When you're choosing a camera sensor, you'll need to consider a couple of things. First, you need to decide
between a Charge-Coupled Device sensor (CCD) or a Complementary Metal-Oxide-Semiconductor sensor (CMOS).
CCD sensors are cheaper to manufacture than are CMOS, but that's hardly the issue here.
Lower Noise Levels
In theory, there's no difference in sensor performance between these two technologies, but CMOS has one huge
advantage: it can incorporate signal processing that's not possible on a CCD sensor. The best advantage for the
camera designer comes from incorporating all the analogue parts of the camera on the same chip, thereby
simplifying the camera design. This makes it possible to lower the noise level (video noise, not acoustic)
considerably, making the pictures more restful and with a considerably higher exposure range. For the user,
the advantage comes from lower cost, and the increased exposure range which allows for the capture of a greater
contrast range. The sensitivity should hardly be changed, since the photo-sites are still based on the
photon-conversion efficiency of silicon.
The advantages of CMOS come into their own when the camera has a single sensor.
Camera sensors: single or 3?
All electronic cameras need to generate three images, R G and B.
In a 3-sensor camera, the light path from the lens to the sensor incorporates a dichroic block, to
separate the R G and B images onto the 3 sensors. The assembly of the block and sensors forms one component
in the camera, the sensors being glued in place to guarantee that they don't move around in use. The
manufacture of this assembly forms a considerable proportion of the cost of the camera, although all broadcast,
professional, and high-end consumer cameras are made this way. The sensors can be CCD or CMOS, although most
are CCD. The resolution of the camera is defined, or rather limited, by the pixel dimensions of the sensors,
the three sensors are always of the same pixel dimensions as each other, and are ideally not less than the
pixel dimensions of the video format. There is a trick that can be played during manufacture, called
"precision offset", which can deliver up to about 50% extra resolution in either horizontal or vertical
directions (or, exceptionally, in both).
Single-Sensor Cameras
In a single-sensor camera, the sensor is almost always CMOS, so that some processing can be incorporated in
the sensor. The generation of the R G and B images must come from the same sensor, so pixels must be covered
with coloured filters (usually the Bayer pattern), and that means lower sensitivity by up to a stop. However,
the video noise level should be lower because all the analogue processing can be done on-chip.
The major difference between single- and 3-sensor cameras is in the level of spatial aliasing they deliver.
A 3-sensor camera can deliver clean resolution at the pixel count of the sensors (more if precision offset is
used, but with mild aliasing), while a single-sensor camera is almost certain to produce coloured aliasing,
because it is impossible to fit a bi-refringent spatial filter best suited to both the resolution of G and of R
and B.
Higher Resolution
Single sensor cameras are cheaper, but unless the pixel count is at least 50% greater (preferably 100%
greater) than the resolution of the video system, there will always be spatial aliasing in the pictures.
Any camera attempting to mimic a film camera by using film lenses, must be single sensor, and must have the
correct lens-to-image distance of 48mm, but it can make good use of wider apertures than can a 3-sensor camera,
which is usually limited to F/1.4 by the dichroic block and has a lens-to-image distance of 48mm.
Contributed by Alan Roberts