How to get the best raw exposure, part 2

I re-read this several times because it didn't seem to make any sense but Jim "liked" it, but I still can't make any sense out of it. How do you "underexpose" by using a higher ISO setting with the same exposure that is not strong enough for ETTR at base ISO? You will be further to the right of the ISO setting that you wind up with.

I think a big part of the inefficient communication on such topics stems from the fact that some people are not distinguishing between the empirical phenomena of exposure, ISO settings, and raw files and the personal approaches to risk taking and optimization that people use. Facts vs editorial. Disagreement in editorial seems to some people to be arguments about the empirical facts, but they are not, and need to be kept separate. The empirical facts are fairly simple; much simpler than the range of "strategies".

There really is nothing contradictory about two people who understand the underlying facts choosing different "approaches".

One problem with a lot of discussions is that many people hear "mandate" when someone is describing an option.

I remember when I used to shoot slide film in the '70s; I sent or took the roll to a lab, and if my ISO exposure index did not match the ASA setting, I got dark or bright slides, unsuitable for aesthetic projection. The message is that exposure of digital cameras with raw is very different, and you get to trade off between SNR and headroom, with no need to even think about final image "lightness" because it is not deterministic at shooting time, if you just move some sliders in your converter to normalize lightness.

Less SNR due to less exposure. What is "greater proportion"? It sounds confusing and does not suggest a useful model of noise.

Actually, the more DR in the scene, the more you could benefit from "ISO to the right" with a given exposure, because you may have deeper shadows where read noise is more visible.

This reminds me of my first year law class in contracts. Was there a meeting of the minds? Was the rescission of the acceptance of the offer valid? I'm breaking out in sweat just thinking about it. 😂 Donald realizes that he's cornered, even if he doesn't fully grasp the technical reasons why.

Please, what's so important in this discussion that you are resorting to tricks? We are just talking, nothing life-threatening.

No, but "Helping David". Since we can try, why not.

I don't remember why I wrote that sentence, but I think what it is meant to convey is that with an ETTR exposure, if noise makes up, say, 1% of the data, it will be a smaller proportion of the data then when you underexpose and it makes up 20% of the data. And thus the greater the exposure, the less the noise will be visible. In effect, strong exposure drowns out the noise. It makes sense to me and thus is a useful model for me 😁

With read noise, raising ISO will only help reduce noise if the particular camera you are using does analogue amplification. And not all cameras employ this technique for their ISO dial. Also, with modern sensors, read noise is such a small component of the overall noise burden that even if you reduced it to zero, it still wouldn't make much visible difference. So not really worth worrying about. That was my take home from previous discussions.

I'm thinking a definition of ETTR may be helpful.

I find that shadow noise in high DR scenes where I'm forced to be well under the ETTR exposure at base ISO is usually less of a problem than highlight clipping, and that for the images that I consider to be of good photographic quality, photon noise is more important than read noise, even in the shadows. I do go to the lowest high conversion gain ISO in many circumstances, though.

Well, for the situations that you shoot in, you may be almost completely out of the woods as far as visible read noise differences over a range of ISO settings is concerned, but it is not happening for everyone. I can see a difference in the depth of the Canon R5's fine banding noise (single-horizontal-line blackpoint offsets with no vertical correlation) with 12-bit e-shutter vs mechanical, and HTP-vs-non-HTP, at ISO 51K.

If read noise were like a software Gaussian noise with no spatial correlation, then this would be moot, but read noise tends to have the have enough correlation that it can be visible, even when from a mathematical standpoint, it is dwarfed in quadrature. So, when I am shooting active birds with fast shutter speeds in the red-starved shade with high magnification with a relatively slow lens, HTP and 12-bit readout are both things that I try to avoid, and I don't consider the camera ISO-invariant for such purposes, even at the highest ISO settings.

I don't use that camera. I've not seen a camera with blackpoint errors like that. I don't think I've ever used ISO that high except for testing. All that can explain how we're looking at thing differently.

Are you photographing black cats in coalbins?

What you describe doesn't strike me as being a high DR scenario, which is the constraint Jim posed. If it's a dark, colorful bird in flight and strongly backlit by the sun so that most of its visible body is shaded and especially the under-wing feathers are deeply shaded but its head is drenched in sunlight, what would you do? It's a tough call for me because I find with my micro 4/3 cameras the character of the noise is visibly different between a pushed base ISO shot vs. a pseudo-ETTR'd version (same exposure but higher ISO). Not so much in any spatially correlated way, just in terms of the look and splotchiness of the "grain". My go-to noise reduction strategies tend to work a bit better on the pseudo-ETTR'd versions but admittedly the difference is usually quite subtle. On the other hand, I seem to find myself in lots of high DR scenarios where pseudo-ETTR'ing is going to be dangerous.

I read both parts of the article. I liked it, until it got to the math part at the end :)

I think I understand the concept you're explaining but is more difficult in practice. It has to be done consciously until it becomes a habit.

If you're happy with the raindrop analogy, you can safely skip the math at the end. It's there for people who want a quantitative feel for what's going on.