I'll elaborate a bit on my previous post.
All camera sensors have one and only one sensitivity, which depends on the Quantum Efficiency, the Full Well Capacity of each pixel and the Conversion Noise.
The ISO value of a sensor is a fabrication which was made in the 90's (I believe by Kodak) to facilitate the transition of photographers from film to digital and still sticks around. Computer vision system cameras do no declare ISO values, because they are of no use to calculate the correct illumination and exposure values to feed the RAW data to some algorithm.
To attain some (made up) ISO value, the camera performs a two stage process:
1) Amplification of the analog signal produced by each individual photodiode in the sensor
2) Normalization, which means multiplication by some numerical constant
There is a practical limit to amplification which is dictated by the maximum signal amplitude that the ADC converter can manage. Beyond that, only normalization is possible.
*BUT* by normalizing, the camera is doing exactly the same procedure than increasing the exposure value in any post processing software (Lightroo, CaptureOne, etc.)
So we can divide the ISO value in two ranges:
1) LOW ISO: Less than ISO invariance threshold (amplification dominates)
2) HIGH ISO: More than ISO invariance threshold (normalization dominates)
The HIGH ISO range is only of interest when shooting JPEG, because all the Post Process is done in camera, and a final developed image is produced. There is no intention to perform PP at a later stage, so the product is final.
But when shooting RAW, using HIGH ISO is of no use whatsoever, because PP in camera does the same operation than PP in a computer.
If you are shooting RAW, of course your are *planning* to perform some Post Process, in which case it is better to never ever use HIGH ISO, because:
1) Trying to conform the exposure triangle, HIGH ISO will probably blow up the highlights
2) HIGH ISO will create noise which could be better handled by a full blown PC than a the small CPU in the camera.
In the PC you can manually enhance the shadows and the midtones without burning up the detail in the highlights. You can create masks (Capture ONE, ON1, PS, Affinity, etc.) and reveal detail of select features of the photograph without introducing noise in the rest. You can then dampen the noise only in the areas where it is visible, without softening the whole photo.
I'd recommend always to stop at the ISO invariance threshold and then PP. Then you-ll have all the headroom to perform PP as you wish... Just set the maximum ISO at that value.
This link has the ISO invariance threshold for most cameras:
http://www.photonstophotos.net/Charts/Sensor_Characteristics.htmJust click on your model or look in the table at the bottom of the page. The "Unity ISO" will be the threshold value which divides the LOW and HIGH ISO ranges. Just don't increase ISO beyond that value if you are shooting RAW and plan to post process later.
Hope this clarifies the subject.
I'll elaborate a bit on my previous post. br br A... (