One of the persistent threads in camera reviews is the discussion of how a certain camera will meter a given scene. This leads to the characterization of various cameras being referred to as either tending to “overexpose” (sounds bad) or to being great at holding onto highlight detail in very bright settings (sounds good). This results in a lot of confusion and misinformation swirling around the internet and, I kid you not, a lot of heated arguments among fanboys of various brands. Metering in general can be a confusing subject as it invariably expands into a discussion of subjects such as dynamic range and JPEG compression vs. the merits of RAW files and on and on. From listening to a lot of the debate one could think that many camera designers don’t know what they are doing and are rather thick. The internet at large, in its collective ‘wisdom’ , isn’t the least bit shy about letting them know it.
In reality designing a camera metering system is very much like approaching any other technical design challenge. The engineers have a goal in mind and work towards an optimal technical solution within the constraints of the technology, packaging, costs etc. On top of that they may have a particular philosophy as to what the best solution is for any given problem. Think cars that are sporty vs. cars that are comfortable. Both characteristics are solutions to the same technical problem of passenger transport. They just have different ways of going about it. Different approaches will appeal to different people. Neither is right or wrong, just different and to personal taste.
Since I have various cameras in my collection I thought I’d do a simple test just to see how a few cameras from different manufactures approach metering the same scene. In this case a rather complex setting of a kitchen on a bright sunny morning. This presented a range of challenges. The window area was very bright while the floor area was in shadow plus the walls and appliances making up a large part of the scene were white which tends to throw off camera meters. To make the test fair I shot the scene back to back with each camera within seconds of each other so the light in the scene didn’t change.
Since designing a good experiment requires holding as many variables as constant as possible while only changing the thing you are most interested in testing I set up three different APS-C (crop sensor) cameras (a Canon Rebel XSi, A Fujifilm XE-1 and a Nikon D7100 ) as follows;
- All apertures were set at F8
- Metering was set to Matrix
- All were set to capture JPEG (fine) + RAW
- All were set to the standard picture mode for JPEG
- All were set to A (aperture priority)
- Exposure compensation was set at 0
To keep the focal lengths as close as possible I used Canon’s EF 28mm/F2.8 lens on the Rebel, a Zeiss Touit 32mm/F1.8 on the Fuji and a Nikon 35mm/F2 on the Nikon. Keep in mind that the thing I wanted to understand here was the way the cameras metered the scene. What approach did each manufacture take with each of these cameras. My goal wasn’t to characterize every camera on the market but to understand three different approaches. I also don’t claim that the results are definitive for each brand as of course the designers can always tweak these things from camera to camera. Though one could speculate that in general a given company might hold to a fairly consistent design philosophy over time.
So lets take a look at the RAW images from each camera with the settings as described above.
First up is the Nikon. This is a camera that in some internet circles has been accused of having a tendency to overexpose. Indeed what we see in this scene is the loss of some detail through the bright window area though there is still a bit of detail there. The overall scene is however very nicely exposed. The actual subject, the kitchen, is rendered quite nicely even at these default settings. It’s bright and the shadows show very good lighting. Clearly the Nikon comes down on the side of deciding that the very bright regions are unlikely to be important to the subject as they make up only a small portion of the scene so in this case expose for the darker regions.
Next up is the Fuji X camera. As you can see the Fuji indeed lives up to its reputation as a master of the highlights. Note how much more highlight detail is captured through the window than with the Nikon. However also note how much darker the rest of the scene is. I would say the overall scene is rather underexposed. Looking at the histogram confirms that there isn’t any special magic at work here. The designers have merely told the cameras metering program that if there are overly bright areas in the scene then underexpose to preserve the highlights. Looking at the histogram confirms this as the whole curve is pushed to the left with more total pixels in the ‘darker’ regions. The shadows can be pushed up in post but by default the image is underexposed.
Lets take a look at the Canon CR2 file. Interestingly it seems to come out between the Nikon NEF and Fuji RAF files. With the Canon there is more preserved highlight detail than with the Nikon but less than with the Fuji. Likewise the shadow regions are more brightly exposed than is the case with the Fuji but less so than with the Nikon. The Canons default exposure seems to be to try to find a balance between the extremes of bright and dark. That sounds like a sensible sort of ‘average’ approach.
So what about the in camera generated JPEG files. Well in the default picture style I set on all three cameras the JPEG files do differ a bit from the RAW files. In general they tend to be just a touch punchier, a little more contrast-y basically. In the case of the Fujifilm XE-1 the in camera JPEG processing ‘pushes’ the histogram both to the left and to the right. The result is a notable increase in visible detail through the window. The impact on in shadow areas in much less noticeable. There is nice highlight detail but again the overall image is a bit underexposed by default.
With the Nikon the in camera JPEG processing in this picture style setting is very conservative. The histogram is virtually unchanged between the RAW and JPEG files. The JPEG has more contrast but both images expose for the shadows as mentioned above leaving the outside-the-window detail overexposed. The overall exposure however is quite pleasing as the kitchen itself is nice and bright.
Looking at the JPEG generated by the Canon we see an approach very similar to that taken by the Fuji. There is a noticeable increase in contrast and highlight detail, through the window, is enhanced though not as much as with the Fuji. As in the RAW file comparison above the Canon takes a middle approach between the other two cameras. This still, however, results in an image that seems a bit dark and underexposed though not to the extent as seen with the Fuji.
Overall, for this particular kitchen scene I’d rank the Nikon the best. Here the Nikon is taking a rather sophisticated approach to interpreting the scene. Instead of just thinking ‘highlights, must preserve the highlights!!!’ like the Fuji or trying to ‘average’ the scene like the Canon, Nikon’s approach is to look at the overall scene, note that while there are some bright areas the overall frame is rather dark. The camera then makes a rather brave decision. It exposes for the majority of the scene and is willing to somewhat sacrifice the brightest areas. In this case it is absolutely the correct decision. The subject of this image is the kitchen, not the details through the window and the Nikon clearly does the best job of metering the actual subject.
Keep in mind that I’m talking about the default metering here. You are perfectly free to adjust exposure compensation, for example, with each camera to favour shadows or highlights. Also keep in might that even in these default images there is more data than there might seem to be. By adjusting the highlight slider in CaptureOne I was able to pull some through- the-window detail out of the Nikon RAW file. This however exposes one significant issue with shooting only JPEG files. In doing the JPEG conversion the Nikon basically says there isn’t any useful information in these extreme bright areas and simply throws the data away. So while doing the highlight adjustment with the RAW file recovers highlight detail this wasn’t really possible with the JPEG since most of that data was discarded in the RAW to JPEG conversion.