What’s Grass Valley showcasing at NAB? See page 63>> and at www.tvbeurope.com
Time to move to CMOS imaging?
CMOS is ready as a technology to challenge CCD in broadcast cameras, says Klaus Weber, director of Product Marketing, Cameras, for Grass Valley
SOME MAY find the question posed in the title of this article strange. To be sure, CMOS imagers are widely used in cameras today. It is certainly true that nearly all still cameras and camera phones use CMOS sensors. So too do the latest breed of 35mm equivalent digital cinematography cameras. The thing they all have in common is that they are single chip designs with colour separation on the chip, usually by means of the Beyer pattern. What I propose for your consideration, however, is the high-performance broadcast camera — the system camera we rely on for crystal clear pictures of The X Factor or rugby internationals or music festivals. These have always relied on three discrete imaging channels — red, green, and blue — and we believe they will for some time to come. But today, the latest generation
of CCDs found in system cameras is the last generation of CCDs. Since its invention in 1969, and its use in our broadcast
cameras since 1987, CCD technology has gone through much development, but it has now reached its practical limit. There are a number of
reasons for this, but just one for our discussion: the design of a CCD means that all of its information — the number of electrons in each photosite — is read from a single point. For a
at 330MHz, and that is impractical. We have reached the end of the line with this aspect of the CCD architecture. The alternative is the CMOS
imager. Its construction has an amplifier and output for each photosite, so the bandwidth problems disappear. All ultra motion cameras used in broadcast today have CMOS
While the CCD has been taken about as far as it can go, the CMOS imager is still blazing a trail and will get better in the very near future
full HD picture, that is around 2.2 million values. Multiply that by 50, for 50fps, and you need a bandwidth at the output node of 110MHz.
Super slo-mo cameras are
now the mainstay of sport and other television. To serve 1080p with triple speed cameras means sampling each imager
imagers for this very reason: there is no practical limitation on imager speed. The history of the CMOS
device actually goes back as far as the CCD, but its development as an imaging device is very much more recent, which means we are at a different point in its evolutionary curve. While the
CMOS with full exposure
CCD has been taken about as far as it can go, the CMOS imager is still blazing a trail and will get better and better in the very near future. In 2007 we developed our
own CMOS imager, the Xensium. It contains 2.4 million photosites, for a full 1920x1080 pixel raster. Physically, the target
CMOS with short exposure
is the same size as the CCD, so fits into the same optical path with no impact on lenses. It is used in our three chip multi- format LDK 3000 HD camera. The way that a CMOS device
is fabricated is very different from a CCD. CMOS devices are made very much as other silicon chips are made, which means that other components can be built onto the same substrate. As already noted, there is an amplifier for each photosite. We also incorporate the
analogue-to-digital conversion on the same chip. To reduce the clock rate two 14-bit A/D- converters are used — one for the odd rows and the other for even rows. For an improved FPN (fixed-pattern-noise) behaviour we use a newly- developed technique called double digital sampling (DDS). This quantises the charge on the photosite at the beginning and end of each frame and subtracts the first from the second. This eliminates any bias in the pixel due to residual or leaky charge, and ensures a much more accurate image.
Powered up This integrated design means that overall the chip uses less power than the CCD with its high-speed switching. In a chip, power means heat and our CMOS imager runs several degrees cooler than our CCD imager. In turn, reduced heat