Currently preparing for my future lens-reviews with the D800 replacing the trusty old D300. Here are some thoughts on what I will do.
I'll be using the same test-target, so Siemens-stars and funny circles will stay
They have the advantage that you can get a very good impression on the various aspects of image-quality:
- resolution: how large (or small) is the grey disk in the center of the star?
- astigmatism: is the gray disk circular or elliptic?
- haloing: are the borders of the black rectangles well defined or is light seeping into them?
- distortion: is the rectangle rectangular or skewed?
- lateral color aberration: do the borders show color-fringes?*
So what you can actually see from these test-charts is much more than you can read from numeric diagrams.
See the following example from the my macro-lens comparison
(AF-S DX Micro-Nikkor 40mm f/2.8G with Nikon D300 at f8.0, 100% crop from corner):
The big question is, can I make the new tests comparable to the old ones?
Short answer: Not really!
And here's why. The pixel-density of the D800 sensor is slightly higher than that of the D300: it's like having a 16MP instead of 12MP on the same sensor-area (APS-C), so the theoretical linear resolution went up by 15%. The unknown factor for the real resolution is the strength of the AA-filter of the D800 compared to the D300 but assuming those are similar one can say that a 15% increase in linear resolution is on the border of being visible. My own rule-of-thumb says that a 20% gain in resolving power can be seen. There are in principle three ways to handle this difference:
1. shoot D800 from the same distance as D300: the target will look 15% larger in 100% view, the grey disk will grow by the same amount so the relative size of the grey disk will remain the same.
2. same as before, but downsize the resulting image by 15%: the target will look the same size from the D800 than from the D300 but we would be messing with the test-image on a pixel-level, effectively binning a row of 7 original pixels into a row of 6 pixels rendered by some algorithm in the post-processing software of choice. I'd never do this for the same reason that I never use any distortion-correcting algorithm on my test-charts: I prefer to leave the original pixels geometrically unaltered.
3. increase D800 distance to test-target by 15% over D300: The target will look identical in size in 100% view but is actually projected 15% smaller on the sensor. This is like you were looking with a 1.15x stronger loupe at the performance of the lens. Thus the grey disk will grow 15% which for a good lens might mean that the grey disk on a 26" monitor will increase from say 7mm to around 8mm.
Personally I prefer (3) as it gives you a clearer sense that the new D800 test-bed is indeed more critically looking at the lenses of my lens-reviews. It also keeps the distance between printer resolution (for the test-target) and resolution of the combined lens-camera system, which in turn makes it more future-proof.
I'll now go back to the bench to build a new test-target that will be used to test lens-performance near the center of the optical axis, near the corner of the DX image-circle and near the corner of the FX image-circle in one setup. Mind you, not necessarily in one shot as I focus on each spot individually to exclude any effects from field-curvature! This also means that you might take the center and DX-corner performance as an indicator to D7000 performance as it has the same pixel-density as the D800. You should be aware though that the effects of different AA-filters and (to a lesser extend RAW-conversion) might make the D7000 perform slightly different to what you will see from my D800 test-shots.
And one thing is clear: The FX corner performance from the new D800 test-shots will differ clearly from the results obtained with the D700 in earlier tests as the linear resolution of the D800 is 73% higher. This might lead to some re-evaluation of older results regarding FX corner performance of some lenses. But we'll see as my testing with the new setup gets along.
I hope to bring you some insights soon and reveal how lenses perform on Nikon's brand new "Über-Body".
*please remember that I always test with automatic correction for lateral CAs ON as this has been a standard in Nikon bodies and related RAW-converters for a long time. So when you'll see lateral CAs in my tests it's what the automatic could not remove.
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews
, My Pictures
, My Photography Blog