tomcat wrote:
I'm going to ask you a stupid question. Why did you change the ISO to achieve the under/over exposure? Why did you not change the shutter speed instead? ...
Absolutely not a stupid question.
There are three ways to move the histogram(s) to the right - by using a longer exposure, by opening the aperture (nobody does that) or by increasing the ISO.
If the scene had been initially metered at ISO 1600 it might have called for a normal exposure of 1/2000 second. To move the histograms to the right by two stops would change the exposure to 1/500 second.
But if the scene has initially been metered at ISO 400 it might have suggested an exposure of 1/500 second. Changing the ISO to 1600 also moves the histogram to the right two stops.
So it doesn't really matter which way you do it. You end up in the same place - ISO 1600, 1/500 seconds at f/11 and the histogram ends up in the same position - as far to the right as it can go without blowing the green channel.
By using the same physical exposure (1/500 @ f/11) the sensor received exactly the same amount of light for each exposure. The same physical exposure means that the same amount of noise (if any) would be recorded.
The three different ISO settings moved the histogram progressively to the right until the third image was actually all the way up to the limit (16000) for the green histogram. You can see the progression in the plots below. Any difference in raw image data will be solely due to the change in ISO, not the change in exposure.
Because the sky was overcast, the images covered fewer steps (a narrower DR) than the full daylight images I posted earlier.
You will also notice that the green channel is further to the right of the blue channel and about a full stop beyond the end of the red channel. This is fairly common but in a scene with a large amount of red or blue color things would be different.
Nevertheless, the whole point of this exercise is to show that there is a lot of latitude in the raw file around what would be considered a "normal" exposure. In this case a range of +/- 2 stops does not appear to make any visible difference.
So
Is ETTR better than ETTL? Not much if at all - within limits.