The development of digital cameras has been strongly influenced by its film ancestors, but to me this legacy is most evident in the methodology of how digital cameras determine exposure.  To ensure photographers were comfortable with digital cameras, most features and concepts were adapted from the film world.  It seems part of that legacy is an image exposed similarly to film will produce a “correctly exposed” jpeg file, even though the camera makers could choose many alternative methods to establish the correct RAW exposure, and design in camera processing to render that jpeg as “correct”.

One thing that became very evident to me when we first made the transition to digital capture in our studios in 1999 was a digital camera was substantially different than a film camera. (duh!)  In those days it was sometimes assumed that a digital camera had more latitude than film so you didn’t need to be as careful.  We soon learned how wrong that was, and found that exposing for digital was akin to exposing for transparency film - if you over exposed the file, you blew out the highlights and there was no way to recover them.  Because of that fear, photographers often erred on the side of under-exposure, since typically in an image highlight detail is much more important than shadow detail.

To understand why using concepts from exposing film to determine digital exposure may not achieve optimum results, it is important to understand that a digital capture is linear,  whereas film is very non-linear.  As we increase exposure in film, we get to a point where the information gradually becomes more and more compressed, resulting in what is referred to as the “shoulder” of the capture. This compression of highlight detail is similar to how our eyes respond to increase in light, and even though we can “clip” the highlights using transparency film, this compression of information still occurs, so the end result is more acceptable than if we clip the highlights in a digital capture.

On the other hand, a digital sensor continues to respond exactly the same until it reaches its limit, when it no longer records any change - normally referred to as clipped.  On the highlight side that means we have reached maximum saturation and the pixel is recorded as pure white.  Add more light, and the value remains the same.    On the shadow side it means we have no light recorded, leaving us with a value of 0,0,0.  The effect of compression at the toe and the shoulder of the capture to make the image appear natural to the eye is one of the functions of the RAW convertor, be it the in camera software when rendering the preview or jpeg file, or a RAW convertor on our computer later.

What does this “linear” nature of digital capture mean in real terms?.  After all, if you expose the digital camera just like film the result normally looks just fine.   There are two primary considerations why using a different method for determining your exposure setting may be beneficial, especially if you are trying to achieve the absolute maximum image quality.

The first is due the very nature of how the sensor records data.  A digital pixel expresses how much light is recorded using discrete levels - a 14 bit camera is capable of recording 16,384 levels.  If the sensor has a 6 stop dynamic range, you end up with a logarithmic   distribution of those levels ... each stop represents a doubling of the required light.  So this conversion from a linear capture device into real world data that is representative of how we see becomes problematic.

This means the brightest stop of light utilizes 8,192 of the levels - one half - to record that data.  The next stop requires one half of the remaining levels, or 4, 096.  Not really a problem .. plenty of levels to record the detail ... far more than the eye can see.  But as you keep going, the challenge is fewer and fewer levels.  When you get to the fifth stop of the cameras dynamic range, you only have 512 levels, and the 6th stop there are only 256 levels.  What if your camera is only a 12 bit device?  This means your fifth stop only has 128 levels to work with.  So if you record your images one stop lower than is possible, you compromise the ability to record shadow detail because you push all of the data down a stop resulting in substantially fewer levels to record the shadow detail. 

Because the capture is linear, the data can be recorded anywhere within the sensors limit, then normalized in the RAW converter.  As long as the scene has fewer stops of dynamic range than the sensor can capture, it becomes advantageous to move the data as far up as possible, so our shadows are being recorded in a range that has more levels.  The camera is unaware of how the exposure was captured, and makes its normal conversion for the preview, resulting in a light and “over exposed” preview.  If the camera is saving a jpeg file to the card, it will have the same problem.  But remember, the over exposed preview has nothing to do with the accuracy of the exposure, just the assumptions of the camera makers when designing the internal firmware.

There is some discussion as to the practicality of this concept, after all, as long as you don’t block the shadow details into pure black, the image will most likely appear fine.  However there is another important aspect of digital capture that comes into play which to most is actually more important ... noise.

There are various reasons and sources of noise in a digital capture.  Sensors have improved dramatically over the years, but noise is still an important consideration.  The data recorded by each photosite on a sensor is either valid information (signal) based on the amount of light, or is compromised information because the noise has corrupted it. What makes this important is noise is very low and is fairly consistent regardless of how much light the sensor is reading.  The more signal in relation to the noise - signal to noise ratio - the better the overall data in the capture.

As you move from highlight to shadow in a capture, you move from higher levels of signal to lower.  Thus in the lowest stop of your capture, the signal to noise ratio is limited.  At some point the noise overcomes the signal, and you reach the limit of your sensors dynamic range.

How serious of problem this can be is apparent when you combine it with the first concept.  If the amount of noise is constant, in your brightest stop you have a small amount of noise, combined with over 8,000 levels of data.  Each stop going down reduces the data captured and levels by 1/2, but the noise remains the same.  By the time you reach the 5th stop of your dynamic range, you are down to only 512 levels to record the image data, with the same amount of noise and only 1/16th of the amount of data.  If your camera has a low enough noise threshold, you may get lucky to get a sixth stop of dynamic range, but now there is only 256 levels, and 1/32nd of the data.  Lets assume the signal to noise ratio in the brightest stop is 100:1 .. this means by your fifth stop of dynamic range this ratio is down to 1:6.  It is pretty easy to see how difficult it is to retain shadow detail as your available levels drop dramatically and the noise pollutes the shadow data.

This begs the question how should we determine the “correct” exposure for a digital capture?   Is there a better way than traditional film based methods which will account for these two factors?

A few years ago, an article by Michael Reichmann on his Luminous Landscape website introduced the concept of Expose to the Right, now commonly referred to as ETTR. The premise of this concept is very simple ... if you expose your data as far to the right of the histogram as possible without clipping, you have increased the signal in relation to the noise  resulting in lower noise, especially in the shadows.  In addition you are able to take advantage of far more of the available levels ... detail that was limited to the number of levels in your 5th stop now is being recorded in the 4th stop ... twice as many levels.  Because of the linear nature of the RAW capture, you can then normalize this capture down to the correct level in the RAW converter, but with the added benefit of capturing more signal vs. noise.

ETTR is now a commonly accepted and practiced technique by many noted photographers.  The late Bruce Fraser discusses what he viewed as the correct way to base exposure in digital capture in an article located on Adobe’s website ... Raw Capture, Linear Gamma, and Exposure.  This article is very concise and well done, worth reading for anyone interested in understanding how ETTR can benefit them.  Here is one quote from this article ...

“Correct exposure is at least as important with digital capture as it is with film, but in the digital realm, correct exposure means keeping the highlights as close as possible to blowing out, without actually doing so. Some photographers refer to this concept as “Expose to the Right” because you want to make sure that your highlights fall as close to the right side of the histogram as possible.”  

Two other articles that should be beneficial are Expose to the Right , the original article by Micheal Reichmann that introduced the concept, and Exposing for Raw, a much more in-depth discussion by Andrew Rodney.

Often I see discussions on various forums which dismiss ETTR, but the reasons are normally a lack of understanding of its benefits, and a lack of a practical way of applying it.  In part 2 we’ll discuss the practical application of this concept and why perhaps many discredit it too quickly.