> Question is, why is underexposure easier to handle? Presumably white is > represented by some number, say 255, and black by 0. You'd think exposure > would fall off either end just as easily. Does it have to do with an > exponential response curve of some sort like film has? Sort of Tri-X like > in the toe and T-max like at the top...
Yes, pretty much. Think of a CCD sensor as a device for counting photons. If you double the exposure time (increasing exposure by one stop) you'll get twice as many photons in the bucket. Halve the exposure time and you only get half as many photons. This means that an image overexposed by one stop would have counts that (after scaling) would range from 0 to 511. This is clipped to the range 0 to 255, with everything above 255 ending up mapped to 255. The visual effect is that the bright areas of the image are all washed out to white. An underexposed image, on the other hand, simply has values in the range 0 to 127. You can get back to the 0 to 255 range by simply multiplying all the values by two, although you lose one bit of precision.
