Pro-Tip #9: Expose to the Right (ETTR)

Normal Metered Exposure

ETTR Exposure of Same Scene

You have probably heard this expression bantered about in photography discussions or seen it in published materials, but what does this mean and why do you want to do it. I'll give you a short answer and a more detailed long answer.  This is for RAW image processing only, as it will not work with JPGs as these images have already been processed in the camera. The short answer is when you make an exposure you want your histogram data positioned as far to the right as possible so it is just kissing the right-hand side of the graph boundary, because it will provide you with the highest quality image possibles. That's it, that's  all you need to do if you want the highest image quality available from your camera. You need not read any further unless you want a greater understanding of why this holds true for digital images as I present the long answer.

The long answer requires some background information. As you may know, camera meters try to expose everything as an 18% or neutral grey scene.  This serves us well when we photograph average scenes as we get a good exposures in that we don't have blown out shadows or highlights, but the exposure may not be the best possible (more on this later).  And when we get into hi-key or lo-key situations the 18% meter reading is going to fail by under exposing the hi-key scene or over-exposing the lo-key scenes,  making our whites look grey in the hi-key scene and making our blacks look grey in the lo-key scenes, and making it very apparent that our exposure is noticeably off and we need to intervene. But why would we need to intervene in a situation where the metered reading results in a good exposure, the following explanation from Martin Baily's, "Why Expose to the Right", (Podcast 381), July 29, 2013, he explains the technical side of ETTR and why we should use this technique even for mid-tone scenes:

Camera sensors have to convert the light that is captured by the photo-diodes on the sensor to a digital value. Most cameras these days save this data for each pixel with 14 bits of data, meaning we can have up to 65,536 tonal values per pixel. Many cameras at the moment record about 12 stops of dynamic range [see www.dxomark.com], some more, some less, and this is the range between a true black and a true white, and with a 14bit sensor, this means we’d have 65,536 tonal values between the two extremes.

Think of this as gradually filling a bucket with water, or even filling a photo-diode on your sensor with light. A totally empty bucket would be zero, or totally black, and then as you pour water into the bucket, or light into the photo-diode, you gradually fill it until you hit the maximum it can hold, which would represent a pure white. Anything after that is just overflow. The white can’t get any whiter, and that pixel is now over-exposed.

Because digital sensors are linear devices, if you double the amount of light that hits the photo-diode, you double the voltage generated by the sensor, or the amount of water that we pour into the bucket. The result is that data is literally halved with each stop or EV (Exposure Value) that is recorded, so the brightest stop of light the sensor can record has to be double that of the second brightest. This means that to use the full 65,536 tonal values across the entire image, we have to half the amount of data that can be used for each exposure value from pure black to pure white. For a 12 stop dynamic range sensor, this means your data is broken up into 12 steps, as follows…

Darkest-------------------------------Mid Tones-----------------------------------------Brightest
EV	1	2	3	4	5	6	7	8	9	10	11	12
#Tones	16	32	64	128	256	512	1,024	2,048	4,096	8,192	16,384	32,768

This is why the darker parts of the image are more noisy than the brighter parts. The noise is always there, but because we have less data to record the darker areas of our image, the noise is much more visible. I’m sure you’ve noticed that even a nicely exposed image often has a bit of noise in the shadows. This is why.

When we are shooting hi-key scenes or lo-key scenes, we know the camera's metered normal exposure is going to be off and we can use exposure compensation or manually adjust our exposure to record the scene properly.   But what about the mid-tone scene that our camera can easily meter and record the full dynamic range of the scene? If we leave the exposure up to the camera, the camera will record the scene in the middle of the histogram, so you will essentially be recording your image with between  256 and 2,048 tonal values per stop as opposed to performing  ETTR and recording the scene between 4,096 and 32,768 tonal values per stop in the brightest end of the histogram.

When you capture more data the result is higher quality more date rich image files that will result in the highest quality images your camera is capable of recording and because we have more data noise will be less apparent adding to the quality of the image.