dante stella stories photographs technical guestbook
|A proposito della civiltà di apparenze... why does an expensive digital camera come in a box with recycled paper inserts?!|
|Sizing up the Nikon D2x and the Kodak DCS Pro 14n|
It is now about year since the release of the Nikon D2x. There are a lot of reviews out there, and they cover a lot of technical information. I have been a bit dissatisfied with these reviews, because they completely fail to capture the practical aspects of use and output. This has caused me to surmise that:
-- Many people who write about digital cameras have little or no experience with analog cameras. They can review an EOS-1Ds or anything else, but where it says f/stops in the viewfinder, it might as well be ergs, knots, or any other mysterious unit.
-- Many people who write about digital cameras take two or three shots at a time, immediately plug the cameras into their computers and examine the pictures of their cats at "actual pixels" size to see which lenses are "good" and which one have "chromatic aberration" (as if they knew what that was).
-- Many people who write about digital cameras have never used a really efficient output system, whether it be a semiautomatic analog enlarger or a good piece of RAW conversion software and a calibrated dye sublimation printer.
I would caution the reader to rent anything that costs this much money before buying it. Numbers and statistics tell a very small part of the story, and there is only so much even a practical review can convey.
At least for the purposes of this review, I think the solution is not to look at the D2x in a vacuum, but instead to evaluate its functional aspects against the other high-resolution Nikon-based SLR: the Kodak DCS Pro 14n. This review will concentrate on three key areas: (A) imaging characteristics, which describes the hardware; (B) handling characteristics (how it works); and (C) image processing (how much work it requires on your part to get to paper).
My short conclusion is that the D2x is fantastic - and I only wish that its firmware was a little more informed by Kodak's far more practically oriented processing systems. Put another way, I think that Nikon employs better engineers and Kodak actually talks to photographers.
A. Imaging Characteristics
1. Image quality: draw. Let's get this one out of the way first. The Kodak has no antialiasing filter. The Nikon has a very weak one. They hit moire occasionally with different types of subjects. But across a wide range of prints, I can't identify any real difference in image quality. For a given angle of view, there is no palpable difference in image quality.
By the numbers, it's not hard to see why. The Kodak has 13.8 million pixels in a 36x24 area; the Nikon has 12.2 million in a 24x16mm area. 4500 x 3000 vs. 4288x2848. This not a huge difference in absolute count and less than 10% difference in the individual vertical and horizontal dimensions. Testers have discovered that the extinction resolution is identical. In terms of usable area, consider that the Kodak shows you about 90% of the frame in the finder and the Nikon 100%. So in terms of shooting for composition, you get the same overall resolution for what you actually see in the finder.
Both cameras will challenge your lenses in different ways. The Kodak will show you which 35mm lenses suffer from aberrations at the edges of their fields (but not as badly as a Canon 1Ds, which adds chromatic aberration from the microlenses on the sensor, too). Because 35mm lenses were designed around a medium that has a 3-dimensional sensing surface, the Kodak is sensitive to lenses where light does not hit the sensor more or less head on. This means that retrofocus zooms work better than wide primes at the edges of frames (ok, since the 17-35/2.8 is better than almost any prime Nikon makes in that range). Look at Kodak's optimization list - their orientation is toward high-end lenses, zoom and prime. Kodak doesn't spend a lot of time writing algorithms for cheapo optics, nor should they.
The Nikon is ruthless. Although it technically uses the "best" part of 35mm lenses on its DX-sized sensor, the pixel pitch is so small that you will never get the full benefit of 12.2MP at wide apertures and huge magnifications unless your lens is getting 90 lp/mm resolution or better - a performance figure not representative of SLR lenses shot at their widest apertures. Many lenses can do this stopped down, but only a very small number can hit this wide-open. In the real world, the D2x has fantastic image quality, even with cheap lenses wide open, on enlargements up to 8x12. It's only past that, or if you are a pixel sniffer, that you need to adjust your technique or the quality of your optics to get the best results.
The Nikon will also challenge you because you need higher shutter speeds to account for shake. Instead of 1 / focal length, go 1 / focal length x 2. You will be surprised. You can also use the AF-on button to initially focus, which makes the shutter release one step and vibration less likely.
2. Dynamic range: draw. I won't belabor the point. The Nikon has 9.5 stops of DR at base ISO; the Kodak 11.5. Although the Kodak technically exceeds the Nikon by 2 stops, I think that either figure is more than you realistically need. It's really a question of translating the camera's range into an output - and how you do it (and hence how much detail you need) is largely dependent on how you photograph. You either apply a curve in the input or a curve in the output.
One school of thought is the all-inclusive school (applying the curve in the output). With film, these are the T-Max 400 shooters. Record everything literally, from the deepest shadow to brightness well beyond what the human eye can tolerate. Then try to figure out how to translate 214 tones in capture into 28 in output. Don't delude yourself into thinking this is particularly easy; there is a lot of arbitrary curve-fitting that goes into this. You need a good handle on how you want shadows to transition to midtones to highlights, and you can bet that some part of the picture will be rendered contrary to the way it exists in nature. Either that or you just let the tones roll off the low end of the output, in which case the extra dynamic range is for naught.
The other is to use a more limited capture medium and apply a curve in the capture. Most films do not not respond to light in a linear fashion (like T-Max 400 would) and have "only" about an 8-8.5 stop range (absent developing heroics like N+2 developing). Rather, they respond more to certain ranges of light levels than others. Their response curves may have less slope at the ends (aptly called the "toe" and the "shoulder") meaning that shadows or highlights produce less response on the capture surface per unit light increase than midtones. This helps to compress scenes of, say, 2500:1 light:dark ratio onto a capture medium that can only record, say, 1000:1. Many digital cameras can be set to do the same thing in JPEGs by selecting (or creating) curves. The Kodak does this via "looks;" the Nikon lets you do this via downloading a curve from Nikon Capture into the camera.
If you record less than a 256:1 range with either method, you use an output curve to expand the data that is there to fit a 256:1 range. This would be contrast control on black and white paper. Or the levels control on Photoshop.
Both cameras have great DR - and it is amazing that with even after taking incredibly inconvenient steps (essentially the s-l-o-w compositing of two separate exposures), the Fuji S3 only gets 1/2 stop more range than the Nikon does at up to 8 frames per second.
3. Noise: who cares. It may come as a shock to some people, but digital imaging chips all have blue channel noise. You see less of it on chips with bigger pixel pitch (or lower resolution), but it is always there. Noise can be mitigated by in camera electronics , in-camera software noise reduction, and noise reduction in RAW conversion or later. The Kodak adds one neat new trick: allowing you to make smaller RAW files, averaging several pixels together to "average out" the noise.
A lot of people complain about noise in digital pictures. It is my sincere belief that anyone who complains about noise in a Kodak 14n or a D2x either (a) has never printed a photograph with one of these cameras; (b) has less than 10 years' experience with photography, i.e., never shot film; or (c) just likes the plastic look of strong anti-noise algorithms like the D50 or the Canon DSLRs use.
Honestly, though, if you know what 400-speed film grain looks like, either of these cameras at ISO 400 has less noise than scanning and printing negatives (this includes Frontier prints. Native analog prints are another story - grain is about equal). Neither of these cameras creates any appreciable grain in 8x12 prints. Stop your pixel-sniffing. And either camera at low ISOs is noise free. And if you do landscapes, grain and noise may in fact be desirable. Why? Because human perception of distance depends in part on atmospheric haze at distance (this is now a feature of all CGI movie animation). Grain and noise helps simulate distance. If you didn't have it, your pictures would look a little bit weird. Oh, wait. That's why most digital pictures look so weird.
B. Handling Characteristics
1. Viewfinder quality: Nikon. The D2x has a nice, bright 100%, high eyepoint finder with LED focus indicators (a la F100). It also has a nice display on the right side showing color balance and selected ISO. The focusing brackets are in the finder, not on the screen. The focusing screen is all but useless for manual focus operation. If you want gridlines, you have to buy the E screen ($32). The viewfinder is nicely enlarged with the DK-17M magnifying eyepiece, but it is more difficult to see the displays below and to the right of the viewfinder image without "panning" your eye.
The Kodak has a typical 35mm finder that crops about 8% of the frame. It has a high magnification and a low eyepoint. The focus indicators are on an LCD overlay on the focusing screen, as are the "gridlines on demand." When the battery is out, the finder "blacks out." The Kodak's finder lacks the clarity of the D2x finder, I suspect due to the presence of the LCD panel (this was also the case with the F5 and its LCD screens). Look for the same overall quality on the D200, which has a similar overlay.
I find it very surprising that Nikon has no manual focus-capable screens for the D2x, considering that it once again allows you to use AI and AI-s manual focus lenses. The electronic rangefinder has excellent frame coverage, but sometimes the AF brackets are too big for effective manual focus. While I understand that the ultrabright screens came in with slow consumer zooms, a lot of people using the D2x are using lenses with maximum apertures of f/2.8 or faster - and these are certainly bright enough to accommodate standard focusing screens.
2. Overall speed: Nikon. I don't know how many people need 5-8 frames per second, but the D2x's instant-on, instant-fire capability is impressive. The shutter lag of 37ms is commensurate with (or less than) mechanical film SLRs. It also creeps up on the (alleged) 20ms of a Leica M6. The AF speed with the Multi-CAM 2000 module and D2x CPU is mind-bending. It even works quickly with screw-drive lenses. The Nikon also has a huge burst capability. The Kodak is not designed to be used quickly; it takes about 15 seconds to power up and shoots 1.5 frames per second, with a much smaller RAW buffer.
3. Autofocus: Nikon. The D2x's Multi-CAM 2000 has eleven focusing zones; 9 cross-type in a grid covering 80% of the frame and one line-type on each end. It can be configured to look at one, some, or all sensors for tracking. It is very fast. The Kodak has four sensors: one cross type (center) and four line-type (top, bottom, left, right). These are clustered around the 12mm metering reference circle - an arrangement which Nikon has elaborated in the D200. The Kodak is excellent with AF-s lenses and good with those primes and zooms with low moving mass. It is adequate with other AF lenses, but it would not be a first choice to cover dog races at close range.
4. Metering: draw. Has matrix metering has come too far? Originally introduced with the Nikon FA in the early 1980s, it had a good idea: take readings from five discrete areas of the frame and compare them to a few thousand image types to guess the subject and thus the correct exposure. It was supposed to make exposure compensation obsolete, and Nikon disabled AE lock on the FA because, hey, you don't need that either. The original system (called AMPS) was fine with negative film and tended to produced washed-out slides. It was elaborated slightly in the F4 a few years later. It took a big step forward with the F5, when the number of segments jumped to 1,005 and the meter started taking light temperature into account (important because photocells respond differently to different colors of light). Both the Nikon and the Kodak use Nikon's system.
The Nikon uses Color 3-D Matrix Metering II, which is basically the F5 system with 300,000 image types (10x the number that the F5 had). This is the most sophisticated metering system in the universe and by most accounts, the best. In contrast to the original matrix system, this one tends toward less exposure. In practice, it produces decent exposures most of the time. Its failure modes are high-contrast scenes and scenes where a highlight is nowhere near the selected focus sensor. Then you need exposure compensation (which still isn't recommended). The base compensation of the system seems to be to underexpose by 2/3 stop ambient and 1 stop flash (D-TTL). Although some people rationalize the 3D Matrix II system's steep learning curve as "learning what it can and cannot do," the whole point of matrix metering is that the user does not have to memorize 300,000 different exposure compensations for as many types of scenes. Of course, you can always use the centerweighted metering and just compensate based on past experience developed with other centerweighted systems. Or use the spot meter, which is astoundingly precise.
The Kodak uses the N80's 10-segment monochrome 3-D Matrix Metering with 30,000 scenes. It does a surprisingly good job, but it may be that the Kodak's sensor has a huge range and its post-exposure correction system are incredibly good. If you set the camera body to +1 for ambient and +1 for flash (at ISO 160), it is difficult to get a bad exposure. I mean, really tough.
5. AI, AI-s and non CPU Lenses: Nikon (all the way). Yippee! One of Nikon's big (alleged) cost-saving moves was to drop AI and AI-s support on its autofocus bodies, particularly eliminating the meter coupling allowing the camera to tell how many stops off wide-open the lens is. On cameras like the Kodak, you have to set the lens ring manually and shoot in manual mode. This is time-consuming but better than nothing. You could get a Matrix chip put into an old lens, but it is really expensive compared to the value of most AI or AI-s lenses.
The D2x (wonderfully) has the coupling, meaning that you can plug in your AI and AI-s lenses and get aperture priority and matrix metering. You can plug in the correct maximum aperture and focal length,* and the camera will display the correct f/stop and record both the shooting aperture and focal length in the metadata. You can't do shutter-priority or program, though, because the camera doesn't know if it is dealing with an AI or AI-s lens (because the camera sets the aperture "blind," it needs to know whether the stopdown is linear or logarithmic. In aperture-priority mode, the lens stops down until it literally hits the stop (what the aperture ring on the lens is set to).
*The one real gripe here is that an F6 allows you to program up to six AI or AI-s lenses; the D2x lets you program one. For the price difference, Nikon should allow you to program twelve lenses with the D2x!
The AI/AI-s capability will mean different things to different people. By the numbers, most Nikon AF primes are at least as good as their AI and AI-s predecessors. But some lenses (like the original chrome 105mm f/2.5) have special image characteristics that the newer lenses can't match. Alternatively, if you like focusing for depth of field, the old 24mm f/2 AI-s lens is light, compact and has a real focusing scale. Finally, you might just want to expand your range with the AI and AI-s lenses being fire-sold on Ebay lately.
6. Durability: draw. Durability is not intuitive. The D2x is an impressive, tank-like, all-metal camera. The 14n, which has a magnesium shell on a composite body,* feels like a lightweight toy by comparison. This superficial description aside, it's important to remember the maxim that metal bends and plastic bounces. This is why you see so many dented metal Nikon F3 pentaprisms and so few cracked plastic F4 prisms. The 14n's durability will likely be enough for anyone, and it has a significant weight advantage. The D2x has some durability advantages, though. Well-anchored internal battery contacts, extensive weather sealing, and flush control buttons mean that this camera is not going to be put off by rough handling or bad weather (although NB that the specifications only provide for operation at up to 85% humidity, non-condensing).
*According to Kodak personnel, the composite body and N80 platform were dictated by the fact that that film plane aperture of the F5 was too small to accommodate a 24x36mm sensor, and it would have involved massive labor costs to enlarge the opening on a metal frame camera.
7. Power management: Nikon. In spades. Both of these cameras have huge batteries. The Kodak uses 7.4v, 2000MaH batteries; the Nikon uses 11.1v, 1900 MaH. According to their respective manufacturers, the Kodak should do 300-400 pictures; the Nikon should do over 2,000. In real life, the Kodak hits about 220 (a full 4Gb microdrive), and the Nikon appears to do about 600. The Nikon can be switched on and off; it is not so easy with the Kodak, since there is a startup lag. To be fair, though, the Kodak has to support a monstrously powerful DSP to support the full-frame chip, and this is a battery killer.
That said, one annoying thing about the D2x is that the battery cap (external cap on the camera) and the battery are two separate pieces. While this saves some incredibly marginal amount of space when you travel, it also means that you have to struggle with moving the cap from battery to battery - or buy more caps and attach them, which in turn defeats the space-saving aspect of the battery in the first place.
Another Nikon issue is the insulting idea of omitting an AC adapter with the camera. This is a part that probably costs Nikon $10 to make (consider that the camera costs $5,000) - and inexplicably, you have to buy it separately. This is inexcusable in light of the fact that you need the AC adapter to perform foreseeable maintenance (sensor cleaning).
8. Storage: draw. The Kodak is designed for CF and SD cards. It works especially well with microdrives because it is set up to take advantage of their internal buffer. Write-accelerated CF cards operate at the same speed. Backup is something you can do on the fly, writing the same (or different) images to the second card. RAW files are nicely (and losslessly) compressed, giving you 220 RAW + 6MP JPEGs on a 4Gb microdrive. The Kodak cannot be mounted on a desktop like a USB drive - it is a Firewire-only show. The Kodak has two features that can really come in handy. One is recovering deleted files on the CF card (even after a quick format) and the other is doing a secure erase (writing to the entire card to prevent recovery).
The D2x is designed for CF cards only, and it does take advantage of WA technology. You can use it with a microdrive, but the preview time is 1/4 sec longer, the power consumption is palpably higher, and review is somewhat slower (still better than the Kodak on all counts). None of these is a deal-killer for most people - who would rather buy storage based on the price of hard drives (i.e., microdrives) than on the cartel pricing of CF solid state cards. One really nice feature of the D2x is that because it can be used in Mass-Storage mode, you can plug the camera right into a 5G Ipod and back up all of your pictures fairly effortlessly. You get about 140 RAW+6mp JPEGs per 4Gb card/drive.
9. White balancing: Kodak. Digital is great in that it has dumbed down users so much that they worry about individually white-balancing shots. With film, you really only have two choices: daylight and tungsten. That's all you needed, since electronic flash is corrected to daylight temperature. Some people use 80A and 81A filters with color to fix color at sunrise/sunset, but that is really the extent of it. Digital, on the other hand, gives you a panoply of microscopically-different white balance settings and cameras that always seem capable of choosing the wrong one.
Nikon gives you white balancing using a combination of TTL color metering and a "third eye" (I like to say, "bindi") that acts as a control for ambient light. The Nikon automatic system generally works well, but it is often flummoxed by situations where there are two different light sources: an incandescent light overhead while you are shooting daylight or flash in front of the camera will cause the system to pick the wrong balance. Wander outside of the automatic mode or the various presets (tungsten, daylight, flash, etc.)* and you either need to know (a) the color temperature or (b) be able to shoot a grey or white card (full frame).
*Nikon gives you fine-tuning on all of its broad categories through its menu system, but these are expressed in +/- integer settings, not exactly intuitive.
Kodak has a different tactic - you can choose "A," which is TTL or various presets (flash, daylight, etc) which have subcategories (warm, normal, cool). Alternatively, you can take a test shot and use an eyedropper on-screen to pick a neutral-colored object. This creates a calibration for subsequent shots, which you can save to the CF card and reload as necessary. This is a more sensible solution, because it does not require you to stop what you are doing and shoot an entire frame of grey card. You just need to make sure there is a white or neutral grey object in the shot.
9. External Displays: Draw. Both cameras have pretty complete information on their monochrome LCD panels. The Kodak carries more information but it is entirely devoted to different things the Kodak does with file formats and cards.
10. Menus: Kodak. There are two things that decide this (at least in my view). First, the Kodak has a set of menus written in plain English. They are relatively good in Italian, too. I'm sure they are good in other Western languages as well. Second is the Kodak's use of "Ok" and "Cancel" keys, rather than the strange left=cancel, right=accept on the D2x.
C. Image Processing
1. Imaging efficiency: Kodak (by a mile). This has nothing to do with quantitative aspects of sensors, buffers, files or dynamic range. This is a question of time and motion. You can drop a roll of film off at a lab, and assuming that you have not blown the exposure, you can expect to get good prints that at least show you the potential of what is on the negatives. It has always been my contention that digital often fails a very basic test: can you go from camera to print without heroics?
Nikon and Canon in particular have failed to grasp the concept of making camera firmware that generates good, printable images from the camera more than 90% of the time. The current emphasis on RAW files among digital users is symptomatic of this. In the vast majority of cases, it is highly preferable simply to be able to plug a CF card into a Frontier machine and print 200 4x6 proof prints than to sit in front of a computer screen working on a frame-by-frame basis or struggling to come up with a curve that works for some number of those images - especially when you know that they won't all sell (or be usable for your purposes, etc). You can always use RAW files to generate enlargements or tweak that $10,000 shot - but RAW should never have to be a file format of first resort for proofing.
Take a look at the process differences between the D2x and the 14n. Without my making any qualitative statements on the differences (or the host software), you might be able to see for yourself why giving up some speed and convenience on the taking end might be worth it in the context of the whole process. A lot of this is just firmware - but something like Digital Exposure Correction would be a HUGE thing for the D2x because it would raise productivity levels considerably.
Is the efficiency argument a fair one against Nikon? Probably, although practical considerations drove the development of Kodak's Digital Exposure Correction. Although it is hard for people to appreciate, Kodak invented the CCD Bayer-pattern sensor (he was an engineer at Kodak) and all of the basic technology used in digital cameras today. It managed to field its first digital cameras when the Nikon F3 (DCS 100, 1990) and N8008 (DCS 200, 1992) were top-end products - and at that time, there were no cost-effective color LCD displays. Digital Exposure Correction was a way to ensure good images when you couldn't verify them on a screen or look at a histogram. As long as your exposure was in-bounds, you would get something usable. This only became better with time. Nikon was largely spared this, because its first production DSLR (the D1) followed Kodak's by nine years. By that point, the color LCD was a much more viable technology, and there was no need to use software intelligence to fix exposures. Kodak kept on with it, and even to this day, you can shoot with a DCS 760 or a DCS 14n and spend very little time checking histograms.
2. The great Adobe scam: Kodak. It is now notorious how Nikon encrypted the D2x white balance information and forced everyone to rewrite their RAW converters. But only Adobe really stepped up to the plate and used it as an opportunity to milk its customers. Adobe implemented D2x support in Camera Raw 3.x - which forces you to upgrade to Photoshop CS 2.0 ($$$) (and don't start me on Photoshop CS - very poorly integrated in light of the price). The only alternative is to buy Photoshop Elements 3.0 ($99) - but the catch there is that Camera Raw 3.x no longer includes controls such as chromatic aberration control (these have been shifted into the main program), and Elements does not include those controls.
3. The great Apple oversight: Nikon. As of this writing (November 2005) Apple's new Aperture RAW conversion software does not include support for the Kodak FF cameras (DCS 14n, SLR/n and SLR/c), but it does include support for a number of obsolete Kodak DCS medium format backs. Odd.
D. The upshot
I know, you are expecting a final recommendation. I can't honestly say that one of these is clearly better than the other. The Nikon is beefier but it requires a whole lot more attention to what you are doing. The Kodak is built more like spun glass and is slower, but it is capable of getting a good picture with far less effort. Or put another way, it is like the difference between shooting slide film and shooting negative film. But since the Kodak has been discontinued, it looks like the future is all Ektachrome.