ETTR, when and why

[SrMi]

Advertisement (gone after registration)

[In the other thread that I started (SL2, ISO50, and overexposed JPEGs), ETTR discussion threatened to overpower the issue of live-view/image-review discrepancy.
Hence another thread.]

Wikipedia (link) has a good summary of ETTR (exposing to the right).

The point of ETTR is to increase the signal to noise ratio: 
a) maximize the signal by bringing more light to the sensor. 
b) minimize the noise by using base ISO.

Bringing more light to the sensor means increasing the exposure without clipping the relevant highlights.
Without a RAW based histogram/clipping, it isn't easy to know how much to expose (specific camera settings can reduce the discrepancy between JPEG and RAW histograms).
I have a conservative approach to ETTR, where I expose less rather than blow out the highlights.

The increased noise of ISO amplification makes ETTR practical at base ISO only.

Today's sensors are much better than they were in 2003 when the ETTR method was proposed. It is not as much a necessity as it used to be. However, I still find 'properly' manually exposed images better than letting the camera compute exposure automatically (no EV compensation).

[BernardC]

One limitation of ETTR, which I mentioned in the other thread, is that it leads to exposure variation from shot to shot.

This is not an issue if you are aiming for a single final print (or processed digital image), but it can be if you intend to process/print multiple shots from the same scene. Each shot might be perfectly exposed, by ETTR standards, but they will not match each other, and you won't be able to apply the same corrections.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Speeding]

Over time you learn from each camera when/where they tend to clip.  Some have adjustable compensation per metering mode you can set and forget.  With The SL, I tend to under-expose at ISO 50 because, as you noted in your other thread, it's too easy to blow highlights.  The only time I find myself ETTR is in the absence of bright/specular light in the scene.

[jaapv]

6 hours ago, SrMi said:

[In the other thread that I started (SL2, ISO50, and overexposed JPEGs), ETTR discussion threatened to overpower the issue of live-view/image-review discrepancy.
Hence another thread.]

Wikipedia (link) has a good summary of ETTR (exposing to the right).

The point of ETTR is to increase the signal to noise ratio: 
a) maximize the signal by bringing more light to the sensor. 
b) minimize the noise by using base ISO.

Bringing more light to the sensor means increasing the exposure without clipping the relevant highlights.
Without a RAW based histogram/clipping, it isn't easy to know how much to expose (specific camera settings can reduce the discrepancy between JPEG and RAW histograms).
I have a conservative approach to ETTR, where I expose less rather than blow out the highlights.

The increased noise of ISO amplification makes ETTR practical at base ISO only.

blanket syToday's sensors are much better than they were in 2003 when the ETTR method was proposed. It is not as much a necessity as it used to be. However, I still find 'properly' manually exposed images better than letting the camera compute exposure automatically (no EV compensation).

With present-dat EVF cameras I find it much more practical to blend the histogram into the viewfinder and shoot for the "fullest" histogram (using the EV comp wheel) , keeping a wary eye on the righthand border. I think that ETTR as a blanket system has been made obsolete by technology.

[SrMi]

28 minutes ago, jaapv said:

With present-dat EVF cameras I find it much more practical to blend the histogram into the viewfinder and shoot for the "fullest" histogram (using the EV comp wheel) , keeping a wary eye on the righthand border. I think that ETTR as a blanket system has been made obsolete by technology.

Where do you think your approach is different than ETTR?

BTW, on SL2, the live-view histogram and blinkies are more reliable than the one shown in the image review. I just wish Leica cameras had large live RGB histograms that can be quickly blended over (Fuji's implementation).

[jaapv]

It is different from ETTR because it pays more attention  the overall exposure instead of fine tuning the highlight clipping.  A matter of emphasis. To put it overly aggressive: ETTR only pays attention to the division of highlights - which can be sacrificed, which can be kept, whereas the method I describe puits the emphasis on the bulk of the exposure, more akin to the zone system. 

[LocalHero1953]

Advertisement (gone after registration)

I also don't use ETTR explicitly in this Liveview age. I use the histogram and aim for a balanced exposure. I see where the clipped bits are - if they are specular highlights or light sources that can never be 'unclipped' (e.g. the sun) then I don't worry about them - or accept that they may wreck the photo whatever I do. So I keep an eye on clipping, but also on blocking and the overall light distribution. ETTR might be what I'd do if I was still using film, or a more primitive digital camera, but there is no need now that we can get fuller info, easily displayed as a histogram.

The histogram is like carrying a spotmeter around that allows you to check the whole scene. Perhaps one day we'll get LV displays that will show a Zone pattern distribution across the scene as well. Or has that been done already?

Edit: I should have added that 99% of the time I use A mode, AutoISO, and adjust exposure with a dial set to Exposure Compensation.

Edited August 28, 2020 by LocalHero1953

[jaapv]

If the image will be wrecked however you expose, one can always fall back on an exposure bracket and HDR.

[pgk]

ETTR leads to inconsistency, variation in tonality, tricky extreme highlights with banding and colour shifts. IMO its better to pay attention to the exposure in terms of what you are trying to achieve (appropriate aperture and shutter speed) and then adjust for underexposure in post - this isn't a perfect methodology either but it works well for me shooting at low ISOs and avoiding over-exposure. The point today is that exposure is rarely something that you can be as really definitive about as you could in the past with film, if that is, you are shooting RAW (I don't shoot JPEGs - to me its a waste of time). So thinking that ETTR or any other fixed way of working is a solution to all situations is wrong, it isn't. ETTR has its place and is good for chatting about, lots of discussion with examples and being generally technical over. The good thing about digital is that its easy to take several different exposures and adjust them to find out which works best in the situations that you shoot.

[Chaemono]

I have to admit, after doing some test shots with the S3 - on a tripod - at base ISO, there was only very little room to ETTR, 1/45 sec., before colors changed and couldn’t be made to look the same in post processing as in the picture captured to protect highlights at 1/90 sec. The shot at 1/30 sec. already was too far to the right. The red color was irrecoverable.

I’m at work now - to pay for the S3 - and will post links to the DNGs later but you can see the JPGs including crops to show the difference in noise if one shoots to protect highlights all the way down here: https://www.smugmug.com/gallery/n-JkgxSt/i-Fm2Cmzr

After this test, I don’t think ETTR is worth it with the S3, at least.

[jaapv]

That is why I would like to have an RGB histogram. It is all too easy to blow one colour channel without noticing.

[Jared]

I think ETTR made sense a few years ago when shadow noise could be quite visible even at base ISO and dynamic range was comparatively limited.  Now? Not so much. Depending on the composition, I am probably more worried about losing highlights than I am about shadow noise in most situations.  That wasn’t true a few years ago, but it is now. In situations where that isn’t true, I’m likely to bracket and fall back on HDR in any event.

[SrMi]

20 hours ago, BernardC said:

One limitation of ETTR, which I mentioned in the other thread, is that it leads to exposure variation from shot to shot.

This is not an issue if you are aiming for a single final print (or processed digital image), but it can be if you intend to process/print multiple shots from the same scene. Each shot might be perfectly exposed, by ETTR standards, but they will not match each other, and you won't be able to apply the same corrections.

I do not understand the reasoning.

Why would you have different exposures with ETTR and not with automatic exposure?

[SrMi]

18 hours ago, Speeding said:

Over time you learn from each camera when/where they tend to clip.  Some have adjustable compensation per metering mode you can set and forget.  With The SL, I tend to under-expose at ISO 50 because, as you noted in your other thread, it's too easy to blow highlights.  The only time I find myself ETTR is in the absence of bright/specular light in the scene.

You need to underexpose at ISO50 only when using automatic exposure and high contrast scenes. ETTR should be applied only at low contrast scenes.

[SrMi]

9 hours ago, LocalHero1953 said:

I also don't use ETTR explicitly in this Liveview age. I use the histogram and aim for a balanced exposure. I see where the clipped bits are - if they are specular highlights or light sources that can never be 'unclipped' (e.g. the sun) then I don't worry about them - or accept that they may wreck the photo whatever I do. So I keep an eye on clipping, but also on blocking and the overall light distribution. ETTR might be what I'd do if I was still using film, or a more primitive digital camera, but there is no need now that we can get fuller info, easily displayed as a histogram.

The histogram is like carrying a spotmeter around that allows you to check the whole scene. Perhaps one day we'll get LV displays that will show a Zone pattern distribution across the scene as well. Or has that been done already?

Edit: I should have added that 99% of the time I use A mode, AutoISO, and adjust exposure with a dial set to Exposure Compensation.

ETTR is not something to be applied for film, AFAIK. It is much easier to do with live-view than with shoot-chimp-correct approach.

Edited August 28, 2020 by SrMi

[SrMi]

6 hours ago, jaapv said:

That is why I would like to have an RGB histogram. It is all too easy to blow one colour channel without noticing.

Yes, I really like Fuji's implementation.

[SrMi]

I believe that ETTR, its application, and the reasons behind it are still misunderstood.

ETTR is a technique to be applied in low contrast images when the DR of the scene is smaller than the DR of the sensor. 

The basis of ETTR is to expose the sensor to more photons than with 'regular' exposure. More photons equal better SNR. Better SNR is preferable, regardless of sensor generation.

I believe that ETTR can be applied correctly without image deterioration.

[LocalHero1953]

20 minutes ago, SrMi said:

ETTR is not something to be applied for film, AFAIK. It is much easier to do with live-view than with shoot-chimp-correct approach.

For film, perhaps I should have written ETTL. My intended point was that with modern Liveview one has no need to rely on single exposure measurements (e.g. of highlights) when more detailed data is available in the histogram.

[Chaemono]

With sensors that are prone to color banding in pushed shadows at high ISO in high contrast scenes, ETTR may be the lesser evil than blowing a color channel.

Anyway, here are three DNGs at same settings but with shutter speeds of 1/90 sec., 1/45 sec., and 1/30 sec. The last one blows the red channel and the color cannot be recovered in post.

S3 + 70 Summarit-S exposed to protect highlights. DNG here: https://cc2032.zenfolio.com/img/g774922268-o750076470.dat?dl=2&tk=aAPElt7y9RgU2hunDK8mcXMKuGf1TddpFCVX50XVNUg=

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

ISO 200 f/4 @1/90 sec.

S3 + 70 Summarit-S exposed somewhat to the right. DNG here: https://cc2032.zenfolio.com/img/g610177191-o750076470.dat?dl=2&tk=l8jMc3imecOfKa-sD0dAE4nRd_cVifN71kTl4iR3bN8=ISO 200 f/4 @1/45 sec.

S3 + 70 Summarit-S pushed even more to the right with the results that the red cannot be recovered in post. DNG here: https://cc2032.zenfolio.com/img/g891700046-o750076470.dat?dl=2&tk=oYn1a9P7APaAq52FXzyqM1Cu-ARy7tx0DaAe943y1mI=ISO 200 f/4 @1/30 sec.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[SrMi]

24 minutes ago, Chaemono said:

With sensors that are prone to color banding in pushed shadows at high ISO in high contrast scenes, ETTR may be the lesser evil than blowing a color channel.

Anyway, here are three DNGs at same settings but with shutter speeds of 1/90 sec., 1/45 sec., and 1/30 sec. The last one blows the red channel and the color cannot be recovered in post.

<snip>

Those are very high contrast images that are, therefore, not good candidates for the ETTR strategy.