Exactly. So why are you argueing against me? Obviously you don't understand your own statements. It does absolutely not make any difference, technically, whether you set your light meter to ISO 200/24° and, ignoring the metering, over-expose deliberately by one stop on your own devices, or set your light meter to ISO 100/21° and let it control the exposure automatically. In both cases you're effectively using ISO 100/21°, in both cases you're sacrificing dynamic range, and in both cases you're getting less shadow noise. So what's your problem?
I do understand my own statements, but you apparently don't. One thing you are missing. When one increases the ISO of a camera such as the D3 over base two things are done: 1) the metering causes less exposure to be given and 2) the amplifier gain is increased so that the full scale of the ADC is used. For example, if you raise the ISO set on the camera to 400 from the base of 200, only half as much exposure is given and amplification is doubled so that the the raw data values are recorded at the same level as before.
If you set the camera to Lo1 (ISO 100), the metering changes but the amplification does not. There is no need to change the ISO. Merely expose to the right. There is only one optimum ISO for the D3, contrary to your assertions. I don't know how Canon handles this situation, but I suspect that it is the same as with the D3 (according to the DXO ISO data). Rather than confuse the op with your double talk about two optima, I would merely advise him to shoot at base ISO and expose to the right.
If a higher ISO is needed due to shutter speed or f/stop considerations, as pointed out by the DigitalDog earlier in this thread, one should be aware of the characteristics of the camera as explained by
Emil Martinec. Often, one can increase exposure in the raw converter and preserve highlight headroom. However, the histogram would not be to the right. It is the number of photons collected and not the appearance of the histogram that is important. That is one thing that Michael does not point out in his essay.
Regards,
Bill