How does EOS evaluative metering work?

[tomfrh]

Hi,

I have a number of EOS bodies (3, 30, 30v, 500n, 70D) and I've found they all meter slightly different in evaluative mode. They're normally within 1/2 a stop, but I've seen as much as 1 1/2 stop difference depending on the scene/lighting. They all agree very well when looking at a white wall, so it would appear to the evaluative pattern/logic that is different.

Does anyone have any technical information about how the evaluative metering works in the different canon bodies?

 

[wiltw]

Canon has always been very secretive about how its evaluative works. Unlike Nikon, they do not use brightness patterns based upon real photos that are in the database. So Canon's evaluative is not fooled if you hold a camera upside down and aim at a beach scene. And while evaluative gives priority to the auto focus zone metering area, it always factors in all of surrounding zones as well,

 

[Sim2]

Hallo,
I think that as the evaluative metering is partly based around the selected focus point if there are more focus points there may be more 'segments' of metering for the camera to take into account. An early EOS could have, for example, nine focus points and eleven or so metering areas but a later EOS with 40 or so focus points may have 50 or more segments of metering areas: the early EOS could average severe light differences that appear across one metering zone whilst the later EOS with more, smaller metering areas may 'see' only part of this light difference giving it a 'correct' exposure and this could give the exposure shift more or less significance than the early EOS. Tricky to explain I think but does that help at all?
http://www.techradar.com/how-to/pho...ectly-exposed-images-in-any-situation-1320895
This page gives a visual example of focus points and size of reference...

 

[tomfrh]

Thanks for the comments.

I'm learning!

I didn't know that evaluative gave priority to the selected AF points.

My simplest EOS is the 500n which I've learned has 6 segment evaluative.

My EOS 3 has 21 segments. Same as the EOS 1v

Yet funnily enough 30/30v (and even the lowly rebel 300 series) has most of all - 35 segments!

Any idea why the best cameras have less metering segments than cheaper model?

 

[wiltw]

Hallo,
I think that as the evaluative metering is partly based around the selected focus point if there are more focus points there may be more 'segments' of metering for the camera to take into account. An early EOS could have, for example, nine focus points and eleven or so metering areas but a later EOS with 40 or so focus points may have 50 or more segments of metering areas: the early EOS could average severe light differences that appear across one metering zone whilst the later EOS with more, smaller metering areas may 'see' only part of this light difference giving it a 'correct' exposure and this could give the exposure shift more or less significance than the early EOS. Tricky to explain I think but does that help at all?
http://www.techradar.com/how-to/pho...ectly-exposed-images-in-any-situation-1320895
This page gives a visual example of focus points and size of reference...

ALL the zones are used, the AF zone(s) are simply given priority. That is why Evaluative can be badly fooled, as in this shot of an 18% grey card against the background sky.

A spotmeter would expose it like this, so you see the true tonality of the 18% card

Having more zones would not make them any more accurate than fewer zones. Increasing the NUMBER of zones is pretty related to how recently that model was launched...more recent, more zones.

 

[tomfrh]

That is why Evaluative can be badly fooled, as in this shot of an 18% grey card against the background sky.

That's an interesting example. Ultimately the camera has no idea what it's looking at and what's important. Maybe Nikon would have done better with their 3,000,000,000 photo lookup tables. "BEEP BEEP BEEP, grey card sitting on white paling fence against blue sky detected BEEP BEEP BEEP"

Having more zones would not make them any more accurate than fewer zones.

It would to a point. A 2 zone evaluative meter will be more accurate than a 1 zone. A 3 more accurate than a 2, etc. It would be interesting to know where it plateaus and additional zones bring negligible benefit. I know modern matrix/evaluative meters have thousands of zones!

My pentax 645n has 6 evaluative zones and it meters great. I rarely misses when shooting velvia.

 

[wiltw]

It would to a point. A 2 zone evaluative meter will be more accurate than a 1 zone. A 3 more accurate than a 2, etc. It would be interesting to know where it plateaus and additional zones bring negligible benefit. I know modern matrix/evaluative meters have thousands of zones!

That example was shot with a Canon 40D (2004) and I dare say that its 9 AF points/metering zones would not be appreciably more correct in that shot using the 45 AF points found in the new 80D.

 

[Sim2]

@wiltw - that is a great example you posted, before seeing that I would have reckoned a camera of that era would have done better than that, not discounting what you have shown, I wonder how a newer model e.g. 80D would render exactly the same scene at the same time. However, it does go to show that film era EOSs (what is the plural of EOS - EOI? ) being earlier models, would be easily fooled by the same scene; I dread to think how my EOS1n's would render that scene! Good examples of the issues, thanks.

 

[Poisson Du Jour]

Refer to the mir.com.my EOS website for more technical information on the EOS metering systems, particularly the 1 / 1N configuration that was well-proven as a stepping stone before being updated and expanded for the EOS 1N's replacement, the 1V.

http://www.mir.com.my/rb/photography/hardwares/classics/eos/EOS-1n/metering/index.htm

{ Particularly a) Evaluative metering Sensor zones Divisions and their output characteristics }

The reason the grey card experiment did not come up correctly is due to the size of the key object compared to the size of the bright area behind it and the distance from the subject to the camera. The evaluative system has therefore exposed the scene correctly based on the ratio of brightness of the background taking priority to the insignificant size of the grey card. The alternative is to use partial or spot metering when a small subject against a bright background is required to take precedence.

___________________________________________________________
• I contributed to the editorial content of the mir.com.my reference
with Philip Chong from 2004 until 2008 (EOS 1, 1N, 1N RS)
http://www.mir.com.my/rb/photography/hardwares/classics/eos/EOS-1n/metering/index.htm

 

[wiltw]

Refer to the mir.com.my EOS website for more technical information on the EOS metering systems, particularly the 1 / 1N configuration that was well-proven as a stepping stone before being updated and expanded for the EOS 1N's replacement, the 1V....
The reason the grey card experiment did not come up correctly is due to the size of the key object compared to the size of the bright area behind it and the distance from the subject to the camera. The evaluative system has therefore exposed the scene correctly based on the ratio of brightness of the background taking priority to the insignificant size of the grey card.

I enlarged the screen when looking at the article and calculated that the area encompassed by the color zones in the lateral area metered is about 15-20% of total frame area. In my test shot, the card filled 6.5% of the total frame area. When it is daylight tomorrow, I will run the test again, but try to fill the frame about 15-20% and publish the result in this thread.

 

[Poisson Du Jour]

It is daylight here — blinding, blue temperate southern light. What is the delay...?

 

[wiltw]

It is daylight here — blinding, blue temperate southern light. What is the delay...?

Your skies may be bright, but mine are still dark...sorry but it will have to wait.

 

[Poisson Du Jour]

Your skies may be bright, but mine are still dark...sorry but it will have to wait.

Sorry... got my wires crossed; I see you're in 'cisco. I somehow thought I was addressing a respondent in Sydney!

 

[wiltw]

Sorry, have not been in Sydney in about 10 years! Skies just now getting brighter, maybe do the test in about 2 hours.

 

[wiltw]

OK, I have overcast skies, unlike my earlier examples, but I shot anyway. In this example, I have quanitified the 18% gray card as filling about 17-20% of the frame area, more in line with the estimates of EOS 1N metering pattern in Evaluative.

Much closer, but Evaluative still underexposes a bit (shot 2) because it is fooled by the sky. but much better that the original test.

 

[tomfrh]

These are pretty mean tests though, aren't they! How is the camera supposed to know you want the dark thing rendered as middle grey?

 

[wiltw]

These are pretty mean tests though, aren't they! How is the camera supposed to know you want the dark thing rendered as middle grey?

A meter wants EVERYTHING that it sees to be rendered the midtone!!! It is even trying to make the surrounding sky to be midtone, which is why it is underexposing the main target at the AF zone. So the meter is simply TRYING -- even without you telling it -- to make the AF zone midtone, but it is also averaging in the surrounding zone but secondarily in priority to the AF zone (primary priority). I wasn't 'being tough' in this test, the meter simply did what it ALWAYS DOES.

In the example below, I use a one-degree spotmeter to read the white square (lowest row, all the way to the left) and it suggested a reading to make THAT target 'midtone'...the left shot shows it did exactly what it normally tries to achieve! The right shot was spotmetered on the 4th (midtone) square, and it suggested a reading to make that square to be 'midtone', its inherent brightness.

METERS TRY to render ALL TARGETS that it is pointed at to be 'mid tone'.

 

[tomfrh]

A meter wants EVERYTHING that it sees to be rendered the midtone!!! It is even trying to make the surrounding sky to be midtone, which is why it is underexposing the main target at the AF zone. So the meter is simply TRYING -- even without you telling it -- to make the AF zone midtone, but it is also averaging in the surrounding zone but secondarily in priority to the AF zone (primary priority). I wasn't 'being tough' in this test, the meter simply did what it ALWAYS DOES.

Sure. I agree with all that. That's a what a meter does.

What I was referring to as "being tough" was you guys saying the meter was "badly fooled" when it failed to spot meter the grey card.

 

[wiltw]

Sure. I agree with all that. That's a what a meter does.

What I was referring to as "being tough" was you guys saying the meter was "badly fooled" when it failed to spot meter the grey card.

We each have our own 'standards' that we judge things by. There is nothing wrong with considering anything other than 'perfectly exposed' to be flawed. We need to recognize the limitations of technology, and compensate them (which is why we should use the concept of 'exposure compensation'...which all too few photographers even understand the concept!) I still say the shot in post 5 was 'badly fooled' and would result in a near unusable backlit portrait silhouette, and shooting with color neg, the post 5 Evaluative shot would have poorly saturated muddy color! Whereas the shot in post 17 was not so badly fooled and would be usable. So, to me, post 5 is indeed 'badly fooled' and would result in a shot I would not present to a client.

an example of a poorly (under)exposed shot on color neg, and its effects on contrast and poor color saturation and muddy color in the darker areas:

 

[tomfrh]

I still say the shot in post 5 was 'badly fooled'

I understand what you mean by it "being fooled", however I in my opinion the meter did its job and got the exposure right. If the camera gave me the second shot, with half of the image (including the centre) blown out, I'd consider it mistaken. It's just a meter, not a mind reader!