Digital Camera Home > Digital Camera Reviews > Kodak Digital Cameras > Kodak DCS Pro 14n Digital SLR

Kodak DCS Pro 14n Digital SLR

Kodak's latest digital SLR brings full-frame, 13.7 megapixel resolution to market for under 5,000.

<<Video, Power, Software :(Previous) | (Next): Reference: Datasheet>>

Page 12:Test Results & Conclusion

Review First Posted: 03/23/2003


Test Results
Given that the particular Pro 14n unit that I tested for this review wasn't a final production model, I didn't run it through all my usual standardized tests. While I did shoot some of my standard test targets, I focused more on shooting some side by side comparisons with the Canon EOS-1Ds, its most natural competitor. (Both are full-frame digital SLRs, with resolutions considerably higher than the now-standard six megapixels, although the 1Ds is a much higher-priced camera.)

I've assembled a number of analysis pages that present the shots I snapped with the Pro 14n and EOS-1Ds, as well as fairly detailed comments on what I saw in them. If you're seriously considering one or the other of these cameras, you'll want to examine these photos and read the commentary on them in detail. To jump to these pages, see the temporary picture-index page, or follow the links below:

Following is a summary of my comments on the analysis pages, as well as some general observations about the two cameras:

Resolution
There's no question about it, the 14n shows the highest resolution I've yet found in a digicam. Besides its very high pixel count though, some of its exceptional sharpness derives from Kodak's decision to leave out the antialiasing (lowpass) filter. This had the (not unexpected) effect of producing artifacts such as jaggies and color aliasing in areas of very high detail. These effects were quite evident in the shots I snapped using my Nikkor 105mm f/2.8 Micro (Nikon parlance form"macro") lens. Given the tiny dimensions of the pixels relative to conventional film grain structures, it's very likely that the use of less-sharp lenses could avoid these artifacts. - But resolution is one of the primary reasons a photographer would purchase a 14n in the first place.

Kodak does offer special anti-moire processing in its Photo Desk software, but I found it only partially successful at removing color "twinkles" around fine, high-contrast detail. (See the "Silos" shot analysis page.) I personally would much prefer to have an an antialias filter and give up some of the cutting edge of sharpness, if it would mean losing the camera's tendency to produce color artifacts. Others may well choose differently though: I'm reminded of the response of many of my readers to the Sigma SD-9 SLR, which also lacked an antialias filter. I personally found that camera's jaggies and tendency toward luminance moire unacceptable, but many people don't seem to be bothered by it, instead exclaiming over the unusual sharpness of its images. I strongly suspect that the 14n will find a similar response among prospective owners.

In contrast to the 14n, the 1Ds pretty much never produced a moire or other aliasing, but at the cost of somewhat softer images with slightly lower absolute resolution. Not to overemphasize the point, but this is going to be an area of personal preference: Examine the test photos closely and make up your own mind.

My comparison of the 14n and 1Ds did emphatically underscore the multidimensional nature of "resolution" though. While the 14n would consistently out-resolve the 1Ds when dealing with high contrast subject matter (res target elements, tree limbs against the sky, etc.), the 1Ds just as consistently blew it away when dealing with low contrast subjects (interior detail in the clumps of pine needles of the Silo shot, detail in Marti's hair in the studio shot, etc.). This is a result of the 14n's sensor noise issue, and the software legerdemain used to cope with it. - See my comments on image noise below.

Color
Very good to excellent, although I felt it sometimes was more "pleasing" than accurate. Looking at the results from the 14n and 1Ds side by side, I felt that neither camera really nailed the color on all (or even most) of the shots, but found the 14n's color bobbles less objectionable. The 1Ds tended to produce images with a slightly greenish color cast under most circumstances, while the 14n tended toward slight oversaturation, and blues that were a bit deeper than the original subject. (A good example of this is the sky in the Silos shot - The actual sky color that day was somewhere about halfway in between the ways that the two cameras rendered it. The 14n's sky blue was too deep and had too much magenta in it, while that of the 1Ds was too pale and had too much green. If you literally averaged the two colors, you'd end up with the sky as it actually appeared that day.

For a professional camera like the 14n and 1Ds, color rendition is possibly less important than it is in consumer-level cameras, simply because pros are much more likely to employ some sort of a color-managed workflow. Both cameras seemed very well-behaved (that is, consistent) in their color handling, and both support color spaces with wide gamuts that are well suited to use in color-managed environments.

Exposure and Tonality
I didn't challenge the 14n with any especially difficult subjects, in terms of tonal range and contrast. (This is the function of my standardized, horribly-lit "outdoor portrait" shot, which deliberately uses terribly harsh direct sunlight to evaluate how well cameras manage to hold onto detail in very strong highlights.) That said, I've observed that Kodak's RAW file format (and the associated ERI JPEG format) preserves noticeably more highlight detail in overexposed areas than any does any other RAW format I've seen. The 14n did seem to offer a really unusual ability to retrieve data from overexposed highlights, via the Photo Desk software.

Tonality on the 14n generally seemed to be quite good. Its native contrast level struck me as about right for general shooting, and its "Portrait Look" setting further reduces contrast to preserve more subtle tonal gradations. In comparison, I felt that the default tone curve of the 1Ds was somewhat harsh, particularly as evidenced in the studio portrait test. Before you write off the 1Ds for portraiture though, it must be pointed out that Canon's host software for it gives you an unprecedented ability to create your own tone curve, embodying whatever characteristics you choose. (Albeit via an excessively tedious user interface. Still, most users would only need to create one or two custom tone curves to meet their shooting needs with the 1Ds, and the software/camera combination really does give you complete control.)

Exposure metering on the 14n was generally pretty accurate in the limited tests I performed, although none of the subjects I shot (apart from the Las Vegas night scenes) presented much of a metering challenge. I really didn't like the limitation to 1/2 EV exposure steps though. Half-EV steps are really too coarse for digicams, IMHO. (The too-large exposure steps are somewhat mitigated by the fact that the combination of the DCR format's excellent highlight headroom and Photo Desk's post-shot exposure correction capability make it easy to bring back "lost" highlights. I'd much rather have the control on the camera to avoid losing the highlights in the first place though.)

Image Noise
Image noise in the 14n is a little hard to discuss in objective, quantifiable terms, simply because so much of the camera's performance in this area is a function of how the images are post-processed in the Photo Desk software application. Furthermore, it's very much up to the photographer to choose how he/she wants to make the tradeoff between image noise and detail in areas of low subject contrast.

The issue has to do with how short-exposure noise reduction algorithms work in general, and how those employed by Kodak work in particular. The approach involved is to look at the contrast between picture elements in a given area, and make an assessment as to whether the inter-pixel variations are due to noise or subject detail. The general idea is that variations below a certain magnitude, and involving areas smaller than a certain minimum are considered to be noise. If the camera thinks its seeing noise in an area where it believes there isn't any significant subject detail, it "flattens" the image, reducing the inter-pixel variations.

If the scene at that point consisted of a flat patch of color or tone (a blank wall, MacBeth(tm) target swatch, etc.), this approach works very well: It reduces the image noise without any apparent deleterious effects. On the other hand, of there's a lot going on in the subject at that point (res target patterns, lots of fine, contrasty subject detail), chances are the noise won't be noticeable anyway, so leaving the image alone lets the viewer see the subject detail without them being aware of the image noise overlaid on it.

The problem comes when you have significant subject detail with relatively low contrast between scene elements. In such situations, depending on where the thresholds are set for discriminating noise, the noise-reduction processing can turn into detail-reducing processing. This is what happened with the 14n in dealing with the detail in Marti's hair in the studio portrait shot and with the inner detail in the clumps of pine needles on the Silos shot.

You can also see the 14n's noise processing "backing off" as it gets to an area of strong subject contrast, by looking around the edges of the color swatches on the MacBeth chart in the Davebox shots. There, you can see "fuzz" around the edges of the color swatches, which is the underlying sensor noise showing through. What happened is that the high-contrast edge with the black surround of the color swatch told the camera that there was strong subject detail in the area, so it needed to back off on the noise processing to preserve the subject information.

Properly setting the parameters for this sort of noise reduction processing is a ticklish and time-consuming business, as it's a matter of constant tradeoffs and experimentation. This is evidently what Kodak has been doing leading up to their shipping announcement for the 14n, and the process is apparently still continuing, based on what I've heard from other reviewers about more recent firmware releases.

The problem is, once there's noise in an image, there really isn't any way to get it out, other than having some a priori knowledge about the structure of the scene that's being photographed. (A bit of the old purse-from-a-sow's-ear story.)

While the image noise levels in flat-tinted areas are quite similar between the 14n and the 1Ds, it's obvious that the sensor in the 1Ds has much better noise characteristics than that in the 14n, reducing the need for this sort of tricky trading off of subject detail and image noise.

Having focused so much on the noise characteristics of the 14n so much though, I feel that I'd better not let the 1Ds off the hook totally, or I'll have readers emailing me crying foul, etc. So I'll comment briefly on a frequently observed but (to my knowledge anyway) never explained noise characteristic in images from the 1Ds. I refer to the little single-pixel flecks of black or white that appear in res target images shot with the 1Ds. If the camera is locked down and multiple shots taken, these tiny imperfections will stay put, but the least movement of the camera between exposures will cause them to disappear and others appear in different places. What's going on?

The answer is that this is Canon's bad-pixel substitution algorithm at work, swapping in data from adjacent pixels to take the place of that from pixels which are bad. This sort of thing happens all the time in digicams, it's just that we usually don't get to see it. (And we can only see it on res target shots snapped with the 1Ds because artificial targets like that represent something of a worst-case scenario for basic bad-pixel substitution scheme.

There's nothing too remarkable about there being a significant number of bad pixels on the 1Ds' sensor array: The economics of manufacturing image sensors mean that it'd be prohibitively expensive (and how) to require 100% pixel yield from the sensor manufacturing process. The solution is to come up with clever ways of substituting data from the surrounding area to take the place of missing data from dead sensor elements. A complete discussion of approaches for doing this is way beyond the scope of a simple camera review like this one, but most methods use some variant of a "median" filter (for those interested in looking it up in image-processing texts) to sort through the data and automatically swap-in relevant data for the bad pixel. Where Canon gets caught out on the 1Ds (and other recent cameras of theirs but not, apparently, the 10D) is when a bad pixel happens to be in either a light or dark area, with a highly contrasting region immediately adjacent to it. When this occurs, the algorithm sometimes chooses the wrong data value to substitute, so you end up seeing a light speck in a dark area, or a dark one in a light region. (Look closely at res target images from the 1Ds, and you'll see that you never find these wrong-colored specks in the middle of large areas of similar tone.)

All this just to make the point that, as good as it is, the 1Ds isn't perfect: It shows these bad-pixel-swapping errors, while the 14n doesn't seem to. (Note too though, that while this problem shows up in res target shots, you'll be very hard-pressed to find it in natural scenes, where the subject tone generally varies more gradually.)

Night/ High ISO Shots
Not to put too fine a point on it, "fuggeddabouddit." ("Forget about it," for those not indoctrinated by American TV shows.) While more recent firmware updates apparently help its long-exposure image noise quite a bit, the Pro 14n just doesn't look like a good choice for low-light or high-ISO photography. In normal lighting, its ISO stops at 400 for full-res photos (increasing to 800 for lower-res ones), and exposures at any ISO in light levels of a foot-candle (11 lux, representative of average city street lighting) or so are either very noisy or show very poor tonal gradation, depending on how the noise-reduction parameters are set.

Shutter lag and cycle time/general responsiveness
The 14n doesn't do too bad in the shutter lag department, with a manual-focus lag time of only 128 msec (0.128 seconds). This is much faster than most consumer cameras, but isn't as fast as the best D-SLRs (I've seen lag times as low as 58-59 msec) and the equivalent lag time on the 1Ds is 83 msec. Experienced photographers would probably describe it as "a little sluggish," but I doubt anyone would miss a critical shot as a result of it.

I do have some complaints over the 14n's buffer memory though: It just doesn't feel big enough, and takes way too long to empty. Depending on the shooting mode, the 14n's buffer can hold either 4 or 7 shots before having to wait for the memory card to catch up. Once full, the buffer takes a minimum of 30 seconds to fully clear, even with the fastest memory cards I tested (including Lexar "Write Accelerated" ones), and slower cards can take more than two minutes.

By contrast, the 1Ds can fire off 10 shots without stopping, and even after that point is reached, can continue to snap an additional 7 shots at 2-second intervals before really slowing down.

I'm told that the 14n has a 256 megabyte buffer memory, but based on my experience with the prototype camera, it still felt too small. (NOTE though, that this is an area in which the production models could do much better, as the firmware is tweaked further.)

AF Performance/Flexibility
Another win for the 1Ds I'm afraid, although perhaps to be expected, given the $3K price difference. I don't have any way of measuring AF system performance quantitatively, so I generally refrain from making comments about it unless there's a glaring deficiency or a dramatically better than average "feel" to the camera. The 14n really doesn't fall into either camp, at least based on my relatively limited testing with it. I didn't experience any significant focusing problems with it, and it was overall pretty workmanlike in its operation. Comparatively speaking though, the AF system in the 1Ds is in another category entirely, with 45 AF sensors vs 5, and much faster, more fluid operation. Probably a non-issue for studio use, but if you're thinking in terms of fast-breaking action, you'll probably need a faster and more capable AF system than what's in the 14n.

Battery Life
Battery life on the 14n was oddly short. I say "oddly," because the power consumption I measured at the external AC adapter terminal really wasn't all that high. (At least in the camera's quiescent capture mode.) The current really spiked whenever I fired the shutter, and the AF system drew a lot of power with certain lenses, but even given those factors, I really expected to see the batteries last longer than they did. - In heavy use, I could drain a battery in a bit over an hour, and even in light use, two hours of run time was unusual. I've since heard from some reviewers that the camera appears to draw current from the batteries even when its power switch is off, so it's possible that I was using partially drained batteries when I thought they were full. Whatever the case, definitely plan on buying an extra battery pack or two along with the camera, if you plan any extended shooting sessions.

Summary
Phew! That came out *way* longer than I'd intended, but somehow there just kept on being more to say. Apologies for the length, but on reviewing it, I really didn't feel I could leave anything out without missing important information.

Here's what I see a the pros and cons of this camera:

Pro:

  • Fantastic resolution
  • Full-frame sensor permits true wide angle D-SLR photography
  • Excellent color
  • Very good tonality
  • Excellent highlight headroom in DCR RAW files
  • ERI (Extend Range Imaging) JPEG format combines power of RAW files with compatibility of JPEG
  • Dual-card (CF and SD) supports flexible workflows
  • Simultaneous dual-mode (RAW/JPEG) file recording is a nice touch
  • Compatibility with Nikon lenses is a plus
  • Fully field-upgradeable firmware is a big plus
  • Much lower cost than the competition (Canon EOS-1Ds)
  • All-magnesium alloy frame should be very rugged

Con:

  • Noisy sensor forces tradeoff of low-contrast subject detail
  • Lack of antialiasing filter results in artifacts and color aliasing
  • Poor high-ISO performance
  • Poor long-exposure performance
  • Too-short buffer length (max 7 frames, only 4 in JPEG mode)
  • Slow buffer-empty times
  • Short battery life
  • 0.5 EV exposure steps are too coarse
  • Unsophisticated AF system
  • Despite magnesium frame, camera feels more prosumer than pro.



Conclusion

Free Photo Lessons

Check out the Free Photo School program for lessons and tips on improving your photographs!
Learn how to take stunning photos with simple pro lighting tips, in our free Photo School area!
The 14n is a remarkable achievement in that it brings to market the highest resolution yet for a 35mm-format D-SLR, at the same time offering full-frame imaging for true wide-angle digital photography using existing lenses. It also does so at a price point that's dramatically lower ($3,000 as of this writing in late March, 2003) than its only serious competition, the Canon EOS-1Ds. It delivers excellent color, and an unusually flexible post-exposure workflow, thanks to the capabilities of Kodak's DCR "raw" file format. Unfortunately, it also has a number of limitations, many of which appear to stem from overly high noise levels in the sensor itself. No matter how much tweaking Kodak may do with their image processing algorithms, they're really at the mercy of their sensor supplier to deliver chips with better noise performance.

I think the 14n will find a lot of application in studio photography environments, where there's plenty of light available (strobes), and where its relatively sort buffer length may not be seen as a limitation. It's sad though, that it'll probably miss a lot of applications that it could have found if its sensor performance had been more on a par with that of its competition. If you need full-frame imaging, high resolution and very nice-looking color but have to stay within a budget, the 14n could be just the camera for you. If cost isn't an object, the Canon EOS-1Ds is a lot more camera (with a little less resolution) for a lot more money.

Your purchases help support this site!
Did this information help you decide between these two cameras? Support our site by buying through our preferred resellers! Your purchase will help keep more reviews like this coming!

NOTE: Neither Ritz Camera nor State Street Direct charges a fee to get on their waiting list for these cameras, so there's no reason not to get on both waiting lists to increase your chances of getting a camera sooner.

Kodak DCS Pro 14n

Click to order from Ritzcamera.com
Canon EOS-1Ds

Click to order from Ritzcamera.com

Click to order from State Street Direct

 

Reader Comments! --> Visit our discussion forum for the Kodak DCS Pro 14n Digital SLR!



<<Video, Power, Software | Reference: Datasheet>>

Follow Imaging Resource:

Photo of the Day from Daily Digital Photo Today’s Photo of the Day
Enter your photo