dbreview on the D800 and switching from Canon to Nikon
Of course, dbreview stands for Daniel Browning's Review. ;)
I've been shooting Canon since I took an interest in photography seven years ago. Last week I decided to switch to Nikon, so I waltzed into a local store, grabbed a D800 off the shelf, and proceded to sell off all my Canon gear. So far I am very pleased with the D800.
I switched camera systems for three main reasons: dynamic range, resolution, and video. The first of these is that Canon's dynamic range is limited primarily by pattern noise (AKA "banding"). A large number of dynamic range tests that ignore pattern noise show Canon to be trailing the competition by a significant margin (e.g. DxOMark), but when you factor in the effect of pattern noise, the disparity grows even wider, to an order of magnitude difference.
So why did I switch now? Why not wait and see if Canon addresses at least the dynamic range problems in a future camera? Mostly because I'm tired of waiting. Unlike their competitors, Canon hasn't improved it much over the last 7 years. I looked at some 5D3 raw files to see if Canon finally fixed the problem and found they had not:
For example, just opening this ISO 125 raw file ( http://www.kleptography.com/dl/5diii/raw/02.cr2 ) with default settings in LR4 shows slight (to me) line noise in the dark tonal levels. Enable lens corrections and add a 2/3-stop push and it becomes very obvious. That is a woefully inadequate performance compared to the competition.
The degree to which pattern noise occurs on my 5D2 varies with color temperature, ISO tweeners, lens & SAOIR correction, etc. With other DLSR cameras, it becomes possible to photograph subjects that have a greater range between highlights and shadows, to emulate a film-like highlight rolloff, or even to keep an image that had the the optimal exposure but a sub-optimal gain (AKA ISO) setting -- a great heresy among the in-crowd.
The increased pixel count also factored into my considerations. A lot of photographers are characterizing this jump to 36 MP as a sort of tremendous, earth-shattering, never-seen-before leap in pixel count. But it's only a 72% increase (32% resolution) over the 5D2. Compared to the D3X, A850, and NEX cameras, it's only 50% more pixels (23% more resolution). When you consider it from the perspective of a relative increase, it's the same ballpark as going from a 30D to 450D (50%) or 40D to 50D (50%). It's still a noticeable increase of course, which made it a factor in switching.
It seems to me that most photographers only concern themselves with attaining the bare minimum resolution necessary for a good looking print. I look at it from a different perspective: what is the highest possible resolution that is needed before further increases make no difference in the print. Some photographers say that if you are only printing an 8x10, there is no benefit in the increased resolution of the 5D2 (21 MP) over the 5D (13 MP). But in my tests I found that to be untrue. Even in a print as small as an 8x10, 13 MP is visibly worse than 21 MP due to lower contrast from the OLPF, post processing (distortion, horizon, and other corrections), a small amount of cropping (e.g. 5% from each side), and other factors. See http://photography-on-the.net/forum/...d.php?t=747749
So while I do consider 21 MP to be capable of delivering the maximum possible detail and contrast in an 8x10 (if there is very little cropping), I'd prefer to be able to do the same with larger print sizes such as 12x18, 16x20, 24x30, etc., and with non-trivial degrees of cropping. Compared to the 5D2, the D800 should be capable of delivering the maximum possible contrast in a print that is 31% wider (e.g. 8x12 -> 10x15), but only if all other sources of diminishing returns are avoided.
Video was another significant factor. Perhaps about a third of my time with a DSLR is shooting video (99.9% going by disk space used). While I do also have an XH-A1 video camera and sometimes bring it along on my shoots, the video feature on my HD-DSLR has a number of significant advantages over it: superior low light capability, more control over DOF, takes up 0% of the space in my camera bag, weighs zero pounds, and uses all my existing still camera lenses. That is why video functionality factors into the decision so heavily.
With the clean HDMI feature on the D800, it is now finally possible to shoot for longer than 10 minutes at a time and with a *huge* increase in compression quality. Before now, my only option was the XHA1 with 1080p over component. Sometimes I setup to record from uncompressed 1080p directly to a high quality intermediate format using my BMD Intensity Shuttle. That way I can skip over low-quality compression (e.g. 25 Mbps HDV or 50 Mbps AVC-Intra) and go right for the good stuff (500 Mbps wavlet such as Cineform). Plus there is no time limit (not even tape changes every hour). I find it baffling that Canon did not provide clean HDMI out.
I've also been asked why I didn't spring for the D800E. The reason is that I strongly dislike aliasing. Most people think the only downside to removing the Optical Low-Pass Filter (OLPF) is moire, but that's only the worst manifestation of aliasing. There's also jaggies, stair-stepping, sparkling, wavy lines, bands, fringing, popping, strobing, and false detail. To me the overall look is very displeasing and has an obvious "digital-ness" that resembels a TRON-like computer world, not the natural world I see with my eyes. But I'm in the minority. Many people consider aliasing to be a desirable "crunchiness".
I think aliasing taints every image that has any sort of fine, high-contrast detail. Of course, images with fine, high-contrast detail are not really all that common, especially with a 36 MP sensor. A blue sky is not affected because it has no detail. Some photographers have so much motion blur, missed focus, or thin DOF that there is more than enough anti-aliasing to go around. It's only when circumstances provide high contrast fine details that aliasing can be an issue.
The majority opinion is that aliasing artifacts are not objectionable in photographs of natural subjects (i.e. organic, not man-made) such as landscapes and portraits. But I see the aliasing artifacts even there. It bothers me when blades of grass and branches grow in perfectly straight horizontal and vertical lines. I see those types of differences in the following example of the same subject taken both with and without an OLPF:
http://www.maxmax.com/images/Cameras...parision_6.jpg
I'm looking forward to the release of the D800E so that better comparisons can be made.
Lenses were a significant concern I had about swiching. I first considered it four years ago when the D3 came out, but at that time Nikon didn't have any meaningful competition for the 24L, 35L, 50L, 85L, 135L, 70-200 f/4L, or 17-40L. Now they at least have the first two and the last one covered. As for the 50L, they still don't have anything that can touch it for bokeh or aperture, let alone do it with a nice focus ring and fast AF speed. Their 85 is close enough for my purposes, but they're not really the same. The Nikon 135mm f/2 DC can't even come close to the 135L in detail or AF speed. But Nikon also has a few options that blow Canon away, like the 14-24 f/2.8. In the end I decided that there was enough parity for me to make it work.
There was also a host of smaller factors that influenced the switch. Nikon addressed a lot of them in the D800 (e.g. LV mirror cycling, headphones, etc.). The most galling offenses, to me, are those particular camera changes that would be both *extremely* useful to me and yet require only the *tiniest* modicum of effort by the manufacturer to effect. I have a long list of these for both Canon and Nikon:
Things that they should *stop* doing (i.e. very easy)
- Both: stop throwing away up to 5 stops of highlights by using numpty ISO and idiot ISO. See http://photography-on-the.net/forum/....php?t=1081982
- Both: stop warping the raw data to cover up the poor sensor angle of incidence response.
- Both: stop hiding important features such as the dead pixel remapper and shutter count.
- Both: stop encrypting parts of the raw files (it's not like it's hard to crack anyway).
- Both: Stop limiting the camera (e.g. 30-minute video) just because Europeans like paying extra taxes on everythihng.
- Nikon: stop masked-off pixel removal.
- Nikon: stop white balance preconditioning.
- Nikon: stop black clipping.
- Nikon: stop star killer (HFS) algorithm.
- Canon: stop bloating files with useless 14 bits until your sensors can actually use it.
- Canon: stop papering-over hardware flaws such as gain and offset imbalances by blurring the raw conversion (e.g. 7D+DPP).
List of easy feature changes/additions:
- Both: Raw histograms and blinkies
- Both: Break the 4GB video barrier by daisy-chaining files.
- Both: Focus stacking
- Both: Video features - zebras, peaking, waveforms, etc.
- Both: Allow us to control the lens aperture (to work around aperture activation variance in your lenses).
- Both: Allow exposure duration limits higher than 30 seconds.
- Both: Allow in-camera calibration for gain and offest imabalances.
- Both: Reduce bit depth with increased ISO
- Nikon: update Compressed NEF LUT for 14 bits
- Canon: clean HDMI out.
- Canon: Add some intelligent compression
- Canon: Improve AutoISO (needs to work in manual mode, with flash, and have "IC" - ISO compensation)
- Canon: Stop changing shutter speed during video just because of zoom.
- Canon: Stop limiting the user to just exactly 1 stop of Smart ISO (HTP).
- Canon: Add interval and timelapse
Here are some features that would take some real development effort:
- Both: Tilt-swivel LCD (without compromising weather sealing, if possible).
- Both: Wavelet compression like RED
- Both: Sensor-based PDAF like the Nikon One system.
- Both: Improve video moire and rolling shutter.
- Both: Allow liveview magnification while recording un-magnified video.
- Both: Faster CDAF and FDAF
- Both: Raw liveview (sRGB gamma, optional RGB balance and peaking)
- Both: Raw false color (each color represents a number of stops below clipping)
You'll notice that the first two lists are mostly just simple software changes. Sometimes I wonder if the reason they exclude some features is because they have incredibly low estimations of their customers' intelligence. That idea is reinforced when I see that Nikon's manual has gems like "care should be taken not to put your finger in your eye accidentally." (p.15).
When I decided to make the switch, I weighed the pros and cons of each manufacturer's list and decided that I could live with Nikon's particular set of braindamage just as I'd lived with Canon's for so long. So having decided that I could make do with the lenses and known issues, I picked up a D800 and started selling all my Canon stuff. The only lens I had to use with the D800 at first was a used Micro-Nikkor 55mm f/2.8 that I picked up for around $100.
My first impression with the camera is that I love that new camera smell. :D Second, that Nikon has a lot more buttons and doo-hickeys. The 5D2 has 18 buttons and 4 selectors/wheels, while the D800 has 26 buttons and 8 selectors/wheels. But they are still a lot more similar than they are different. Despite having never used a Nikon DSLR, I was able to pick it up and immediately start shooting. It was very intuitive to figure out how to set the AE mode, ISO, WB, EC, focus point, etc. without ever having to look at the manual.
Having now shot with it for a little while, I really like Nikon's controls/ergonomics better. I can make more changes to the settings without moving my hands or eyes from the shooting position. I love having two custom function buttons right under my middle and ring fingers. I still prefer Canon's AF point selector, AF-ON button, and zoom placement. I can no longer shoot and zoom with one hand. But when I shot with the 5D3 a little bit, I saw they moved the zoom over to the same spot as Nikon anyway, so I would have lost that anyway.
As far as build quality goes, I don't really know how to evaluate that. I did notice that the Nikon seems to have a lot of nice little after-thoughts, like springs in the CF door, larger eject button, and a viewfinder eyepiece shutter. The Nikon camera strap is uncomfortable and makes me itch, but I use a blackrapid anyway.
I love the optical viewfinder. The eye relief is a little better, and having all the options of the LCD (crop marks, AF point selection) is fantastic. That said, it is unfortunate that I cannot get a higher precision viewfinder screen like the Eg-S I had for my 5D2, but I guess that is part of the price I pay for the other benefits. The second part is that the viewfinder does not work at *all* without power. The manual only tells half the story: "When the battery is totally exhausted or no battery is inserted, the display in the viewfinder will dim. This is normal and does not indicate a malfunction. The viewfinder display will return to normal when a fully-charged battery is inserted." (p.33). The other half is that it also cannot achieve focus (using manual focus, obviously) without power -- the viewfinder is always out of focus until a battery is inserted. Personally, that doesn't affect me at all.
I like the liveview digital zoom. Unlike the two fixed choices you get from Canon, there is a wide variety of magnifications to choose from. Plus the highest magnification is much higher. The frame rate is very good until you get to high magnifications, where it slows down noticeably. Achieving critical focus is still hindered by the JPEG processing engine just like it is on Canon. Entering and exiting liveview seems to be slower than the 5D2.
I like the built-in AF-assist lamp, though I would have preferred it on the pop-up flash (even if that meant it was slightly larger). I hated putting the bulky 430EX on my 5D2 just to get the AF-assist lamp. I like having a pop-up flash. Works great for adding a catchlight and in other situations. I'm looking forward to trying the IR commander feature with Nikon's flashes. The FEC only goes from -3 to +1, I would have liked to be able to use +2 in some cases.
The dynamic range really is as good as I expected. It feels truly liberating. I no longer have to choose between preserved highlights and clean shadows -- I can have both highlights and shadows in the same raw file. The dynamic range is so good that I can expose for ISO 6400 even when the camera is set to ISO 100; then when I bring it back up in post, I still get a good result. Ironically, that means the first time I get a camera with a working AutoISO and the camera doesn't even need it. A working AutoISO on the 5D2 would have been incredibly useful because it has such poor dynamic range, but it was too severely braindamaged for me to ever use.
Diffraction is really not as bad as it's made out to be. See http://photography-on-the.net/forum/...d.php?t=747761 for more of my opinions on the matter. I took this shot on a $100 lens at f/11, but with some sharpening, it still looks sharp enough for me. Full size image:
http://thebrownings.name/images/2012...ample-full.jpg
100% crop:
http://thebrownings.name/images/2012...ample-crop.jpg
In fact, even with both the softening from diffraction and from the OLPF, I encounted a lot more aliasing than I was expecting. Here is an example of some moire:
http://thebrownings.name/images/2012...re-example.png
That is one thing I dislike about the D800: I was expecting less aliasing than with my 5D2, but I'm actually getting more, so I think the OLPF is weaker relative to Nyquist.
I got a chance to shoot the autofocus in anger during a martial arts practice. Compared to the 5D2, I was blown away. I never considered using any AF points except the center with my 5D2, because the others were so inaccurate that they were useless for anything shot at f/2.8 or faster, not to mention that they wouldn't work at all in low light. But on the D800 I was able to use *any* AF point and still turn in a high percentage of keepers with a Sigma 85mm f/1.4 in low light. I can finally say goodbye to AF-and-recompose forever (which itself introduced inaccuracies).
I haven't got to shoot any timelapse with it yet, but I'm looking forward to it. So far as I can tell, I will be required to continue using the same risky and suboptimal workaround I did on Canon: untwisting the lens in order to stop the camera from opening the aperture stop after every frame (which causes flicker due to aperture activation variance).
I like being able to use all the old F-mount glass with no adapter. I like the NEF compression and the option to use just 12 bits when I don't need the full dynamic range. I like having a built-in intervalometer. There's one less device and one less battery I have to carry around when I'm shooting timelapse.
I think the D800 is pretty good evidence that the War On Pixels (the long-standing antagonism against smaller pixels) is largely unjustified, something that I have been trying to convey for a while now -- e.g. http://photography-on-the.net/forum/...d.php?t=706255
Overall I'm very happy with the switch so far and I hope that the competition will spur Canon on.
