Update May 1, 2012: I’ve published a preliminary report on the extent of the issue I’ve seen on my camera. Though not as scientific as I wanted it to be, it’s good enough to draw some conclusions about how and what situations could cause metering issues.

Update April 23, 2012: Canon has updated the service notice regarding the light leak situation. Canon will be offering a free inspection and adjustment on a voluntary basis if you feel your camera is operating improperly. I have to say this is about what I expected them to do. In the mean time I’m in the process of trying to put together a write up based on my findings under various conditions. Unfortunately due to limited time and equipment, I was unable to investigate to quite the level of rigor I was hoping to be able to, but even with the wider error margins from what I could do, the picture is vary clearly one of a non-issue at best, and a potential issue for a very small minority of users at worst.

Update April 13, 2012: I don’t know if I should be surprised or not, but Canon has officially acknowledged the top LCD backlight affecting the meter in certain dark environments and that they are working on a fix for it. The official service notice has been posted by Canon USA, and I’m sure if they haven’t already the rest of Canon’s global sites will have similar notices posted soon.

Update April 9, 2012: I have spent some time playing with the range where the back light causes a metering error, and so far as my limited testing shows as long as you’re operating in the EV 1 to EV 20 metering range, the meter behaves correctly. I used a 45W lamp, on an X10 lamp module, and shot at f/4 ISO 3200. The test goes something like this.

1. Set the camera up pointed at a constant scene lit that’s lit by your dim-able lamp.
2. Pick an aperture. Any aperture will show the change in metering, however the max aperture is what you need to look at to determine what light level exists in front of the camera.
3. Set the camera to ISO 1600 or ISO 3200 (I used 3200, but in 1600 would probably work as well).
4. Dim the light or adjust the angle of the camera until you meter EV 3. I used 1/15th f/4 ISO 3200.
5. Read the meter, and switch between backlight on and off. Note if the meter changed.
6. Dim the light until the metered shutter speed drops 1-stop. Repeat step 5.

For my test, I’ve summarized the the settings and whether the meter was affected by the backlight in the table below.

Even though the meter seems to work correctly inside the designed/specified EV 1-20 range, I still think that Canon should address this.

Update April 8, 2012: I’ve been keeping up with the discussion on Canon rumors forum and my own testing and the thing is there doesn’t seem to be any real impact in metering when actually shooting. Moreover, there are a number of people reporting that they can reproduce the behavior on a huge verity of cameras including Nikon bodies. Given that, and the fact that I see virtually no difference in metering between my 5D Mark 3 and my other bodies in actual use and under conditions that are within the design parameters of the light meter (i.e EV 1-20 with a 50mm f/1.4 lens as ISO 100) I’m even more comfortable in my conclusion that this is even more of a non issue that I had initially believed.

Originally Posted April 7, 2012: Canon Rumors is reporting that there’s a potential light leak in the 5D mark 3 LCD into the viewfinder. Subsequently, at least for some it seams, the world has ended, and we should all send our cameras back to Canon, switch to Nikon, flail our hands over our heads and generally freak out.

That said, there doesn’t seem to be much of a problem in actual practical usage of the camera.

If you have a 5D mark 3, or really any other camera, you can test it for this light leak by following the procedure outlined in the Canon Rumor’s article.

I sat down and ran the test, and yes, turn the backlight on and my 5D mark 3’s meter changes when the backlight is on and both the viewfinder cover and body cap are on. Out of curiosity, I figured I’d test this out on my other bodies as well. My 40D doesn’t exhibit this behavior, even when shining a light on the LCD. However, the meter on my 1D mark 3 does change, though it requires shining a light on the LCD not the backlight.

But the real concern for me, is if the “light leak” is leaking to the sensor during an exposure or just the meter, since this would be a real problem. Fortunately this doesn’t appear to be the case at all.

Both images are uniformly black and have no noticeable localized lightening. Moreover, at ISO 25,600 even a small amount of light, never mind a 3 D-cell Maglight at point blank range if there was a light leak that effected the sensor there would be a something in those images.

So all this has the potential to do is affect the metering in practice in the real world?

Maybe.

Under normal use, I’m having a hard time finding a situation where there’s an change in metering that 1) I can attribute directly to the top LCD leaking and not a change in the composition, subject, eye-placement at the viewfinder, or lighting, and 2) where said error in metering is appreciable enough to affect the captured image in any appreciable way.

So far as I can tell, so long as the light levels at the camera and subject are in the same neighborhood, there’s really no impact what so ever. The only case where I can make an appreciable change by shining light on the top LCD is when it’s basically the only light the camera’s meter is seeing.

In any case, I’m certainly not all that concerned. As I already mentioned, if I shine a light on my 1D mark 3s top LCD under the same kinds of ambient conditions and I can get the meter to change as well. However, in more than two years of using that camera, I’ve never had a single exposure related problem.

So what should you do?

Well that’s up to you, but I’m going to continue to use my camera as I normally would. Moreover, whenever Canon gets around to addressing this—and I’m assuming they will—I don’t expect to be real quick in sending my camera back unless it’s also convenient for me to do so. While this certainly shouldn’t happen and it would be good to have fixed, it also doesn’t seem to be impacting my images either and that matters more than any potential problem that doesn’t quite manifest itself in real world usage.