gmitchel wrote:
>
> Yes. No kidding the math is correct. The question is the
> application of the mathematical model to how CCD/CMOS devices work.

Yes, the application is correct also.

ETTR is what I think is the correct exposure model for digital sensors--as bright as possible, WITHOUT clipping. The "without clipping" part is critical. It also requires the use of a raw format--JPEGs destroy too much highlight information, which cannot be recovered.

This model is different than the "18% gray" exposure model used for analog film. This is due to the linear nature of CCDs and CMOS sensors, vs. analog film's non-linear response curve.

ETTR results in the best noise performance, since both the quantization noise, and the ISO noise are mimimized.

Use of this exposure model requires some care, since most cameras don't display color histograms, so you need to be careful to avoid clipping brightly colored highly saturated objects.

ETTR also sometimes requires a longer exposure time and/or a wider f-stop. It is basically using a lower ISO than what the camera is set to. With moving subjects, or handheld shots, this can result in blurred images. In these cases, it is almost always better to have a sharp image with a high level of noise, rather than a blurry image with a lower level of noise.

ETTR also is highly dependent on the subject matter's dynamic range. If the subject has a very wide dynamic range, ETTR will result in a DARKER image than an "18% gray" exposure image. This is because of the "without clipping" part of ETTR.

Thomas Knoll