The CMOS Sensor
Early on in the development process, Kodak settled on CMOS as the sensor technology that would enable them to build the smaller, less expensive camera they envisioned the 14n to be. CMOS sensors are cheaper to manufacture than CCD sensors of a similar size and resolution, while their analog-to-digital conversion and related circuitry is significantly more compact. Given Kodak's direction for the 14n, CMOS was the logical choice. Says Kodak's Steve Noble: "For the manufacturing of this camera, CMOS was just nicer to work with."
But, emphasizes Noble, CMOS' lower cost and smaller circuitry size is only half the story. The image data rolling off a CMOS sensor needs a lot more massaging than the Full Frame-type CCD sensors Kodak was accustomed to working with. Colour shifts from edge to edge and significant shadow noise are among the gremlins Kodak has had to tackle. In addition, Kodak switched from an RGB colour filter array (CFA) composed of layered CMY dyes (two filters combined over each pixel to get R, G or B) to single-layer, pigment-based RGB filters.
In short, a new sensor from top to bottom added up to a wealth of engineering challenges that Kodak's DCS group hadn't faced before, using a sensor technology in which their primary high resolution SLR competitor - Canon - had a several year head-start dating back to the Canon EOS D30.
Kodak didn't have to start from scratch, though. Starting in 2001, they teamed up with FillFactory, a company known in engineering circles as a pioneer of ultra-high speed CMOS sensors. The 14n's sensor is truly a global affair: Developed by Kodak in the US and FillFactory in Belgium, it's fabricated by Tower Semiconductor in Israel.
With an 8.0 micron x 8.0 micron pixel size and an active area of about 24 x 36mm, the 14n's CMOS sensor has more pixels than any other digital SLR today. It's 13.85 million total pixels translate into a final photograph of 13.5 million pixels. Hanging off the sensor's almost 14 million pixels (that's the 14 in 14n) are four 12-bit analog-to-digital converters. Having four such converters allows, among other things, for the camera to fire at 1.7 fps, despite the huge quantity of image data being generated each time the shutter button is pressed. Unlike the CMOS sensor in the EOS-1Ds, the 14n's pixels do not include microlenses. In theory, keeping pixels microlens-free means less risk of sensor-induced chromatic aberration, but reduced light-gathering capability as well.
Introduction to Image Quality
Looked at strictly as a piece of technology, there's little doubt that the DCS Pro 14n's CMOS sensor is way cool. But what about the quality of the photos it captures, after Kodak's pixel massaging is complete?
In the pages ahead you'll find an analysis of 14n detail, noise and colour. To date we've been able to shoot with four different DCS Pro 14n bodies; files from three of them are spread through this report. Despite rumours to the contrary on the Web, 14n image quality has actually been fairly stable since the first editorial review cameras began making the rounds earlier this year, even through several iterations of firmware.
One notable exception is long exposure noise. Only relatively recently did Kodak enable long exposure noise reduction in the 14n's firmware, and even then only for exposures up to 2 seconds. As such, the 1/2 second and longer photos we've included were taken after the long exposure-enabled firmware was released to us. Otherwise, we're confident that the 14n photos presented here are representative of a 14n that you might buy from a Kodak Professional dealer tomorrow.
We've chosen to look at 14n image quality on its own and in comparison to photos captured with Canon's EOS-1Ds. Released late in 2002, the EOS-1Ds is Canon's flagship digital SLR. It produces photos with amazing detail and impressive photographic smoothness. Like the 14n, it features a high-resolution CMOS sensor about the size of a 35mm film frame. It's pixels, however are slightly larger, at 8.9 microns, and there are fewer of them: 1Ds photos contain about 11 million pixels. There are a number of key differences that separate the 1Ds and 14n, one of them being price: at about US$4800-US$5000, the 14n isn't exactly cheap, until you compare it to the cost of a 1Ds with the must-have Capture One DSLR software. That combo tips the scales at about US$8000-8500 in the US.
While the 14n may be less costly than the 1Ds, its files should still look expensive. Part of the reason for comparing the files from the two cameras so closely in the preparation of this report is our belief that the 14n should be able to meet or exceed 1Ds picture quality, even if its lower cost means the 14n isn't as well-appointed as a camera. We didn't set out to prepare a 1Ds-14n image quality comparison, but it has essentially turned out that way.
All photos in this report originated as RAW format DCS Pro 14n or EOS-1Ds files. DCS Photo Desk 3.0 handled the conversion duties for the Kodak camera; File Viewer Utility 1.1.1 and 1.2.1 were used to process files from the Canon camera.
Kodak DCS Photo Desk 3.0
Photo Desk is the only software that officially supports the 14n, so choosing it was easy (the unofficial option is Photoshop Camera RAW, which can process 14n files on the sly, but in our testing produced particularly poor results).
On the Canon side, there are three noteworthy options right now: File Viewer Utility (or third party software that is built around the same Canon conversion engine), Capture One DSLR and Photoshop Camera RAW. We opted to use File Viewer Utility and Capture One DSLR in all internal comparisons, since these are the two EOS-1Ds RAW format conversion tools we use most often.
Canon File Viewer Utility 1.2.1
In real life, we use Capture One DSLR more than File Viewer Utility, but we opted to make File Viewer Utility the exclusive converter for this report so that you would get a sense of what EOS-1Ds colour, tone and noise processing looks like right out of the box, without the aid of extra cost software. But we maintain that if you're considering the purchase of an EOS-1DS, and you intend to shoot RAW format photos, life is just better with Capture One DSLR.
Phase One Capture One DSLR
Photos run through Photo Desk were not sharpened. Of the settings that affect overall colour and tone, only the basic white balance controls were used. All photos have been processed on the default DCS Product Look settings, which we determined was really the only usable 14n colour look included with Photo Desk. The quality of the final image is affected greatly by the Noise Reduction function. Care was taken to find the optimum setting for each photograph, one that balanced noise and/or moire reduction with the retention of key image detail.
In File Viewer Utility, only basic white balance controls were used and no sharpening was applied. Most photos were processed on Color Matrix 4, which is the supposed low contrast, low saturation setting intended for viewing in Adobe RGB, or so indicates Canon in the materials describing the camera's five Color Matrix options. In fact, the EOS-1Ds' Color Matrix 4 is actually the camera's highest contrast setting.
While this level of contrast was suitable for a number of the photos that we happened to shoot for this report, Color Matrix 4's high contrast, which is all but impossible to counteract with a Custom Tone Curve, makes mincemeat of portraits. For that, our testing shows Color Matrix 1 to be the best all round choice when people figure prominently in the photo (though we prefer the contrast, and contrast controls, of Capture One DSLR). So, for the photos in which Color Matrix 4 was too contrasty we opted for Color Matrix 1, which is tuned for sRGB viewing.
All photos were processed out to Adobe RGB, except for Color Matrix 1 EOS-1Ds photos, which were processed out to sRGB. All photos were then saved as TIFFs. Copyright and shooting information was added, and reduced resolution versions prepared, in Photoshop 7.01. Any photos available as full-resolution downloads were saved out of Photoshop as Quality 11 JPEGs. They remain in their original colour space, either Adobe RGB or sRGB. All reduced resolution versions have been converted to sRGB and saved as moderately compressed JPEGs for viewing in your web browser.
The analysis of what's hot and what's not in DCS Pro 14n image quality is based on an examination of its photos on a Sony Artisan 21 inch CRT monitor, as well as 10, 18 and 24 inch wide (equivalent) enlargements printed with a Canon S9000 inkjet. As always, we place the greatest emphasis on printed output. We avoid discussing image quality defects in particular if they appear onscreen but are masked by the printing process, even at significant enlargement sizes.