We all know that digital pictures are different from film, but most of us don't, I think, usually sit down and think about exactly how they differ very often. We notice fairly quickly that digital photos have gorgeous saturation and highlights that are reminiscent of slides film, we adjust our technique (or our subject matter) accordingly, and we go on our way. But are we really getting the most out of our digital cameras? Probably not. One technique those of us shooting RAW can use to squeeze a little more our of our cameras is what Michael Reichmann of The Luminous Landscape calls "exposing right". He's not talking about correctness here, but about where exposures fall on a histogram. His wonderful artcle can be a bit daunting for photographers who don't consider themselves technologists as well—it's drawn from a conversation he had on the road with Thomas Knoll, the original author of Photoshop—but this gist is this:

Unlike traditional film, which absorbs light fairly equally across the entire dynamic range of the film and "rolls off" smoothly at both ends of the range, the sensitivity of digital sensors falls off linearly with decreasing intensity. You've probably noticed this: highlights tend to come through better than shadows; dark areas in pictures can have a tendency to posterize. What you probably haven't realized is just how bad it is.

Digital cameras break the range up into five or six f-stops, and fully half their sensitivity is in the first stop, with each successive stop having half the sensitivity of the one before it. This means that a 16-bit sensor (which really only records 12 bits of data), for instance, can capture about 2048 levels of the brightest highlights, 1024 levels of less intense highlights, 512 levels of true mid tones, 256 levels of dark tones, and only 128 levels of the darkest shadows. Even worse, most compositions have a lot of mid tones, and many don’t even use the top 20% of the range (like the top histogram above), so they lose fully half their sensitivity before the shutter even snaps: 16-bit pictures become 8-bit pictures; 32-bit pictures become 16-bit pictures.

The solution? Shoot RAW and overexpose a little. Push the exposure until it’s just short of “blooming”. IF you camera has a red overexposure warning light in the viewfinder, open the aperature or slow down the shutter until it blinks, and then back off one click. This will ensure that you’re storing lots of information in the senstive highlight ranges (the bottom histogram), and then you can use a tool like dcraw or Photoshop’s Camera Raw to adjust the levels laterwithout losing detail.

Don’t ry this with JPEGs, though: you camera has already performed white point and other adjustments, and the information has been lost.