I realize there is much more than simply resolution in evaluating the differences such as dynamic range, tonality, etc. This test can only attempt to simulate resolution due to the limit of using a single camera at different levels of optical zoom. Even with resolution there will be difference due to the presence or absence of an anti alias filter and the lens quality. But hey, at least it helps answer one of my questions.
The methodology used was to shoot two different scenes using varying amounts of zoom to simulate the difference between 12, 24, 31, and 39 megapixels. No attempt was made to change the framing between the typical 4:3 ratio of a medium format back vs the 3:2 ratio of my Nikon D3.
In the first test I show a landscape scene using a Nikon D3 using a Tamron 28-75 f/2.8 lens at f/5.6. The first photo shown displays the overall scene at 28mm.
Now for the various zoomed crops.
First the 28mm crop which represents a 12 mpix sensor:
A 40mm crop representing a 24 mpix sensor (I choose 24 mpix since the rumored D3x is expected to have 24 mpix resolution and existing medium format backs are at 22 mpix):
Now a 45mm crop to simulate a 31mpix sensor:
And finally a 50mm crop to simulate a 39mpix sensor.
Now for an indoor scene with side lighting to highlight the texture in the pillow and text in the scene. This time I used a Nikon 70-200 f/2.8 VR zoom at f/6.3.
Once again we show the overview of the scene showing the 12 mpix resolution staring with 70mm:
Now for the various crops.
First at 70mm showing 12 mpix resolution:
Now 98mm simulating 24 mpix resolution:
Now 111mm simulating 31 mpix resolution.
And finally 125mm simulating 39 mpix resolution:
I've made available the full scenes and crops at the following link for your reference:
http://picasaweb.google.com/gseitz/PhotoRamblings
