After a couple more days of testing, I’ve got more observations on the D800’s autofocus system:
1. I think we’re reaching the limits of accuracy for CAM3500FX, and in fact, any phase detect based AF system. There are just too many parts that have to be precisely perpendicular and in exact alignment to achieve focus accuracy – the AF sub mirror assembly, the AF sensor itself, and the main imaging sensor. If any of these is out of plane by a few microns, then you’re going to see some softness. We’re now getting enough resolution that the planarity of the lens mount relative to the sensor becomes an issue – to say nothing of perfect alignment of optical elements. I believe there was an article posted a while back on the Luminous Landscape about shimming a sensor and how much resolution improved by both on-center and especially in the corners of the frame.
2. Future AF systems will have to be hybrid – i.e. use some form of contrast detect or phase detect embedded into the imaging sensor in order to work around these limitations. It doesn’t however solve the problem of mount planarity or lens element alignment.
3. There are some things you can do as a photographer to counter these limitations, chief of which is use live view for critical focusing, or stop down – or better yet, both. Live view eliminates problems of AF sensor/ sub mirror alignment. Stopping down covers slight sensor misalignment with depth of field.
4. AF fine tune is an absolute must to get the most out of the AF system.
5. Bad news for manual focus fans. I did my mirror alignment and calibration this morning – it was almost perfect from factory, which is a first; however, my joy died after removing the focusing screen. The focusing screen in the D800 is a different size to anything Nikon has yet produced. Worse still, it’s the largest one I’ve ever seen, so you can’t even cut something down to fit – it’ll just drop out. This is a real shame; I can only hope a third party produces replacement screens for MF aficionados.
6. Finally, lenses you thought were fantastic on the previous 12MP FX cameras may now only be mediocre or average on the D800 – you have been warned. MT
A couple more Nikon D800 images, and some commentary
Shadows. Nikon D800, 28-300VR
One of the big questions that’s been running around the internet is whether the D800 can replace medium format – the answer isn’t quite as simple as you might think. Firstly, the resolution is definitely there to compete with the lower pixel count options – and even more so once the D800E joins the fray. However, a very large part of the medium format ‘look’ is a product of the interaction between lenses, sensor size (i.e. angle of view) and tonal response of the sensor. Remember that almost all medium format sensors are CCDs, which have a very different – and less linear – tonal response to CMOS sensors. For comparison at the 35mm/FX size, the D700 and D800 both have a different tonal palette to the CCD-based Leica M9. Undoubtedly the look is different already due to no other factors.
However, once you consider the angle of view vs DOF equation, then things look different again – it’s similar to the difference between APSC and full frame – for a given angle of view and aperture, you’re going to have ever decreasing depth of field (and quicker transitions between in-focus and out of focus areas) with the larger sensors.
Some of the other things which have been of concern to shooters – like having enough light and decent support – are nothing new in the medium format world, especially to users of very high resolution backs like the IQ180 or H4D-MS; this is perhaps why a lot of people are crying ‘so what?’. Even so, good support pays off for both medium and smaller formats – even compact shooters have something to gain. The tradeoff is always weight and flexibility.
A quick note on color reproduction: the D800 has the most accurate color I’ve ever seen. This seems to be corroborated by the recent batch of DXOMark tests, which rank the D800 first (!). Most cameras have trouble accurately reproducing the blue-green-cyan tone of glass reflections (due to UV transmission/ reflection issues) – but the D800 required only minimal corrections. Furthermore, the very fine repeating patterns between the squares on the textured floor panels do not show visible moire, but plenty of detail. It’s also worth noting that nothing is blown out, and I crushed the shadows for the visual effect – there weren’t any blocked up blacks, either.
The look. Nikon D800, 28-300VR
Again, I was very pleased with the quarter tone reproduction quality here. I’m actually finding that for a lot of images – especially those with less contrast – you need to apply a curve with a very long, shallow shadow tail in order to get the right ‘look’. This is undoubtedly due to the D800’s huge dynamic range at low ISOs. As for detail, you can see the weave in the man’s shirt and my reflection in his glasses.
Bike traffic. Nikon D800, 28-300VR
This shot was a test in many ways – extreme highlights and shadows; lots of fine detail; tonality of B&W conversions. The short answer is, the camera passes. The image (which looks much better on a large monitor at full size, by the way) retains good detail even into the extreme corners, despite being shot with the weakest end of the 28-300VR, and has nice rich quarter and half tones, which make for a good B&W image.
The reader and the thinker. Nikon D800, 28-300VR
This portrait is a little deceptive, because there was a lot less light than it appears – 1/125s at ISO 1800, f5.6 and 150mm. Yet the camera held on to shadow details well – with remarkably little noise, I might add – and I can read the text in the newspaper at full size. I don’t think the files make as outright punch B&W conversions as say the Leica M9-P, but black and white conversions from the D800 seem to have a unique signature of their own – perhaps best characterized by subtle tonal gradations and deep shadow detail. In short, I like what I’m seeing, but I’m going to need to adapt the processing style a little to get the most out of it. MT