Why is auto exposure so useless?

[Justinr]

When I've been faced with similar circumstances, I've studied my subject to find a spot that nearly approximates 18% grey, spot metered it, and reframed for the shot.  Shoot only raw.  I have never submitted images to anyone that I have not processed to my satisfaction . . . except once many years ago.  I had a client approach me with a job for his website.  He had a limited budget and said his web design guy could process the images if only I could shoot them for what he could pay.  That was a learning experience I will never repeat.  I don't think there's a viable shortcut for processing.

Quite so on the release of images. I certainly don't let them out until I'm happy with them and this has caused some upset when companies just expect me to hand over the pictures willy nilly, but that's a subject for a different thread.

Studying the subject when it is proceeding towards you at trotting pace with flails and knives slicing away at the grass and anything else it their path is an exercise that is best kept to the bare minimum. Sure, we all know that would be the ideal but we are not talking about a static situation here, rather it is one that is fast moving and dangerous to the unwary.

[spidermike]

Apparently not it seems.  It's probably best to recap at this point.

I was covering an event for a couple of magazines back in July where there were numerous tractors and forage harvesters strutting their stuff. Also present were two colleagues working for other publications. We got talking to each other and we all despaired of the fact that AE was pretty blo*dy useless in our line of work as it was likely to deliver less than optimal results, quite a bit less on frequent occasions. It was not always wrong but certainly not consistent enough to trust.

The work entailed taking pictures of machines as they worked the harvesters and so our positions in relation to them changed constantly, if only to avoid having our legs chewed off!

Personally I have been taking pictures professionally for over 12 years with a variety of cameras and formats and I know the other two as  competent photographers. Whatever tricks and settings I tried on the day the camera would deliver a variety of exposures, some of which were fine, some of which would be acceptable after adjustment but many were instantly deleted mainly because of gross under exposure.

If it were just me then perhaps we could settle on my stupidity but my experience was mirrored by two others who's work I respect, so it's either us poor lads in the field not having a clue or posters on here lacking in grey matter.

Take your pick!

That sounds not me not so much a matter of 'consistency' by the camera because you are continually moving position and at different times presenting the camera with different balances of light and shade, colour tone and glare. Some of those situations it can cope with (being within its programming limits) and other times it gets fooled. It is acting totally consistently within itself and how it meters but it is inconsistent as to how far from (or how near to) your desired output its picture is. This is why I kept asking what you meant by consistency which you derided me for.

I have a similar problem all the time with birds in flight where the bird is passing clouds, then dark shadow, then well exposed foliage and it is those times I switch to manual. In AE the camera is being consistent, but how acceptable it is to me (the exposure on the bird) depends on where in the flight path the picture was taken.  

[Slobodan Blagojevic]

Ok, ok... time out, everyone! Time to pause and see what is going on here. In one corner, we have a clearly disappointed, yet without doubt experienced photographer, and in the other all of us making fun of him.

So let me switch corners and come to my friend's defense. I think I know what is causing this misunderstanding.

It is called Matrix metering. Also known as Evaluative metering and several other names. While all other metering modes are indeed utterly predictable (as long as you assume 18% gray), matrix metering may not necessarily be. Here is what Wikipedia says about it (emphasis mine):

... the camera measures the light intensity in several points in the scene, and then combines the results to find the settings for the best exposure. How they are combined/calculated deviates from camera to camera. The actual number of zones used varies wildly, from several to over a thousand. However performance should not be concluded on the number of zones alone, or the layout. In general, the most advanced metering is found on single-lens reflex cameras.

Many manufacturers are less than open about the exact calculations used to determine the exposure. A number of factors are taken into consideration, including the following: Autofocus point, distance to subject, areas in focus or out of focus, colors/hues of the scene, and backlighting. Multi-zone tends to bias its exposure towards the autofocus point being used (while taking into account other areas of the frame too), thus ensuring that the point of interest has been exposed for properly, (this often means the subject area being exposed for as a mid-tone). A database of many thousands of exposures is pre-stored in the camera, and the processor can use a selective pattern to determine what is being photographed...

There is considerable variation from different manufacturers as to how multi-zone metering is implemented, and even from the same maker in their model range, and how much "priority" is given to the AF point itself...

However, some photographers may be uncomfortable with multi-zone metering. This tends to stem from a lack of clarity about "how" the camera reacts in certain situations...

Some users have problems making wide-angle shots with high contrast, due to the large area which can vary greatly in brightness...

So, for instance, while Justin is sure he is photographing a bright orange combine, his camera might think it is a sunset instead.

[Justinr]

Ok, ok... time out, everyone! Time to pause and see what is going on here. In one corner, we have a clearly disappointed, yet without doubt experienced photographer, and in the other all of us making fun of him.

So let me switch corners and come to my friend's defense. I think I know what is causing this misunderstanding.

It is called Matrix metering. Also known as Evaluative metering and several other names. While all other metering modes are indeed utterly predictable (as long as you assume 18% gray), matrix metering may not necessarily be. Here is what Wikipedia says about it (emphasis mine):

So, for instance, while Justin is sure he is photographing a bright orange combine, his camera might think it is a sunset instead.

Now there is an interesting article and may well start to explain what is going on. All we need now is some way of opting out of this memory bank consultation process that seems to be going on. Can it be done?

[spidermike]

Than you Slobodan. I assumed people knew how matrix metering worked, especially experienced professionals (not a dig at all, just an assumption that may well not be valid) .

All we need now is some way of opting out of this memory bank consultation process that seems to be going on. Can it be done?

I don't think you can. In the old days (maybe the Mamiya you had) it would assume the whole scene was 18% and gave the appropriate reading and people overrode it. But in many scenes this leads to under/over exposure of the true subject so manufacturers developed the matrix metering and it databank to try and second guess what the user wants. So now if the cameras will see a scene with white on the lower half and from its database works out it is a snow scene and it is programmed that for that particular type of scene it knows it needs to increase exposure from 18% to keep faces from underexposure and to keep the snow from turning grey.

[Slobodan Blagojevic]

Now there is an interesting article and may well start to explain what is going on. All we need now is some way of opting out of this memory bank consultation process that seems to be going on. Can it be done?

Ha! That is a very succinct and accurate way to describe it!

It can be done, but not within matrix metering. You'd need to switch to any other auto or manual metering mode (they all rely on the 18% gray assumption) and then adjust accordingly. Or rely on post-processing.

[Colorado David]

Quite so on the release of images. I certainly don't let them out until I'm happy with them and this has caused some upset when companies just expect me to hand over the pictures willy nilly, but that's a subject for a different thread.

Studying the subject when it is proceeding towards you at trotting pace with flails and knives slicing away at the grass and anything else it their path is an exercise that is best kept to the bare minimum. Sure, we all know that would be the ideal but we are not talking about a static situation here, rather it is one that is fast moving and dangerous to the unwary.

I'm really not trying to be difficult.  My point was that there is no viable option other than processing your files.  What I mean, and in my experience, is that modern professional level (most consumer level even) equipment will get the exposure within a window that can be easily corrected in post processing.  I have recovered highlights or shadows that would have been impossible a few years ago.

[Justinr]

Than you Slobodan. I assumed people knew how matrix metering worked, especially experienced professionals (not a dig at all, just an assumption that may well not be valid) .



I don't think you can. In the old days (maybe the Mamiya you had) it would assume the whole scene was 18% and gave the appropriate reading and people overrode it. But in many scenes this leads to under/over exposure of the true subject so manufacturers developed the matrix metering and it databank to try and second guess what the user wants. So now if the cameras will see a scene with white on the lower half and from its database works out it is a snow scene and it is programmed that for that particular type of scene it knows it needs to increase exposure from 18% to keep faces from underexposure and to keep the snow from turning grey.

I would consider myself quite familiar with the concept of matrix metering but the existence of a look up data bank is certainly new to me; you? It is not dissimilar to the idea of engine mapping and I suppose we should not be surprised if methods used in one area of technology are deployed in another.

The Mamiya I refer to is a 645AFD, a model which is still pretty much with us today although it bears the Phase One badge. The ZD back is not perfect but it's upgrade is expensive, however, I doubt that either bothers with the data bank.

[Justinr]

I'm really not trying to be difficult.  My point was that there is no viable option other than processing your files.  What I mean, and in my experience, is that modern professional level (most consumer level even) equipment will get the exposure within a window that can be easily corrected in post processing.  I have recovered highlights or shadows that would have been impossible a few years ago.

I'm not so sure TBH. I do find that colours and saturation levels, or perhaps the relationships between colours, start to deteriorate the greater the depths or heights from which you try and recover the image. Maybe on much more recent equipment it is different but the D3 I have  (age  4 or 5) isn't particularly generous in this quarter and anything more than about two stops isn't worth bothering with, but that's subjective.

[spidermike]

I would consider myself quite familiar with the concept of matrix metering but the existence of a look up data bank is certainly new to me; you?

Yes, I was familiar with it. It has constantly amazed me how the computer in that little plastic box can not only meter and adjust settings in the time it takes to press a shutter, but that in doing it so it uses a library of (in the link below) some 30,000 images! There's some fancy algorithms gong on there...

https://support.nikonusa.com/app/answers/detail/a_id/13774/~/what-is-the-difference-between-spot,-center-weighted-and-matrix-metering%3F

[Justinr]

Yes, I was familiar with it. It has constantly amazed me how the computer in that little plastic box can not only meter and adjust settings in the time it takes to press a shutter, but that in doing it so it uses a library of (in the link below) some 30,000 images! There's some fancy algorithms gong on there...

https://support.nikonusa.com/app/answers/detail/a_id/13774/~/what-is-the-difference-between-spot,-center-weighted-and-matrix-metering%3F

From the link -

This meter instantly analyzes a scene’s overall brightness, contrast, and other lighting characteristics, comparing what is sees against an onboard database of over 30,000 images for unsurpassed exposure accuracy, even in the most challenging photographic situations.

My reply to that is a most genteel 'Bollux'.

Interestingly it is also a claim by the manufacturers that AE is accurate and far more than just an approximation or guide to exposure, so those who suggested that no such claims have ever been made might take note of this marketing ploy.

[Slobodan Blagojevic]

... Interestingly it is also a claim by the manufacturers that AE is accurate...

Strictly speaking, it isn't (the claim). They do not claim it is "accurate," just that it is more accurate than other (auto) methods.

Coincidentally, most grown-ups understand that "truth in advertising" is an oxymoron.

[Justinr]

Strictly speaking, it isn't (the claim). They do not claim it is "accurate," just that it is more accurate than other (auto) methods.

Well spotted, but the inference is there. It is part of what a friend recently described to me as the "Bullshit economy". There's a lot of it about!

[telyt]

In shooting basketball, with light and dark toned jerseys, you will drive yourself crazy trying to use autoexposure.  The light is not changing, but the camera sees different tones based on what you are focusing on and tries to compensate.

Your best choice with constant lighting is to take test shots to establish exposure, set on manual, and the click away.

Also with birds moving against various backgrounds, particularly birds with extensive areas of white or black plumage.  I gave up on auto-exposure many years ago and now with a live histogram and zebras in the viewfinder most of the time I even ignore the meter reading, using only the image in the EVF to set exposure.

[Justinr]

Err... No, because I have deleted them.


I try not to delete all the unsatisfactory photos but keep some as-examples; note what I think's wrong with them and what I think I could have done better.

Nice to have the time.

The camera doesn't know what picture you want. Some photographers know how to use the AE modes to make the camera expose for the picture they want.

Wonders will never cease.

While matrix/evaluative metering is indeed the most sophisticated metering mode. let me once again say that meters do not necessarily give the correct answer for proper exposure. They give "an" answer but "correct" is another matter. However, as I'll discuss… matrix metering is what I use almost all the time. … Because digital cameras also offer one of the best-of-all exposure aids: the histogram. (John Shaw's Guide to Digital Nature Photography.)

But why bother with metering if you are going to resort to the histogram anyway? Cut out the middleman and all that.

I have never used Nikon, but I read about this in a Joe McNally book.

It's not exclusive to Nikon

The manufacturers also say that you can do stuff wrong -- "Matrix metering will not produce the desired results with autoexposure lock." page 129 D3 Manual pdf

It's called covering their ass and TBH it's fairly self evident that if you lock the autoexposure then it can't work if the conditions change.

But otherwise very good.

[spidermike]

Well spotted, but the inference is there.

There are the old sayings relating to 'Assume' and 'assumptions'....and I think the word 'inference' is close enough for the same logic (and before you go all semantic on me I am aware of the difference between them).

[Justinr]

There are the old sayings relating to 'Assume' and 'assumptions'....and I think the word 'inference' is close enough for the same logic (and before you go all semantic on me I am aware of the difference between them).

This meter instantly analyzes a scene’s overall brightness, contrast, and other lighting characteristics, comparing what is sees against an onboard database of over 30,000 images for unsurpassed exposure accuracy, even in the most challenging photographic situations.

So are we to assume or infer from the above that Nikon believes it's  AE system  works well or not?

[Slobodan Blagojevic]

...So are we to assume or infer from the above that Nikon believes it's  AE system  works well or not?

Yes, and not just Nikon, most of us also believe it works freakishly well... most of the time. Just not perfect, not all the time, and not in every situation. That's what the gray matter behind the camera is for.

[Justinr]

Yes, and not just Nikon, most of us also believe it works freakishly well... most of the time. Just not perfect, not all the time, and not in every situation. That's what the gray matter behind the camera is for.

Well that's me buggered then, even yer man Isaac is on my case pointing out the multitude of my wicked shortcomings. Time for the whiskey and revolver!

[Justinr]

Do you have time not to learn from your mistakes?

(You do make mistakes, don't you?)


You photograph from a low camera position looking upwards, and then wonder why AE exposes for the sky instead of those bits of machinery that interest you :-)


The photographers who know how to use AE modes to make the camera expose for the picture they want -- still get the picture they want when the light keeps changing.

If you say so me old mucker, if you say so.