Digital Camera Data
Publication date: November 1998 Last modified 03-Dec-2011.
CCDs decoded
Pretty much all currently available digital still cameras use one or more Charge Coupled Devices (CCDs) to create their images. Complementary Metal Oxide Semiconductor (CMOS) sensors are quite popular in cheapo digitals and a few expensive exotics as well, but once you look beyond the quantum level, the two sensor types work in much the same way. All consumer digital cameras have only a single CCD - some of the big boys have three, not unlike top-flight video cameras. What’s the difference?
An ordinary colour CCD, as used in consumer camcorders, has a resolution of 600 by 450 pixels. This is 270,000 pixels total. But the CCD is not, in and of itself, a colour sensitive device. All the litle cells on the CCD respond to is light intensity - they don't care what colour it is. To make a CCD work in colour, the cells are tinted with a dye layer that only lets them respond to light of one colour. A plain vanilla 270,000 pixel colour CCD thus has 90,000 pixels for each colour, arranged in horizontal or vertical stripes of the different colours.
Yet this CCD, which can only produce a 346 by 260 pixel image in any of the primary video colours, produces a 600 by 450 colour image, which isn’t composed of interwoven stripes of colour. It does this by means of interpolation - figuring out what colour each pixel should be based on what it knows about the one colour that pixel really detects and the colours really detected by the pixels around it. This system works, but it is of necessity not as accurate as a full colour CCD, where every pixel actually detects red, green and blue.
Unfortunately, full colour CCDs don’t quite exist. Foveon have managed to make one, but as I write this there's no camera on the market that uses it yet.
Fortunately, there are several clever techniques used to improve image quality. The dumbest of these is simply to make the CCD bigger and/or denser, so you’ve got more pixels to play with. This is all very well, but as soon as you stop using the cheap mass-produced video camera CCDs and decide to go for a better unit, the loss of economies of scale spikes the price up notably; this is why PC webcams and still image cameras with similar resolution are so cheap, but "proper" digital cameras are so much more expensive. Add to that the fact that larger CCDs have exponentially higher manufacturing failure rates, and it rapidly becomes apparent that super-resolution single chip designs, while technically possible, rapidly become very expensive.
A more lateral way of boosting CCD resolution is to vary the ratio of filter colours, so there’s more green and less red and blue. Since green comprises 59% of average luminance, because our eyes are more sensitive to it, it makes sense to give green a larger share of the CCD. Every decent digital camera these days does this. Most of them use a 2:1:1 ratio of green to red to blue pixels. This makes the interpolated picture about as good as it’s going to get, but fine detail can still be spoiled - interpolation is another word for guessing what the right value should be.
The layout of filter colours in many single-CCD cameras. Note that there are as many green lines as there are red and blue lines put together.
A more sophisticated way of doing the same thing.
Canon's approach to the problem, using a grid of cyan, yellow, green and magenta filters. The interpolation process is more complex this way, but it gives a more accurate image.
Once you’ve got your pixel colours optimal, the next step is, or at least was, to increase the number of CCDs. Going to multiple CCDs allows you to use low resolution (read: cheap) CCDs, yet achieve higher output resolution - the incoming light is split between CCDs, but they can be treated as if they’re interleaved like a single chip design, and the same sort of interpolation applied. Pixel colour weighting can be used in multi-chip cameras, too; the old Agfa ActionCam/Minolta RD-175 used two offset green CCDs and a red/blue combo chip. Nowadays, multi-chip still cameras are relics; eveybody's cameras use single-chip sensors, because improved technology's made very high resolution sensors far easier to manufacture, and single sensors make the optics much simpler.
For ultra-high resolutions, though, you have to abandon the ability to grab the whole scene in an instant. The less extreme way of grabbing high resolution colour is the three-exposure strategy, familiar to anyone who’s got a three-pass flatbed scanner, or who got into video digitising when a monochrome camera and a spinning colour filter wheel was the state of the art. The idea is simple - do away with the tiny filters on the pixels and use the whole CCD for red, then for blue, then for green, using a big filter for each colour. Three shot studio cameras can work with conventional flashes (they have to fire them three times, of course) and can thus take pictures quite quickly, though not of moving objects.
Want more resolution per dollar? Can do.
A CCD that genuinely captures a 6000 pixel square image costs a fortune, even for a monochrome version. But a trilinear CCD array such as is used in a single pass flatbed scanner can capture a 6000 dot wide colour image one line at a time for a lot less. In digital photography, this is done by putting a scanner array in the back of a large-format studio camera. Unfortunately, scanning-back cameras, as they’re called, can take minutes to photograph a scene, which makes them utterly useless for action photography or subjects that don’t like being very brightly lit for that period of time. Conventional flashes are no good.
Ingenious design can allow humans some input into the interpolation process, which scanning back cameras still have to deal with. Leaf’s Catchlight studio camera has four different coloured filters in a random pattern on the scan head. The user can choose from four interpolation algorithms and see which one makes the shot look best.
If you don’t need colour (hello, Artistic Photographers!), all of the digital studio cameras can accommodate you, and often with better resolution than they can manage in colour.
Dicomed’s BigShot 4000, built on a Hasselblad body, might provide a glimpse of the future - it’s a three-pass camera that uses a superfast liquid crystal filter (removable for monochrome) to take three 4000 by 4000 exposures fast enough for use with a single flash. It sells for a mere $US54,900, but the technology is likely to trickle down.
Photographic terms
Here's a quick glossary of the basic photographic terms you need to understand to make sense of digital and film camera specs. Read it when nobody’s looking so you’ll look more knowledgeable when they are.
Aperture: The size of the hole that lets the light through the lens and into the camera. The larger the aperture, the more light gets through.
F stops: The F stop or f-number of a lens is the focal length divided by the aperture, and it indicates how much light is actually getting into the camera. The brightness of the image on the film - or CCD - is inversely proportional to the f-number squared.
Film speed: See Sensitivity.
Focal length: Lenses are referred to as being of x many millimetres. A 60mm lens, for instance, produces an image of a distant object on a camera’s film or CCD that’s the same size as the image that would be projected by a pinhole 60mm from the film or CCD. Longer focal lengths give more magnification, but let less light into the camera.
All but the most expensive digital cameras have image acquisition areas which are smaller than the area of a 35mm film frame, and often MUCH smaller. This means that the "consumer" models’ lens focal lengths are equivalent to larger focal lengths for 35mm cameras - the 7mm focal length of a digital camera is equivalent in field of view to a 38mm lens on a 35mm camera.
Pro digital cameras with 35mm fronts also generally have CCDs smaller than 35mm film. They still use 35mm film camera lenses, which still throw a 35mm film frame sized image on the back of the camera, but that image spills over the edge of the small sensor - it only sees the middle of the image. This increases the effective focal length of lenses; a 50mm lens on a Minolta RD-175 will have the same field of view as a 100mm lens on a regular 35mm camera.
Sensitivity:
Known to film photographers as "film speed", this is how sensitive a given film or digital camera CCD is to light. The International Standards Organisation has laid down a standard measure, the ISO number, so film is described as, for example, "ISO 100". Digital cameras typically have only medium sensitivity, and give poor results in low light. Film cameras have the same problem, to a lesser extent; the higher the film speed, the grainier the film becomes. Low speed films capture fine detail better.Shutter speed: How long light is let into the camera to form an image. The slower the shutter speed, the more light gets in, but slower settings also make moving objects more blurred. Digital cameras don’t need to have a physical shutter, though some do.
