Peter van den Hamer, a Dutch physicist, has written a very interesting four-section (see the navigation buttons at the top) article looking at DxOMark’s sensor tests. This independent analysis indicates that the DxO guys are doing a pretty good job. My big Canon system, sadly, ends up in the “losers” category along with medium-format sensors. Nikon, Sony, and other users of the Sony sensors are crushing everyone else. van den Hamer notes that Olympus is beginning to close the gap. What he didn’t say is that they did this by buying sensors from… Sony (source).
[Separately, DxOMark has tested a new Nikon D5200 APS-C (smallish) sensor camera and the result was a 14 f-stop dynamic range (Canon SLRs manage between 11 and 12). What’s unusual about this test is that the sensor is made by Toshiba. So now there are at least two sensor manufacturers that can easily outperform anything that Canon makes in-house. As DxO says in their conclusion: “The new sensors comfortably out-perform the current Canon offerings in practically every metric.”]
That Canon doesn’t excel in sensor technology is not surprising, considering their DNA as an optics company. Nikon at least has enough (short-term) smarts to outsource their sensors from someone with a clue, but alas, is not smart enough to do the same for their software (e.g. camera firmware), which at times can be gratuitously abysmal.
Canon’s approach may pay off in the long term, though, if they can eventually develop something no one else has, they’ll have something uniquely compelling.
I was a classmate of one of the DxO guys, they are all solid engineers, physicists and mathematicians, not amateurs as on far too many photo review sites.
Until a couple of years ago, roughly up to the 5DmkII, Canon sensors were better (the D3 had better low-light performance, at the cost of far fewer pixels). They rested on their laurels and opted not to update their fab and are still building their sensors on 0.5 micron technology, when Sony is using 0.18µ technology. This means the transistors used to read photon data take more surface area, space that cannot be used for actually capturing pixels and thus improving SNR.
For more details, see:
http://www.chipworks.com/blog/technologyblog/2012/10/24/full-frame-dslr-cameras-canon-stays-the-course/
Either Canon makes the investment in upgrading its fab, or they will also have to buy their sensors from someone else or end up completely uncompetitive. One short-term trick they could try is introduce backside illumination (BSI) to their DSLR sensors, a technology more often used in cameraphone and compact digicam sensors, where the readout electronics do not share the same surface as the pixels.
Interestingly, Toshiba has entered the space, and the Nikon D5200 uses one (the D3200 uses Sony). Toshiba uses copper rather than aluminium in its sensors, and that could give it an edge over Sony:
http://www.chipworks.com/blog/recentteardowns/2013/01/08/inside-the-nikon-d5200-dslr-toshiba-found/
Even more interestingly, the new Leica M will be built with a Belgian-designed (CMOSIS) and French-made (ST Microelectronics) sensor on a 110nm/90nm copper process, which should be better on paper than Sony’s 180nm aluminium process.
http://www.cmosis.com/news/press_releases/new_leica_m_uses_cmosis_24_mp_cmos_image_sensor
How do Nikon’s optics compare to Canon’s? I’ve never used Nikon SLR equipment, but have long been under the impression that their lenses were superior.
Trevis: Canon and Nikon lenses are pretty comparable, though I think that Canon may have an edge in the super telephoto lenses (look at what working sports photographers are using, for example, and you’ll see that it is mostly Canon (this could be due to Canon’s historical lead in AF technology, though)). I think I wrote earlier about how a lot of fairly inexpensive Nikon and Canon lenses have beaten Zeiss and Leica lenses in objective tests, such as those at DxOMark. Check out http://www.dxomark.com/index.php/Lenses/Camera-Lens-Ratings and then restrict by Brand to Canon and Nikon and you’ll see that they are interspersed in a rank by DxOMark Score.