Evaluative metering - now I get it

Some years ago, I went to a Best Buy store to see a Canon EOS Digital Rebel (300D). All the pundits were saying that Canon's introduction of a serious dSLR under $1000.00 would, instantly, make all other digital cameras irrelevant. I rejected this notion. I told Carla that we needed to veer on our trajectory to actually look at the great silver hope.

She handled the display model a bit, and really enjoyed the fast response. (We had been using a spiffy EVF camera, a Fuji S602. But it has some real speed of response problems, partly due to EVF latency.) At lunch, she said, "I think we should go back and buy that Canon - it's what we need for the rodeo outing of my Red Hat Society chapter tomorrow."

Thus began our long and complicated interaction with The Mind of Canon. The first Great Mystery was Evaluative metering.

The generally accepted explanation was that it uses some kind of intelligent algorithm to determine the appropriate exposure, and that the algorithm "placed emphasis" on . . .

Now, here, you had your choice of several expressions to complete the sentence (I will state them more precisely than we usually heard them):

a. [the metering segment at] the active AF point (based on the notion that only one was "active" even though several might be "lit").
b. [the metering segments at] all the active AF points (based on the notion that all the "lit" AF points were "active").
c. [the metering segments at and surrounding] the Active AF point.
d. [the metering segments at and surrounding] all the active AF points .

I myself subscribed to interpretation "d".

Now just recently I found out (from Chuck Westfall of Canon USA, in a discussion thread on the ProPhoto Home forum) that only one AF point is ever "active"; that one controls the AF, and the other lit ones are only showing that they have the same opinion of subject distance. (And we are never told which is the really active one.)

And whatever it is that Evaluative metering does "at" the active AF point, it only does it at the "really active" one, not at the "me toos".

I already linked here to this disclosure.

But this still leaves the question, OK, what does Evaluative metering do that is special with the metering segment at the "active" AF point, or at the segments at and surrounding the "active" AF point (and which of those is it)? What is "emphasis", anyway?

Today Chuck Westfall (in a prolongation of that same thread) gave an explanation that is the best I have heard in all my years of Canon fancy:

By way of introduction, I had (by way of provocation!) asked whether one of two views of the matter were correct. He said, in effect, well, not really either, and then proceeded (I will bold what I consider key passages):

Doug:

Perhaps a better way of expressing the issue is as follows:

1. Evaluative metering reads all segments of the metering sensor.
2. Primary emphasis is placed on the metering segment overlaying the active focusing point, or the center focusing point when the camera is set for manual focusing.
3. Secondary emphasis is placed on the metering segments directly adjacent to the primary metering segment.
4. Tertiary emphasis is placed on all other metering segments.

Evaluative metering uses a variety of algorithms to determine if there is a need to apply exposure compensation to the reading obtained at the primary segment. The algorithms are proprietary and confidential for competitive reasons, but we know that they look at various factors including overall light level, backlighting, spotlighting, and an estimate of relative subject size, as revealed by a comparison of brightness values for each metering segment. They may look at other available data as well, but if so we're not sure exactly what that data may be or how it is used. We know that there are limits on the maximum possible range and gradation of exposure compensation steps that evaluative metering can apply, and we know that the specifications for these factors vary somewhat according to the camera model in question, but we do not know the exact specifications for either.

That's just about as much technical info as I can provide on this topic.

Chuck Westfall
Technical Advisor/Professional Products Mktg. Div.
Consumer Imaging Group/Canon U.S.A., Inc.

So, the truth is:

•The basic exposure measurement is based on the metering segment at the (always single) active AF point (in manual and some other modes, that is always the center one).
•This may then be adjusted ("exposure compensation", a phrase we did not formerly hear in connection with Evaluative metering) based on an intelligent analysis of the metering segments surrounding the active AF point, and (further down the priority chain) all the other metering segments.
•The scope of the subject is not estimated from the cluster of "lit" AF points (as I used to believe, thus my belief in interpretation "d" above), but rather based on an evaluation of the brightness profile.

So, while there is still a lot that we don't know about evaluative metering, I am today gigantically more comfortable with my understanding of it.

Thanks so much to Chuck Westfall in bringing this to a good conclusion. And my ass is worn out from fishing.

So what does this new "knowledge" mean to you in practical terms?

Ken Tanaka said:

So what does this new "knowledge" mean to you in practical terms?

Click to expand...

That the active focus point determines exposure, and is somewhat modified by the other metering segments. I used to believe that it was a subject weighted average based on a library of common scenes. Apparently the active focus point not only weighs heavy in the calculation, but it dominates.

That's not just semantics. It means that a bright sky which is at some distance from the active focus point will probably be more overexposed that one would think. In such a case, an average exposure weighting would more likely save the sky.

Of course, there are other methods of metering complex scenes, if one has the time to do e.g. multiple spot meterings (I don't think all models support that, the 1 series do, maybe other single digit models as well?).

Bart

Hi, Bart,

Bart_van_der_Wolf said:

That the active focus point determines exposure, and is somewhat modified by the other metering segments. I used to believe that it was a subject weighted average based on a library of common scenes. Apparently the active focus point not only weighs heavy in the calculation, but it dominates.

That's not just semantics. It means that a bright sky which is at some distance from the active focus point will probably be more overexposed that one would think. In such a case, an average exposure weighting would more likely save the sky.

Click to expand...

That's how it seems to me.

Again, I am struck by the shift in Canon's presentation of Evaluative metering over the years. When I first encountered it, Evaluative was presented as the real "King of the Road", with other modes for specialized circumstances.

Now it is essentially presented on the basis that "in some situations it will do a really good job".

Best regards,

Doug

Another realization (this came in the first wave of this enlightenment) is this.

I had often wished I could force AF to be on the center AF point, but still leave all the other AF points potentially active so they could control the full flower of Evaluative metering. But there was no mode that would do that.

Well, in fact if "automatic" AF point selection is in effect, and a particular AF point takes over AF, then all the other "lit" AF points play no role in exposure determination at all.

Therefore, if I preselect the center AF point (for focus control reasons), and thus no other AF points can "light up", I have done nothing to emasculate the Evaluative metering scheme. It will work just as if automatic AF point selection was in effect and the center AF point happened to take command of AF.

Best regards,

Doug

Bart_van_der_Wolf said:

That the active focus point determines exposure, and is somewhat modified by the other metering segments.

Click to expand...

Doesn't it say something like that in the manual? I swear I read that while I was learning my 1D years ago? or maybe it said that the active point weights the exposure, or something. I do remember reading something about the active focus point playing some sort of roll in the exposure calculation though, but I don't remember exactly what roll it said was played.

Hi, Daniel,

Daniel Buck said:

Doesn't it say something like that in the manual? I swear I read that while I was learning my 1D years ago? or maybe it said that the active point weights the exposure, or something.

Click to expand...

Yes, or something.

Best regards,

Doug

See pages 35-36 of the Canon 1Ds Mark III White Paper. Direct link to PDF.

Ken Tanaka said:

See pages 35-36 of the Canon 1Ds Mark III White Paper. Direct link to PDF.

Click to expand...

From White Paper link above said:

The new evaluative metering algorithm is based on the algorithm for the previous 21-zone
and 35-zone metering systems. With the optimized 63-zone metering sensor and
improved algorithm, more consistent and correct ambient and flash exposures are
obtained with less influence by the subject. The basic concepts for the evaluative
metering algorithm are:
1. Metering is weighted on the linked AF point.
2. If there is a very bright object in the picture, the exposure will be increased.
3. In backlit scenes, the exposure will be increased. With dark backgrounds,
the exposure will be reduced.

Click to expand...

What Chuck says, seems to contradict point 2 to a certain extend. Only when the bright object is in the immediate vicinity of the active focus point will it influence the exposure significantly, which contradicts the White Paper a bit. The White Paper doesn't indicate that distance to the active focus point has an influence.

Bart

Hi, Bart,

Bart_van_der_Wolf said:

What Chuck says, seems to contradict point 2 to a certain extend. Only when the bright object is in the immediate vicinity of the active focus point will it influence the exposure significantly, which contradicts the White Paper a bit. The White Paper doesn't indicate that distance to the active focus point has an influence.

Click to expand...

Notwithstanding my recent euphoric response to the latest revelation, I fear that our grasp of this will remain murky.

For example, to me there is quite a conceptual difference between:

•An exposure measurement (presumably otherwise averaged) that is weighted at the location of the active AF point

and

•An exposure measurement that is principally based on a point measurement (a single metering segment) at the active AF point, but subject to adjustment based on an evaluation of other indications.

But of course, maybe these are just two ways to describe, intentionally-vaguely, the same actual truth.

Best regards,

Doug

Hi Doug,

I was just talking to Bart on the phone I have come up with the following question.
If all this above is true, then how does it work when I make use of the CF4-3 option? I do that always. So my focusing is detached from the exposure which is done later when I half squeeze the shutter button. Even the DPP does not show which AF point has been used for CF4-3 purposes. So how does this then work, I wonder?

Cheers,

Hi, Cem,

Cem_Usakligil said:

Hi Doug,

I was just talking to Bart on the phone I have come up with the following question.
If all this above is true, then how does it work when I make use of the CF4-3 option? I do that always. So my focusing is detached from the exposure which is done later when I half squeeze the shutter button. Even the DPP does not show which AF point has been used for CF4-3 purposes. So how does this then work, I wonder?

Click to expand...

I'm pretty sure I determined the answer to that a while ago, but I forgot my conclusion! I'll try and reconstruct it.

My best recollection right now is that the active AF point is saved (though never disclosed to us) and is used in connection with later exposure determination (at half press). I'm pretty sure that it is saved even for a future shot (if one has not invoked AF anew with the button by that time).

I'll see if I can dig up my report on those findings. I'm pretty sure I reported it to some forum or another.

Best regards,

Doug

Doug_Kerr said:

Hi, Cem,


I'm pretty sure I determined the answer to that a while ago, but I forgot my conclusion! I'll try and reconstruct it.

My best recollection right now is that the active AF point is saved (though never disclosed to us) and is used in connection with later exposure determination (at half press). I'm pretty sure that it is saved even for a future shot (if one has not invoked AF anew with the button by that time).

I'll see if I can dig up my report on those findings. I'm pretty sure I reported it to some forum or another.

Best regards,

Doug

Click to expand...

Hi Doug,

It that is true, and I believe you as it sounds utterly logical, then most of my exposures are wrongly measured. I use the central AF point in CF4-3 and then recompose before I shoot except when the dof is shallow such as in portraiture. If the assumptions are correct, then this would be totally silly. Don't you think?

Cheers,

Hi, Cem,

Cem_Usakligil said:

Hi Doug,

If that is true, and I believe you as it sounds utterly logical, then most of my exposures are wrongly measured. I use the central AF point in CF4-3 and then recompose before I shoot except when the dof is shallow such as in portraiture. Silly, isn't it?

Click to expand...

Well, almost certainly if you have preselected the center AF point, that will be the "key" one during evaluative metering (since it is the only one that is "operative" - with regard to either AF or evaluative metering). Thus your procedure should be OK with respect both to AF and metering. (I assume that before you half press, you lay the center AF point on the place in the scene on which you wish to meter, and them maybe move again for the actual shot.)

My point is more telling in the event of "automatic" AF point selection. There, if AF was done on a certain AF point, I think it likely Evaluative metering "emphasis" will still be done on that point when metering is done. That might not be the point that was "on your metering target" when you half press for metering (we wouldn't even know which point to lay on the metering target).

If this is not the case, then (with AF and metering separated) when you half press to meter, the AF system would have to "cycle" again to choose a new AF point only for the purpose of metering, but not change the lens focus setting based on that operation. That seems unlikely.

The other possibility would be that, with AF and metering separated (and with Automatic AF point selection in effect), during metering the "Evaluative metering emphasis" would always be placed on the center AF point (regardless of which point was chosen by the camera for AF).

I will try and do some basic testing today to see if my recollection is at all on the mark. I will use the EOS 40D (the earlier determination was done with the EOS 20D).

Best regards,

Doug

Hi, Cem,

Forget what was here - it is a bum steer.

Testing continues!

Best regards,

Doug

Hi, Cem,

Well, now that I have things set properly!

The testing reported here was a little agricultural, and I can't really "certify" the conclusions. But for what it's worth:

It seems that, with the EOS 40D, with CF4-1=3 (AF and metering separated):

•With an AF point preselected, Evaluative metering (takes place for real at half press) is "linked" to that AF point.

•With automatic AF point selection, Evaluative metering (takes place for real at half press) is linked to the AF point selected as active during the most recent AF operation (caused with the AF ON button). (If I recall correctly from my 20D tests, that would even be if that last AF operation was last week.)

This is consistent with what I now think was my earlier finding with my EOS 20D.

Best regards,

Doug

Doug_Kerr said:

Hi, Cem,

Well, now that I have things set properly!

The testing reported here was a little agricultural, and I can't really "certify" the conclusions. But for what it's worth:

It seems that, with the EOS 40D, with CF4-1=3 (AF and metering separated):

•With an AF point preselected, Evaluative metering (takes place for real at half press) is "linked" to that AF point.

•With automatic AF point selection, Evaluative metering (takes place for real at half press) is linked to the AF point selected as active during the most recent AF operation (caused with the AF ON button). (If I recall correctly from my 20D tests, that would even be if that last AF operation was last week.)

This is consistent with what I now think was my earlier finding with my EOS 20D.

Best regards,

Doug

Click to expand...

Doug thanks a million for this testing, much appreciated.
I (used to) shoot with an AF point preselected (central point). I then use the Focus button to obtain focus using that point. Mostly, I then shoot with my subject off centre. In situations the centre point is considerably lighter or darker compared to my subject in the final composition, this method will then result in a wrong exposure. Talk about an eye opener! So I have to ponder this and how to overcome it in the future. Possibly, I'll have to revert to using the * key to lock the exposure as well.

Cheers,

Quite frankly, I find the outlook that Evaluative metering essentially starts with a "spot measurement" at the active AF point and then decides how to modify it based on an intelligent evaluation of surrounding conditions (to "correct" for conditions that could make the spot reading itself inappropriate to guide exposure) a bit hard to accept.

It seems as if a still simplistic but more meaningful outlook would be that it starts with some type of overall measurement "weighted" around the active AF point (with greatest weight given to the collocated metering segment, and less weight given to the immediately-surrounding ones), with further intelligent analysis thrown in (to "correct" for conditions that could make the weighted reading itself inappropriate to guide exposure).

Now maybe this is a meaningless artificial distinction.

Of course the bottom line is that if we really don't know more-or-less exactly what happens, then almost any murky explanation can be said to be equally "reasonable".

It's a little like arguing about, if there were saloons on Mars, what time could they open on Sunday.

Best regards,

Doug

Hi, Cem,

Cem_Usakligil said:

Possibly, I'll have to revert to using the * key to lock the exposure as well.

Click to expand...

That one way to do it if these three locations are (potentially) not the same:

a. AF target
b. Metering target
c. Shot aim point

It's like Reverse Polish Notation:

a. Aim at focus target, press AF
b. Aim at metering target, press *
c. Aim at shot aim point, full press shutter release

Here is another way to accomplish the same thing with CF4-1=0 (normal behavior) (assume center AF point preselected):

a. Aim at focus target, half press (and hold until step c) [This does AF, which of course is held, and starts metering, which "tracks" as we change aim.]
b. Aim at metering target, press * [This holds the metering as it is at that place]
c. Aim at shot target, full press.

This is my normal practice.

It is better for me than CF4-1=3 since if I just "grab, point, and shoot", I get AF. (Saves a lot of "oh shít"s!)

Best regards,

Doug

Doug_Kerr said:

It's like Reverse Polish Notation:

a. Aim at focus target, press AF
b. Aim at metering target, press *
c. Aim at shot aim point, full press shutter release

Click to expand...

Or as has been done since the invention of the camera:


a. Focus on desired element(s);
b. Meter for the scene and appropriately adjust camera;
c. Press shutter

So in turning it into rocket science, we have come a really long way...

:-},

Gentlemen.

I can see that you're enjoying this research micro-project and don't really mean to spoil your fun.

But...the subtleties of this don't really matter a whit in practical terms with today's digital imaging. Getting in the exposure ballpark is a very quick simple matter while shooting. Refining to taste is even simpler later.

So whether "evaluative" metering weights 70% to the AF points or 75% ... <shrug>.

Doug_Kerr said:

Hi, Cem,



That one way to do it if these three locations are (potentially) not the same:

a. AF target
b. Metering target
c. Shot aim point

It's like Reverse Polish Notation:

a. Aim at focus target, press AF
b. Aim at metering target, press *
c. Aim at shot aim point, full press shutter release

Here is another way to accomplish the same thing with CF4-1=0 (normal behavior) (assume center AF point preselected):

a. Aim at focus target, half press (and hold until step c) [This does AF, which of course is held, and starts metering, which "tracks" as we change aim.]
b. Aim at metering target, press * [This holds the metering as it is at that place]
c. Aim at shot target, full press.

This is my normal practice.

It is better for me than CF4-1=3 since if I just "grab, point, and shoot", I get AF. (Saves a lot of "oh shít"s!)

Best regards,

Doug

Click to expand...

Hi, Doug...

Now that you have the evaluative ambient metering " nailed down" have you thought about the implications, in focus/recompose, on flash metering, particularly evaluative. It whould seem that when/where the ambient exposure is locked will have impications to the accuracy of the flash exposure, since a lot is based on comparing the pre-flash reflections to the ambient exposure....zone-by-zone....and if the ambient zones are no longer centered over where their values were set, what happens...??

John

Hi, John,

John_Schwaller said:

Hi, Doug...

Now that you have the evaluative ambient metering " nailed down" have you thought about the implications, in focus/recompose, on flash metering, particularly evaluative. . . .

Click to expand...

Indeed. E-TTL II flash metering remains to me an even greater mystery than Evaluative ambient metering, notwithstanding the various nice explanations we've gotten from Chuck Westfall over the years.

Of course one big problem here in communication is the fact that there is so much "trade secret" involved.

As a result, all the explanations have been "sanitized" (or "concocted"). Sometimes the pieces don't seem to fit together, and that sets me off on a trail of investigation.

I'm not real optimistic that I have enough energy left to try and "reverse engineer" E-TTL II.

Best regards,

Doug