Normally, we look at a camera from a holistic point of view and compare it to the competition or the class leader. This unfortunately doesn’t make sense for extreme outliers like the DP2Q; we’ll have to do something a bit different. This review will look primarily from the point of view of image quality, and whether we can live with everything else. This is the opposite from every other review I’ve written to date, and the reasons for this will become clear soon enough. The other big change will be considering workflow and software as part of the camera package: it’s impossible to do anything else, since unlike every other camera, there is no universal workflow we can apply. Those of you who do not like caveats, are unable to look at something objectively, or are not open minded, I suggest you save yourselves some angst and stop reading now.
You’re probably wondering why this DP3M doesn’t look anything like the press release photos – my friend attached a RRS grip and plate to it, and rightly so; without it, the camera is not very comfortable to hold.
I eventually caved to both pressure and curiosity, and borrowed the Sigma DP3M from master printer and good friend Wesley Wong – who has the DP Merrill in all three flavors. In case you’ve been living under a rock for the last few years, Sigma has been going their own way with the DP series of large sensor compacts; all of their cameras now share the same 14.7 MP (effective) three-layered Foveon sensor, with a 4.99 micron pixel pitch and true color/ true resolution information across all photosites. In a nutshell, the difference between Foveon and Bayer sensors is that the former records actual RGB values for each pixel, but the latter only records R, G or B, interpolating the other values from neighboring photosites. It’s difficult to determine precisely just how much resolution loss Bayer interpolation causes, but in my experience it seems to be around 50% or so. Sigma claims that the camera has the equivalent of 46MP (being 15.3 total MP x3 layers) – but this is really pushing it; images upsampled this far simply do not have the pixel-level ‘bite’ of a good Bayer file.