Debunking ETTR

<p><strong></strong><br /> <strong></strong></p>

<h3><strong>SUPER-XX</strong></h3>

<p><strong></strong><br /> In the old days if one shot glassware with alot of highlight detail one PURPOSELY chose a film that "kept on going" in the highlight region; like Kodak Super-XX.</p>

<p>The DlogE curve had little or no shoulder at all; it just kept on going upward with more exposure.</p>

<p>Super-XX did NOT "top out" much at all compared to todays films.</p>

<p>Thus one had this HUGE dynamic range; GREAT for not loosing subtle highlight details.</p>

<p>There was thus not all this boxed in narrower range like todays films; or sensors either.</p>

<blockquote>

<p>No, not in my use (and historically many others) of the language based on years of analog and digital photography. <strong>Over expose</strong> is just that; you blew out highlight detail you didn’t want to. It was true for shooting film, I don’t know why anyone needs to modify this for digital. </p>

</blockquote>

<p>I'm afraid that I have to contest this definition. Overexposure does not necessarily mean that the recording medium's upper limits have been exceeded. It only means that the image was rendered brighter than a nominal exposure, i.e. Zone V looks like Zone VI or Zone VII instead of Zone V. Mr. Schewe's waterfall image was clearly overexposed, but the over-exposure did not exceed the limits of his camera's sensor, and the adjusted image did not suffer at all. Yes, sometimes over-exposure crosses the point of no return and highlight detail is lost, but this is not necessarily the case.</p>

<p>I'll try to phrase my point more accurately: ETTR relies on exposing to the upper limit of the sensor's RAW highlight range. This might result in an overexposure OR an underexposure depending on the brightness of the highlights in the scene. But either way, over-exposure or under, the act is a deliberate, calculated attempt to minimize noise by maximizing the signal-to-noise ratio of the sensor.</p>

<p>A simple yes or no to the following will allow me to move on with a clearer understanding of what was said.</p>

<p>In context of ETTR, is it ever meaningful to keep the same exposure and increase ISO to push the image data to the right on the histogram?</p>

<p>Yes.</p>

<p>The ISO should be kept as low as possible, but *not* so low that the histogram peak is leftwards.<br>

That just results in underexposure-associated increases in noise.</p>

<blockquote>

<p ><a href="../photodb/user?user_id=891451"><em>Michael Young</em></a><em> </em><a href="../member-status-icons"><em><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></em></a><em>, Mar 18, 2010; 01:53 a.m.</em></p>

<p><em>A simple yes or no to the following will allow me to move on with a clearer understanding of what was said.</em></p>

<p><em>In context of ETTR, is it ever meaningful to keep the same exposure and increase ISO to push the image data to the right on the histogram?</em></p>

</blockquote>

<p>Mike, by "meaningful" I'm guessing that you mean "helpful".<br>

Your one word answer is....</p>

<p>No.</p>

<p>For those people who might like a more in depth answer, here is my take....<br>

Not to be argumentative, but to begin with, ETTR stands for EXPOSE to the Right, not ISO up.<br>

Why is that important?<br>

ISO in the context of electronic photography, refers to the gain factor of the analog signal chain before the data gets digitised, or quantised.<br>

If you increase the gain, the noise factor increases, and that shows up along with the data you want to quantise. At some point that noise becomes visible and annoying.<br>

So the idea is to shoot with as low an ISO (I still want to say "ASA") as possible to keep noise out of the photo.<br>

To get the exposure to the right, you can open up the aperture and expose for longer periods of time.<br>

This increases the "Signal" in the "Signal to Noise" equation, assuming constant noise at the given ISO.<br>

Of course, both methods (f-stop and exposure time) bring a host of headaches, but that's for another thread.<br>

Adrian likes to compensate by boosting the ASA, hence his answer of "Yes".<br>

Maybe I'm missing something here, but I keep ASA as low as possible (200) and always shoot around it.</p>

<p>Bill P. vs. the Forces of Doom</p>

<p>Actually I'd like to thank Ben for his post.<br>

It led me back to a review of 'pseudoscience', which is always interesting to say the least.<br>

Here's a link that pretty well summed up things for me...<br>

<a href="http://www.quackwatch.org/01QuackeryRelatedTopics/pseudo.html">http://www.quackwatch.org/01QuackeryRelatedTopics/pseudo.html</a></p>

<blockquote>

<p><a href="../photodb/user?user_id=4564852"><em>Mike Hitchen</em></a><em> </em><a href="../member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub2.gif" alt="" /></em></a><em>, Mar 17, 2010; 07:01 p.m.</em><br /><em>So using ETTR is, by definition over-exposure IF your definition of 'overexposure' is that it does not represent the original scene.</em></p>

</blockquote>

<p>Mike, I learned it the way you did.<br />Overexposed meant just that, overexposed.<br />The AMOUNT of ovexposure would lead to "blowout", but that was a degree, not a definition.<br />The same would apply to "UNDEREXOSED"<br />Here's what Photoshop Support dot com folks have to say about it.....<br>

<a href="http://www.photoshopsupport.com/tutorials/jennifer/fix-overexposed.html">http://www.photoshopsupport.com/tutorials/jennifer/fix-overexposed.html</a></p>

<p>Bill P. vs. the Forces of Doom</p>

<blockquote>

<p ><a href="../photodb/user?user_id=361342"><em>Andrew Rodney</em></a><em> </em><a href="../member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub9.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></em></a><em>, Mar 17, 2010; 07:15 p.m.</em></p>

<p><em></em></p>

<blockquote>

<p><em>But isn't that what ETTR is basically doing? </em></p>

</blockquote>

<p><em>No, not in my use (and historically many others) of the language based on years of analog and digital photography. <strong>Over expose</strong> is just that; you blew out highlight detail you didn’t want to.</em></p>

</blockquote>

<p>Andrew,<br>

I learned it this way.<br>

Overexposure means that there is more light in the photo than was desired based in the original scene.<br>

The AMOUNT of overexposure would lead to "Blowout".<br>

Overexposure is a phenomenon, "Blowout" is a DEGREE of overexposue,<br>

As I pointed out earlier, UNDEREXPOSURE is the smae thing, in the opposite direction, ie, not enough light to reasonably depict the original scene.<br>

The corollary, if overexposure meant "Blown Out", would be "Hard Black", and that is clearly not the case.<br>

Numerically, "Blown Out" is a string of "255"s, unrecoverable data, and "Hard Black" would be a string of unrecoverable "0"s.</p>

<p>Bill P. vs. the Forces of Doom<em></em></p>

<blockquote>

<p ><a href="/photodb/user?user_id=361342"><em>Andrew Rodney</em></a><em> </em><a href="/member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub9.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></em></a><em>, Mar 17, 2010; 05:25 p.m.</em></p>

<p><em></em></p>

<blockquote>

<p><em>If Ben is comfortable doing his work his way and he feels that ETTR is more hassle than it's worth, I want to hear what he's got to say.</em></p>

</blockquote>

<p><em>That’s all good and fine and I think your attitude is a good one. But lets not forget what the thread title is and what the OP at the top proposes! Then look at some of the serious misunderstandings in his theories (debunked by more than a few members). See the summery about submission of theories to a peer group for review? Isn’t that the idea here? Otherwise why the post in the first place?</em></p>

</blockquote>

<p>Andrew, in the grand scheme of things, I learned a lot from this thread. <br>

Sure, Maybe Ben came on like Maynard Ferguson and you're more in tune with Chet Baker, but given the chance, I'll listen to and enjoy both.</p>

<p>Bill P.</p>

<p>In the <a href="http://www.photo.net/digital-darkroom-forum/00W1Rn">other thread on ETTR</a>,

Andrew inadvertently brought to my attention another fundamental problem with ETTR that is

unresolvable with current tools.</p>

<p>It’s actually the same problem of color shifts that we’ve discussed above. Follow

the bouncing ball.</p>

<p>If you shoot a gray card with a correct incident exposure, that will cause the camera to record a

certain pattern of pixels in the RAW data. (It’s good to remind people that RAW data is

nothing like the RGB data we’re familiar with; most striking, the pixels are in single-channel

Bayer array patterns.) The RAW converter will render that RAW data into RGB and apply an ICC

profile. Assume for the moment that you use a correct white balance and the converter’s

most neutral settings. If you then convert that ICC-tagged RGB file to the Lab color space, the gray

card will be, within the precision of the metering system of the camera and the color accuracy of the

card, L* = 50, a* = 0, b* = 0 — otherwise known as middle gray. Other real-scene colors will get similarly mapped to their

colorimetrically-correct values, with the caveat that highlights are compressed so as to provide some headroom to

accommodate objects that are emissive, specular, and fluorescent. A 100% white object will be

rendered, typically, as about L* = 90.</p>

<p>But what about ETTR? Your gray card is no longer L* = 50, it’s more like L* = 60, or

maybe 55 or 75 or who-knows-what. Your scene is tagged with a wildly-incorrect ICC profile. You

now have to correct that. And, while you can certainly bring any single color or small range of colors

back to where they should be, as anybody who has ever used an ICC-managed workflow can tell

you, there’s simply no way a human can possibly create an accurate color map of the entire

spectrum (or even come close). The transformations aren’t even close to linear and

don’t even pretend to be.</p>

<p>Of course, creative photography isn’t usually concerned with capturing a colorimetricaly-correct rendering of the scene, but the same problems arise. You’ve now caused all sorts of non-linear shifts across the entire spectrum and tone range, you have no reference to judge the type or amount of changes, and only the crudest of tools to correct them. With ETTR you have, in essence, created an untagged RGB file, assigned an entirely arbitrary profile to the file, and you’re now in the quagmire of trying to manually reproduce color transformation math using a sledgehammer while wearing oven mitts.</p>

<p>So, to all who have wondered why ETTR causes all sorts of color shifts, that should give you an

idea as to the true source of the problem, just how deep it goes, and just how perverse ETTR is for

those who care about accurate color rendition.</p>

<p>Cheers,</p>

<p>b&</p>

<blockquote>

<p>I learned it this way.<br />Overexposure means that there is more light in the photo than was desired based in the original scene.<br />The AMOUNT of overexposure would lead to "Blowout".</p>

</blockquote>

<p>I would agree with that yes. If I stated that over exposure = blown out (fill in the blank), that was incorrect or unclear. <strong>I would however argue that ETTR is over exposure since the results of the processing (we always have to link exposure and development, film or digital) does not result in over exposure </strong>(more light in the photo than was desired based in the original scene).<br>

If you truly over expose to the point of clipping, you didn’t practice ETTR correctly or, like the film days, exposed past the point of a preferred rendering with the development stage of the process. Anyone recall snip tests and pulling (or pushing, to a small amount), the development? In such cases, we used meters and sometimes did or sometimes didn’t follow them to a T (a Polaroid then a snip or clip test was sometimes required). We were able to control exposure and development to over come under or over exposure using your definition above. If we screwed up exposing the film, no amount of darkroom work would resolve the rendering issues. Its the same with digital. </p>

<p>So when you practice ETTR correctly, apply the development correctly, and the results are NOT more light in the photo than was <strong>desired</strong> based in the original scene, do you consider this over exposure? I don’t. </p>

<p>If the definition is, <em>over exposure is exposing more than the</em> (sometimes stupid) <em>meter recommended, despite the final rendering (development)</em>, then yes, ETTR <strong>is</strong> over exposing. I don’t think that’s a very good definition do you? </p>

<p>Andrew,</p>

<p>As you repeatedly point out, RAW data is untagged linear data. It’s the job of the RAW

converter to interpret that data and produce an ICC-tagged RGB file. If the scene exposed properly,

a RAW converter in its most neutral settings will produce a surprisingly good colorimetric rendering

of the original scene.</p>

<p>How do you propose one should create a good colorimetric rendering using ETTR?</p>

<p>For cases when a more creative rendering is desired, how do you propose one should replicate

the intended function of the creative adjustments?</p>

<p>Or, as you make clear in your article, in this thread, and elsewhere, do you simply not care about color rendition so long as it doesn’t look too bad at first glance?</p>

<p>Cheers,</p>

<p>b&</p>

<blockquote>

<p ><a href="../photodb/user?user_id=361342"><em>Andrew Rodney</em></a><em> </em><a href="../member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub9.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></em></a><em>, Mar 18, 2010; 10:15 a.m.</em></p>

<p ><em>So when you practice ETTR correctly, apply the development correctly, and the results are NOT more light in the photo than was <strong>desired</strong> based in the original scene, do you consider this over exposure? I don’t.</em></p>

</blockquote>

<p >Andrew, you're right. The desired outcome is what we strive for. If you ETTR, get a "hot" file without blowing anything out, and then DTTL to get the proper result with less noise, how can I disagree?</p>

<p >That is why I've been mentioning DTTL as the necessary corollary.</p>

<p > </p>

<p >Interesingly enough, the Dolby and dBX noise reduction systems popular in the '70's used a similar (note the keyword "similar") process to reduce noise in audio tapes. Let's leave it at "similar" to avoid going off-topic.</p>

<p > </p>

<p >Wait, who? us? off-topic?</p>

<p > </p>

<p >Bill P. vs. the Forces of Doom</p>

<blockquote>

<p>

<p>How do you propose one should create a good colorimetric rendering using ETTR?</p>

</p>

</blockquote>

I don’t know why I continue to bite and educate you, but what you’re saying is (again) incorrect in terms of a colorimetric rendering since you don’t understand (yet) the differences between scene and output referred. Here’s a piece by the ICC (that group that has a little to do with profiles) you really need to read!

http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf

<blockquote>

<p>

<p>That is why I've been mentioning DTTL as the necessary corollary.</p>

</p>

</blockquote>

No question that its like one hand clapping to discuss ETTR without DTTL (DTTL I think is a new acronym, perhaps the best net result of this mess of a post).

<blockquote>

</blockquote>

<blockquote> But what about ETTR? Your gray card is no longer L* = 50, it’s more like L* = 60, or maybe 55 or 75 or who-

knows-what. Your scene is tagged with a wildly-incorrect ICC profile.</blockquote>

<p>Huh? Your gray card should be wherever you put it in your raw processing. There is no RGB, LAB or ICC profile (wildly-

incorrect or otherwise) until after raw processing. If you shoot a gray card, expose to the right, and then end up with a gray

card at L=60 when you wanted L=50, you are doing it wrong. </p>

<blockquote>

<p>Huh? Your gray card should be wherever you put it in your raw processing. There is no RGB, LAB or ICC profile (wildly- incorrect or otherwise) until after raw processing. If you shoot a gray card, expose to the right, and then end up with a gray card at L=60 when you wanted L=50, you are doing it wrong.</p>

</blockquote>

<p>Absolutely correct. </p>

<p>Mark,</p>

<p>You miss the point, as does Andrew.</p>

<p>Once more, with feeling.</p>

<ol>

<li>Determine the correct exposure using a spot meter on a gray card, an incident meter, or

whatever.</li>

<li>Process that file using the flattest, most neutral settings (including the Camera Faithful profile)

in ACR (other RAW converters behave similarly).</li>

<li>Convert the resulting RGB file to Lab.</li>

<li>A gray card will be L=50; other values will be close colorimetric matches, with the exception

that highlights will get compressed such that L=100 is mapped to roughly L=90.</li>

</ol>

<p>Attempt the same with ETTR, and the colorimetric match will degrade in close correspondence

with the amount of ETTR applied.</p>

<p>Andrew, if you actually understood that introductory-level ICC paper you’ve got

bookmarked, you would understand how the process I’ve repeatedly outlined addresses

your misunderstandings of scene <em>v</em> output referencing.</p>

<p>Cheers,</p>

<p>b&</p>

<blockquote>

<p>

<p>Andrew, <strong>i</strong><strong>f you actually understood that introductory-level ICC paper you’ve got bookmarked, you would understand how the process</strong> I’ve repeatedly outlined addresses your misunderstandings of scene <em>v</em> output referencing.</p>

</p>

</blockquote>

Funny statement considering I coauthored it with Jack Holm the chief color scientist at HP for digital cameras and both of us are on the ICC photo committee which requested this article. But of course, you are correct again, Jack, Mark, Schewe and everyone else, especially me are wrong. Fortunately we live on a different planet than you do....

<p>Andrew,</p>

<p>If you’re on the ICC photo committee, then you should have no trouble answering my

question.</p>

<p>Yet again, how is one to achieve at least the degree of colorimetric match attained with a correct

incident exposure when applying ETTR?</p>

<p>Cheers,</p>

<p>b&</p>

<blockquote>

<p ><a href="../photodb/user?user_id=361342"><em>Andrew Rodney</em></a><em> </em><a href="../member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub9.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></em></a><em>, Mar 18, 2010; 10:37 a.m.</em></p>

<p><em></em></p>

<blockquote>

<p><em></em><br>

<em>That is why I've been mentioning DTTL as the necessary corollary.</em><br>

<em></em></p>

</blockquote>

<em>No question that its like one hand clapping to discuss ETTR without DTTL (DTTL I think is a new acronym, perhaps the best net result of this mess of a post).</em>

</blockquote>

Andrew, I remember DTTL from way back in the "old days" of wet photography, when ETTR-DTTL was a common practice, used for the same reasons, reducing noise, which in those days was "graininess", as you know.

I'm amazed at the lack of use of the term in recent times.

Here's a guy that still uses it, and has an interesting video to boot.

<a href="http://fleetingglimpseimages.wordpress.com/2009/08/21/expose-right-develop-left/">http://fleetingglimpseimages.wordpress.com/2009/08/21/expose-right-develop-left/</a>

Your thoughts?

Bill P. vs. the you-know-what

<p>Andrew, just to be a wee bit more specific, one of the stated goals of the <a

href="http://www.color.org/groups.xalter">DPWG charter</a> is to “[identify] limitations of

ICC color management with respect to digital photography use cases, and [develop]

recommendations to the digital photography community to address these limitations.”</p>

<p>In your article waaaaay up above, you identified a limitation of ICC color management with

respect to a digital photography use case when you observed that your implementation of ETTR

produced color shifts.</p>

<p>What is your recommendation to the digital photography community to address this limitation?

</p>

<p>Cheers,</p>

<p>b&</p>

<p>Ben, I was a little intrigued by your assertion that ETTR leads to nonlinear color shifts so I went out and did a casual test. I'm attaching a GretagMacbeth chart I just shot. </p>

<p>This is shot on a Phase One P65+ back, 645 DF with an 80mm lens. The left is exposed at the suggested incident meter reading of f/8 1/15. The left is one stop "over exposed" I set the white balance in capture one and brought the 'over exposed' shot down one stop with the exposure slider.</p>

<p>Since the attached is a screen grab converted to an sRGB jpg you'll have to take my word on the rests to a certain extent. The results: identical colors. Not a single color chip varies by more than 1 RGB value. The is no color shift whatsoever. The sensor is showing just what we expect: a linear response. I see no evidence of any color problems let alone wildly incorrect profiles. There is no reason to judge one as a 'correct' exposure over the other.</p><div></div>

<p>Mark,</p>

<p>Is it safe to assume that you’re using something other than ACR for RAW conversion?

</p>

<p>Even Andrew described color shifts in his own experiments using ACR in his linked article. It

could very conceivably be that your RAW converter does things differently, in a manner much more

conducive to ETTR.</p>

<p>Would you mind doing two more experiments?</p>

<p>First, if you have ACR, could you process the two files in ACR and see if it behaves differently?

</p>

<p>Second, would you mind fiddling around with any creative settings, such as tone curves,

saturation, shadow / highlight recovery, and the like to see if those introduce differences?</p>

<p>Thanks,</p>

<p>b&amp</p>

<p>A simple yes or no to the following will allow me to move on with a clearer understanding of what was said.</p>

<blockquote>

<p>In context of ETTR, is it ever meaningful to keep the same exposure and increase ISO to push the image data to the right on the histogram?</p>

</blockquote>

<p>Yes.</p>

<p>At low enough ISO the signal will probably increase by more than the noise does. It will depend on the camera, but on a typical DSLR moving from ISO 100 to 200 will not increase the noise by anywhere near 1 stop.<br>

Increasing exposure time or shutter speed would offer greater benefit, though.</p>

<p> </p>

<blockquote>

<p>But what about ETTR? Your gray card is no longer L* = 50, it’s more like L* = 60, or maybe 55 or 75 or who-knows-what. Your scene is tagged with a wildly-incorrect ICC profile.</p>

</blockquote>

<p>A jpeg will be messed up, yes, though most if not all of the damage has nothing to do with the ICC profile. Raw images can completely ignore ICC profiles until the output file from the raw converter. The exposure can be corrected <em>first </em> in post processing and then the gray card will have whatever L* value it was supposed to (roughly 50, but not all gray cards are identical).</p>

<p>I don't even think that the exposure change would affect the accuracy of standard ICC profiles like AdobeRGB or sRGB anyway, but I would have to look into the details. The color shifts are caused by nonlinear tone curves combined with exposure changes, and would appear even in images not tagged with any ICC profile.</p>