Exposure to the right (ETTR) advantages & disadvantages.

[Michael Fryd]

I have achieved better results from 5 stop exposure bracketing, +2, +1, 0, -1, -2 , as compared to ETTR when it comes to noise, color and tones with better shadow and highlight detail.

It's impossible to achieve better results with bracketing than ETTR. If one of the bracketed shots accidentally matches or gets close to the ETTR exposure, you can do as well as ETTR, but not better.

Only if you limit yourself to discarding all but one of the bracketed shots. In which case bracketing makes it more likely that you will get the ideal ETTR shot.

Plus it provides an opportunity to select a single frame when needed thereby providing an opportunity to go ETTR or ETTL at same time.

Does that imply you're not always selecting a single frame but using the bracketed shots for HDR merge?

On DPR I can often see people do HDR merge from their pretty powerful cameras such as yours D750, or Sony A7III, A7IV etc. If they enjoy the process it's fine, I understand it. But technically it's a waste of time and often a struggle with ghosting.

Sometimes I do HDR merge and sometimes I bracket in very tricky situations, but the vast majority (>90%) of my daytime landscape shots are just a single exposure by ETTR.

Bracketing allows you to capture far more dynamic range than ETTR. Therefore, if absolutely maximizing quality is your goal, then bracketing and merging is far better than a single ETTR capture.

So here are some of the available choices in decreasing order of quality, and increasing order of convenience:

1. Multiple captures with bracketed exposures, and then combining them in post
2. ETTR
3. Normal exposure as per the camera’s metering system

So the question is not which is better (that’s multiple exposures with bracketing), or which is more convenient (normal exposure as per the meter). The question is which is the right compromise for the situation at hand.

Is ETTR the best of both worlds (not very inconvenient and not too bad)? Or the worst of both (not convenient enough, and not good enough).

ETTR is helpful when you need just a little bit less noise than what a normal exposure would provide. If you need a lot less noise, then ETTR won’t suffice, and you need multiple bracketed exposures.

 

[Mako2011]

I have achieved better results from 5 stop exposure bracketing, +2, +1, 0, -1, -2 , as compared to ETTR when it comes to noise, color and tones with better shadow and highlight detail.

It's impossible to achieve better results with bracketing than ETTR.

That's not always the case? After all, the method that uses spot metering to facilitate ETTR is often made even more accurate (results in the best possible ETTR exposure solution) by adding small step bracketing to the workflow. I'll use -.3, 0, and +.3 bracketing in that case. After all, with a few exceptions, even with histogram based ETTR one is limited to changing exposure in limited steps of aperture/shutter speed. Now if we had a perfect RAW histogram to work with...that might change things. Even UNI-WB doesn't get us a perfect RAW histogram with regards to where clipping falls all the time

 

[Quarkcharmed]

I have achieved better results from 5 stop exposure bracketing, +2, +1, 0, -1, -2 , as compared to ETTR when it comes to noise, color and tones with better shadow and highlight detail.

It's impossible to achieve better results with bracketing than ETTR. If one of the bracketed shots accidentally matches or gets close to the ETTR exposure, you can do as well as ETTR, but not better.

Plus it provides an opportunity to select a single frame when needed thereby providing an opportunity to go ETTR or ETTL at same time.

Does that imply you're not always selecting a single frame but using the bracketed shots for HDR merge?

On DPR I can often see people do HDR merge from their pretty powerful cameras such as yours D750, or Sony A7III, A7IV etc. If they enjoy the process it's fine, I understand it. But technically it's a waste of time and often a struggle with ghosting.

Sometimes I do HDR merge and sometimes I bracket in very tricky situations, but the vast majority (>90%) of my daytime landscape shots are just a single exposure by ETTR.

ETTR is guesswork, because the ETTR limit is not accurately predictable.

ETTR, if executed properly, is definitely not a guesswork as you usually follow a simple well defined routine.

There are a few methods of ETTR that give you different accuracy, or different margin of error against an ideal ETTR exposure.

• Flat/neutral picture profile in camera plus RGB histogram (can still be up to 0.5-1 stop under the ideal exposure)
• Luminosity histogram with UniWB (very accurate, that's what I use most of the time)
• Highlight zebras (my cameras don't have it in Live view, but people use it successfully. I don't know how accurate this method is. Perhaps it can be used with UniWB too).
• Bracketing by a small step and then choosing a proper ETTR exposure when postprocessing. If you bracket by 1-stop steps, it's not very accurate, but most importantly, very time-consuming and space-consuming. If you bracket with a 1/3-stop steps, it can be very accurate but tremendously boring.
• Spot metering with compensation by ~3ev, it was discussed recently in another thread.

Using 1-stop bracketing is guesswork. But, you can use 1/3 stop bracketing once you have adjusted the exposure to where highlights are blown.

Also, you can't shoot any motion in the scene with bracketing, when you're trying to catch a 'decisive moment'. E.g. when shooting seascapes and surf. Also it gets frustratingly slow if you bracket long exposures (say longer than 3-5 seconds).

Exposure bracketing should be the last resort when the default exposure method fails.

--
https://www.instagram.com/quarkcharmed/
https://500px.com/quarkcharmed

 

[Quarkcharmed]

I have achieved better results from 5 stop exposure bracketing, +2, +1, 0, -1, -2 , as compared to ETTR when it comes to noise, color and tones with better shadow and highlight detail.

It's impossible to achieve better results with bracketing than ETTR.

That's not always the case? After all, the method that uses spot metering to facilitate ETTR is often made even more accurate (results in the best possible ETTR exposure solution) by adding small step bracketing to the workflow. I'll use -.3, 0, and +.3 bracketing in that case. After all, with a few exceptions, even with histogram based ETTR one is limited to changing exposure in limited steps of aperture/shutter speed. Now if we had a perfect RAW histogram to work with...that might change things. Even UNI-WB doesn't get us a perfect RAW histogram with regards to where clipping falls all the time

I've been using UniWB for a long time, and I've never had the actual clipping when the UniWB histogram was showing no clipping. I got clipping a few times because of my laziness, mostly when I wasn't tweaking the ETTR exposure when the lighting conditions were changing.

If you bracket with a small step, 1/3 - 1/2 ev, you may get good results but with a lot of effort and time spent. But it won't be better than ETTR with UniWB. It'll have roughly the same accuracy but achieved through a much greater effort.

 

[Mako2011]

I've been using UniWB for a long time, and I've never had the actual clipping when the UniWB histogram was showing no clipping.

Yes...combining Uni-WB with ETTR really helps. Only seen clipping with it once or twice in rare/hard cases

ETTR with UniWB.

How are you changing exposure in that case?

 

[Quarkcharmed]

Sometimes I do HDR merge and sometimes I bracket in very tricky situations, but the vast majority (>90%) of my daytime landscape shots are just a single exposure by ETTR.

Bracketing allows you to capture far more dynamic range than ETTR.

It doesn't, unless you do an HDR merge. But that's not not bracketing as such, that's HDR merge that gives you the high dynamic range.

Bracketing itself only gives you a number of versions with different exposures.

Therefore, if absolutely maximizing quality is your goal, then bracketing and merging is far better than a single ETTR capture.

HDR merge means ghosting and misalignment artifacts. Also it gets very tricky if you shoot seascapes with moving water.

There's also so called exposure blending where you selectively combine two or more exposures in Photoshop with masks.

So here are some of the available choices in decreasing order of quality, and increasing order of convenience:

1. Multiple captures with bracketed exposures, and then combining them in post
2. ETTR
3. Normal exposure as per the camera’s metering system

That sounds right, and they also go in increasing order of applicability. HDR merge is very limiting.

So the question is not which is better (that’s multiple exposures with bracketing), or which is more convenient (normal exposure as per the meter). The question is which is the right compromise for the situation at hand.

There's an additional question, whether you can actually get the shot. Using HDR merge rules out many shooting scenarios and techniques, including long exposure, moving subjects etc. ETTR can be used whenever you have enough time to set it up - which gets problematic in action shooting, but not a big problem in landscape photography.

Is ETTR the best of both worlds (not very inconvenient and not too bad)? Or the worst of both (not convenient enough, and not good enough).

ETTR is helpful when you need just a little bit less noise than what a normal exposure would provide. If you need a lot less noise, then ETTR won’t suffice, and you need multiple bracketed exposures.

'A bit less' or 'a lot less' are very subjective metrics. In my experience, ETTR gives quite a significant noise improvement, on average.

 

[Quarkcharmed]

I've been using UniWB for a long time, and I've never had the actual clipping when the UniWB histogram was showing no clipping.

Yes...combining Uni-WB with ETTR really helps. Only seen clipping with it once or twice in rare/hard cases

ETTR with UniWB.

How are you changing exposure in that case?

By histogram, and normally through the shutter speed. The issue with UniWB is of course the awful strong greenish tint in the images. So I set my camera up so that I can switch to UniWB (which sits on a custom white balance) in a few clicks, set the exposure and then switch back to a daylight WB. It takes a few seconds, so if I'm time-constrained, I just ETTR under a normal white balance. It's less accurate but still about 2-2.5ev higher than evaluative metering would provide.

 

[Mako2011]

I've been using UniWB for a long time, and I've never had the actual clipping when the UniWB histogram was showing no clipping.

Yes...combining Uni-WB with ETTR really helps. Only seen clipping with it once or twice in rare/hard cases

ETTR with UniWB.

How are you changing exposure in that case?

By histogram, and normally through the shutter speed.

Then you are limited to steps of exposure adjustment. Not unlike bracketing to a degree. If it's truly critical...a manual iris controlled lens can get one as close to perfect as possible. Even then...if one uses RAW Digger to check, they may find they were off a hair...with no practical detriment. "Close enough" really does work at times

 

[BertIverson]

and adjust EV to your liking.

...

I agree. One would not want the birds any brighter, the trees are fine and the black background certainly does not need to be brighter.

Well done,

Bert

 

[The Ghost of Caravaggio]

The best way to maximize your chances that you will get the ideal exposure is to use the histogram or blinkies to adjust shutter speed to just above overexposure indication and then bracket shutter speed.

This method has flaws.

Both the histograms and blinkies do not reflect the in-camera raw data. The histograms and blinks are estimates from rendered, in-camera data. The in-camera Display Menu rendering JPEG parameters will affect the histograms and blinkies as well.

--

"The belief that ‘randomness’ is some kind of real property existing in Nature is a form of the mind projection fallacy which says, in effect, ‘I don’t know the detailed causes – therefore – Nature does not know them."
E.T Jaynes, Probability Theory: The Logic of Science

 

[Photodog2]

The only disadvantage to ETTR is it can be difficult to get it just right. It is a bit amazing that camera manufacturers haven't exactly been focused on making it easier to get it right while on the field. Should be something AI can be used in programming the in-camera software. So which camera companies do you think have made it easier than other companies to get ETTR just right in-camera?

 

[ggbutcher]

The only disadvantage to ETTR is it can be difficult to get it just right. It is a bit amazing that camera manufacturers haven't exactly been focused on making it easier to get it right while on the field. Should be something AI can be used in programming the in-camera software. So which camera companies do you think have made it easier than other companies to get ETTR just right in-camera?

This one:

https://fujiframe.com/articles/natural-live-view-ettr/

I keep posting this link, but it would seem no one is reading my stuff...

 

[Photodog2]

Thanks ggbutcher. That was a good read. Are other other companies doing that?

 

[ggbutcher]

Thanks ggbutcher. That was a good read. Are other other companies doing that?

Don't know, got the link from someone else. I'm not a Fuji shooter, but that might have convinced me to switch if it was back when I just had the D50...

IMHO, It's the sort of thing that, if talked up enough, could show up in firmware updates. It wouldn't require a hardware change.

 

[rb61]

Since it seems that it is not always clear how to properly execute ETTR in practice, I propose an alternative acronym which hopefully resolves this ambiguity:

ETFAMIPOTSITPDRATAMTTRAPWCACC - Expose To Fit As Many Important Parts Of The Scene Into The Photographic Dynamic Range And Then As Much To The Right As Possible Without Clipping Any Color Channel

I have to admit that it is a bit of a tongue twister. Does anybody have any idea for improvement? ;-)

I miss the experience of exposing for shadows and utilizing plus or minus development with B&W film,

but not all the space and materials required.

I still have my Zone VI spot meter and full library of newsletters even though I may never fully utilize them.

 

[ggbutcher]

Since it seems that it is not always clear how to properly execute ETTR in practice, I propose an alternative acronym which hopefully resolves this ambiguity:

ETFAMIPOTSITPDRATAMTTRAPWCACC - Expose To Fit As Many Important Parts Of The Scene Into The Photographic Dynamic Range And Then As Much To The Right As Possible Without Clipping Any Color Channel

I have to admit that it is a bit of a tongue twister. Does anybody have any idea for improvement? ;-)

I miss the experience of exposing for shadows and utilizing plus or minus development with B&W film,

but not all the space and materials required.

I still have my Zone VI spot meter and full library of newsletters even though I may never fully utilize them.

Me too, just a little bit...

A corollary, I've recently taken up scratchbuilding HO model railroad locomotives. In brass, I did a bunch of research and trying-out of machine shop tools and techniques. Recently though, decided to scrap all that and buy a 3D printer. A lot easier...

 

[ggbutcher]

I was working an image for another reason, got to thinking, it might be good fodder for this general discussion...

We all run into this at some time, making an image in a relatively dark place with really high light-energy surroundings. This is an old sawmill under a canopy, and the surrounding meadow and treelines are in the broad daylight. The first rendition I'll show is of the linear white-balanced RGB, with the only tone curve being the one in the sRGB export profile so's you can see what's going on:



The camera mode was Manual, and the metering mode was center-weighted average, so you can pretty well tell where the exposure was anchored. The magenta areas in the well-lit outside are blown highlights; all those RGB values just pile up at the top of the histogram, and when white balance is applied, those three piles are separated, causing this cast.

So, what did I see in the scene? The machinery pretty well looked like what you see here, but the outside was freaking bright! Essentially, the blown highlights in the camera were similarly sensed in my head, so I really don't care what detail they might have in the rendition. I don't want to HDR this. So here's my final rendition:



IIRC, that green cast you see on the mill's lumber and metal is from green plastic corrugated panels in the shed roof, what a white balance mess. But note the outsides; I applied a highlight-reconstruction essentially to just get rid of the magenta cast, but there's enough of the bright greenery to tell what that area is.

The ETPH way would have been to scooch down the EC to bring those blown areas into sensor-resolvability, and then pull up the shadows. I currently do such with highlight-weighted metering, but in such cases as this I don't see the benefit. Indeed, the subject was anchored squarely in the sensor without tone curve shenanigans, so no down-dark noise to deal with.

So, YMMV,C (Your Mileage May Vary, Considerably)...

 

[Kameratrollet]

The only disadvantage to ETTR is it can be difficult to get it just right. It is a bit amazing that camera manufacturers haven't exactly been focused on making it easier to get it right while on the field. Should be something AI can be used in programming the in-camera software. So which camera companies do you think have made it easier than other companies to get ETTR just right in-camera?

Canon with Magic Lantern installed has made things easier https://www.magiclantern.fm/forum/index.php?topic=12096.0

 

[The Ghost of Caravaggio]

Advantages

• ETTR optimizes information content. In fact, it is the only way we can maximize information content.[1, 2]
  • Because shutter time and lens aperture alone determine exposure, ETTR maximizes raw-file information content. This is so for in-camera JPEGs as they are rendered from raw data.
  • Exposure generates the analog signals we need to render the image. The signals contain the information that constitute an image.
  • Noise decreases the information content. The noise represents the uncertainty for the signals. We have no control over the noise generated by the camera. The best we can do is optimize exposure.

• ETTR minimizes the effect of photon noise.
  • The photoelectron creation in photosites exhibits a type of quantum noise called photon noise.[3] The loss of information content due to photon noise decreases as exposure increases.
  • At high exposure photon noise is the dominant noise source but it has the least effect on information content.[4] At low exposure camera electronics noise is the dominant noise source but photon noise has the greatest effect on information content.
• Camera operation might be simplified. Camera ISO setting is not a variable.
  • Cameras with ISO-invariant data streams offer a simplified exposure technique for raw file users.[5]
  • Camera ISO is set to the base (native) value.
  • A shutter time and, or lens aperture appropriate for the scene at hand are set.
  • In low light, the sensor will be underexposed to minimize camera and, or subject motion and to optimize depth of field.
  • Auto-bracketed exposures is useful because the image with optimum exposure can be retained and the others can be deleted.

Disadvantages

• Over exposure destroys information content.
  • At camera base ISO setting, the shutter time and, or lens aperture can result in excessive energy reaching some sensor photo sites. These photo sites' full-well capacity is exceeded. When this occurs, all useful information content for these photo sites is destroyed.
  • The mathematical models used to compute the rendered image assume all the photo sites have useful information content. When this is not true the rendered image can not correspond to reality.
  • The most common image symptom of unintentional overexposure are unrealistic sky hues.
• Optimum exposure is inconvenient or even impractical.
  • Optimum exposure means regions all the important highlight regions in the scene are not over exposed when the shutter is open.
  • Not every highlight region is important. Two examples of unimportant highlight regions would be specular reflections of sunlight (chrome trim on a vehicle in the background and bright point source lights in night scenes).
  • It takes time and effort to optimize exposure. Not everyone is interested in the benefits.
  • In many situations times is of the essence. An underexposed image is vastly preferred to no image at all. In action photography.
• Over brightening destroys information content.
  • Over brightening occurs after the shutter closes.
  • Above camera base ISO setting, the sensor is intentionally underexposed.
  • The metering system estimates the signal gain requited to achieve acceptable image brightness.
  • Above camera base base ISO setting, the ISO value, shutter time or lens aperture parameters can result in exceeding the maximum analog voltage input to some of photo sites' analog-to-digital converters. Clipped signals result in the loss of all useful information content for these photo sites is destroyed.
• Optimum exposure is least useful when using in-camera JPEGs.
  • This issue has nothing to do with exposure. The problem is JPEG compression destroys data that could be useful to remedy image perception problems caused by small degrees of sensor over exposure or over brightening.
  • However, in-camera JPEG image perceived quality will benefit from optimum exposure as long as camera ISO setting, shutter time, and, or lens aperture do result in signal clipping.

1/ The relationship between image MTF50 (information bandwidth) and signal-to-noise ratio is a fundamental tenet of Shannon Information theory.

2/ This assumes all other factors that affect information content are equal. Some of the most important of these are sensor surface area, sensor assembly efficiency, lens transmission efficiency and scene brightness.

3/ The noise in this discussion is limited to time-dependent (uncorrelated) sources. In photography it is common for photon noise to be mis-named as shot noise. Shot noise represents quantum noise for electrons. In photography, photon noise is greater than electron shot noise from camera electronics (read noise). Camera electronic noise is typically dominated by photo-diode gain and further electronic signal processing.

4/ The ratio of mean signal levels to photon noise levels increases as exposure increases.

5/ ISO invariant cameras have data streams where the read noise level is independent of camera ISO Setting. For fixed shutter time and aperture, at base ISO setting the in-camera JPEG could be too dark compared to using an higher camera ISO setting. But when the base ISO, rendered image is brightened in post-production, its perceived image quality is essentially identical to an image made with a higher camera ISO setting and no post-production brighting.

--

"The belief that ‘randomness’ is some kind of real property existing in Nature is a form of the mind projection fallacy which says, in effect, ‘I don’t know the detailed causes – therefore – Nature does not know them."
E.T Jaynes, Probability Theory: The Logic of Science

 

[Michael Fryd]

I was working an image for another reason, got to thinking, it might be good fodder for this general discussion...

We all run into this at some time, making an image in a relatively dark place with really high light-energy surroundings. This is an old sawmill under a canopy, and the surrounding meadow and treelines are in the broad daylight. The first rendition I'll show is of the linear white-balanced RGB, with the only tone curve being the one in the sRGB export profile so's you can see what's going on:



The camera mode was Manual, and the metering mode was center-weighted average, so you can pretty well tell where the exposure was anchored. The magenta areas in the well-lit outside are blown highlights; all those RGB values just pile up at the top of the histogram, and when white balance is applied, those three piles are separated, causing this cast.

So, what did I see in the scene? The machinery pretty well looked like what you see here, but the outside was freaking bright! Essentially, the blown highlights in the camera were similarly sensed in my head, so I really don't care what detail they might have in the rendition. I don't want to HDR this. So here's my final rendition:



IIRC, that green cast you see on the mill's lumber and metal is from green plastic corrugated panels in the shed roof, what a white balance mess. But note the outsides; I applied a highlight-reconstruction essentially to just get rid of the magenta cast, but there's enough of the bright greenery to tell what that area is.

The ETPH way would have been to scooch down the EC to bring those blown areas into sensor-resolvability, and then pull up the shadows. I currently do such with highlight-weighted metering, but in such cases as this I don't see the benefit. Indeed, the subject was anchored squarely in the sensor without tone curve shenanigans, so no down-dark noise to deal with.

So, YMMV,C (Your Mileage May Vary, Considerably)...

Yes, your result was good enough for your needs. As photographers we often have to find the best compromise for a situation.



In your situation, if you had wanted to maximize quality, multiple exposures and some sort of tone mapping or blending would have yielded the bet result.

Yours was a reasonable compromise, as the subject was the equipment, and the background wasn't critical.