By Ctein
Got your attention, didn't I?
Unless you don't know what I'm talking about. Lucky you.
For those who haven't been force-fed this bit of dubious dogma, "Expose to the right" is a rule that asserts that to get the best quality in your digital photographs, you should push your exposure as far to the high side (the right side of the histogram) as you can without clipping the highlights.
Once upon a time, this rule made a certain amount of sense, although not as much as its proponents ever claimed. Once upon a time, digital cameras were pretty noisy beasts, and suppressing that image noise was one of the more important ways to improve image quality. Increasing exposure increases the number of photons counted, which improves accounting statistics. Hence, reduced noise. If you can do this without clipping highlights, it's a win.
These days, noise is really not a big source of image quality loss, unless you're that particular kind of photo-fetishist (see "The Photo-Fetishistic League"). Cameras and sensors are so much better. Clipped highlights, as Mike and I discussed last week, haven't gone away. It's still a big issue when trying to get real quality in a digital photograph.
The thing is, digital behaves like slide film—slide film with a really, really sharp toe. The toe of the film curve is low in contrast, so there is not a hugely abrupt transition from no detail in the highlights to a little detail. Digital is abrupt. When you hit the wall, you know you've hit the wall.
The worst thing you could do with a slide film was to blow out the highlights. Many professionals routinely underexposed their slide film to avoid this. Pictorially, the results weren't as great, but you could fix that in printing and reproduction. You couldn't fix blown highlights.
In theory, you can still use the dubious right-hand rule. Just be careful to never blow out any pixels. In practice, much easier said than done. Histograms and camera-back displays are only an approximation of what's actually in the file. Even when they aren't, highlights are frequently small enough regions of the photograph that they don't contain a statistically significant number of pixels. You may not even notice them in a histogram, and on that little screen on the back of the camera that shows you maybe one pixel in ten, highlight warnings may not show.
1. The original photograph was underexposed by more than a stop to maximize the amount of highlight detail it held. A straight conversion in ACR, unsurprisingly, looks quite underexposed.
2. For the sake of clarity in illustration, I've converted illustration 1 to black-and-white in ACR so that color doesn't distract from the important discussion.
Unless you're sure you're dealing with a low contrast subject, pushing your exposure to the high side makes it likely you'll blow highlights. If you're trying to improve your odds of getting a good exposure, pulling away from the right is a much smarter thing to do. If you know your subject is really high in contrast, pull far, far away from the right. Keep those highlights under control and let the shadows go where they may.
Of course, if you don't take steps to correct the tonal placement when you process your file, it'll look lousy. It'll be very dark, with middle-tones that look more like charcoal. Not a pretty sight. Kind of like illustrations 1 and 2.
This photograph was substantially underexposed to ensure that the highlights didn't blow out. Illustration 2 is just a desaturated version of illustration 1, because it's easier to see what I'm talking about without the confounding effects of color. Don't think this is just about black-and-white, though; blown out highlights are at least as annoying in a color photograph. Nighttime scenes like the one shown are especially high in contrast, which is why I'm using this as an illustration for the article, but the rules apply equally to normal daytime photography. This is not a special technique to be used in unusual circumstances; it should be your normal way of working.
3. Even with underexposure, there are some blown highlights, shown in red.
Even with this underexposure, there are some blown-out highlights. Illustration 3 is a screenshot from ACR with the highlight warning (bright red) turned on. In a really contrasty scene or one with glaring or specular highlights, something is likely to blow out, no matter what the exposure range of your film or sensor it is. Underexposing minimizes but doesn't always eliminate that. We can deal.
4. The straight-line curve and the accompanying histogram on the left are from the default conversion in ACR. The curve and histogram on the right are my custom conversion designed to compensate for the underexposure in illustration 1 and 2.
The left side of figure 4 shows the characteristic curve in ACR that I used to get illustrations 1 and 2. It's my standard default ACR setting. Below it is a histogram for illustration 2. Notice how everything is piled up near the shadow end, far from the right side. Overall, way too dark. To fix that, I used a curve setting like that on the right side of figure 4. It produces the histogram below that curve, and illustration 5 shows what the picture looks like with that curve applied.
5. The same photograph as converted with a custom curve. Good tonality, good shadow separation, and no harsh highlight transitions.
That curve has three major benefits. Obviously, it restores the overall tonality of the photograph to something normal without adding more blown highlights. Next, it produces good contrast and separation in the shadows, where it is hard for the human eye to see tonal differences. We see shadow differentiation in prints poorly, so some extra contrast down there is a good thing.
The third benefit, and it's a huge one, is the rolloff in the shoulder of the curve, lowering highlight contrast. That means that what blown highlights are left don't jump out as harsh white blotches; the highlight detail just gradually fades out into white, the same way it does in a well-made darkroom print.
If you really can't stand the idea of any additional noise in your photographs, run a noise reduction plug-in on the image. You won't need a very strong setting; the noise difference between a normally exposed and an underexposed digital photograph is not all that great. Just a whisper of noise reduction will take the noise level down to where it would have been if you'd exposed normally or even to the right.
Just, whatever you do, don't expose to the right unless you're absolutely positive there are no highlights to get blown. It was a questionable rule to begin with; these days I call it downright dangerous.
Ctein
Ctein, who is a high-end custom printer among other things, writes a regular weekly column on TOP that appears on Wednesdays.
Editor's Note: Several readers have referenced an article I was not aware of on The Luminous-Landscape about ETTR. That article was published in 2003, when, as Ctein acknowledges, ETTR made more practical sense. From 2003 to 2011 has been a very long time in this particular stretch of the history of photographic technology.
Send this post to a friend
Please help support TOP by patronizing our sponsors B&H Photo and Amazon
Note: Links in this post may be to our affiliates; sales through affiliate links may benefit this site. More...
Original contents copyright 2011 by Michael C. Johnston and/or the bylined author. All Rights Reserved.
Featured Comment by Joe: "In my opinion, digital photographers have gotten way too used to using Photoshop to bring out detail in the shadows.
"A few years ago I visited a wonderful exhibition in Tucson, at the Center for Creative Photography, of some of the great New York City street photographers, mostly vintage prints from the mid 20th Century. And it turned out to be an eye-opening experience for me, because it reminded me that photo printers didn't used to fear the shadows. There was lots of detail in the highlights, but those blacks were black. Whole swaths of those prints were completely without detail. And they were just stunningly gorgeous.
"That was the end of my expose-to-the-right period."
Featured Comment by Niels Geuze: "I have learned to expose the subject right."
Featured Comment by Bill Pierce: ""I used to feel terribly guilty not exposing to the right, just exposing the same way I did shooting an ungodly amount of slide film over the years. Thanks to Ctein my sin has been washed away and I feel whole again (and I’m often using a handheld incident meter that favors exposure for the highlights)."
Pierce, as he's known to his friends, wrote the "Nuts and Bolts" column at The Digital Journalist from 2006 to 2010. —Ed.
I did some tests of my own and what I discovered was there is no way to determine the presence of 'small' specular highlights on the histogram - in camera as well as in DPP/Photoshop. If these are missed during an ETTR exposure, the highlights look a lot worse. I agree with what you've written - most amateurs (like me) shouldn't try ETTR. I'm not sure most pros should either.
In case anyone's interested, my results are here: http://sareesh.com/2011/08/ettr-expose-to-the-right/
Posted by: Sareesh Sudhakaran | Wednesday, 19 October 2011 at 10:24 AM
Ctein, close but no cigar — close for high contrast subjects, that is, which you basically have to expose for the highlights just like slide film. But for anything with less contrast you do better by exposing to the right, for the reasons described in the Luminous Landscape article for this subject. Not just because of less noise in the shadows but also because of better gradation. The trick, if you want to call it that, is to then to pull back on the Exposure Slider in your raw processor, as otherwise the image will look overexposed.
Anyone who has a Mac would do well to download the free raw developer called RPP, which is optimized for digital exposure, as opposed to film exposure. When you open a file in RPP most files that have not been exposed to the right will look underexposed; and looking at the files in this way will give you a good indication of how much you need to pull back on the Exposure Slider in Aperture of Lifghtroom for files that have been exposed to the right. Also, read the documentation that comes with RPP and you'll see why digital exposure is different than slide film exposure, if you still need to be convinced of this at that point.
—MItch/Bangkok
Posted by: Mitch Alland | Wednesday, 19 October 2011 at 10:33 AM
Thanks Ctein. This matches my experience and understanding of the pertinent physics. See for example this article I wrote a while ago: http://lagemaat.blogspot.com/2009/08/high-iso-is-bad-right.html There is an outstanding analysis with lots of actual data of ETTR available here that you might be interested in: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#ETTR (linked to the part about ETTR but the article goes far further than that)
Posted by: Jao | Wednesday, 19 October 2011 at 10:36 AM
Amen to that brother. To be honest, when I first started with digital photography, this was my instinct - broadly speaking, treat the sensor like slides. I mentioned this to someone who I respect(ed), and they very nearly exploded with condescension, telling me that what I was doing (exposing to the left slightly) was idiotic, and everyone who was anyone exposed to the right. It took me a very long time to get over that, but happily I now know what I need to do to get the pictures I want, which is exactly what I started out doing. I might just send them a link to this article.
Posted by: Rowan | Wednesday, 19 October 2011 at 11:02 AM
Après Ctein, le déluge.
I can't wait to see what comments this one will attract!
Will
Posted by: Will Frostmill | Wednesday, 19 October 2011 at 11:03 AM
Ooooh its on now. I suspect to see some ruffled feathers for sure. I, and I think every other photographer who has ever shot a wedding know exactly what your talking about. In my portraiture I have always taken ETTR as unpractical and dangerous to employ. A bride spends $10,000 + on a bright white dress she wants to be able to see each bead and stitch. Thank you for coming out and saying it, a brave man you are.
Posted by: Richard | Wednesday, 19 October 2011 at 11:05 AM
I am not sure I understand this article. But, FWIW, I believe the theoretical benefit of exposing to the right was because the number of bits used to differentiate the higher luminance areas of the image is greater than the number of bits recording the darker regions of the image. So, underexposing on purpose would underutilize the available data points.
Here is an explanation that I read some time ago regarding this:
http://www.luminous-landscape.com/tutorials/expose-right.shtml
Posted by: James | Wednesday, 19 October 2011 at 11:09 AM
Thanks Ctein, nice, to the point article
A post to recommend
-Sylva
Posted by: Sylvain | Wednesday, 19 October 2011 at 11:10 AM
I don't think any of the proponents of ETTR were advocating: "expose to the right, and don't worry about the highlights being blown". If you take the issue of blown highlights off the table, then it's much easier to dig decent quality information out of the shadows if you are biased to the right.
Posted by: Tim Gray | Wednesday, 19 October 2011 at 11:10 AM
Absolutely right on! That's exactly how I handle it. And quite often I can even bring back all the highlights, in Photoshop Bridge. Of course, the color saturation and image contrast may also need tweeking at this point, but this is of less importance to me since I primarily work in B&W.
Posted by: Richard | Wednesday, 19 October 2011 at 11:12 AM
I love your photo-fetishist post you linked to. "[P]erfectly-tuned pianos playing Three Blind Mice" is perhaps the best description of the photographs they tend to produce I've ever seen.
Posted by: pepeye | Wednesday, 19 October 2011 at 11:20 AM
To clarify, I should note that contrast always requires some adjustment in my B&W work, so it's of no consequence for me to include it here.
Posted by: Richard | Wednesday, 19 October 2011 at 11:22 AM
Oh, you've done it now. I sure wish I could set up a lemonade stand here, because I think you'll be seeing a lot of traffic.
BTW, for the little it is worth, I don't disagree. This the approach that works best for me, but the semantic and technical debate to follow should be a good show.
Posted by: Mark Crabtree | Wednesday, 19 October 2011 at 11:38 AM
Here is my reply to Ctein's proposition about the practice of ETTR being a bunch of bull, and that is to quote Jeff Schewe one of the advocates of ETTR.
ETTR (Expose To The Right) doesn't mean Over Expose To The Right.
Quoting from Jeff's article about the topic:
"This is not intended to be used in every photographic situation, only those situations where the scene's contrast range is lower than you camera's sensor dynamic range. Clearly if you have highlights whose texture and detail is important, you wouldn't want to increase the exposure to the point where the highlights are clipping."
More here:
http://schewephoto.com/ETTR/index.html
Posted by: Don Bryant | Wednesday, 19 October 2011 at 11:40 AM
Hi Ctein,
thanks very much for explaining this, because last week I really wasn't clear on the difference between ETTR and expose for highlights. At first glance they look quite similar.
I also enjoyed reading the article on fetishists you linked at the start. Now if only one of your future articles could explain how I can measure the Flummox Coefficient of my photos, because that would make a great fetish!
Kostas
Posted by: Kostas | Wednesday, 19 October 2011 at 11:49 AM
Expose to the left now? :)
Posted by: toto | Wednesday, 19 October 2011 at 12:05 PM
I've been wondering about this again, myself, of late.
There's no denying that getting as much of the image away from the noisy end at time of shooting does make for a much nicer image. That much the EttR-ists have in their favour.
And of course your warning about the highlights is reasonable - although I'd say we *have* spot-meters, we *have* histograms, we *have* highlight-recovery[0], we even have braincells and experience spotting contrasty scenes and a knowledge of our sensors' finite range (DxO tell me I have 10.bit EV to play with), so to a fair degree these things can be established whilst shooting, too.
There's also a school of thought that says you should place the tones as near as where they're going in the final image, to avoid crunched-comb histogram syndrome. I guess, for normal to high-key subjects, that plays into EttR a bit as well - again, more of an 8-bit problem.
With the above logistics all pushing and pulling the exposure around, it's no wonder I've got used to semi-auto exposure modes and started concentrating on composition, because frankly it's far more important to point the darn' thing in the right direction than to fuss about how you're going to gratuitously prolong the photo-making process.
What I don't hear enough of is the idea of stacking. As the number of input frames doubles, simply averaging them together halves the effective ISO (as far as signal:noise ratio is concerned). You can see a huge quality difference just by shooting everything twice, and IME it's quite easy - just hold the shutter button twice as long in burst mode.
[0] for those of us not shooting ourselves in the foot by using ACR for raw conversion
Posted by: Tim | Wednesday, 19 October 2011 at 12:06 PM
I wish something could be done about the histograms. I remember being told they're generally based on the jpeg form of the image. Possibly they're based on the scaled-down-for-the-LCD form of the image as well, or something. (One of the many things that open-source firmware for the cameras would make fixable. On the other hand, bricking your Canon 1Dx would be kind of expensive. On the third hand, proper architecture and such can make it possible to always install new firmware no matter how bad the existing load (put that function and enough hooks to get to it in ROM)).
Posted by: David Dyer-Bennet | Wednesday, 19 October 2011 at 12:20 PM
In my opinion, digital photographers have gotten way too used to using Photoshop to bring out detail in the shadows.
A few years ago I visited a wonderful exhibition in Tucson, at the Center for Creative Photography, of some of the great New York City street photographers, mostly vintage prints from the mid 20th Century. And it turned out to be an eye-opening experience for me, because it reminded me that photo printers didn't used to fear the shadows. There was lots of detail in the highlights, but those blacks were black. Whole swaths of those prints were completely without detail. And they were just stunningly gorgeous.
That was the end of my expose-to-the-right period.
Posted by: Joe | Wednesday, 19 October 2011 at 12:20 PM
Thanks Ctein. I don't know from physics and I don't check my histogram very much, but this confirms my skepticism about "expose to the right," as I've always though of "don't blow the highlights" as my prime exposure directive.
And fortunately we've gone this far in the comments and nobody's mentioned HDR yet. ;-)
Posted by: Ed Hawco | Wednesday, 19 October 2011 at 12:26 PM
What's a histogram? Some kind of correspondence delivered from the past?
Posted by: photogdave | Wednesday, 19 October 2011 at 12:27 PM
OMG, Ctein, I trust you own a heavy flack suit! :) All this expose to the right, expose to the left... How 'bout the histogram simply reveal exactly what our files contain, and then we could expose exactly CORRECTLY?!? OK, sorry, heading to the closest mirror to slap myself like Mike did!! :) :)
Posted by: Alan Huntley | Wednesday, 19 October 2011 at 12:27 PM
Hello-
I'm sure Michael and the gang from the Luminous Landscape will be checking in soon...
http://www.luminous-landscape.com/tutorials/optimizing_exposure.shtml
http://www.luminous-landscape.com/tutorials/expose-right.shtml
MS
Posted by: MichaelS | Wednesday, 19 October 2011 at 12:29 PM
Until someone comes up with a sensor that gets at least 14 useable stops of range, I'll be exposing for the highlights.
Posted by: Harry Lime | Wednesday, 19 October 2011 at 12:32 PM
Yea, Ctein. Ettr - one of those great misunderstandings of digital color representations. Coming from a background in computer graphics, I just had to shake my head at how ettr was considered gospel. Thanks for setting the record straight (though I suspect you'll get some push back).
Almost as bad is what the histograms call on the right. For a giggle (or a frown) just take a look at what the histogram looks like for a gradient from red (255,0,0) to white looks like. Pretty much any pure hue gradient histogram will look pretty strange. An orange to white (255,128,0) is particularly peculiar. For a real thrill do the same thing in LAB mode.
Cheers, Michael
Posted by: Michael Gerald-Yamasaki | Wednesday, 19 October 2011 at 12:40 PM
Is this a conspiracy of bloggers?
Mr. Reichmann's, Luminous Landscape post for 18 October 2011 is titled:
"Understanding Criticism, Part 1 of 3"
;~))
Seriously, did you invite Mr. Reichmann to participate in this debate?
Cheers! Jay
Posted by: Jay Frew | Wednesday, 19 October 2011 at 12:45 PM
I think the whole argument boils down to subject matter and that both sides are correct. Shots taken in overcast conditions, or at night without artificial light sources, lend themselves to ETTR. Cumulus clouds lit by the sun, light filtering through leaves, or night time shots with artificial light, demand more care of the highlights. My camera on matrix mode tries to walk down the middle. I tend to nudge it one way of the other depending on what is in front of the lens.
Posted by: Pieter de Bruin | Wednesday, 19 October 2011 at 12:46 PM
Tim, could you expand your dig at ACR highlight recovery, or point at existing articles explaining your issue there?
Posted by: David Dyer-Bennet | Wednesday, 19 October 2011 at 12:47 PM
"And it turned out to be an eye-opening experience for me, because it reminded me that photo printers didn't used to fear the shadows. There was lots of detail in the highlights, but those blacks were black. Whole swaths of those prints were completely without detail. And they were just stunningly gorgeous."
Two factors to consider:
A lot of printers back then expected a lot brighter display lighting than is common today.
If the prints were for reproduction , engravers used to like darkish prints because they could dig detail out of the shadows of prints on fiber base fully developed paper.
Posted by: hugh crawford | Wednesday, 19 October 2011 at 01:08 PM
Ctein, you nailed it!
I have been wrestling with this for a while and now, finally, someone with the required gravitas has come out on the side of reason.
I can only justify my position empirically, but I know I shoot to prevent blow-out and damn the shadows.
Expose for highlights, develop for midtones and let the shadows fall where they will. Should be printed on the box of every digital camera sold!
(Now, please help with my other pet peeve - the erronious belief that you can "get it right in camera" and avoid PP. To me, getting it right in camera means giving me the best starting point for PP, not trying to avoid it altogether).
Posted by: Steve Jacob | Wednesday, 19 October 2011 at 01:11 PM
I've slowly adopted your approach over the last few years, not worrying about the shadows as much, and my photos have become much better to my eyes. And for me, in low contrast, "it depends" is the rule. If you are talking low contrast mostly white scenes, fog or snow, you had better expose right if you don't want gray (most of us know that). But if you are talking low contrast dark scenes with some midtones here and there, for me the photos still turn out better exposed for those midtones, as midtones, so expose left, and keep the shadows how they look, nice and dark.
Posted by: John Krumm | Wednesday, 19 October 2011 at 01:16 PM
There are limits to the scene brightness range that any and all photographic imaging systems can capture; There will always be scenes that cannot be captured without blowing out the brightest areas. In digital, the combined sensor/on-board processing/post-production process that yields the prettiest, or least-objectionable, blown highlights wins.
Posted by: Keith B. | Wednesday, 19 October 2011 at 01:20 PM
David Dyer-Bennet: Better, grab yourself a copy of RawTherapee and spend an evening twiddling impulse versus luminosity+chroma NR, USM versus deconvolution sharpening, saturation limiting, AmaZe demosaicing versus anything else, dark-frame subtraction, ... and anything else that takes your fancy that it offers. I have done so, for several ISOs, and am sufficiently happy with the results to use it as raw converter of choice. YMMV but the potential for knocking ACR into a cocked hat is definitely there.
Posted by: Tim | Wednesday, 19 October 2011 at 01:56 PM
David Dyer-Bennet wrote: "I wish something could be done about the histograms." My suggestion is to turn them off; they aren't that useful. Better to spot meter textured highlights just like you would have with slide film. Just like with slide film, you need to figure out how many stops you can open up before you clip. It's fairly simple, really.
Posted by: latent_image | Wednesday, 19 October 2011 at 02:04 PM
Dear Pieter (and Mitch) (and several others),
Absolutely, with some subject matter and some cameras, ETTR is a successful strategy. With most subject and cameras, it is not. Furthermore, the failure mode of ETTR is far more objectionable than the very modest increase in noise that comes with current cameras. That makes ETTR a terrible rule of thumb-- it fails more than it succeeds, and the failures are more objectionable.
BTW, note that the original ETTR was written in 2003. Back then, it may have made more practical sense. Today, sensor noise and data issues are NOT a predominant problem.
~~~~~~~~
Dear Steve,
Can't help you too much with your second pet peeve. They're not much different from the folks who thought that using Kodachrome was a virtue because there was no way to alter the results in processing or presentation.
So silly and of so little consequence that it's not worth a column to shoot it down.
~~~~~~~~
Dear James,
You are not likely to be able to to refute what you admit you don't understand. You are also not likely to accomplish anything useful by directing me to the very article(s) I'm refuting.
~~~~~~~~
Dear Tim Gray,
If you take the issue of blown highlights off the table, then you're living in a fantasy world that doesn't correspond to what happens in real life. No one, including me, ever said the THEORY of ETTR was wrong. The practice, with current equipment, is what fails (see comment above to Pieter et.al.). A useful rule of thumb is about practice, not theory.
~~~~~~~~
Dear Don,
I know Jeff's argument. Much as I respect the guy, and that his discussion is much more nuanced and sensible than the original, the method still fails more often, and in a worse way, than concentrating on doing what is necessary to avoid blowing out highlights. It is simply focusing on the wrong problem.
~~~~~~~~
Dear Tim,
Well, no, the in-camera and in-computer tools for avoiding blown highlights actually are inadequate to the problem. Currently. Quite possibly, even probably, in another eight years my advice will be as bad as ETTR (which dates from 2003) is now. But right now, highlight problems are hard to avoid if you follow ETTR.
~~~~~~~~
Dear Jay,
I didn't for a couple of reasons. First, Michael didn't write the original article(s), so he's not a party to the technical argument-- he's merely the publisher. Second, I'm not debating them. I am saying they're wrong, here and today. Maybe not eight years ago (although I am dubious). Very maybe not in in more years, But right now, wrong.
The essay on criticism is about artistic criticism, not technical. Not the same subject. It's a really nice article, BTW.
pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================
Posted by: ctein | Wednesday, 19 October 2011 at 02:10 PM
Just so I understand, Ctein, the adjustment curve is applied to luminance only after the image is opened in PS. I assume that you prefer this to bringing up the exposure or fill-light slider or otherwise adjusting the image in ACR.
Helpful post - Thanks.
Posted by: Bill Poole | Wednesday, 19 October 2011 at 02:26 PM
Sensor physics may change too fast for this topic to keep up :-) But as general rules, sounds like for digital we should still expose for the highlight (e.g. take a spot meter on the highlight that you care about, and place it at zone VIII) and for B&W film (I shoot both), we should still expose for the shadow.
When in doubt, shoot first and ask questions later :-)
Posted by: Richard Man | Wednesday, 19 October 2011 at 02:26 PM
I don't agree that ETTR is a bunch of bull, although it is when it's presented as a kind of photographic ideology. You just have to use your head and decide that sometimes, you need to ETTR -- it's basically an aesthetic decision. And it's not something to do if you always shoot jpgs.
Because you (Ctein) are interested in a certain kind of look, what you've done in your examples is "blown the lowlights." You have great blacks, with no detail.
On the Luminous Landscape, which is one of the more prominent advocates of ETTR, they take that position because they're basically *landscape* photographers. That means they're shooting in daylight (usually) and they're looking for a lot of detail in the midranges and the lower ranges, because that's what they *do.*
ETTR doesn't mean that they're willing to blow the highlights, but they are generally willing to crowd them if it means better detail in the lower and mid-ranges...because that's what they're interested in. They want, for example, to differentiate a lot of greens in trees. But greens can get quite dark, and if you don't ETTR, the masses of highly detailed greens at the lower end of the exposure range can take on a nasty, curled look in prints, especially large prints. As an accomplished printer, you may know how to fix that, but I, for one, do not, except to ETTR. Put another way, you're willing to lose dark detail sooner than many other landscape photographers...an aesthetic position, not a technical one.
I personally do most of my shooting on the street, in support of painting. I want a lot of detail everywhere, so I can see how things fit together. So I may take three or four photos of the same scene, at different exposures, and print all of them, and look at them sequentially. If you have, for example, a skateboarder on Venice beach, wearing dark clothes, I'll first expose for his clothing and skin, which usually aren't too far apart, to capture his gesture. Then I'll way ETTR to get all the dark aspects of the surrounding scene (if there are any -- the backs of buildings, cars, etc.) and then ETTL to pull details out of the clouds off the beach. Works for me...it just doesn't produce one full-range photo.
I think this was an exceptionally interesting article because it made me think more about what I was doing (as the LuLa articles did as well.) I would like to see more articles from you on curves, and how they might best be used to smooth out the shoulder and toe.
Posted by: John Camp | Wednesday, 19 October 2011 at 02:29 PM
The big issue is not ETTR per se, it's that histograms as they are implemented today are poor tools to judge blown highlights. Even more annoying is the fact that it's based on the JPEG produced, even if you shoot in raw.
Then there is the issue of white balance that can makes you blow just one color channel without knowing it. Some people use the UniWB technique to avoid that.
I would vote for an ETTR mode on the camera which would be smart enough to avoid the blown highlights. It would save me a lot of time doing bracketing, histogram chimping, looking at blinking highlights/shadows, etc.
I'd wish also for better tools in-camera to judge the exposure. Lightroom has these nice triangles at each side of the histogram to tell you if any channel is blown or black.
And a true raw histogram, please.
Posted by: Emmanuel Huybrechts | Wednesday, 19 October 2011 at 02:45 PM
What a great article, Ctein.
I've never actually heard the 'expose to the right' rule/dogma and have always done what you suggest, without really knowing why — it just seemed right. After reading your article, I know why it seems right!
Great stuff.
Posted by: Ben | Wednesday, 19 October 2011 at 02:49 PM
Ok, then, expose to the left only until the highlights are no longer blown. Same thing; less dangerous mindset, perhaps.
Posted by: Jeff | Wednesday, 19 October 2011 at 02:58 PM
For the past couple of years — has it really been that long? — I've set my Canon 5Dm2 to underexpose everything by one and a third stops by default.
Yes, sometimes that's overkill, and sometimes it's not even enough (in extremis I've underexposed by as much as two stops or more) but most of the time it seems about right for what I do.
And the results are really visible!
An additional advantage is the gain in shutter speed and/or ISO: I've come to love low ISOs, ISO 100 or even ISO 50, and at those ISOs a one and a third stop advantage can be telling in terms of shutter speed.
And of course the low ISOs mean less noise in evidence when I bring up the shadows in post-processing.
What I haven't done in all this time is tell anyone, or at least any photographer friends, what I'm doing (the exception being a dyed in the wool slide film shooter who seems to understand it all perfectly): underexposing just seemed so contra the conventional wisdom about ETTR. And sometimes I've even felt a little guilty about underexposing, like it was a secret vice or something.
So this post, and the many comments, have been very affirming. Maybe I'll even have the nerve to up the underexposure by another third of stop now …
Posted by: Richard Howe | Wednesday, 19 October 2011 at 03:13 PM
The only way you are going to prove your point to me, Ctein, is if you make two exposures of a subject, one exposed to the right, one exposed to your liking, and then processing each optimally in PS, printing, and demonstrating a discernible difference.
Which I do not think will turn out in your favor. Why?
Physics. My exposure has a lot more protons being counted than yours.
And because there is little reason to be afraid of blowing highlights these days unless you are shooting only jpg. ACR offers us localized exposure control by brush or by gradient; 4-slider parametric control of highlights, lights, darks and shadow; highlight recovery, fill light, and exposure sliders in the basic ACR module.
There is almost no chance, short of shooting straight into the sun or powerful light source, to have blown highlights - which had escaped notice in either histogram or flashing highlight warnings - that would not be easily recovered with these tools.
I have my camera jpg settings set at mid values so my histogram (and I assume my flashy highlight clipping warning)correlate well with my RAW exposure. I welcome small amounts of flashy warnings in review mode - it means I have effectively pushed my exposure far enough to the right and therefore maximized the number of protons my sensor has counted. Because I know how darned easy it is to recover those small areas of overexposure in post.
And like it or not, more protons means more information, and more information is better than less information. Especially at higher ISO settings, where your scheme of adding positive exposure in post +/- external noise reduction can in no sense be expected to outperform exposing to the right.
Yes, modern sensors are a lot better than older sensors, and one does not need to expose to the right to get perfectly good results, especially at lower ISO's. But Adobe Camera Raw (and other RAW developers) are also a lot better than they once were. Clipped highlights that can not be successfully recovered, and recovered locally at that, from exposures that are not wildly overexposed, are becoming rare as hen's teeth.
Starting with more protons yielding a brighter exposure (and therefore a higher signal/noise ratio) and then artistically making the image darker locally or globally will (must?) offer slightly better results than the other way round. Surely, that is simple physics at work? And why I think you have your work cut out for you if you are going to convince this photographer. :)
Posted by: Gingerbaker | Wednesday, 19 October 2011 at 03:21 PM
Bravo Ctein. I learned my lesson regarding ETTR the first time I had an image with the sun included. Instead of the graceful transition to complete saturation yielding a circular image of the sun that even slide film exhibits, I got an amorphous blob that looked like a fried egg without the yolk. Noisy information in the shadows is a heck of a lot better than no information in the highlights
Posted by: Tom | Wednesday, 19 October 2011 at 03:40 PM
Hey, what about those of us still shooting with cameras from 2006? My sensor ain't getting any better with age!
I'll ignore your advice for now :-)
Posted by: Miserere | Wednesday, 19 October 2011 at 03:53 PM
I loved how the applied curve mimics somehow the shoulder of film. This is reason enough to take this into consideration. I remember photographers (10 years ago) recommending to underexpose, even with lowly Olympus ZC20 or something like that.
This article seems synchronized with the announcement of the EOS X. If this camera deserves to exist, it ought to yield substantially better clean shadows than the 1Ds IV or bust.
Posted by: Eduardo Cervantes | Wednesday, 19 October 2011 at 04:01 PM
Great article/lesson. If I would have taken a shot similar to the example and seen a histogram with a spike in the middle and the edges flattened out I would have thought "I got everything I need in that exposure" Now I know that might not be so. Thanks.
Posted by: Mike Anderson | Wednesday, 19 October 2011 at 04:17 PM
You can also use Camera Raw / Lightroom's "Brightness" control to increase exposure without clipping. It has a very broad highlight "shoulder" similar to what Ctein has shown in Figure 4 (right side).
More generally: I think a more balanced view is that the more light you capture with a digital sensor, the better the signal-to-noise ratio. This is the reasoning behind ETTR. But like Ctein said, there is a hard wall on the highlight end: if you set the exposure too high, you will clip (possibly important) highlights. Conversely, the shorter your exposure, the more likely you'll be able to preserve important highlights. But you'll also have lower signal-to-noise, so you may have noisy shadows. This is the fundamental tradeoff.
For many scenes, there is a single exposure setting that will optimize both: maximize signal-to-noise while preserving the important highlights. The real problem, in my view, is that current cameras makes it really, really hard to determine what that exposure setting is. (The feedback/preview mechanisms in current cameras are optimized for JPEG capture, not raw capture.)
Posted by: Eric Chan | Wednesday, 19 October 2011 at 04:20 PM
Good one, Ctein.
I too read the Luminous Landscape article and jacked up the exposure - now I'm back to either 0 or -1/3 compensation for most subjects of average to high contrast.
And I was pleased to see your curves remedy for the high contrast scene - just as I use.
In addition, I sometimes lower the top of the graph slightly down the Y axis (it will rescue one or two channels for persistent blowout), and is MUCH better than using the highlight rescue slider in ACR (less banding etc).
Someone commented above with the now cliched "no cigar" but I reckon you can takes as many puffs as you like.
Posted by: Dohmnuill | Wednesday, 19 October 2011 at 04:31 PM
Ok I will admit it, I am probably one of those damned Photo Fetishists and throw bricks if you wish but over the years I have developed many ways to squeeze just a little more from those captures and and one of those is ETYE. Expose to your experience.
I ran many tests to find out exactly what exposures gave optimal results for all sorts of conditions with my camera and lenses and settings. Now all I do is put the camera on manual, set the base exposure and fire away. I adjust aperture/shutter combinations up or down the scale as creatively desired. No histograms, no exposure compensation issues, heck sometimes even discipline myself to not look at the screen post capture. Very quick and very precise once you have invested the time to work it out, and I don't have to spend time trying to second guess what my histogram and meter are telling me.
It's not a new idea just a refinement perhaps of the "sunny 16" concept but it really works......no BS.
Posted by: Brad Nichol | Wednesday, 19 October 2011 at 04:33 PM
Ctein, if I read you well you are not per se opposing to expose to the right, but warning of the dangers of blowing out one or more channels in the process, mainly due to not precisely knowing when this clipping occurs.
This is when the technical specifications of each camera model come into play.
I currently use two camera bodies. With the D700 it is generally possible go quite to the right without blowing highlights. Often enough there was some blinking at capture, but once imported into Lightroom you realise that all channels are fine. With time I learned to use to cautiously expose to the right (CETTR?). Applying this technique to the recently acquired G3 proved to be a disaster. Not only would the highlight clipping warning mean just that: irrevocably clipped highlights, but all too often stark clipping is already taking place when you thought it safe to conclude from the histogram that the picture was still underexposed.
As a result I use ETTR with the D700 and a method close to the one you described with the G3.
I wish the manufacturers gave us the option to display the histogram of a chosen color channel spread out over the entire screen. Better histogram legibility would greatly reduce the hazard stemming from ETTR.
Posted by: CMS | Wednesday, 19 October 2011 at 04:46 PM
If I might humbly offer up one maybe-not-so-counter point...
I agree that most of the time, modern dSLRs are not noisy beasts. However, for those of us who shoot, say performance, where low light and high ISOs are the rule, but who don't live in the rarified D3s or 1D brackets, it is still pretty meaningful to avoid underexposure in important areas of the photograph. Situational choices and horses for courses, I guess.
Nonetheless, enjoy the read and the active thinking it encourages.
Posted by: Marshall | Wednesday, 19 October 2011 at 05:08 PM
@Richard: that same bride doesn't want see her beautiful white dress as grey either. So that dress would be a perfect example where you do want to expose the right. It's very easy to meter it, with your histogram. Expose the dress to the right (if it's white it should be to the right to be right) and your other tones will be spot on as well. Underexposing wedding photographs is the worst thing you can do. Nobody likes muddy, dark wedding photos, they should be nice and bright.
@Ctein: you seem to be advocating not overexposing highlights. I couldn't disagree more. Some photos benefit from (sometimes large) overexposed parts, especially when it means correctly exposing other parts of the photographs. I.e. when you shoot backlit, you will always have overexposed parts. Protecting those highlights would ruin those photos, and no shadow recovery on earth could fix them.
So let's just expose right. Whatever that means for you. Like it dark, expose the left if that's the result you're looking for.
Also, exposing to the right in a low contrast situation seems kinda wasteful to me. Yes, you can restore the photo to the proper tones by pulling it in raw conversion, but to me that's like cropping. I like to get it right when I make the photo, both framing and exposure (that doesn't mean I won't change things in post if I have to).
Posted by: Jan | Wednesday, 19 October 2011 at 05:16 PM
My version is: expose as far to the right as you can. If highlights are unimportant in a casual shot, blow 'em out and enjoy the midtones. If highlights are critical, don't take chances. But it's more important to me at high ISO. The more light you put on the sensor, the less noise, so when shooting indoor events (ice hockey, basketball, school concerts, plays, dance recitals) I find it ideal to set my exposure (exposure meaning aperture and shutter speed, since those - and NOT ISO - control how much light you "expose" the sensor to) and then vary the brightness as needed. Ideally, my camera would do this with Auto ISO, but as I don't shoot Nikon or Canon it doesn't. So I end up setting a low ISO (low enough that I'm not overexposing anywhere I might be shooting) and adjusting exposures as needed in Lightroom. It's a nuisance, but I don't shoot these events too often. But this way, I'm effectively getting ISO 1600 when needed, and ISO 800 or 1250 when it's not. Exposing as far to the right as possible.
Posted by: Dennis | Wednesday, 19 October 2011 at 05:40 PM
Based on your post title I was expecting you to tell us that exposing to the right was technically a bad thing but really, all you are saying is that when you do expose to the right, you risk blowing the highlights. That seems fairly obvious to me, so I guess the point of this article is to say, be careful.
Posted by: steven palmer | Wednesday, 19 October 2011 at 05:50 PM
Actually ETTR proponents usually specifically warn against blowing out of highlights. So it is not a fair criticism.
However the reality is that most of the pictures usually have DR broad enough that ETTR is just not possible. The example that Michael@Lula gives is that of a Black cat over Black Background. Well, I never take pictures of black cats on black backgrounds, most people don't. So IMHO, ETTR is a good idea, but mostly moot.
Another thing to remember is that when you have to compromise between shadow details and highlight details, compromise in favor of highlights. The amount of shadow detail that can be recovered in post is amazing, to say the least. I can underexpose by four stops, and bring up the shadows with a simple pull of Fill slider in LR. It will be noisy but details will still be there. But if I overexpose even by one stop, highlight detail is simply gone.
Nevertheless, when going for the ultimate image quality, if the scene allows for it, ETTR WILL allow to eke out the last ounce of IQ from sensor.
Posted by: Z.K. | Wednesday, 19 October 2011 at 07:26 PM
Emmanuel,
The problem is that a very accurate histogram doesn't solve the problem that specular highlights are sometimes so far off to the right that exposing to encompass them in the histogram makes the rest of the photograph unusable. If you'll refer back to my Halloween picture of the graveyard from a few days back, which combines several different exposures, the exposure that captured detail in the moon has almost no detail anywhere else--not recoverable in post. I think a histogram that "acknowledges" the reality of specular highlights (headlights, the sun, etc.), however dumbly, is probably just necessary. Even though it's not very technically elegant.
Mike
Posted by: Mike Johnston | Wednesday, 19 October 2011 at 07:40 PM
"There is almost no chance, short of shooting straight into the sun or powerful light source, to have blown highlights - which had escaped notice in either histogram or flashing highlight warnings - that would not be easily recovered with these tools."
Gingerbaker,
That's...crazy, man.
I do so very totally disagree with you.
Mike
Posted by: Mike Johnston | Wednesday, 19 October 2011 at 07:46 PM
I came late to digital photography. I started out determining exposure just like I would for 4x5 chrome. It worked. I never saw a reason to change. Like my Dad always said " if it ain't broke don't fix it"
Dan
Posted by: Dan Berry | Wednesday, 19 October 2011 at 08:07 PM
@Gingerbaker: "Physics. My exposure has a lot more protons being counted than yours."
Physics has come further than I thought. Where can I get one of these proton-counting cameras?
Posted by: expiring_frog | Wednesday, 19 October 2011 at 08:45 PM
Dear Jay,
Oops, I was wrong. Michael did write the original article. Dunno how I messed up on that. Sorry. Still, not an invitation to debate.
~~~~~~~~
Dear Bill,
Yeah, it's the curve shape that's important. You want to get a nice rolloff in the highlights, so the near-whites don't slam into the wall.
There may be other combinations of tools that will do this, it's just that curves is the one I know the best.
~~~~~~~~
Dear JC,
No, you're trying to read way too much into an illustration. I haven't lost the shadows at all. The histogram shows that.
In fact, the file I used to create this article's example is the same one I used for this column:
"It Doesn't Matter How You Get There If You Don't Know Where You're Going"
http://theonlinephotographer.typepad.com/the_online_photographer/2011/02/it-doesnt-matter.html
Are you REALLY unhappy with the shadow detail in that figure 6???
The difference is that in this article I wasn't trying to make it look really good, only restoring the shadow-midtone value placement to what it would have been if I hadn't "under"exposed. In other words, a "properly" exposed file would still have had those apparently-but-not-really inky shadows.
For that matter, this is how I work all the time. I don't recall you noting any obvious lack of shadow detail in the prints I've shown you.
Have you also not noticed that I, too, am basically a landscape photographer?
There are certain kinds of highlights one does let blow out, whether film or digital, B&W or color. When clear skylight filters through the gaps between branches and leaves in forest trees, we're all pretty OK (even Mike) with letting those go to pure white. In fact, we pretty well don't have any choice, regardless of the medium. Rules of thumb are only rules of thumb.
Your method of making digital photos makes a HUGE amount of sense for a painter. It's great. You should write a more detailed article for Mike about this. Seriously.
pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================
Posted by: ctein | Wednesday, 19 October 2011 at 09:41 PM
Dear Ginger,
Ummm, you do understand that arguing physics with me is not playing to my weak suit, right? [grin]
But if you insist, I shall take the first exceedingly cheap shot.
Your camera detects photons, not protons. Hydrogen is made of protons (99+% by mass/energy), light is made of photons (100%, no added ingredients).
I know, it was almost certainly your auto-spell correct, but how could I resist?
(What? Rules? In a physicists' fight?!)
S/N improves with the photon count, but, so what? The S/N is so low, in practice, that the visible image degradation from less exposure, at low ISO's, is far, far less objectionable than blown out highlights. It's simply not a major visual factor, with decent cameras used at low ISOs.
(High ISO's don't enter as much into the discussion because exposure range drops with ISO. E.g., if your favorite speed is 3200, you ain't hardly ever gonna have to worry about ETTR, because you've got no spare room to the right to begin with unless the scene is quite low in contrast.)
"Clipped highlights that can not be successfully recovered, and recovered locally at that, from exposures that are not wildly overexposed, are becoming rare as hen's teeth.."
You'll have to define what you mean by "wildly-overexposed?" But on the face of it, ummm, no. Maybe for you, with your camera, and your choice of subject matter. For the typical good photographer with a good camera, ETTR is more often a fail than my method and the failure modes is much worse.
"There is almost no chance, short of shooting straight into the sun or powerful light source, to have blown highlights - which had escaped notice in either histogram or flashing highlight warnings - that would not be easily recovered with these tools."
Even more strongly, ummm, no. And furthermore, no. 'Cept maybe for you and some other scant minority of the photo universe.
But, ummmm, no.
~~~~~~~~
Dear Eric,
Thanks for the tip. As I said to Bill, it's likely there's more than one way to get a highlight rolloff. It's just curves are what I know. And I like that I can see exactly what's happening (what starting values end up as what end values). But it doesn't mean it's inherently superior.
I suppose one day in my copious spare time I should really do a systematic and methodical study of how the different controls correspond to different transforms. Might learn something (been known to happen).
~~~~~~~~
Dear ZK,
Fairness has nothing to do with it. It's bad PRACTICAL advice. Theory be damned.
Yeah, they warn against blowing highlights. Of course. What they don't say is that it's mostly unworkable and that its failure mode is a lot more objectionable.
pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================
Posted by: ctein | Wednesday, 19 October 2011 at 09:48 PM
In my two or so years of exposing to the right while avoiding blown highlights, I have routinely found that avoiding blown highlights in most cases (virtually all pictures except very low contrast scenes) actually resulted in underexposure.
Nowadays I tell myself: expose just enough.
The point is not more exposure. The point is not less exposure. The point is the right exposure. (A witticism offered with apologies to Mike :) )
Posted by: Zeeman | Wednesday, 19 October 2011 at 09:53 PM
Dear Pierce,
Just returning the favor. Years ago, you blessed me with your Prone System* of B&W exposure and development. It saved me much anguish. The negs still look gorgeous.
Grace under pressure. Always.
pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================
(* Overexpose by a stop. Underdevelop by 30%. Get on your knees and pray. **)
(** Zones? ZONES?! We don' need no stinkin' ZONES!))
Posted by: ctein | Wednesday, 19 October 2011 at 09:54 PM
@ Jan. Your right I wouldn't make her wedding dress go gray, my point is a bride in a white dress standing with others in black tuxedos and maybe dark dresses there is going to be a whole lot of information on the left side of the histo just to ensure that the dress doesn't blow out. More importantly the formal portraits are only a few hours of a long day and the rest of the time is spent reacting very quickly in constant changing lighting and one needs to employ an exposure strategy. In my strategy I would rather brighten a gray dress then pull it out of blow out. As Ctein says its much easier to lighten something and add a noise reduction then have no hi-lite info at all. Thats what ETTR is, an exposure strategy, but in my opinion a more dangerous one especially when you don't have five min to examine a histogram and re-expose for each shot.
Posted by: Richard | Wednesday, 19 October 2011 at 11:03 PM
Citrin,
I get what you are saying, but an inexperienced person could easily simplify your argument to "always protect your highlights from clipping," which is just as bad a rule of thumb as ETTR.
The truth is, we need to understand proper exposure, proper exposure is a complicated thing, and no rule of thumb is useful more than it is not. If we are unable to learn exposure, we're as likely to be successful trusting the camera's meter with no exposure compensation as we are following any dogma.
Posted by: Will | Wednesday, 19 October 2011 at 11:27 PM
I can't wait until my photographic skills have developed to the point where arcane topics like ETTR are the final frontier. Until then, I will continue to work on composition, narrative, voice, subject matter, concept, etc... You know, the things that normal people look at.
Posted by: JohnMFlores | Wednesday, 19 October 2011 at 11:33 PM
I use ETTR all the time. Not because Michael Reichmann told me so but because I think my pictures look better when I pull the exposure down in my RAW converter. The pictures look richer somehow.
I guess sensors have advanced a lot since 2009 because pulling out the shadows in a 5D Mark II doesn't look so hot IMO.
The new Canon 1D X has built-in HDR, so you can expose AWYP (Any Way You Please) henceforth. Of course I always thought it was expose AWYP since forever.
Posted by: TrinDreamscape | Wednesday, 19 October 2011 at 11:54 PM
Dear expiring,
Me, too, I want one! I'm positive we'd get a real charge out of it.
(Oh lord, it's down to bad physics jokes. Readers, run, run fast. Save yourselves while you can!)
~~~~~
Dear Will,
No. No. Really, really, really NO.
An "inexperienced person" doing this is going to have a much, much higher percentage of failure with ETTR and the consequences of the failures will be more serious.
All rules of thumb are not created equal.
pax / Ctein
Posted by: ctein | Thursday, 20 October 2011 at 12:03 AM
Ctein,
One of the cleverest (probably unintentional) witticisms in this exchange is the guy who addressed you as Citrin, which would suggest you are some sort of lime, or perhaps a grapefruit...I almost hate to think it was a simple misspelling.
I certainly don't think you missed too much in the way of values in your shot #6, referred to in your answer to my previous post.
[But] it has always seemed to me that one of the hardest things to get right is massed foliage in color, which is what they mostly deal with on LuLa. You have extremely complex detail in a variety of subtly different greens, sometimes of the same and sometimes of different values and trying to maintain both the separation of the greens AND the logic of the detail is difficult. I'm sure you've seen the kind of granulated noise that breaks up that subtle detail. (I say 'logic" of the detail because there are lots of ways to represent foliage that the eye rejects, because the eye knows what it really looks like.) That problem is much muted in B&W where you eliminate the problem of color, and the close-value greens are simply lumped together, while maintaining the logic of the detail.
...If that makes any sense.
As you know, I've actually seen a good deal of your work and I don't consider you a landscape photographer, not in the conventional way that phrase is used. (And I've been using it conventionally.) I've never seen one of your photos where I thought what you were shooting was *really* the landscape. They all appear to me to be photographs of an intellectual or emotional condition, as with the images of the Minneapolis waterfall.
Posted by: John Camp | Thursday, 20 October 2011 at 12:29 AM
Ctein is this what built-in "high dynamic range" settings in advanced cameras do (eg. Nikon SLRs, Fuji X100 etc)? Underexpose then drag out the values in the raw processing?
I noticed that the wide dynamic range setting on my X100 leads to shooting at ISO 800, even in broad daylight.
Posted by: Kelvin | Thursday, 20 October 2011 at 02:21 AM
so, 8 years ago reichman said: "expose to the right, but make sure you don't blow the highlights". Today Ctein comes and says: don't listen to that, because you will blow the highlights... what changed?
Posted by: Dan | Thursday, 20 October 2011 at 02:25 AM
It's amazing how much misunderstanding this topic brings to light...
Some food for thought :
1 One of the causes of misunderstanding might be that one of the main arguments for ETTR, under-quantization, does not seem valid - the other one, noise, just renders it moot. You just don't need high-bit quantization to render correctly noisy parts of an image.
That said, noise is a very good argument for ETTR in higher-than-a-few-stops contrast scenes.
I completely agree that for low contrast scenes where there ain't any problematic noise in the shadows, ETTR does not yield much gain (I mean proper ETTR without blown highlights of course).
2 Another one that I observed here is that, as said, ETTR is not Over-ETTR : for me, ETTR just means expose for the highlights, which is quite often under-exposing rather than over-exposing, and which is very close to what I did for slides back then too.
Blown highlights, including the blown individual color channel, are a failure according to the ETTR theory too.
As for a quality difference between almost-but-not-clipped highlights and unederexposed highlights, as long as any individual color channel is not clipped (which of course does belong too to the ETTR theory), I personally don't see any, and having one would not be very coherent with the way digital sensors work (linear up to saturation). Some experimental evidence of such a difference could help the debate.
3 As said here too, ETTR would be much easier, and therefore more successful, once a clever manufacturer implements a true raw histogram showing the 3 color channels (and not only the L one, and not something coming from a processed jpeg).
All that makes me think that ETTR is still a more valid method of exposure than the XXth-century middle-gray-based one used in many of our XXIth century cameras.
I'm very thankful to Olympus for the configurable Spot-Highlight metering mode of my E-PL1 - it could even be better (taking into account raw color channel values rather than luminance), but it helps me having more detail into the shadows - which is very useful for the kind of images I like, especially with a 4/3rds sensor with less dynamic range as a K5's one.
Posted by: NikoJorj | Thursday, 20 October 2011 at 02:32 AM
I'm really glad that someone else thinks the same about exposing to the right. But I'll also add that "it depends..." I think that different exposure meters will favour different exposures.
For example, my GF1, seems to prefer exposing to the right, and my K-5 so far seems to be slightly left of that. But that's just my gut feel after a couple of hundred exposures on the K-5. But I always had a problem of clipped highlights on the GF1 -- drove me insane.
Pak
Posted by: Pak-Ming Wan | Thursday, 20 October 2011 at 02:33 AM
—Mitch/Bangkok
I did that in a reaction to the cliched and overwrought title of this article, which is, in reality making only a minor point: when exposing to the right be careful about blowing the highlights — and, perhaps implicitly, there is the secondary point that it isn't always possible to expose to the right. It seems to me that a more valuable point, is made well by the documentation for the RPP raw developer, that the optimum digital exposure is often different from the optimum film exposure, although sometimes the two may coincide.Posted by: Mitch Alland | Thursday, 20 October 2011 at 03:18 AM
Advice such as ETTR, like all such rules of thumb, can't really be "right" or "wrong" advice in my view. They are a descriptive way of categorising explanations for what went belly-up in a particular situation, after the fact.
Take the two equally valid proverbs "look before you leap", and "he who hesitates is lost". A wise person retains both rules in parallel. When something happens or fails to happen, one of these two proverbs is wheeled out and pointed to, with the condemning words - "you see? I always told you so!" Of course the other one remains quietly out of sight, so as not to spoil the drama of the moment. The fact that "he who hesitates is lost" was never meant to recommend that people jump off a cliff, is forgotten - when someone has just jumped off a cliff. "Look before you leap" is instead selected as both simpler and more fitting... since we are temporarily not thinking about other situations where opportunities are missed through dithering, and that justify the other rule.
Posted by: richardplondon | Thursday, 20 October 2011 at 04:53 AM
JohnMFlores,
No-one likes a smart-ass!! :)
(But I take your point!)
William
Posted by: William Walker | Thursday, 20 October 2011 at 04:53 AM
Good article, with enough comments to cover both sides of the issue.
I have been trying to execute on the ETTR, with poor results, especially Weddings and Action Events where I have to shoot quickly and have extremes in contrast.
Slightly underexposing has been my fallback position recently and I am glad to see support for that.
ETTR may have its place for certain photographic genres but from a practical standpoint slight underexposure has given me great results in what I do.
Posted by: Stan Semuskie | Thursday, 20 October 2011 at 08:24 AM
In photography, the hard and fast rules regarding exposure are subject to interpretation. My mentors taught me that previsualizing the scene before taking the picture is the way to find the path to proper exposure. Whether it requires exposing to the right, left, or the middle, is a matter of art and science. ACR, particular camera sensors/processing engines, lens characteristics, ISO and signal-to-noise ratios, dynamic range,and the mind's eye all fall into the mix. I neither agree nor disagree with Ctein. He used a method that rendered the result he intended or fortuitously stumbled upon. I would have exposed the subject he used to illustrate his point differntly. The biggest thing that I have learned about taking pictures in the digital age is that flexibility regarding the triangle--shutter, f/stop,and ISO, affords more elasticity now more than ever.
Posted by: Bob Rosinsky | Thursday, 20 October 2011 at 08:34 AM
Bravo Ctein.
Posted by: Fernando Terrazzino | Thursday, 20 October 2011 at 08:38 AM
I believe that these highlight-expansion modes that you find in most DSLRs these days are essentialy doing what was explained here, underexposing 1 stop and pulling up the shadows.
Posted by: Fernando Terrazzino | Thursday, 20 October 2011 at 08:42 AM
Having tried ETTR in the past, I've now settled on getting the exposure close to where I want it in the final print, unless blown highlights are an issue in which case I underexpose. In my case it hasn't been just about retention of highlights- the other reason has to do with colour. For reasons I don't entirely understand (perhaps someone with a little more technical knowledge could shed some light on this) I find that if I take two exposures, one at roughly the correct final exposure and the second overexposed (without clipping highlights) and then pulled back, I strongly prefer the colour from the first case. The colour seems to come out flat and lacking in vibrancy in the overexposed file, even when it has been brought down to the same exposure.
Posted by: Cary Seipp | Thursday, 20 October 2011 at 09:00 AM
Is ETTR really so popular and well known that it requires this sort of response? I've been shooting digital since 2001 and I've always thought the conventional wisdom was to underexpose by up to one full stop because, as noted above, digital behaves more like slide film than negative. The biggest problem with negative film is/was underexposure -- if the shadows are too dark, there's nothing on the negative to work with. With digital, the biggest danger is overexposure -- if the highlights are blown, there's no information in the file to work with.
Posted by: Robert S. | Thursday, 20 October 2011 at 09:26 AM
Thanks so much for this post. I was unclear before what curve should be applied in this situation, but your example is a model of clarity, and I'll certainly be trying your custom curve in future.
Posted by: rrockley | Thursday, 20 October 2011 at 09:58 AM
re: "Citrin"...
Yeah, I'm normally able to catch the iPad's autocorrect, but I missed that one. I also frequently miss when it replaces "Mamiya" with "Kamiya."
re: my assertion that one rule of thumb is as bad as another...
The point I was trying to make is that sometimes the highlights need to blow out. Specular reflections on glass, artificial light sources, and a host of other things one regularly encounters, particularly in urban settings, are going to blow out. Preventing them from doing so will result in an image that is severely underexposed, and just as irrecoverable as a shot where some highlight areas blow out unintentionally. If a person dogmatically tries to protect all highlights or ETTR and place the "meat" of the histogram to the right, they're simply not thinking about exposure properly. "Expose for the highlights" is correct, insofar as it means "blow out only the parts of the frame that would properly be rendered as pure, detail-free white in an ideal print." But exposing in this way requires more care and thought than merely following any rule of thumb. Which is why I suggest that a person is at least as well off letting the camera try to figure it out and trusting in that. It is unlikely to provide an optimum exposure, but it applies for more situations than any other rule of thumb I can think of.
Posted by: Will | Thursday, 20 October 2011 at 10:25 AM
""There is almost no chance, short of shooting straight into the sun or powerful light source, to have blown highlights - which had escaped notice in either histogram or flashing highlight warnings - that would not be easily recovered with these tools."
Gingerbaker,
That's...crazy, man.
I do so very totally disagree with you.
Mike
"
Please notice the qualifier - blown pixels that escaped notice of the over exposure warning. These are very small areas indeed, and a quick swipe with the adjustment brush in the ACR module set to (-) exposure will rectify these small areas easily.
Do you normally use the ACR adjustment brush? It's a miracle.
Posted by: Gingerbaker | Thursday, 20 October 2011 at 10:32 AM
I don't know what SLRs do but the HDR mode in the iPhone actually takes two exposures and composites them to compress the scene contrast to something that the final image can hold.
I think the point of all of this discussion should be to point out the fundmental truths of exposure: there are no hard and fast rules. To do well you need to
1. Understand what you want the final picture to look like.
2. Understand where you have latitude and where you do not.
If the important part of the picture is in the highlights, you don't want to blow the highlights, so you sacrifice the rest of the frame.
If the important part of the picture is in the shadows and you want to avoid the potentially bad effects of pulling up the shadows, you put more exposure into the shadows and maybe you end up with blown highlights.
If the important part of the picture is in both places, and the contrast is too great, then you need to figure out how to compensate for that.
Any hard and fast rule is going to lead you the wrong way some of the time. It's always been curious to me why people search for these things.
That said, I *do* think that treating digital like slide film is a good overall heuristic. It gets you into trouble less often than treating it like black and white or color negative film, which are really different.
Finally: most of the time and with most modern sensors, you have enough latitude to not think about this too much. Trouble only happens in the edge cases, IMHO.
Posted by: psu | Thursday, 20 October 2011 at 10:42 AM
It seems like none of the featured comments make the counter argument that blowing highlights is just fine if you know what you're doing.
"Just, whatever you do, don't expose to the right unless you're absolutely positive there are no highlights to get blown."
This sort of makes sense if your goal is to capture from just under 255 to 0. But, if you're planning on manipulating the image later it would seem likely that the quantization (errors? for lack of a better term - I'm no scientist) that may occur in the darker areas later would be exaggerated compared to someone who exposed to the right.
So, it really comes down to how you intend to use the photo. I think both ideas play very well with each other.
Pretty soon someone is going to say - always bracket and don't worry about the original exposure. The HDR argument that has been mentioned before.
I used to use a Fuji S5 - you were much better off exposing to the right. So, no rule is hard and fast.
Posted by: James | Thursday, 20 October 2011 at 10:50 AM
Dear JC,
I am not sure about the landscape distinction you're making, but I know a compliment when I hear it, so I'll shut up now.
Dark foliage is interestingly problematical in so many ways. In B&W, in color, with film, with digital. Even on the printing end, as dark blues and greens are parts of the color space where computer printer/ darkroom paper manufacturers feel comfortable letting things go a bit wonky because it's not a part of the spectrum we're especially sensitive to, physically or psychologically.
Complicated and interesting. Makes me wonder if I could study that enough to *understand* it. I suspect not.
~~~~~~~
Dear Dan,
Let me reframe a bit. What Michael said (shoving words in his mouth) was that poor S/N ratios were enough of an issue that it was photographically useful to push exposures as high as possible to improve the signal.
That's technically true, although I think that was overpromoted even back then.
What's changed, very quickly, is that S/N has improved so much that noise in low ISO photos, even in the shadows is NOT a major obstacle to high quality photography for most good photographers with good equipment (not great-- just good) under most conditions. So, as an important rule of thumb, it's not very.
If it had no downside, it'd still be better than not. Problem is that it has two big ones-- first, it is usually hard to do this in practice without blowing highlights (again, on average) and blown highlights are much more damaging to the photo's appearance than a smidge more overall noise.
Since this is about practice, not physics, it makes it a less-than-desirable rule of thumb.
Does that help?
~~~~~~~~
DDB and I are off on a two day photo trip, hunting wily, feral foliage. My comments may be very sporadic through to Saturday. You're mostly on your own.
Do good exposures. Don't do evil exposures.
pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================
Posted by: ctein | Thursday, 20 October 2011 at 11:15 AM
I quite agree. The problem with ETTR is that with a very few exceptions, highlight recovery on current cameras absolutely stinks. What's worse, the histograms are not always particularly accurate. My Olympus E-P2 at ISO 100 will often display a histogram with no blown highlights despite significant portions of the image being completely unrecoverable.
(That said, the beauty of the E-P2 is that in order to get a semi-valid histogram, you have to shoot at ISO 200+, which automatically nets much noisier shadows.)
Posted by: Michael Bernstein | Thursday, 20 October 2011 at 11:27 AM
I'll just say that I find in-camera histograms to be all but useless, especially on cameras with less than nine stops of usable DR. As the original ETTR article was largely about histogram-based exposure determination, I think that Ctein has driven the nail home here. At moderate ISO's with decent sensors, noise (usually) isn't a huge issue if you're making prints rather than pixel-peeping.
I will say that on marginal images, the choice of RAW developer can have significant influence (particularly with respect to posterization), and that for these images I'm increasingly impressed by Raw Photo Processor 64, an impressive piece of shareware that uses floating-point rather than integer arithmetic for several stages of its RAW conversion.
Posted by: Semilog | Thursday, 20 October 2011 at 11:43 AM
Are we discussing film photography?
Just bracket and you save about 2000 words.
On a more serious side this is probably sponsored by a company offering a noise-reduction software.
Ok I'm kidding again now seriously:
Sorry I can't do it, the article is too funny.
Posted by: Luis | Thursday, 20 October 2011 at 11:44 AM
That referenced "ETTR" article on Luminous-Landscape.com was updated just last month (September '11), and the 'expose-to-the-right' advice originally given in 2003 was once again reaffirmed.
Although Ctein is quite correct in that any overexposure of highlight areas leads to unrecoverable loss of detail, what he doesn't mention is that the 'shadow side' of the Histogram box represents those least-capable 'areas' of any digital sensor for capturing pixels. From the 255 available tonal values (from black to white in a final 8-bit file), fully half of those (127 tones) are captured in the far right-hand Highlight zone of the Histogram, and the remaining 4 vertical Histogram zones (of the 5) moving left toward Shadows (left side of the box) can only capture 64, 32, 16, and (finally) 8 tones each. (And though 8 tones still remain after doing the math, I've always assumed they are tones too close to either black or white to be individually distinguished via printing.)
So if an exposure piles most pixels against the left side of the box as Ctein posits, then (theoretically) only 10% (in the 2 most left-side zones) of those 255 available tones will be captured for creating that image. Theoretically. From all I've read to-date, and after 37 years behind a camera.
As a great admirer of Ctein's knowledge, experience and photographic skills, what am I missing here? (And than you for always pushing us all toward a greater understanding of our craft.)
Posted by: Joe Boris | Thursday, 20 October 2011 at 11:52 AM
Ctein:
"Your camera detects photons, not protons. Hydrogen is made of protons (99+% by mass/energy), light is made of photons (100%, no added ingredients).
I know, it was almost certainly your auto-spell correct, but how could I resist?
(What? Rules? In a physicists' fight?!)"
LOL! You need to download the updated software. Next year, they promise tachyon detection. :D
Posted by: Gingerbaker | Thursday, 20 October 2011 at 12:09 PM
Ctein:
"S/N improves with the photon count, but, so what? The S/N is so low, in practice, that the visible image degradation from less exposure, at low ISO's, is far, far less objectionable than blown out highlights. It's simply not a major visual factor, with decent cameras used at low ISOs."
Thank you for confirming my point - that ETTR improves S/N ratio. :) But why do you keep insisting that "blown highlights" are so objectionable? As I said, with my camera, a Canon 5D with jpg picture settings set to middle values, my overexposure flashies provide very useful information with regard to recoverable highlights. Unless the exposure is truly too high - producing truly blown skies, for example - small amounts of exposure flashies are not a problem at all.
This means that after using ACR adjustments sparingly - I have no blown highlights any more. Really, the discussion should be about recoverable highlights and truly blown (unrecoverable) highlights. With proper in-camera jpg settings, small amounts of flashies mean *virtually no* unrecoverable highlights (Lighting fixtures in long-exposure night photography might be the exception). And a much better S/N ratio in the shadows, which allows more shadow detail to be recovered, if one wants, before noise gets to be a problem.
Those small areas of red-highlighted over-exposure in your lovely night photograph of the domed building, for example, would be perfectly amenable to localized correction, allowing you greater latitude to reveal more shadow detail if you wanted. Forgive me, but I can't help but remark that neither you or Mike make mention of the these new tools in the ACR RAW develop module - the adjustment brush or the parametric curves sliders. Have you rejected their use?
My Canon 5D is getting fairly long in tooth and is a pretty noisy beast compared to more recent offerings, so this ETTR stuff, while helpful to me, may not be so important to other folks.
ETTR seems to work quite well with the new tools available to us. I truly see no downside to its implementation, outside of more time in post, although I do wonder if all this manipulation might lead to drawbacks in regard to color rendition,etc, perhaps?
Posted by: Gingerbaker | Thursday, 20 October 2011 at 12:10 PM
Expose to the right seems like one of those photography rules such as the rule of thirds and the golden rule which we should all follow except for when we don't want to.
Posted by: B Grace | Thursday, 20 October 2011 at 12:50 PM
Yes, it is a bunch of bull. When the first voices saying expose to the right did arise in web sites and blogs, I did some tests, fairly similar to yours, never getting acceptable images
Regards
Posted by: Marcelo Guarini | Thursday, 20 October 2011 at 01:52 PM
What about clipped shadows. Are we worried about that at all. By exposing to the right in a high contrast scene you may be blowing out your shadows, but you are also collecting more pixel data that allows for greater adjustment later. In this example you are clipping far more blacks that you would whites by over exposing.....a digital sensor DOES NOT work just like film...
Posted by: Sam | Thursday, 20 October 2011 at 05:19 PM
Incidentally, my recommendation in my previous comments here for Mac users to look at "RPP" is for the same shareware raw processor linked by semilog above: Raw Photo Processor 64.
—MItch/Bangkok
That would certainly be the result for anyone who thinks that is "should be all done in the camera", because exposing to the right will result an initial image in the raw processor that won't look good until you pull back on the exposure slider in your raw processor. It is clear that exposing to the right is not for anyone who only shoots JPGs or who doesn't know how to, or doesn't want to, do post-processing on RAW files.Posted by: Mitch Alland | Thursday, 20 October 2011 at 08:14 PM
An expert I'm not, but my understanding is that the idea behind ETTR (WITHOUT overexposing, which everyone agrees is bad) has less to do with shadow noise and is mostly about maximizing the use of sensor data. Yes, the proverbial white rabbit in the snowstorm and the black cat in the coal mine are both extremes, and every image must be considered individually.
Assuming that as given, from this article (July 2011) http://luminous-landscape.com/tutorials/optimizing_exposure.shtml:
"A typical consumer DSLR recording 12 bits per sensel is able to record up to 4,098 separate tonal values.
If we assume a 10 stop dynamic range this is how this data is distributed...
* The brightest stop = 2048 tonal values
* The next brightest stop = 1024 tonal values
* The next brightest stop = 512 tonal values
* The next brightest stop = 256 tonal values
* The next brightest stop = 128 tonal values
* The next brightest stop = 64 tonal values
* The darkest stop = 32 tonal values
As can be seen, each stop from the brightest to the darkest contains half of the data of the one preceding it."
To put a picture around that, imagine several glasses, all lined up in a row, except each glass to the left contains half the volume of the one to the right of it. Now, begin pouring water into the glasses. Obviously the biggest glass(es) can hold more than the smaller ones, and if you continue to pour water into the smallest glass it will simply overflow and be wasted. Therefore, if your histogram is oriented toward the left (i.e. underexposure)then you have effectively wasted the potential information that could have been captured by using a histogram oriented more toward the right.
Part of the problem of course is that the histogram shown on the back of the camera is based on the embedded jpg and not the RAW data, but this can work in one's favour as the RAW file tends to have more 'head room' than the jpg.
Mike.
Posted by: Mike Nelson Pedde | Thursday, 20 October 2011 at 09:02 PM