ISO for digital photographers in 2023

Just set the damn thing on 'auto', with some safe shutter speed floor, and never think about it again.

just use AI to generate your photos ... and

Instead of concentrating on how low the ISO value is, it’s probably best to focus on maximizing the sensor exposure in any way possible and then just accepting whatever the ISO will have to be to get the job done. The noise you’re experiencing is primarily the result of the initial exposure (or lack thereof), not the ISO settling value.

Indeed. If you want to minimize noise, maximize exposure with whatever degree of freedom that is available to you, fiddling with ISO won't help.

If you don't mind motion blur, or only part of your subject in focus, it's easy to maximize your exposure to minimize image noise. Just set the camera to base ISO (often ISO 100), use the widest available aperture, and use a long enough shutter speed.

However, in many situations, there isn't enough subject illumination to let you hit a high exposure, while getting enough depth of field to have the entire subject in focus, without objectionable levels of motion blur.

In these sorts of situations, Auto-ISO can be a useful tool to get the maximum exposure (minimum noise) for the specific situation.

Put your camera in "M" mode. Select the largest aperture that gives you sufficient depth of field. Select the slowest shutter that doesn't yield unwanted motion blur. This will give you the maximum exposure for your conditions. Set the ISO to Auto, and you will get reasonable image lightness for the exposure. This is a particularly useful technique for changing lighting conditions (for instance shooting a theatrical play), as it always gives you the highest exposure for the conditions. If you your camera was in an auto mode, and you had picked a particular fixed ISO, you would not get a higher exposure as the lighting increases.

Your post is totally spot on but I feel it might help if it was explained to anyone interested, why you would not get a higher exposure* as the lighting increases.

With a fixed ISO you are effectively locking in a set final image lightness for the sooc jpeg.

Now say you are in aperture priority, ISO 400, f/2.8 and the meter sets 1/100s shutter speed for the scene lighting. Now lets say for the next shot the light from the scene increases. To maintain the same image lightness set for ISO 400, the camera will set a faster shutter speed to compensate for the increased light from the scene.

This faster shutter speed results in a lower exposure*, causing a lower signal to noise ratio resulting in an image with more visible noise than if the aperture and shutter speed had been maintained at f/2.8, 1/100s and the ISO lowered via Auto ISO to compensate for the increased light from the scene.

* exposure - amount of light striking the sensor per unit area while the shutter is open
** optimal exposure - the maximum exposure* within dof and motion blur requirements without clipping important highlights.
*** under exposed - more exposure* could have been added with the DOF and blur constraints still being met without clipping important highlights.

Up until recently, I always manually set my ISO and strived to keep it as low as possible for the available lighting conditions. When possible, I tried to shoot at ISO 800, 400 or less. On rare occasions I would push to ISO 1600 but that was mostly for indoor family events were the memory was more important than image quality.

My approach changed rapidly when I got my Canon 90D. The out of camera jpegs looked pretty good at ISO 1600 and 3200. When I opened raw files with Adobe ACR, the results were not much better. I was really impressed when I opened raw files with the Canon DPP4 software using the default noise reduction setting. I got great results with sharp details and good noise reduction at ISO 6400. The results were typically pretty good even up to ISO 12800. Use of dedicated noise reduction software can give even better results.

Except for my macro work, I now shoot almost exclusively with aperture priority and autoISO with an upper limit set at ISO6400. Obviously I consider my manual aperture setting and under low light conditions, I shoot an aperture as wide as possible consist with my desired depth of field. The camera's autoISO algorithm does a great job of selecting SS and ISO. I rarely see situations where I would have made even a slightly different selection.

Mostly leave it in Auto ISO. I'd rather get the shot than worry about noise. I'll admit that chroma noise does bother me though. Also we now have these AI noise reduction programs getting better and better each iteration, that do an excellent job.

And do not forget to put the camera on P. :)

P.S.: I usually shoot in Auto-ISO, sometimes in A mode, sometimes in M mode.

I hate ISO.

It was a constant in films

( It's a variable pain in DP)

😁😂😁😂😁😂😁😂😁😂😂🤦‍♂️

This is a technique that gives up headroom in situations where increasing ISO produces negligible noise reduction.

It is not constant if you can swap the back :).
When I was shooting film, I could not afford anything costlier than an OM-2n. My dream was a camera where I could swap the back and thus change the ISO that I am shooting with (or B&W vs. color),

Luckily, most Auto-ISO implementations reduce ISO and not exposure when applying negative EC.

Well, I couldn't afford mistakes and 64 was a little too hard, and 400 way too soft...
... and ended up shooting 90% ISO100 and some 200 :-D

( as old people, nowadays I rant about these modern boxes and their astounding photos)

I am guilty of feeding my ISOphobia.
Always maxing the exposure before touching that nasty settings. But fiddling with the camera recently, I see that the last stop of well capacity is nonlinear, thus exploiting maximum exposure capabilities does not come without price to pay.

AFAIK, the cameras use only the linear part of the sensor. Non-linearity is introduced by post-processors. How did you determine nonlinear behavior?

On all but one of the cameras I've tested, linearity was plenty good enough for normal photography at base ISO. And, on that one it wasn't bad at all. How are you measuring nonlinearity (I calculate a photon transfer curve), and what camera are you talking about?