Are smaller pixels really bad for the image or subject, or are only smaller sensors noisier for the image as a whole?

Same shot, from same distance, with approximately the same *real* focal length (88.1mm for the Panasonic FZ50 and 90mm for the Canon 10D), both set to ISO 1600, and both receiving the same exposure (same f-stop and same shutter speed in manual exposure mode - f/4 and 1/20). I have both cameras at 100%, and I have the 10D at 272% to match the size of the FZ50's 100% crop, and I have the FZ50 binned down to 3x3 (33.33%) to roughly match the 10D's 100% crop.

All 100% crop images are RAW, Bayer-interpolated homogenously for both cameras, with no other processing than "auto-levels" in photoshop to make the images roughly white-balanced. They are still linear, and still in the RAW color space. Before auto-levels, they both had cyan-ish green casts.

IMO, the larger pixels accomplish nearly nothing for image quality, with the same focal length, and the same subject size in the final viewing. This doesn't mean that the FZ50 is as good as the 10D at ISO 1600, by any measure where you get to match the lens to the task. But clearly, I'd rather have image of a small subject with the FZ50's sensor than to crop it from the 10D, if I had the choice. At ISO 100, I think we could expect the FZ50 to look even better, as it cleans up faster than the 10D.

Perhaps the only real issue with super-MP DSLRs would be manufacturing issues, and readout time issues (as well as storage issues). More and smaller pixels is not necessarily the problem it is commonly believed to be, IMO.

Feb 20, 2007 - Since I originally put this composite image here, I have come to discover that the FZ50's ISO 1600 is a bungled implementation; IOW, ISO 100 pushed 4 stops gives less noise than 1600, as 1600's amplification does nothing but add noise. The fz50, therefore, is performing worse in these images than its sensels really need to.

May 25, 2007 - I have also since discovered that the FZ50's ISO 200 is the best ISO to push to higher ISOs, not ISO 100.