As you know, I have been struggling with whether to go to the Pentax 645D. There is a controversial article by Mark Dubovoy on Luminous Landscape at the moment - to the effect that 'real photographers shoot medium format'. Of course, that isn't what Mark said, but it is what he implied - that no matter what the print size, one could see the difference between medium format and smaller formats.
The example he used showed distortion with the dslr image, and greater contrast and thus less well controlled highlights, and better shadows.
Of course, much of this is a function of the particular lens and f stop he chose to use, and I dare say the two images could have been made to look MUCH closer to each other with a bit of care.
My immediate reaction was 'what a load of bull', but then I calmed down and decided to look at far better comparisons made in some of the reviews of the Pentax 645D, in which images are compared to the Canon 1Ds3.
I made prints of reasonable size, 24 inches in the long dimension - something that both cameras should be capable of making. I then looked at the images at 100%.
I have poured over the images, and here is what I have found. If one looks at the finest details, branches on the horizon miles away, there is nothing to separate the Pentax from the Canon at this size of print. The same applies to railings and bricks and anywhere the contrast in tones is significant. Where I can see a difference between prints is in the low contrast areas, branches against bushes or downed leaves. Here there is a noticeable difference in the Pentax images - it's as if large areas (up to 1/4 inch of the print) were smeared, the colour blended, a complete absense of texture and detail and even simple variation in tone, on the Canon image. it isn't about greater resolution (which doesn't really show up until you make very large prints), but if the Canon can smear details over 1/4 inch at this size, it can still do that over 1/8 of an inch in a 12 inch print - that is definitely visible.
I don't know what the difference is - whether there is a lot more manipulation of data to squeeze every bit of accutance (edge sharpness) and low noise out of the Canon at base ISO, or whether this loss is the fuzzy filter, or due to the number of bits per pixel in the pipeline, or what - all I know is that for all I think Mark picked a poor example, he is fundamentally right that there is something different about medium format.
Whether this will still be true when there are 36 megapixel full frame dslrs without fuzzy filters remains to be seen.
It reminds me though of the days of film when people using Technical Pan and special compensating developers could show extremely sharp prints, albeit terrible ones, lacking in the tonalities that were readily visible in an 8 inch print when comparing 6X6 with 35 mm.
Careful examination of the two digital images at 200% clearly shows that no amount of sharpening or local contrast enhancement is going to bring back the missing texture in the smaller camera images - for whatever reason.
I wish I knew the reason. It would help me decide whether to fork out $10,000 for the Pentax, or spend almost 2/3 less for the next generation Canon or Nikon with 36 megapixels. Those cameras will have sensors 3 years newer than the Pentax, but their pixels will be considerably smaller than the Canon 1Ds3 has now - does this mean even more loss of subtle contrast detail?