When set to M mode with Auto ISO, my Sony RX10iv will implement bracketing by changing ISO. Yes, they call it "exposure bracketing". Not my fault.
I would expect that the raw file with the largest DR and (at least marginally) the lowest shadow noise would be the one which has the highest ISO without clipping in FRV.
I would expect also that the optimum raw file (above) would contain as much information as the entire gamut of JPGs, even though the JPGs have a wide range of lightness depending on the ISO used in any individual image. (I'm neglecting any effects of "multi-image noise reduction" here)
Yes, I have noticed that even in A mode with Auto ISO, AE bracketing will attempt to change ISO instead of, as expected, exposure (shutter speed). However, since ISO has little influence on noise, I do not see a point in bracketing where only ISO changes.
The raw file with the largest exposure (shutter speed and aperture) will have the largest DR.
The method did seem rather useless. Even the lower-ISO raws would add nothing to the mix.
I posted primarily in the context of the ongoing exposure discussions
Sherm
I took a set to confirm the described behavior, and the JPGs of course became progressively lighter, but I didn't examine and compare the raw files. That would remove the Gedanken ;-)
Interesting comment. I didn't expect that. Why would I use that feature - except if I wanted the best JPG?
Typically I'd select an aperture and shutter speed to match what I need for the photograph, and use auto ISO, perhaps with a bit of negative EV for safety. Then I'd process the raw.
For use in HDR, the sensor is exposed identically for the full series, so what would be the advantage there? I can just use the "best" raw.
I tried it with my camera. One stop, of course, is nothing but at four stops there’s no question that the high ISO RAW images were way blown out and not recoverable. So on my Canon, yes, the ISO setting definitely matters.
It matters on the Sony also. The best raw is the one which has the highest ISO without areas of saturation, but working below that ISO would be of marginal cost because the exposure is fixed.
The photon noise is fixed, but the post-gain read noise is not with most cameras, and the pre-gain read noise can drop significantly if the highest ISO setting goes into the upper range of dual conversion gain.
OK. Thought you were implying that the low ISO files had something to add. You're saying that the shadow areas of the high ISO/clipped images would have useful (less noisy) information?
You brought up HDR, not me, but I stated why HDR would still be useful with a fixed exposure and a range of ISOs (cleaner shadows from a frame that might blow the highlights). I did not recommend this over variable-exposure HDR.
The primary benefit to shooting RAW at a lower ISO vs.a higher ISO with the same exposure is dynamic range/highlight headroom. As read noise generally improves at higher ISOs, especially beyond the dual-gain threshold, if you don't need maximum headroom, you really don't want to "underISO" for no good reason (and your images will be darker in the EVF and for culling/initial assessment) but, within reason, you will have extra highlight headroom to work with, and there's typically very little penalty to pay in terms of noise if you want to defer much of the "ISO brightening" until later in post.
These are two examples using the Fuji X100V (dual gain threshold, ISO 500), both were the same sensor exposure, but one's at ISO 800 (L) and pushed (3 2/3 stops) in post, and the other is at ISO 10000 (R), identical processing otherwise. There is a slight difference, but it's probably hard to see here (at around ISO 1600+ there's no discernible difference).
I know of a professional photographer who shoots with m43 cameras and uses bracketing with HDR merge to reduce noise in shadows and/or manage high contrast. When shooting with a flash, the aperture and shutter speed must remain constant. Therefore he brackets ISO (manually), and HDR merges in post. His ISO bracketing range covers the dual conversion gain (DCG) point where the biggest reduction in read noise occurs.
Has anyone attempted ISO bracketing that covers the DCG as well? I tried it once quickly (no flash) but did not see any benefits.
The idea is to shoot with an exposure that does not clip highlights and then with an exposure with lower read noise (higher ISO) that clips highlights. The merge should preserve highlights while reducing noise in deep shadows. That noise reduction is much smaller than you gain by increasing exposure, as in typical AE bracketing.
I’m not sure about other cameras, but in ”ISO Bracketing” mode, Fuji cameras produce 3 jpegs of different brightness from a single exposure (and three functionally identical RAW files). As the ISO bracketing is being implemented post-ADC, and the three RAW files are identical, there won’t be any noise advantage there.
I’m not sure about what happens when you’re in normal “Exposure Bracketing” mode above base ISO though, you do get three unique exposures there, I think (and the possibility of misalignment), but I’m not sure how the ISO is being implemented. Sounds like an experiment to conduct.
I was thinking about real ISO bracketing, where the camera's ISO changes with each shot. You can do it either manually or with AE bracketing and Auto-ISO (as long as ISO bound is not reached).
As you wrote, the camera's automatic ISO bracketing typically takes one shot at one ISO and then generates images with fake ISOs.
If you have three "ISOS" generated from the same digitization of the same exposure, and one that is not "0 EC" looks like the "correct" brightness, what would be "fake" about it? Should it not be the actual ISO exposure index?
Seems to me that it's "lightness bracketing" at one fixed exposure (I think even DonaldB would agree) creating files processed at multiple ISOs. None is more correct than the others.
"Fake" may have been too strong a word. I meant that you do not get the same effect and benefits as when changing the ISO manually.
I wish the "fake-ISO" functionality would be available for single shots in the camera whenever the camera is in ISO invariant range (fixed ISO, with a brightness tag).
The only practical value of the concept of "ISO" to me is that a certain exposure of a middle grey object that may or may not be in the scene is at perceptual middle grey in the output, and it doesn't matter how it got there, to be real.
