I'm a Nikon landscape/macro shooter, and am seriously thinking about adding a medium format system with a digital back these days. I'm wondering if the present stagnation in upper-level 35mm (whether full-frame or not) type DSLRs is just marketing, or if we've actually run up against the laws of physics here - nobody can figure out how to add resolution without losing tonality and dynamic range, and vice versa? Another possibility is that the physics-based limit is in lens design - nobody's figured out how to build a lens with decent coverage that resolves better than 6 microns? Lenses sharper than that exist that are able to cover a fingernail sized area (if they didn't, all digital compacts would have effective resolutions of 2-3 mp due to lens blur - some do, but not all!), but they may be nearly impossible to build with broader coverage? Almost all the major players, whether 35mm or medium format, have settled into a pixel pitch range of 6 to 9 microns. Without moving to 4 or 5 micron pixels, we can't add resolution, and smaller pixels mean more noise and less dynamic range. Could it be that nobody's figured out how to deal with this yet, therefore sensors are stuck where they are for right now? A 6 micron full-frame 35mm DSLR would have 21-22 million pixels, and that seems to be at the edge of what present sensors can do - is a 21 MP EOS 1Ds mk III that different from the Mk II? Nikon already has their DX sensors crammed even closer to the 6 micron mark - the D200 is right at 6 microns, and the D2x and D2xs are actually slightly under 6 microns, the smallest pixels on an upper-end DSLR today.
Without cramming sensors tighter than 6 microns, we have image size restrictions of:
APS-C 10 mp
35 Full Frame 22 mp
"Double 35" - cropped 645 sensors - under 46 mp depending on exact dimensions (some are slightly above 36x48mm)
Will it take technologies like Foveon-type sensors or Super CCD to add image quality (whether or not this is accomplished by adding resolution) above this? We've had sensors in the 6-7 micron range since at least August 2004 (EOS 20D), yet nobody's come up with a satisfactory sensor with a tighter pixel pitch... Have we hit the wall? If so, then buy what you need with confidence that nobody else will introduce something that outresolves it for less anytime soon (without a technological breakthrough)? We also haven't seen (apart from the low-resolution Fujis and Sigmas, which, until recently, also meant putting up with an outdated camera body) anything with much better per pixel image quality than a 2004 20d in the 35mm type cameras. From what I've read and the samples I've seen, digital backs ARE substantially better, even pixel for pixel, possibly due to extremely high quality supporting electronics.
One interesting possibility that COULD pop up with current technology is a full frame 35 mm camera using Fuji's Super CCD. With their existing pixel pitch, this would have nearly "15+15" mp (15 million regular pixels plus an equal number of tiny ones). Reports I've read put the per-pixel quality of those Fuji sensors (dynamic range, tonality, etc...) extremely high - well above any other 35mm type camera, and at least competitive with digital backs on a pixel by pixel basis - the problem so far is that the maximum resolution has only been 6 mp. If a 15 mp camera using the Fuji chip showed up, it would probably have stunning image quality well above that of the 1Ds mk II. Due to the two-part pixels and odd orientation, the Fuji sensors also tend to outresolve their primary pixel count by about 50% (a 6+6 mp Super CCD behaves roughly like a 9 mp sensor in terms of resolution and outdoes a 9 mp sensor substantially in tonality and DR). This suggests that a 15 + 15 mp Super CCD might behave like an extremely high quality conventional 22.5 mp CCD - a P25, for example... This would be substantially better than what's expected from a Canon 22 mp sensor using current technology. Of course, this beast doesn't exist and I have no information on whether it is even possible to manufacture a full-frame Super CCD or what the cost would be.
The other wild card is, of course, the Foveon sensor and its allies. While Sigma's cheating by counting each tricolor pixel three times (the SD14 isn't really a 14 mp camera), it does resolve better than other sensors of its base pixel count (and it has excellent color) - does anyone know about the DR? A full-frame 14-16 (real) megapixel version of this sensor could be very interesting, especially if it were in a non-Sigma body. Canon's working on something like that, aren't they?
If Nikon's new pro camera turns out to be full frame Super CCD, and Canon's is some Foveon variant, we could see some real strides in image quality on the 35mm side. Other than that, I haven't seen any evidence of anything but stagnation - the present image quality champ was crowned in 2004...
-Dan