Article to dig into at Luminous Landscape - Optimum exposure

[Nelu]

This is why I almost never ETTR. I prefer to expose for the subject.

Interesting...And when you shoot landscape what is it that you consider your subject? The dark shadowed valley, or the brighter, green hills for example? What about the sky?

 

[Orangutan]

This is why I almost never ETTR. I prefer to expose for the subject.

Interesting...And when you shoot landscape what is it that you consider your subject? The dark shadowed valley, or the brighter, green hills for example? What about the sky?

I'd like to hear more about this as well. From my (admittedly limited) understanding of sensor design, you want the maximum exposure possible short of over exposure. ETTR can mean one of two things, one of which is useless.

• Subjective ETTR: Use the maximum exposure that will not overexpose essential elements of the scene. In many cases (e.g. specular highlights and well-lit fog banks) there's not enough DR in any DSLR to capture the full tonal range in a single frame, so the photographer must make some judgements about which elements to preserve, or whether to use other techniques to control egregious tonal areas.
• Ritualistic ETTR: look at the histogram and make sure the entire tone curve comes close to the right edge without going over. This is a sure way to sacrifice your main subject(s) on the alter of ritualistic technique. #2 works only when the entire scene is within the DR of the sensor.

ETTL is completely unsupported by the principles of digital sensor design.

 

[tcmatthews]

This is why I almost never ETTR. I prefer to expose for the subject.

Interesting...And when you shoot landscape what is it that you consider your subject? The dark shadowed valley, or the brighter, green hills for example? What about the sky?

It really depends on what you want your subject to be. Sometimes the sky is the only interesting thing so you expose for the sky. Other times it is the land or geological feature. The patterns the shadows make in the valley below. In general you must choose a subject. Often you have to balance out the sky and the landscape. Photographs are always more interesting when you have something to draw you in.

The difficulty comes in when you have an interesting scene that includes hard shadows and bright sky. In general you end up in a situation with two much dynamic range. Even ETTR there will be to much dynamic range. So you have to resort to braced shots then pic the best one, HDR in post or Dual ISO ML tricks.

ETTR is a valid technique to maximize DR of the sensor. However what you really want is to control of the amount of DR in the scene.

 

[tcmatthews]

This is why I almost never ETTR. I prefer to expose for the subject.

Interesting...And when you shoot landscape what is it that you consider your subject? The dark shadowed valley, or the brighter, green hills for example? What about the sky?

I'd like to hear more about this as well. From my (admittedly limited) understanding of sensor design, you want the maximum exposure possible short of over exposure. ETTR can mean one of two things, one of which is useless.

• Subjective ETTR: Use the maximum exposure that will not overexpose essential elements of the scene. In many cases (e.g. specular highlights and well-lit fog banks) there's not enough DR in any DSLR to capture the full tonal range in a single frame, so the photographer must make some judgements about which elements to preserve, or whether to use other techniques to control egregious tonal areas.
• Ritualistic ETTR: look at the histogram and make sure the entire tone curve comes close to the right edge without going over. This is a sure way to sacrifice your main subject(s) on the alter of ritualistic technique. #2 works only when the entire scene is within the DR of the sensor.

ETTL is completely unsupported by the principles of digital sensor design.

If I use ETTR at all it is subjective ETTR. In general I prefer to bracket out the shot in that case. I find often I use the bracketed shot just below my ETTR shot anyway. So I sacrifice some possible but hard to recover tonal areas.

True, ETTL is not supported by principles of digital sensor design. But it is a valid technique especially if you want to produce high saturation, high contrast images with out touching those sliders in LR. You will need the entire scene with in the DR of the sensor.

Remember ETTR is a technique. Do not use it ritualistically. Only use it when needed. I just find I do not like to use it much or need it.

 

[climber]

Don't bother with ETTR. Simply make 2, 3 or 5 exposures and blend them manually in PP. Thus, you have each shot properly exposed for each subject (hill, sky, etc...)

 

[heptagon]

Don't bother with ETTR. Simply make 2, 3 or 5 exposures and blend them manually in PP. Thus, you have each shot properly exposed for each subject (hill, sky, etc...)

Do you use RAW or JPEG for this?

 

[BenKing]

I'm so confused. I read the article, and what I'm getting out of it is that we should overexpose our capture as much as possible, and that by doing so, we will have a "cleaner" image when we correct it back down in post. Is that what I'm reading? I'm new to this whole ETTL and ETTR concept, but I admit that many of my images, even at ISO 100 are still noisier that I would like.

However, I look at the "Optimum" example with the black cat, and if I overexposed my capture that much, there's no way I could get the detail back to the "Proper" photo. It looks like the detail on the cat's back is completely gone and blown out. Besides, both of those examples (bunny in snow, black cat against fake black background) are nowhere near as dynamic as the stuff that I always end up shooting, whether it be portraits or whatever.

So is the idea that we should push the exposure until just before the highlights clip? As mentioned earlier in this thread, we should really be mostly concerned with the subject, though, otherwise it may be lost into irretrievable darkness.

Confused.

 

[privatebydesign]

Ben,

Don't be confused, the article is not very good advice and what he is suggesting doesn't actually work.

But gentle ETTR does work, an up to one stop over exposure I find to work well, especially when shooting darker subjects with more emphasis on darker tones, much more than that and it becomes difficult to retain the colours and tonality of the original scene.

Of course you always have to watch for clipping, once your pixels register white they have no detail, but if you meter for a scene and make an EV adjustment for the subject, giving it an extra stop, so long as you don't clip too much, often works well for your darker tones.

 

[climber]

Don't bother with ETTR. Simply make 2, 3 or 5 exposures and blend them manually in PP. Thus, you have each shot properly exposed for each subject (hill, sky, etc...)

Do you use RAW or JPEG for this?

I always use RAW.

 

[RLPhoto]

This article is meant for landscapes primarily but I've noticed ETTR with skin tones can be a very bad thing. I don't like to expose more than 1/2 of a stop over the correct metered exposure if they are light skinned and 1 stop of they are dark skinned. Skin tones get caulky and blotchy if you ETTR too much with them.

ETTR has its place but not everything should automatically be ETTR.

 

[privatebydesign]

This article is meant for landscapes primarily but I've noticed ETTR with skin tones can be a very bad thing. I don't like to expose more than 1/2 of a stop over the correct metered exposure if they are light skinned and 1 stop of they are dark skinned. Skin tones get caulky and blotchy if you ETTR too much with them.

ETTR has its place but not everything should automatically be ETTR.

That is because the red channel clips first and people images tend to have higher red channel levels, the green channel is the lowest because there are twice as many green pixels as the other two and nature shots (generally) have more green channel info, but it is balanced better with the red and blue so the IQ hit is not as noticeable.

Tonality changes with ETTR, the higher you push the ETTR the more difficult it is to get the original tonality back because each channel is affected differently.

 

[Orangutan]

That is because the red channel clips first and people images tend to have higher red channel levels

<snip>

Tonality changes with ETTR, the higher you push the ETTR the more difficult it is to get the original tonality back because each channel is affected differently.

Isn't that going beyond ETTR, though? The point of ETTR isn't to clip, or even necessarily to get past 98% for any channel on your subject; the point of ETTR is that you shouldn't have a chunk of void to the right of the tone curve (for the subject).

 

[privatebydesign]

That is because the red channel clips first and people images tend to have higher red channel levels

<snip>

Tonality changes with ETTR, the higher you push the ETTR the more difficult it is to get the original tonality back because each channel is affected differently.

Isn't that going beyond ETTR, though? The point of ETTR isn't to clip, or even necessarily to get past 98% for any channel on your subject; the point of ETTR is that you shouldn't have a chunk of void to the right of the tone curve (for the subject).

You need to read the whole thread to get the relevance.

But the second part of the first sentence you quote is the important bit "people images tend to have higher red channel levels" not the clipping part.

The original premise is to do extreme ETTR such that nothing clips but everything is as overexposed as possible.

The problem with that is if a pixel has very diferent R,G abd B values they fall at very different places on the gamma curve, doing something as simple as lowering the exposure in post does not alter all three channel levels for each pixel equally, that is, the relationship between the recorded values and the new vales changes, this changes the tone, or colour, of the pixel, it makes it brighter and, generally, less red.

But depending on how discerning you are and the bit depth of your workflow these small differences can be seen on relatively gentle ETTR, though most of the time the improved shadow detail overwhelms the slight loss of accurate colour relationships.

Greys can be raised and lowered in post with impunity, all three pixels have the same value and any adjustment impacts each channel equally, but coloured pixels change tone (colour) as well as brightness when you raise or lower the brightness/exposure/shadows etc in standard post software techniques.

 

[mackguyver]

Greys can be raised and lowered in post with impunity, all three pixels have the same value and any adjustment impacts each channel equally, but coloured pixels change tone (colour) as well as brightness when you raise or lower the brightness/exposure/shadows etc in standard post software techniques.

The biggest challenge of using ETTR in my experience are with strongly saturated colors of subjects like sunrise/sunsets, flowers, or similar subjects. If you push it too far to the right (even watching the RGB histogram), dropping the exposure won't bring the colors back.

For difficult exposures where the subject is based on my interpretation (e.g. landscapes), I tackle this by exposure bracketing (one underexposed shot with fully saturated colors), and then a strong ETTR exposure to the edge of the histogram. In post, I pull the shots up side-by-side and after lowering the exposure on the ETTR shot, I use the sliders to match the colors as close as I can to the saturated shot, referring to the RGB values as I go. It's not perfect, but helps me get cleaner noise levels in sillhouette and other low light photos.

 

[Orangutan]

You need to read the whole thread to get the relevance.

OK, I'll go back and read it more carefully.

The original premise is to do extreme ETTR such that nothing clips but everything is as overexposed as possible.

The problem with that is if a pixel has very diferent R,G abd B values they fall at very different places on the gamma curve, doing something as simple as lowering the exposure in post does not alter all three channel levels for each pixel equally, that is, the relationship between the recorded values and the new vales changes, this changes the tone, or colour, of the pixel, it makes it brighter and, generally, less red.

But depending on how discerning you are and the bit depth of your workflow these small differences can be seen on relatively gentle ETTR, though most of the time the improved shadow detail overwhelms the slight loss of accurate colour relationships.

Greys can be raised and lowered in post with impunity, all three pixels have the same value and any adjustment impacts each channel equally, but coloured pixels change tone (colour) as well as brightness when you raise or lower the brightness/exposure/shadows etc in standard post software techniques.

Without more explanation this doesn't make sense to me. The individual photosites are linear with respect to exposure. If there is a non-linearity with respect to color channel then that's a problem with the raw conversion or other curves applied to the color channels. I would believe that a standard JPEG engine would do better with less than ETTR, but I'm not at all convinced for those who choose raw processing.

The way I look at, ETTR is the digital analog (pardon me) to Adams' zone system. I'll concede to anyone who knows the ZS better than I (that would be lots of people), but the general premise seems to be to look backwards from the final product:

1. How do I want the printed image to look?
2. How do I need to develop the negative to create #1?
3. How do I expose so I have a negative optimally suited to #2?

For digital it's similar, but there are no shoulders or toes, just a precipice at each end. We also know that lower tonal values in the scene will be noisier, and have less tonal gradation. It follows that the optimal "raw material" has as much light as possible for each important pixel, without going over the edge. How that's processed is a different question entirely.

 

[eninja]

Ben,

Don't be confused, the article is not very good advice and what he is suggesting doesn't actually work.

But gentle ETTR does work, an up to one stop over exposure I find to work well, especially when shooting darker subjects with more emphasis on darker tones, much more than that and it becomes difficult to retain the colours and tonality of the original scene.

Of course you always have to watch for clipping, once your pixels register white they have no detail, but if you meter for a scene and make an EV adjustment for the subject, giving it an extra stop, so long as you don't clip too much, often works well for your darker tones.

Thanks..
After I read this article from op, I tried set exposure compensation to above +1 and did check histogram for clipping, its a plus work. (previously I just set to +2/3 at most) - to preserve image quality (less noise) to maximum at high ISO.

Then I get a bit used to +1 and 1/3 compensation. Then on the last shoot, I found my photo clipped.
Tsk Tsk Tsk!!!