DPReview has posted their in-depth review of the new Canon 50D, a 15 megapixel update to the 40D. The 50D also adds a high resolution (640×480) 3 inch LCD screen. Also, Canon has promised that sensor technology improvements — redesigned photo diodes and microlenses — mean lower noise despite the small photosites due to the increased resolution. And, to back up their claims, they’ve bumped the maximum ISO all the way to 12,800, even if that’s in the “extended” ISO range. The 50D is also the first appearance of Canon’s new DIGIC 4 processor, and also adds contrast-detection auto focus to live view. You also get a new sensor cleaning system, automatic lens vignetting correction, HDMI output, better weather-sealing, and a spiffy new user interface.

So, how does it do? Well, for starters, that 12,800 ISO setting isn’t anything worth getting excited about, according to the review:

The 50D produces comparatively clean output that demonstrates a well balanced mixture of chroma and luminance noise reduction. However at ISO 1600 and above it can’t quite keep up with the D300 in terms of detail retention. The Canon’s more heavy-handed luminance noise reduction inevitably leads to some blurring of fine detail. Output up to ISO 3200 is usable, ISO 6400 should be reserved for emergencies and the inclusion of the ISO 12800 setting was quite frankly pointless.

At least it’s not as noisy as the Pentax K20D. Yikes.

And about Canon’s claim that their improved sensor technology (something about gapless microlenses…) means less noise even with the increased megapixel count? Well, it might be less noise than if they hadn’t improved the sensor, but it’s still significantly noisier than the Canon 40D, at least on a pixel-by-pixel basis:

Despite the fact that the 50D is the newer camera it shows visibly more chroma and luminance noise than the 40D. Considering the 50D’s much more tightly packed sensor (4.5 MP/cm² vs 3.1 MP/cm² on the 40D) this comes hardly as a surprise. It would have been unreasonable to expect Canon’s engineers to overcome the laws of physics.

Of course, the natural question — which I don’t think I’ve ever seen a camera review answer — is whether the newer model camera in a situation like this gives you more or less noise at the same resolution. Obviously packing more pixels into the same sized sensor means a noisier image, in general. But that’s comparing it pixel-by-pixel. Since you’re also getting more pixels, then how does the noise compare if you shrink the higher resolution image down to the resolution of the previous model? In other words, is it worth getting the older model to produce less noisy prints, or will the same-sized prints look the same (or even better) with the newer camera, even if it does have more noise when pixel peeping?

One interesting noise-related bit is that setting noise reduction to “off” for JPEGs still lets the 50D do a fair amount of noise reduction, at least at the ludicrous ISO settings of 6400 and 12,800. Looking at DPReview’s samples of unprocessed RAW shots at those settings is almost physically painful.

DPReview also shows that the shadow dynamic range is slightly reduced at higher ISOs on the 50D when compared to the 40D, again due to the increased noise on the newer sensor. The 50D also has less overall dynamic range than most of its closest competitors, although it’s at least fairly well-designed:

The EOS 50D produced slightly less highlight range than the D300 and A700. In terms of shadow range it is on par with most of the competition (but lags behind the K20D and the EOS 40D). Despite the comparatively small measured highlight range the 50D manages a fairly soft ‘roll off’ at the highlight end which results in less hard-clipped looking white highlights.

In other bad news, automatic white balance is still poor under incandescent and fluorescent light. Not a big surprise, as Canon has never seemed to get these right (especially incandescent) but not a huge deal, since I’m assuming most photographers will shoot RAW and fine-tune the white balance in the RAW conversion anyways.

Interestingly, DPReview is uncharacteristically harsh on the high megapixel count in the 50D. Some choice comments:

In terms of detail the 50D is not quite the step up from the 40D that we would have expected. After all the new model’s nominal resolution has increased by approximately 22% in both dimensions. There is only a very small amount of extra detail in the 50D output though (in fact even at 100%, if you scale the 40D’s output up to match the 50D the results are almost indistinguishable). While the new sensor makes the 50D the highest megapixel APS-C DSLR currently on the market it also makes it the one with the highest pixel density and it appears that Canon has reached the limit of what is sensible, in terms of megapixels, on an APS-C sensor (using current technology). At a pixel density of 4.5 MP/cm² (40D: 3.1 MP/cm², 1Ds MkIII: 2.4 MP/cm²) the lens becomes the limiting factor. Even the sharpest primes at optimal apertures cannot (at least on the edges of the frame) satisfy the 15.1 megapixel sensor’s hunger for resolution. The result is images that look comparatively soft at a pixel level and only show marginally more detail than images from a good ten or twelve megapixel DSLR. If all you end up with is a larger image (and file) one starts to wonder what the whole point of pushing the resolution up to these dizzying heights is.

Considering the disadvantages that come with higher pixel densities such as diffraction issues, increased sensitivity towards camera shake, reduced high ISO performance and the need to store, move and process larger amounts of data, one could be forgiven for coming to the conclusion that at this point the megapixel race should probably stop.

Just to make it clear, the 50D’s image quality is (at identical viewing size) and by no means worse than the competition’s but it’s also not significantly better than the 40D’s (Dynamic range and high ISO performance are even slightly worse) and that simply makes one wonder if the EOS 50D would have been an (even) better camera if its sensor had a slightly more moderate resolution.

And, in fact, moving on to the resolution chart comparison, they find that while the 50D does resolve more detail than the 12 megapixel D300 and A700 (as it should), the improvement is much smaller than you would expect.

In the end, DPReview does give the Canon 50D their “highly recommended” tag, but it truthfully sounds like there are very few people for whom this would be the right camera to get:

The EOS 50D has to stand its ground in a highly competitive bracket of the DSLR market. It is currently almost $500 more expensive than the 40D, almost $500 more expensive than the Nikon D90 and for an extra $100 you can bag yourself a Nikon D300. Looking at the specification differences between the EOS 40D and our test candidate it appears you pay quite a premium for the 50D’s extra megapixels and as we’ve found out during this review you don’t get an awful lot of extra image quality for your money. The Canon EOS 50D still earns itself our highest reward but considering its price point and our slight concerns about its pixel-packed sensor, it only does so by a whisker.