Musings on Photography

Meters in Digital Cameras

Posted in equipment by Paul Butzi on April 30, 2007

5d-070428-3921-600.jpg

One of the things that annoys me about the current crop of digital SLR’s (or at least the Canon ones) is the sorry performance of the in-camera metering system.  It’s not just that it’s not very good, it’s that it’s utterly pathetic.

Here’s the way it works in practice.  I set up the camera, compose, etc.  I then set the metering system to zero exposure adjustment, then make an exposure – I pick the aperture, and the camera picks the shutterspeed.  Then, I look at the histogram on the rear display, to see what stupid thing the meter has decided.  Typically, it’s decided that the sky is unimportant and should be rendered as featureless white.  Then I make a guess at how far I need to adjust the exposure up/down, to get it ‘exposed to the right’ without clipping the highlights.  Generally this takes one try, but sometimes it takes two.

As far as I can tell, Canon have (for reasons known only to some engineering group at Canon) decided that the vast majority of users of their $3000 camera body are yutzes who use the camera only in jpg mode, and never shoot raw.  As a result, the camera just makes a wild-ass guess about what exposure will get the lower mid tones in the middle of the range, and then doesn’t worry about clipping.

It doesn’t seem like it should be hard to build a $3000 camera with a metering system which can reliably make exposures I consider correct.  I would understand if it was hard to build a meter which detected the clipping of areas of the size of just one or two pixels – say, sky peeking through the branches of a leafed-out tree.  But when the silly thing gets it wrong when the sky fills 1/3rd of the frame, you have to wonder.

Film cameras with built in meters in the ‘dark ages of photography’ had very sophisticated metering systems.  The metering system in the EOS-5d is inherited from that era – it offers evaluative metering, partial metering, spot metering, and center-weighted metering.  All of these are attempts to figure out the correct exposure for film based on sampling just a few small areas in the scene.  But digital is not film, and I’m holding in my hands a device that is, in point of fact, capable of sampling the brightness and color of some 12.7 MILLION spots in the scene simultaneously.  Why, then, does the camera set the exposure using a metering system and algorithms which were developed decades ago for a different recording medium with quite different properties?

17 Responses

Subscribe to comments with RSS.

  1. Doug said, on April 30, 2007 at 10:29 am

    The fundamental issue is that the “12.7 million spots” are hidden behind the mirrors and shutter during metering. Instead, a separate sensor is used for exposure metering.

    Canon continues to use a relatively simple 35-pixel monochrome metering sensor in all of its DSLRs except the new 1DmkIII. In contrast, Nikon uses a 1005-pixel full-color metering sensor in most of its DSLR line, and even the low-end D40 has a 420-pixel full-color metering sensor.

    Then there is what you already mentioned: that the exposure calculations that are based on that limited metering data are oriented toward JPEG photography. As are the in-camera histograms. Canon really doesn’t seem to provide the tools that Raw shooters could really use… but then, I don’t know of any DSLR manufacturer that does.

    So yeah, there is lots of room for improvement. Alas, the 1DmkIII’s new and improved 63-pixel monochrome metering sensor and its “highlight priority” shooting mode suggest to me that Canon users shouldn’t hold their breath.

  2. Martin Doonan said, on April 30, 2007 at 10:32 am

    I’m glad you posted this, I was beginning to think it was me. Same deal with the 20D, and I seuspect all the rest.
    I find the inability to deal with mixed light incredible – I’m more or less resorting to bracketing everything in the hope of getting a decent shot it. The number of times I’ve ruined otherwise decent shots due to the inability to factor in the sky doesn’t bear counting. I also find that in these circumstances the histogram is useless, dropping highlights off the chart altogether so they merrily blast off to 255,255,255. Not conducive to “expose to the right2 at all.

  3. StephaneB said, on April 30, 2007 at 10:49 am

    I agree with the observation, but I must say I could not care less. I still use my Pentax Digital Spotmeter. I find the brightest part, expose it a Zone VII and that’s it. If the contrast range is larger than 8 stops I decide if it is OK to lose the shadows or not. If not I consider making two exposures. That method never fails.

  4. Chuck Kimmerle said, on April 30, 2007 at 11:00 am

    I can honestly say that I’ve good luck with my Nikon camera meters, but shoot exclusively in manual mode, using the meter reading only as an initial starting point. From that first exposure I look at the histgram, adjust exposure accordingly, and shoot away.

    To be honest, I’m not sure why anyone would shoot in auto mode as it takes THE critical technical decision – exposure – and turns it over to a mindless box.

  5. matt said, on April 30, 2007 at 2:20 pm

    My D80 handles sky and other large bright areas the same way, so it isn’t just canon. In some ways, the center weighted meter in an old film camera would do a better job with this; it at least would try to make that big bright area medium gray. Of course, the subject would end up underexposed, but that’s much easier to fix these days. Unfortunately, the center weighted meter option on my D80 doesn’t seem to mimic the behavior of a film camera’s center weighted meter.

  6. Mark Hespenheide said, on April 30, 2007 at 5:42 pm

    Paul,
    You’re not wrong. I shoot with a 5d and find myself perplexed why the metering seems stuck in the 1980’s or 1990’s. I find myself wishing there were a “RAW/tripod” mode, where the camera would [1]make an estimate of the proper exposure, [2]lock the mirror, [3]make an exposure at a hight ISO, [4] correct iteratively if necessary to get the proper “expose to the right” exposure, and then [5] dial back the ISO to my preset and make the “real” capture.
    If I can think of it, why can’t Canon?
    If I’m working quickly, I usually just dial in -2/3 stop compensation, or tilt the tripod up on two legs such that the framing is 1/2 sky. Crude, but it seems to work.

    To continue the oddities of RAW cature on higg-end cameras, if I have the camera set for automatic exposure bracket (so that I can make a HDR capture) why do I have to press the shutter release multiple times? Why won’t it just chew through the exposures like I’ve told it to? I had to go out and buy a remote release to minimize the chance of disturbing the camera…

  7. Anita Jesse said, on April 30, 2007 at 5:48 pm

    Okay, I am plunging in and risking making a complete fool of myself in the company of the pros, but I thought I would take a deep breath and venture a post. Disclaimer: I know I am a long, long way from being an expert, and the method I use is not my original idea—I read a lot. Fairly recently, I learned that I can get much better exposures with my 20D (provided I am shooting raw) by acknowledging that my camera will allow me only one full stop over and one full stop under the ideal exposure—and I’m better off not denying it.
    I shoot almost exclusively in manual. I meter for the shadows at one full stop under, and then check the sky. If the sky shows more than one full stop above ideal, I know I must choose among: bracketing, some blown highlights, some stopped-up shadows, or recompose to eliminate either the deepest shadows or the sky. Occasionally, I have tried pushing it to expose the shadows to one-third more than that full and found I could live with the result.
    Be kind, guys.

  8. Paul Maxim said, on May 1, 2007 at 3:32 am

    I certainly agree that digital meters could be improved (I shoot with a 5D). But I would also add one thing about the sky, and that is that it depends on what the sky “looks like”. If the sky is blue (or mostly blue), the built-in meter will work just fine. Little additional correction will be necessary. If, however, the sky is that dreadful whitish gray color, then any meter will be “stupid” in evaluative mode when the sky fills a significant portion of the frame. It will always produce blown highlights (the sky). I also believe that the old film cameras (Canon, anyway) had the same problem. They weren’t terribly bright in that regard, either.

  9. Martin Doonan said, on May 1, 2007 at 3:56 am

    Paul M – personally I find the exact opposite. Largely due to contrast or total range – overcast skies I can get nicely butted up to the right, bright blue skies get consistently over-exposed, even if the histogram apparantly just touches the right. Of course, there are the colour differences to take into account but i genrally find that the overexposure is across all channels, not just blue. I’m not even greatly worried about blocking the shadows – there’s usually enough room down there, if only it wouldn’t wash out the highlights. i’m getting fed up of dialling in a conservative -1 (losing have the available info) to get full capture.

    I do effectively work in manual. Actually I use Av & dial in correction, but that’s because I prefer using the thumb-wheel for shutter speed. It amounts to a zone-a-like system but the damn meter isn’t measuring the higlights properly.

    Only time I had problems with film was in very warm colours (deep sunsets, mainly) which needed some compensation.

    HDR 7 other multiple capture techniques don’t always work due to movemetns in the scene – the slightest breeze through the tress and you’re stuffed.

  10. chuck kimmerle said, on May 1, 2007 at 6:31 am

    HDR is a nice option, but I find it more useful, in a high contrast scene, to do a bracketed exposure 1-stop in either direction and to use layers and masks to utilize the best parts of each exposure.

    Again, though, I do not recommend trusting the meter….ever. It’s a great starting point, but that’s all it should be. The histogram to a photographer is like the “force” to the Jedi….you must believe to make it work.

  11. /thm/ said, on May 1, 2007 at 6:47 am

    Don’t DSLRs tend to expose for shadow detail? The Nikon D200 does, and I’ve thought that proper.

    Hmmm.

  12. paul said, on May 1, 2007 at 10:26 am

    I’d have to say that, for the most part, my Nikon D2X’s meter is pretty accurate and I rarely will use manual mode. In tough situations, like having a lot of contrast between sky and ground, I use the exposure compensation button if I don’t like what I see on the histogram, but for 90% of my shots, I let the camera decide and it’s usually correct. Why would I bother going to manual?

  13. Mark Hespenheide said, on May 1, 2007 at 11:13 am

    Anita,

    Your approach is fine. We’re (the Canon users, at least) are whining that, while there have been amazing developments in building chips that turn photons into data, those chips could also be used as extremely high-level light meters. We feel like Canon’s approach doesn’t make the most of their own existing technology. This is not to say that they can’t make amazing pictures, given the proper nuturing… we just want even more than we’ve already got. Even with a relatively static subject like landscape photography, the light can sometimes change very rapidly.

  14. Andrew said, on May 1, 2007 at 1:09 pm

    Given the range in the metering algorithms used–evaluative, partial, and center weighted on the Rebel XT, I would expect a lot of variation in the metering results. Can some of you comment on the metering modes you typically use? I’ve just started really experimenting with mine after reading Understanding Exposure by Bryan Peterson. A great book by the way on where to take your meter reading from.

  15. […] of interesting comments on my post about the bad performance of metering systems in digital cameras that are used in camera raw mode.  This makes me think it would be worthwhile laying out what I think a decent metering system […]

  16. Mike said, on May 3, 2007 at 7:47 pm

    Well, I shoot all day 5-6 days a week using a 1Ds MkII. 90% of the time I’m on Manual exposure setting, especially when I’m indoors shooting in a flourescent room or stage photography. I prefer to use the center weighted traditional metering pattern, spot for stage light and I NEVER use the evaluative or partial. Evaluative tries to think for you and always makes the wrong decision. Even with RAW, you need to be within a half a stop in order to get the proper tonal values throughout the range. Fortunately I cut my teeth shooting Kodachrome, so this half a stop thing is easy. I do have to chime in that a good hand held incident light meter is way faster for me than shooting a few frames and checking the histogram. We can piss and moan that out cameras aren’t perfect, or we can take heart that there is still a place in this world for professionals who can think and use their experience to outwit technology and make compelling and interesting images. That’s what they pay us for. If the cameras were perfect, why ever hire a photographer? If anything, I wish photographey was harder to do than it is, then all of my accumulated experience and knowledge would be worth more.

  17. Wish List « Musings on Photography said, on November 28, 2007 at 10:49 am

    […] do it, either. In fact, I think they could do it and give it out free. I’ve written before (here and here) about the stupidity of metering systems in digital cameras. A lot of the technical […]


Comments are closed.

%d bloggers like this: