposure in the RAW converter.  A jpeg is a baked image designed to look good out of the camera.  Camera makers may very well be able to design a “ETTR” mode in the camera, where the exposure is set to capture the data higher up the dynamic range and then normalize it, but currently I’m not aware of any camera that works this way.  So if you have a jpeg workflow, no need to try ETTR.  In fact, if you really are seeking to maximize your image quality with ETTR, you should be working in with 16bit files, and using a wide gamut working color space, such as ProPhotoRGB. (that’s another discussion for another day).

To begin, any method of determining exposure has to be fast and reliable.  It is more important to get the shot first ... if you lose the shot because you are trying to use ETTR for a perfect exposure, you may wish to re-evaluate your priorities. Michael Reichmann who introduced and brought this concept into prominence makes this very clear in a few of his video tutorials  ... getting the shot is the priority.

Often an ETTR exposed file is no different than a normal one.  While the way that exposure is determined may be different, the end result may actually be the same.  The key difference is what values of the scene are being used to judge the exposure.  The camera is taking a section of the image and “averaging” the brightness, then using that to calculate an exposure value. When using ETTR we are simply looking for one thing ... what exposure will capture our image with all the pixels exposed as high as possible without any being blown.   With ETTR we are assuming that highlight and midtone detail is more important than shadow detail, so we would rather clip the bottom tones of the image rather than the highlight tones.  In reality if our scene has about the same dynamic range as our sensor, the resulting exposure calculation could be very close for ETTR or traditional technique. 

Where ETTR really comes into play is when the dynamic range of the scene is less than that of the sensor.  Here you have the ability to push the data to the right side of the histogram and realize a significant improvement of the quality of the data in your RAW file.   Even in this circumstance,  the benefit may not be worth the additional effort.  I do not practice ETTR when doing a portrait sitting.  The advantage of a correct looking preview image far outweighs the need to reduce noise and  gain levels in the shadows, because I’m rarely recording data in the bottom level of my dynamic range anyway.  But when doing a landscape shoot in the evening, no noise is the goal, as well as terrific detail in shadows where possible.  I shoot a lot of landscapes in very late evening light ... the light is so soft and magic at that time.  The dynamic range is rarely more than about 3 to 4 stops.  If I use ETTR and maximize my data, I can normalize the exposure and even expand the dynamic range in the RAW converter, often achieving a beautiful soft light with plenty of contrast.

The real challenge of ETTR is an effective and quick method to use it.  Often the challenge of getting the shot precludes the need for anything other than a “safe” route ... perhaps even erring on the side of underexposure.  The reality is if the highlight is blown, it is blown, and if the highlight detail is important and you don’t have much time, then just don’t blow it.

But most of the time when using ETTR it is because you are interested in maximum image quality, and have the time to ensure your exposure will offer that.  When you find yourself in that circumstance, how do you calculate an ETTR shot?

The most obvious is to shoot a frame and look at the histogram.  Another is to enable the highlight “blinkies” ... the feature that flashes any pixels that are “blown” on your preview image.  Both of these work pretty well, but aren’t really accurate.  The good news is using this you will almost always err on the safe side.  The reason is histogram and highlight warning indicators are based on the preview image is a “rendered” image of the RAW data.  Often the “blinking” (blown) pixels are not actually blown in the RAW data and a clipped histogram may not actually be clipped in the RAW file.  The bottom line is if you have nothing blinking, then nothing is blown.  The downside is if you have nothing blinking, you may still have plenty of room to increase exposure by as much as 1/2 to a full stop before blowing any pixels.  ACR is very good at reconstructing highlight data from pixels that have at least 1 channel of valid data,  so even if 2 of the 3 channels are blown, ACR may very well be able to recover that information.

Often this is more than enough ... just look at the histogram or preview,  fine tune the exposure by adjusting the exposure compensation dial so the information is as far to the right in the histogram as possible or nothing is blinking and shoot.  You may wish to experiment, you may find that you can open up another 1/3 to 1/2 of a stop all of the time when you first see blinking pixels, and not have any blown highlights in the RAW data when you open it up in your software of choice.

Sometimes it is just easier to bracket.  Once you determine an approximate exposure, it may be much simpler just to adjust the exposure by 1/3 to 1/2 stops a couple of times.  Let face it ... you are taking the pictures to check the exposure anyway, may as well just bracket, and decide which capture maximizes exposure without blowing pixels in the RAW converter later.

What about when the scene contains more dynamic range than the camera can record?  Good question.  ETTR is still often the best way to nail the exposure ... remember blowing highlights is often worse than not recording detail in shadows.  But here experience and desired outcome rule the day.  It is quite possible that the highlights are so intense there isn’t any chance of not blowing them.  So you have to decide where to draw the line .. which will be the most important parts of the image in its final version?  Here again, bracketing is your friend, and you may find a quick layer in photoshop with an appropriate mask will let you combine two exposures into a single image which pushes the scenes dynamic range into your working dynamic range.  Of course, you can even take 5 or more bracketed exposures, and using various tools combine them as an HDR capture.   That’s another topic, and while I often layer two images and use manual masking to achieve a workable dynamic range, I haven’t done much with HDR (yet).

I think using ETTR will get easier over time.  For example, the new Canon G10 makes it a snap.  It has a dedicated mechanical dial that allows exposure compensation to be adjusted dynamically as you view a histogram on the live viewfinder.  Just dial until the histogram almost clips.  Fast, and for a digicam very useful since noise is a real problem.  As more and more dSLR’s employ live view, tools similar to this may begin to appear.  In fact, it might even be possible for camera makers to employee an ETTR metering mode, reading the data hitting the sensor while in live view, and calculating what exposure will be just below clipping.

I have found ETTR to be very valuable in achieving higher image quality.  I’m not sure where technology is headed, but for now noise is still enough of a problem that using ETTR allows a higher quality capture.  I encourage you to try it for yourself.  Bracket some exposures, and instead of choosing the one where the preview image looks the best, choose the one that looks washed out but has no clipped pixels.  Pull the exposure down in the RAW editor, and then examine the detail and noise in the shadows.  On some images the difference will be dramatic.