Camera Labs

I just finished my new test-bench for future lens-reviews with the D800:


test setup by Thomas, on Flickr

It is assembled from 1.5 A3+ printouts of my fav test-target and a mirror-tile on a 1 x 0.7m KAPAfix panel. The mirror (and the small target on it, green rectangle) is only there to aim the camera at and make sure that the alignment is perfectly perpendicular. I set the distance from the camera so that the lower right corner can just be seen in the image. The 100% crops that you see in my reviews come from the red rectangles. They are 4mm, 12mm and 20mm off the optical axis and are used as representatives for (near) center-, DX-corner-, and FX-corner-performance.
Magnification is at 1:40 and the little black squares that contain the small Siemens-stars have a real size of about 23mm. This magnification also means that the camera+lens has to be set up about 40x focal length away from the test-target. That is OK for testing a 85mm lens at home but it also means that for testing a 400mm lens like the venerable Sigma 400/5.6 APO macro I need a distance of 16m for the test-shots. So some tests are not for rainy days.
Btw.: the test-target covers only 1/4 of the sensor, so I'll be only testing one quadrant of a lens. But I do check beforehand whether the lens is reasonably well centered and also focus separately for the center and the corner(s) to make sure that the influence of any defects or field-curvature are minimized.

Side note:
If you revers your thinking from projecting the test-target onto the sensor to projecting the sensor - or a pixel of it - onto the test-target, you come to the conclusion, that one pixel from the D800 sensor (or a D7000 sensor) is equivalent to a 0.2 x 0.2mm spot on the test-target. And as there are 36 line-pairs in each Siemens-star on the test-target you can calculate that the width of each black or white stripe reaches the same 0.2mm at a distance of 2.3 mm from the center of the star. So it should be near impossible to see well-defined lines within a disk of 4-5mm diameter at the center of the Siemens-stars. If you use a 26" 1920x1200 monitor to view your images that translates into a disk of approximately 7mm diameter on your screen (6mm on a 24" monitor). So any lens producing a gray disk close to this diameter has a very high resolution.

_________________
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews, My Pictures, My Photography Blog
D800+assorted lenses

I've now fully stumbled across the intricacies of different RAW development. After getting some spanking on a German DSLR forum for my bog-standard CaptureNX2 RAW-developments I did some more testing with Lightroom4 with a different kind of sharpening from normal. The respective values for amount/radius/details/mask are
Normal: 25/1.0/25/0
Special: 70/0.5/36/10 (a hefty amount of sharpening but only on the radius of half a pixel)

After having a look at quite some images I must confess that the CNX2's conversions start to look rough and a bit brutal while the LR4-conversions look refined and in the end more natural to me - at 100% viewing. Here are some examples:









Sorry for the large 100% crops here (two of which are slightly over the legal limit for Camera Labs - sorry Gordon) but it saves you the time to go to flickr and select original size. On all these images the CNX2 conversion is on the left and the LR4-conversion on the right. Now judge for yourself which conversion is best to judge the sharpness of a lens!

_________________
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews, My Pictures, My Photography Blog
D800+assorted lenses

To me (& my old, untrained eyes) the images on the right look a lot `softer` & less contrasty.
I prefer the left

_________________

Nikon D700 & D7100 bodies + Nikkor 300mm F2.8 VR11, 1.7TC, 50mm F1.4G, 16-35mm f/4G VR, 105 F2.8 VR macro, 70-300mm lenses. A couple of filters, Giotto tripod & ballhead. Lowepro Slingshot 302 AW

http://www.flickr.com/photos/oldcarlos/

Yeah, that's the problem:
- Some value contrast over "fine-lining"
- Some like to sharpen the images until artifacts appear

As a lens-tester you have to select your parameters in a way that
1. shows the potential of a lens
2. doesn't introduce artifacts / moire
3. is stable to compare results over longer periods of time and changing converter development

Honestly I don't believe one can ever reach (3) but at least over the past years CaptureNX was a pretty stable converter.

_________________
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews, My Pictures, My Photography Blog
D800+assorted lenses

After restlessly testing Lightroom's RAW-developer against CaptureNX 2's I'm now almost sure that I'll switch to LR for future testing. See the following:



The image shows some small detail from a test-chart at 200% (!) magnification. On the left LR4 developed RAW with 70/0.5/36/10 sharpening. On the right CNX2 developed RAW with standard sharpening @3.
See haloing (1) around dark areas to artificially increase contrast and artifacts (dotted lines, 2) from CNX and false colors from moire (3) with LR. The false colors from moire can be eliminated with LR4's new anti-moire brush as demonstrated in the following image:



This image also proves that the anti-moire brush in LR4 does not brush over the luminance signal and makes the image less sharp.

The only thing that I've seen so far that CNX2 still does better is the elimination of longitudinal CAs (no correction for that at all in LR4) and some similar magenta colored lens aberrations. This will show in my 85mm prime shootout as the Nikon AF-S 85/1.4G is unfortunately capable of producing quite strong magenta CAs.
Now there are two possibilities: Show the non-corrected images as anybody could produce them out of LR4 or go for the extra step of processing in CNX2 to eliminate those CAs.
Currently I#m thinking of showing LR4-processed images as the standard in my tests plus the occasional CNX2 developed RAW to show what you could do against some nasty CAs.
---
Btw.: Close scrutiny showed that there is indeed no perceptible difference between jpgs developed in-camera and the results from RAW-conversion with CNX2. But the benefit of using CNX2 on the RAWs is the possibility to alter any parameter of that development post-shot. E.g. reduce standard-sharpening from a value of 3 to 1. Which eliminates the halos in the first image (1) but does nothing against the artifacts (2) - besides of making them look less sharp.
---
Footnote: don't forget that the above images are presented here at 200% magnification. This is Uber-pixel-peeping and should be avoided under normal circumstances

_________________
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews, My Pictures, My Photography Blog
D800+assorted lenses

Currently with the new D800 I'm re-shooting test-scenarios with the different lenses I have at hand. The idea is to complement my reviews based on results on the D300- and D700-bodies with results from the D800-body and give you some update on their performance. As the D300 pixel-density is pretty close to the D800 sensor I would not expect to find exciting new insights regarding the DX image-circle. But the D800 certainly poses some serious challenge for FX-lenses outside the DX image-circle.
Today I was lucky to use the sunny morning light to do some real-life comparative shooting of the "Unremarkables" near where I live.


Unremarkables 51474 by Thomas, on Flickr

The idea is to present a scene shot within only a few minutes so that the lighting has not changed too much.
Interesting: Even at 8:30 in the morning heat-shimmer already poses a problem as it distorts some subjects quite visibly if you look at 100% (click through the image).

_________________
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews, My Pictures, My Photography Blog
D800+assorted lenses