Digicams: the truth
Digital cameras are coming close to the quality expected from traditional photographic cameras, but there are some very good reasons why they may soon come up against a technical and commercial brick wall. It's got nothing to do with the number of pixels crammed on to the CCD, and everything to do with basic optical theory.
Optical lenses are a bodge. There's a list as long as your arm of optical aberrations and inaccuracies that lens designers have to dial out, using multiple lens elements, different glass formulations, refractive indices and complex aspherical profiles. Often, you can't fix one lens aberration without introducing another. It may prove impossible, for example, to gain high resolution in the centre of the image without sacrificing sharpness at the edge.
Worse, people want fast maximum apertures and huge zooming ranges at bargain prices. All lenses are compromises, and whole books have been written on the subject. But while it's bad enough designing good quality lenses for 35mm cameras, it's harder still when you get down to the scale of digicam lenses.
CCD size
Ian Watts, digital imaging expert at Olympus cameras explains the problem: "Optical quality has got to be in the region of three times greater than 35mm to achieve the same effective image quality." Lens resolution is traditionally measured in 'lines per mm'. Ian explains that while a lens with a resolution of 50lpmm is regarded as quite satisfactory in a 35mm camera, a digital camera lens needs to be nearer to 150lpmm - and still with the wide apertures, long zooming ranges and highly-controlled aberrations we all expect. This is because the CCD in a digital camera is a fraction of the size of the 36mm x 24mm negative in a 35mm camera. The image formed on the CCD has to be enlarged far more than a 35mm negative for a same-sized print. It's all very well building CCDs that can record this kind of detail, but you'll also need lenses that can resolve it in the first place.
So how big, exactly, is a digicam's CCD? The manufacturers are oddly evasive here. Check the specs in the back of a 3.3-megapixel digicam's manual and you'll find the measurement is 1/1.8 inches. What's evasive about that? First, they're measuring the diagonal rather than the long edge of the CCD. Second, can you get any more arcane than an imperial measure expressed as the reciprocal of a decimal?
Even knowing that the CCD's diagonal measures 0.56 inches (1/1.8-inch) isn't much of a guide. This measurement relates to the CCD chip, not the size of the imaging area it contains - there's a wide rebate around this, meaning the actual digital 'negative' size is a lot smaller. Around 8mm x 6mm, to be precise.
You can arrive at this figure by a separate route. If an 8mm lens on a digicam equates to a 38mm lens on a 35mm camera, simple geometry indicates that if a 35mm negative is 36mm across, the digicam's imaging area must be no wider than 8mm. To produce a 9 x 6-inch print, a 35mm negative has to be enlarged by a factor of six. To produce an 8 x 6-inch print (most have a 4:3 aspect ratio rather than 35mm's 3:2), a digicam image must be enlarged by a factor of 25. Which is why digicam lenses have to be capable of such high resolutions.
Diffraction
But there's a limit to the resolution that lenses can produce. It doesn't matter how well designed a lens is, because ultimately its performance is limited by the effects of diffraction.
All light is diffracted as it passes through a hole, the hole in this case being the aperture in the lens. Instead of being focussed as a single point, diffraction effects mean light is focussed as a tiny disc. The bigger the disc, the greater the loss of sharpness due to diffraction.
The point is, though, that both images will need to be enlarged for viewing, and the 35mm image much less so. Engineers at the renowned camera maker Leica long ago concluded that a resolution of 8lpmm on the final print would be perceived as being sharp to the average viewer, and some simple maths reveals that you could enlarge a 35mm negative with 200lpmm resolution up to a theoretical maximum of 36 x 24 inches.
In reality, the maximum enlargement size is likely to be a lot lower, because of the limitations on lens design and the resolving power of photographic material. Apply the same calculations to an 8 x 6mm digicam image, and you get a print measuring approximately 8 x 6 inches before diffraction effects start to become significant. Not very big, is it? Remember that this isn't an economic problem, an engineering issue or a technological hurdle. This is basic optical physics.
One solution would be to produce digicam lenses designed to work at much wider apertures. This, though, would mean increasingly esoteric and expensive lens designs - plus incredibly fast shutters so that lenses didn't have to stop down to smaller apertures in bright outdoor lighting. It's not going to happen. The solution? Bigger CCDs. Not necessarily denser pixel arrays, just bigger imaging areas. The problem here is cost. Ian Watts again: "The cost of CCDs the size of 35mm negatives would be prohibitive. It's cheaper to squeeze more pixels on to a smaller chip than it is to make larger chips."
The only digital cameras with CCDs approaching 35mm imaging areas are professional models costing thousands, not hundreds of pounds. The new Nikon D1 models have CCDs measure 24 x 16mm - half the area of 35mm - and are set to cost £3,500 for the body only.
Depth of field
Diffraction-limited resolution isn't the only problem faced by the current crop of digicams. Keen photographers will know the benefits of creatively juggling depth of field. In 35mm, it can be difficult getting a whole scene sharp where it contains both near and distant objects. Just as often, though, throwing the background out of focus can enhance your main subject. Try doing that with a digital camera. They simply have far too much depth of field. This is the distance in front of and behind your main subject (the subject the lens is focussed on) which still looks sharp. Slightly out of focus points are reproduced as a disc of light on the film (or the CCD) called the 'circle of confusion'. As with diffraction, if this disc is small enough it still looks like a point of light.
This circle of confusion value is taken to be 0.03mm for most general-purpose 35mm photography and typical levels of enlargement. Given that the enlargement factor for digicam images is around five times greater, this gives us a maximum circle of confusion size of 0.006mm. Now depth of field calculations can be very complex. But the simplest is the one that gives us a lens's 'hyperfocal' distance. This is the minimum focussing distance we can set where objects at infinity are still sharp.
At the wider end of their zooming ranges and everyday angles of view, digicam lenses hardly need to be focussed at all. It's why fixed-focus cameras like FujiFilm's FinePix 2300 are so effective. Digital cameras only need focussing mechanisms at wider apertures and longer focal lengths, and it's very difficult to achieve differential focus effects without resorting to the far end of the zooming range and hoping that the light levels are low enough for you to use wide apertures. Digicam makers can continue cramming more and more pixels into their CCDs, but until they start making the CCDs physically larger, they (and we) face diminishing returns. The numbers may continue to grow, but it's a smokescreen that disguises a far more pressing problem.
Tonal range
There's another area where digital cameras have had an easy ride, and that's in their ability - or maybe their lack of it - to record wide tonal ranges. How do they compare with conventional, silver halide-based photographic film? The trouble is that no one really knows, as Ian Watts explains: "CCD data is not generally available to the consumer because the expected lifespan of a digital camera in this market is much less than that of a conventional photographic film." It's likely that by the time this kind of information filters down to the end-user, the CCDs it relates to will be out of date.
Conventional films have a so-called 'characteristic curve', which indicates their response to light. The bottom of the curve, or the 'toe', shows how they react to the deepest shadow detail, while the top of the curve, the 'shoulder', shows how they react to the brightest highlights. In between is a straight line section that reflects a film's essentially linear reaction to increasing light intensities.
But it's what happens in the toe and shoulder areas that interests many photographers. In these areas, the curve flattens out, and it's often possible to drag extra image information out of dark shadows and bright highlights at the expense, admittedly, of image quality. CCDs, in our experience, don't react this way. When highlights block out, they're gone for good. When shadows fill in, they're filled in. Sometimes you can tweak the levels to recover highlight and shadow detail, but this introduces another problem.
CCDs produce 256 discreet shades, or 8-bit images, for each of their Red, Green and Blue colour channels. This isn't a problem with images that require little or no manipulation, where 8-bits per channel is enough to give the impression of a continuous tone image. But once you start expanding the shadow, midtone or highlight areas, the pixel values start to get a little bit too far apart, resulting in prominent banding or dithering. You could argue that the same thing will happen when you scan (digitise) a conventional photo. But with certain film scanners, like Canon's FS2710, for example, you can alter the speed of the scanning head to effectively increase or reduce the exposure at the scanning stage.
In addition, most scanners can produce greater bit-depths than digital cameras, offering up to 12, 14 or 16-bits per channel. This gives you much more leeway at the manipulation stage. And you also have the opportunity to rescan a photographic original as many times as you need to in order to achieve the optimum result.
Storage
Which leads us into a final major issue with digital cameras. The better the quality becomes, the bigger the files get. Are you really going to spend £800 on a digital camera and then use it at anything less than its best quality? Do this with the 4-megapixel Olympus E-10 and each picture will be a 2.5MB JPEG. Use FujiFilm's FinePix 6900 Zoom at its top quality 6-megapixel setting and your files will be even bigger.
How many memory cards do you plan to buy? When you go on holiday, how many photos do you expect to take? Do you really want to spend those balmy tropical evenings agonising over which of your saved shots will have to be trashed in order to make room for tomorrow's pictures?
A keen photographer will quite easily shoot a 36-exposure film each and every day. On a seven-day holiday, that's around 250 shots. With a 3-megapixel digicam at a decent quality setting, you're going to be looking at somewhere around 400MB of data. Are you going to buy 400MB of CompactFlash or SmartMedia storage for your holiday? You could take a laptop with you, of course, or make sure you only holiday near internet cafés with broadband connectivity.
Once you've downloaded your images to your hard disk, you delete them from your digicam's memory card. Is that the end of your problems? That depends on whether you like to keep backups. How would you view the prospect of losing your entire photographic collection to a hard disk failure or a corrupted FAT table?
Regular backups of your digital photographs are a good idea. But for this quantity of data you're really looking at CD-writers. More hassle. But of course if you work from halide-based originals rather than digicam images, you've already got a backup, haven't you?
The purpose of this investigation is not to act as devil's advocate for conventional 35mm film technology, but to identify and quantify some of the very serious restrictions that face digital cameras, and to point out that both media have significant advantages. So which type of camera do we use? Both.