AE and ETTL Algorithm: Why not use ETTR?

I don't like the current AE system. It adds the relative brightness levels of all the metered zones, averages them, then sets exposure so that this average falls 3 stops below the maximum recordable tone.

Why is this bad? Because by calibrating this at -3 stops means that only when the scene's dynamic range is exactly 6 does the sensor capture it optimally. When the scene DR is above 6 stops, you get blown highlights. When it is below 6 stops, the image is"underexposed".

An example. Let's say that a sensor has only 4 metering zones: Z1-Z4. For simplicity, each zone is a single pixel. Here are the three scenarios with numbers representing the relative brightness:

1) Metered values in a scene with a 6 stop dynamic range

• Z1 = 2
  
• Z2 = 1
  
• Z3 = 7
  
• Z4 = 6
  

• Z1 = 5
  
• Z2 = 4
  
• Z3 = 10
  
• Z4 = 9
  

• Z1 = 3
  
• Z2 = 1
  
• Z3 = 7
  
• Z4 = 9
  

What happens to the argument when there are 63 metering zones?

TeamSpeed wrote in post #16370226
What happens to the argument when there are 63 metering zones?

There could be 1000 metering zones and the concept would still apply. Calibrate the meter off of the brightest zone.

frugivore wrote in post #16370198
I don't like the current AE system. ... What do you think?

I am pretty sure that you are not REQUIRED to use it. Feel free to do what flips your bippy.

That's not the point. The point is that there is an AE system, and potentially it could be better than it is.

I like the idea. Frugivore, have you tried it out? What do the images look like? Using the multi-spot metering (1D series only, I know) and selecting the zones with different brightness manually would give an idea, right? One can easily aim for the brightest, the darkest and some "typical" zones i the image. The scale in the viewfinder would give an indication of relative brightness of the zones selected, and also allows you to see how you move the whole scale with the exposure compensation button.

If nothing else, they could provide a custom function where you enable ETTR AE instead of the normal method.

I would agree with some new configuration options. The 1d already has AE microadjustment, they could very well add a new option off that allows you to set AE off the brightest point that is utilized for whatever metering mode you are employing.

apersson850 wrote in post #16370625
That's not the point. The point is that there is an AE system, and potentially it could be better than it is.

I like the idea. Frugivore, have you tried it out? What do the images look like? Using the multi-spot metering (1D series only, I know) and selecting the zones with different brightness manually would give an idea, right? One can easily aim for the brightest, the darkest and some "typical" zones i the image. The scale in the viewfinder would give an indication of relative brightness of the zones selected, and also allows you to see how you move the whole scale with the exposure compensation button.

If nothing else, they could provide a custom function where you enable ETTR AE instead of the normal method.

Regardless of whether the selected metered area is the entire sensor (evaluative), a small portion of it (spot), or multiple portions of it (multi-spot metering in the 1D series), the current autoexposure system always makes the assumption that you are photographing a 6-stop DR scene. I've always assumed this and I did some testing last week to confirm, using a uniformly lit foam core board shot with stoppped-down lens (to reduce vignetting. The result was this:

TeamSpeed wrote in post #16370698
I would agree with some new configuration options. The 1d already has AE microadjustment, they could very well add a new option off that allows you to set AE off the brightest point that is utilized for whatever metering mode you are employing.

But if that method gave you a perfect exposure every time, why would you need the old method, which almost always needs EC?

BTW, what is AE microadjustment? Does it apply a certain EC value all the time?

It seems that I do not understand the problem. It also seems to me that the answer being pursued here will only further complicate the problem. The shooter still has to know HOW the meter is reading the scene and suggesting a setting. The shooter still has to adjust that suggestion to his/her own perception of the scene and desired outcome. Adding another metering mode, making another suggestion as to exposure, solves what problem, exactly?

IMHO, having one more metering mode to understand doesn't help much. If I am shooting in Evaluative, which I am most of the time, I have to know how that system works, considering which AF point I used and all that. Or I can shoot Spot, CWA or Partial, which all key off the center of the frame, which may not be what I am shooting.

No matter which metering mode is employed, there is still only one proper exposure for the scene in the VF. It is also still up the the shooter to accept or reject the meter's suggestion. How is one more metering mode going to change any of that? I guess my question is, if the current metering system can be used to arrive at the proper exposure, what else is there?

frugivore wrote in post #16370870
Regardless of whether the selected metered area is the entire sensor (evaluative), a small portion of it (spot), or multiple portions of it (multi-spot metering in the 1D series), the current autoexposure system always makes the assumption that you are photographing a 6-stop DR scene.

...

But if that method gave you a perfect exposure every time, why would you need the old method, which almost always needs EC?

...

I read this post immediately after making my previous post. If you are already intimately familiar with the metering system, and how it works, what exactly is the problem you are trying to solve?

I don't care what metering system you use, or what new one someone comes up with, it is NEVER going to give you a perfect exposure every time. Doesn't the best hope for that lie in understanding how the current system works, and applying your knowledge of it, and the scene at hand, to adjust accordingly?

YashicaFX2 wrote in post #16370888
It seems that I do not understand the problem. It also seems to me that the answer being pursued here will only further complicate the problem. The shooter still has to know HOW the meter is reading the scene and suggesting a setting. The shooter still has to adjust that suggestion to his/her own perception of the scene and desired outcome. Adding another metering mode, making another suggestion as to exposure, solves what problem, exactly?

What I am describing is a metering system that will never need EC. What I am describing is automatic ETTR.The different exposure modes right now just change what part of the scene is metered, not how the system works.

The capture of the scene into RAW and the output onto screen/paper should be separate. So the RAW would have highlights at just before clipping. You can then reduce the exposure in post. Or you could use -EC during capture if you really want to avoid post. But at least you will always have a perfectly captured ETTR exposure.

YashicaFX2 wrote in post #16370888
IMHO, having one more metering mode to understand doesn't help much. If I am shooting in Evaluative, which I am most of the time, I have to know how that system works, considering which AF point I used and all that. Or I can shoot Spot, CWA or Partial, which all key off the center of the frame, which may not be what I am shooting.

No matter which metering mode is employed, there is still only one proper exposure for the scene in the VF. It is also still up the the shooter to accept or reject the meter's suggestion. How is one more metering mode going to change any of that? I guess my question is, if the current metering system can be used to arrive at the proper exposure, what else is there?

Let me be clear, I am suggesting an autoexposure system that will be perfect in every single scenario, every time. Without fail. Photographers would never need to worry about exposure ever again. How useful do you think that would be?

YashicaFX2 wrote in post #16370902
I read this post immediately after making my previous post. If you are already intimately familiar with the metering system, and how it works, what exactly is the problem you are trying to solve?

I don't care what metering system you use, or what new one someone comes up with, it is NEVER going to give you a perfect exposure every time. Doesn't the best hope for that lie in understanding how the current system works, and applying your knowledge of it, and the scene at hand, to adjust accordingly?

Ok, let's take the same example in the OP, where I shot an 8-stop DR scene. In that example, it overexposed because Canon's AE system is calibrated for a 6-stop range.

Here are the same values:

•Z1 = 3
•Z2 = 1
•Z3 = 7
•Z4 = 9

If I base the exposure off of the brightest zone, I have this equation:

10 (sensor's maximum) - 9 (brightest zone) = 1

So if the system then exposes the scene at +1, the resulting zones come in at this:

•Z1 = 3 + 1 = 4
•Z2 = 1 + 1 = 2
•Z3 = 7 + 1 = 8
•Z4 = 9 + 1 = 10

Perfect exposure.

Do you see how if we judge the exposure on the brightest zone we would never get any clipping? Right now, Canon is judging the exposure on the average of all zones. This is the problem. An average does not know the range of EC values in a scene. It does not know if you are photographing a 2-stop scene or a 10-stop scene.

frugivore wrote in post #16370198
I don't like the current AE system. It adds the relative brightness levels of all the metered zones, averages them, then sets exposure so that this average falls 3 stops below the maximum recordable tone.

Why is this bad? Because by calibrating this at -3 stops means that only when the scene's dynamic range is exactly 6 does the sensor capture it optimally. When the scene DR is above 6 stops, you get blown highlights. When it is below 6 stops, the image is"underexposed".

The supposition is that the Canon dSLR somehow meters differently than any handheld meter in existence, when the contrary is true! It can be proven that the Canon meter, when pointed at a purely uniform 18% midtone area, will meter -- regarless of spot or Evaluative or Average or Partial -- virtually the same as...

• A Minolta handheld one degree spotmeter aimed at a purely uniform 18% mid-tone
  
• A handheld incident light meter held at subject position and aimed back at camera position
  
• A film SLR meter aimed at a purely uniform 18% mid-tone

Frugivore,
Effectively what you want is a meter that allows you to identify which zone is 'highlight area' and which zone is 'shadow area'...what the Olympus OM-4 could do 30 years ago and the top of the line Canons can do today, with multizone spotmetering.

Frugivore,
Thank you for the detailed explanation. That makes perfect sense. As far as it goes.

I guess my problem is the implementation of it. Several issues remain. How is the camera going to know which brightness to preserve. If it is ALWAYS going to be the brightest part of the scene, won't spot metering handle that? If the brightest part is NOT what the shooter wants preserved, then what? Say there are several metering zones that I don't care to preserve; I want them blown in order to preserve some less bright detail and get some detail in darker areas. How do you explain that to the camera?

I can see a system where you pick several spots (let's go with 3) to meter, then let the camera decide. If I shoot the scene and pick a wider range of brightness for my 3 points than you would pick for the very same scene, one of us is not going to get what we want. It still comes down to knowing how the camera's metering system works, and applying that knowledge to get the exposure you want.

So far none of this addresses color, which hugely complicates metering and exposure.

The only way I can see around this is a 'open' metering system. You'd have to shoot in live view and select ONE exposure parameter to lock. Say I pick my fixed parameter as f/3.2. I rotate one of the dials and the camera. The LV screen shows a real-time representation of the scene while I rotate the dial, selecting brighter or darker, as the camera adjusts the other two parameters. (Does that sound like Creative Auto?) When it gets where I like it, I fire the shutter. The catch, how to account for screen brightness and the influence of ambient light on perceived brightness.

I grew up with old manual film cameras that used center-weighted-average metering, ONLY. Maybe 10 years ago I bought my first digital. It had several metering modes. I played with them and came to the conclusion that I only need to know how one of them works to get a decent exposure. About 5 years ago I got my first Canon DSLR. I ran into the same problem. If I know how Evaluative works, and can get a proper exposure with that, of what use are the other modes? The only way to get any real use out of the other modes (spot, cwa, partial) is to experiment with them, see how the camera employs them, and adjust accordingly when necessary. If I can already do that in evaluative with some degree of proficiency, and trial & error, why bother switching modes?

Adding another exposure mode only means that you will have to learn even more about how the camera is going to handle a particular situation. It sounds really nice in theory. I don't see how it would ever be implemented, or who it would help if it were.

This whole thing reminds me of a thread a while back where the OP was shooting Av, with safety shift engaged, using ETTL flash and auto-ISO. His complaint was that he was getting underexposed backgrounds. Sometimes the shooter has to take control of the camera. I really think that applies here, too.

Wilt wrote in post #16371163
Frugivore,
Effectively what you want is a meter that allows you to identify which zone is 'highlight area' and which zone is 'shadow area'...what the Olympus OM-4 could do 30 years ago and the top of the line Canons can do today, with multizone spotmetering.

Assume the camera is capable of 6 stops of DR. What does the camera do when you select two points that are 4 stops apart, or 8. It is still a guessing game.