12 stops: How does dynamic range work?

[jimbee]

I'm trying to understand the concept of dynamic range, and especially statements I've seen (in the Zacuto shootout and on Shane Hurlbut's blog) that the Canon DSLRs can get 11 or 12 stops of dynamic range. Does this mean they can accurately record a scene that varies by 11 stops?

When I look at the T2i histogram, an exposure that uses all five segments, it seems to translate to about 6 stops of dynamic range, if I am understanding the histogram. When I add one stop of exposure, it adds slightly less than one segment of data on the histogram. But going outside to the left or right of those five segments just produces crushed blacks or blown out highlights, doesn't it? It seems like the histogram is saying that a scene that has more than 6 stops of dynamic range will be clipped, either in the shadows or the highlights.

 

[Razz16mm]

Dynamic range is the effective range of values a camera can capture between the noise floor on the black end and white clip on the white end. Generally what is acceptable noise on the black end varies with the intended aesthetic result, but at some point you can no longer distinguish detail from noise. In practice you try to find the ideal exposure that captures important shadow detail with as little noise as possible while not clipping important highlight details you would like to preserve. Using the histogram, this is known as ETTR (expose to the right). Generally the more exposure you can give the sensor without clipping highlights the better off you are. Crushing black levels moves the 0 black value cutoff above the noise floor so you don't see the noise, but you also cut off any detail below the 0 level. The art and skill lie in making the best choices for what you are trying to achieve visually under your shooting conditions. Any camera is going to have a "sweet spot" a range of exposure values or ISO speeds that produce the widest possible dynamic range. Moving outside this range in either direction reduces overall dynamic range.