ETTR - Lots of room for push in post

Auto SS, auto f/stop, auto ISO = P&S with interchangeable lenses.

windpig wrote in post #13337920
But that would take all the fun out of it:p

Ha, but that's MY point. I don't have any fun reading and adjusting a histogram, I would much rather interact with my subject! Instead I say, hold on, let me fumble with my camera while I expose your face to the right! Can you smile while I stick my camera in your face since your teeth are whiter than your cheeks? Thanks. I know what I'm doing.

Higgs Boson wrote in post #13342213
Can you smile while I stick my camera in your face since your teeth are whiter than your cheeks? Thanks. I know what I'm doing.

Speaking of cheeks, my comment was meant as Tongue-in-cheek.

Higgs Boson wrote in post #13342213
Ha, but that's MY point. I don't have any fun reading and adjusting a histogram, I would much rather interact with my subject! Instead I say, hold on, let me fumble with my camera while I expose your face to the right! Can you smile while I stick my camera in your face since your teeth are whiter than your cheeks? Thanks. I know what I'm doing.

And that's my point. With a 5d2 and the sort of conditions under which you would be shooting a portrait - good light, low ISO, tonal range not more than, say, 10 stops, print probably not more than 12x18, etc. - the extra quality you get from ETTR would be marginal. A 20x30 landscape is a different matter.

But figuring exposure for ETTR doesn't have to be slow. Spot meter the brightest highlight in the subject and add 2.5 to 3 stops. Do that either by locking exposure and then recomposing in Av or if you prefer M, increase exposure until the needle hits +2 and then give two more clicks and recompose. Or you can do what I do: turn on Live View with live histogram, set the green histogram to the right, turn off LV and shoot.

windpig wrote in post #13342344
Speaking of cheeks, my comment was meant as Tongue-in-cheek.

so was mine.

and I don't ever shoot in controlled settings. 99.9% of my shots are outside, landscapes, environmental portraits, automotive, etc. i don't have a studio.

anyways, i am just the type that wants to maximize potential and enhance performance. if i can ettr even better than i am, i'm going to do it.

One thing that ETTR never considers is the fact that the scene's tonal range often exceeds the ability of the film or sensor to capture that full range. The implication of this failure to capture the full range is: The photographer must decide what fraction of that full range is important within the scene, and then sacrifice the capture of the segment which is less important!
In other words, in some cases it is some of the highlights which contain no important detail, and those pixels fall off the right edge of the histogram...which is what ETTR is trying to fundamentally prevent!
In other cases, the shadow areas do not matter, compared to keeping detail in the brighter areas of the photo, so the deep shadows are allowed to go darker...yet ETTR is fundamentally about better using the 4096 tonal values so that more levels are devoted to better capturing details in the shadows!

One must examine not only the histogram, but also the captured JPG preview, to determine what parts of the scene are blown out and whether or not they are 'worth keeping'. One cannot simply always 'move the pixels to the right of the histogram, and prevent them from falling off the edge'! And since sometimes the full tonal range can be captured, and in other times they cannot, 'proper exposure' is not a simply matter of always adding a fixed amount of offset compared to the meter reading. And because of the need to consider all of the above, exploiting ETTR to the fullest can never be a rushed effort...so often we simply have to ignore ETTR in the interest of simply making a shot when it presents itself on a spontaneous manner!

True stuff, Wilt! It's a bit amusing to hear so many people oversimplifying the idea of a "proper exposure" as if all photography took place in a "properly lit" studio -- it seems like such people have never stepped outdoors with their camera!

But, I'm always happy with a scene where I can, like you said, "ignore ETTR in the interest of simply making a shot when it presents itself on a spontaneous manner!"!

Wilt wrote in post #13347947
ETTR is fundamentally about better using the 4096 tonal values so that more levels are devoted to better capturing details in the shadows!

Wrong. Number of levels are totally irrelevant. When we ETTR, we have more levels, but this means no practical advantage. The only advantage of ETTR is maximize SNR, and this is its only reason of existence.

More levels mean nothing if you don't improve SNR. No matter how much or how little you expose, a RAW file always has enough levels to avoid posterization. Paradoxically, posterization due to lack of levels usually appears on well exposed areas of uniform colour (e.g. skies) when converting to 8-bit output.

The following 2 images have a very different number of levels (just look at the histograms), while both have the same quality and robustness against posterization:

Guillermo, thanks for chiming in! I'll let Wilt respond to this, since he knows more "stuff" than I do!

But, could you post a link to that second article if you by chance have it handy?

Tony

This may be it.

http://www.luminous-landscape.com …optimizing_expo​sure.shtml

_GUI_ wrote in post #13350214
Wrong. Number of levels are totally irrelevant. When we ETTR, we have more levels, but this means no practical advantage. The only advantage of ETTR is maximize SNR, and this is its only reason of existence.

More levels mean nothing if you don't improve SNR. No matter how much or how little you expose, a RAW file always has enough levels to avoid posterization. Paradoxically, posterization due to lack of levels usually appears on well exposed areas of uniform colour (e.g. skies) when converting to 8-bit output.

The following 2 images have a very different number of levels (just look at the histograms), while both have the same quality and robustness against posterization:

[GIFS ARE NOT RENDERED IN QUOTES]

The secret? SNR was low enough not to need more levels. This is exactly what happens in the deep shadows of all RAW files, they have few levels, but noise is sufficiently high to dither any banding.

There has been a long debate about this in the LL, after the second article by M. Reichmann about ETTR.

Regards

Bruce Fraser's analysis of the tonal range and how most of the levels (2048 of them) are used for the highest EV of the tonal range. http://wwwimages.adobe​.com …hop/pdfs/linear​_gamma.pdf

And although the title of this mentions s/n, the article itself dwells more on the distribution of the 4096 levels and how few of them are allocated to the lowest levels in the photo
http://www.luminous-landscape.com/tutorial​s/expose-right.shtml

There is a connection between allocation of digital levels, and SNR. But note that SNR gets only a passing mention in both articles, which both dwell on the distribution of levels in the tonal range.

Tony

This may be it.

http://www.luminous-landscape.com/tu...exp​osure.shtml

Cool, thanks!

Wilt wrote in post #13350377
Bruce Fraser's analysis of the tonal range and how most of the levels (2048 of them)
(...).

If you read carefully my previous comment, I admit there are more levels the higher the exposure (in fact I have plotted dozens of real RAW histograms showing this on the undemosaiced data*), so we do get more differentiated RAW levels through ETTR.

What I say is that those extra levels are totally irrelevant, they will not improve your image's quality or robustness by themselves. The one and only reason to ETTR is because it improves SNR. If ETTR didn't improve SNR, it would be a total waste of time, no matter if it provided more levels. And for the same reason, performing ETTR in an application where noise is not a problem, is a waste of time and effort.

About Michael Reichmann's articles, I think he is quite obnoxious with the 'more levels myth'. He made the same mistake twice, in 2003 and 8 years later in his most recent article, helping spread more the myth that extra levels is one of the advantages of ETTR. The only practical usefulness of ETTR is improve SNR, and that's it.

If you are interested to read a well informed person (he is not a photographer administrating an online forum, but a scientist from the University of Chicago) writing about all these matters: Noise, Dynamic Range and Bit Depth in Digital SLRs by Emil Martinec.

Regards

* RAW histograms (R channel) in EV stops from a Canon 350D (linear 12-bit RAW files) and a Leica M8 (non linearly encoded 8-bit RAW files):

Isn't t the lop sided data distribution about what happens in the gamma conversion?