ETTR: The Ultimate Exposure Technique?

[Ebrahim Saadawi]

After initial testing on how ETTR affects skin tones, I am seeing what I saw earlier, it gives better everything, less noise, more colour thickness.

I am really not seeing any skin issues others claim that supposedly come from placing the skin in the "most compressed area of the curve", in theory yes but to my eyes the ETTR skin looks better.... I've seen a lot of examples when people try to ETTR, they clip the Red channel, and when they try to bring back the exposure in post it turns all funky with loss of tonality. So perhaps that's the reason. ETTR means not to clip anything, not Luma, not colour channels, nothing. Use an RGB histogram if available while shooting.

Here's skin with normal exposure:

Here's with ETTR:

same skin colour, just a better less noisy image and you can push the image much farther due to the lack of noise.

But again, before I declare this my ultimate exposure technique I will run more in depth testing on the Skin issue.

 

[Joshua_G]

Thank you, Ebrahim.
ETTR, done correctly, should improve everything without degrading anything.
ETTR, done correctly, means biasing the exposures so that the histogram is snugged up to the right, but not to the point that the highlights are blown. See Michal Richman's original article about ETTR on his Luminous Landscape website from 2003: http://www.luminous-landscape.com/tutorials/expose-right.shtml and his updated article from August 2011: http://www.luminous-landscape.com/tutorials/optimizing_exposure.shtml . Michal Richman is the one who coined the term ETTR.

There are two points to bear in mind:

1. ETTR is applicable only to raw files for stills, or to files that are intended to be graded/corrected or processed in any way after shooting for video. It isn't intended to files that are to be used as they are produced by the camera, without any processing.
2. ETTR means that the highlights should ALMOST touch the right of the histogram window. Care should be taken not to clip the highlights. As long as the highlights aren't clipped, there is all gain with no loss.

 

[Samuel H]

The skin problem some people mention appears because of what you guessed: there's no luma clipping, but one channel is clipping. If you avoid that, skin will be good.

So, three caveat to take into account regarding ETTR:
* You can have a clipped channel without any indication in the luma histogram.
* It only works for linear or log data. Using ETTR with a camera that doesn't record RAW or log footage is usually a bad idea.
* You don't want to use different amounts of overexposure in shots that will be cut next to each other, because you'll end up with jumps in the amount of noise in the image. Think of the whole scene when you decide how much you overexpose.

 

[Joshua_G]

Hi Samuel,
You are correct.

It's possible that on certain circumstance one of the color channels will clip, while the luma will not. So care should be taken. It can be solved by the histogram showing RGB instead of luminance, or by using waveform monitor set to flat (showing both luminance and RGB). If one is using a monitor that has both histogram and waveform monitor, it's best to use both.

Of course the very same exposure should be maintained between shots (as much as the same white balance).

I'm not clear about what you meant in your second point.
ETTR, done correctly, works well for both linear and log profiles, for both RAW and compressed file formats.
The point to bear in mind is that ETTR is applicable only for files that are to be processed after shooting (in editing or correcting/grading or post).

However I'm somewhat uneasy by the saying "overexposure". Using ETTR is overriding the camera's light-meter recommendation, but by no means it is 'overexposing' - it is actually exposing correctly (for processing after shooting). Overexposure is only when either luma or any of the color channels is clipping.

 

[Samuel H]

My second point is that ETTR is not a good idea if your camera doesn't shoot either RAW or log. Like, when I shoot video with my old NEX-5N, it always has a built-in S-curve, so ETTR is a bad idea because bringing the footage down in post will not always work as expected.

 

[DPStewart]

My second point is that ETTR is not a good idea if your camera doesn't shoot either RAW or log. Like, when I shoot video with my old NEX-5N, it always has a built-in S-curve, so ETTR is a bad idea because bringing the footage down in post will not always work as expected.

ETTR technique seems to work exactly as intended on my GH2 (hacked for very high bit rate).
I agree the best way to describe the practice is by saying: "Exposing to the limits of the camera - not to the light meter recommendation."

 

[Samuel H]

AFAIK, video from the GH2 has a very aggressive S-curve built in. I'd be very cautious about ETTR on that one. But if you like what you get, well, more power to you

 

[Joshua_G]

My second point is that ETTR is not a good idea if your camera doesn't shoot either RAW or log. Like, when I shoot video with my old NEX-5N, it always has a built-in S-curve, so ETTR is a bad idea because bringing the footage down in post will not always work as expected.

Possibly it's unique to the NEX-5N. Tomorrow I'll experiment with Canon 600D, which has neither RAW nor Log.

 

[Joshua_G]

AFAIK, video from the GH2 has a very aggressive S-curve built in. I'd be very cautious about ETTR on that one. But if you like what you get, well, more power to you

It doesn't need to be dependent on personal taste. The set of monitors in Premiere Pro can give accurate answer.

 

[Ebrahim Saadawi]

More accurate tests coming soon. Will test for both LOG, linear RAW and Rec.709 and how each responds to ettr. I hope everyone would do a few simple tests to confirm too, please!

 

[Samuel H]

It's definitely not a NEX-5N thing: anything with a built-in s-curve requires a complex transformation in order to re-expose in post in an accurate manner. You have to undo the s-curve, then re-expose, then reapply the s-curve. If you don't, well, you end up with a darker image but contrast distribution will be different from what you would have got if you shot with a different exposure to begin with. This bothers me, but it may not bother you.

So let's leave it at "use ETTR only if your camera shoots RAW or log, or if you don't care about the side-effects of re-lighting footage that has a built-in s-curve"

 

[Ebrahim Saadawi]

Samuel might be right because all my tests were in "kind of log" formats, I.e, not comtrasty rec709. Canon with Cinestyle, a7s with slog, c100 with clog, and gh4 with cineD. All of these log curves are different but kind of behave the same to ETTR, makw the image better. All of these are an inverse luma curve, so perhaps doing ettr with rec.709 standard would not work. But to be clear I completely don't believe it only works with Raw, it definitely works with compressed Log footage, to my eyes and tests anyway...

 

[Joshua_G]

Indeed, Samuel is correct.
As I wrote earlier:

…

1. ETTR is applicable only to raw files for stills, or to files that are intended to be graded/corrected or processed in any way after shooting for video. It isn't intended to files that are to be used as they are produced by the camera, without any processing.

…

Thus, I'd add to what Samuel wrote: "use ETTR only if your camera shoots RAW or log…" also if you use "kind of log" formats, like Cinestyle or flat.

 

[Samuel H]

Yes, I bunch all those options under "log".

This is a7s+slog2

This is T2i with Flaat_10

This is T2i with CineStyle

No s-curve in any of those.

This is Portrait with Contrast=-1 on the T2i, which builds in a mild s-curve:

 

[Ebrahim Saadawi]

Samuel, Confirmed. ETTR doesn't work with contrasty rec.709 images. Not much room to increase exposure without clipping to make a diffference, so don't bother. For example, it's magical with A7s Slog2 but doesn't work with normal picture profiles, and it works great with Cinestyle but not with normal profiles. Weird but my tests show this very clearly, if you're not shooting LOG/flat do not ETTR!

 

[Samuel H]