Canon EOS D30 Digital SLRCanon's first digital SLR packs 3 megapixels of CMOS sensor into a speedy, compact body! (Smallest/lightest digital SLR as of August, 2000)
<<Internal Design(!) :(Previous) | (Next): Viewfinder>>
Page 4:CMOS versus CCD & what's it all mean?Review First Posted: 8/27/2000
CMOS versus CCD & what's it all mean?
Earlier in this review, we promised readers a look at the CMOS image sensor used in Canon's EOS D30 digital camera. This is certainly not a standard section in our reviews, but we felt that Canon's choice of CMOS was very significant! To understand what CMOS sensor technology can bring to a digital camera, first of all you need some understanding of how CCD and CMOS sensors work, and what they do differently.
CCD, or Charge-Coupled Device image sensors, were invented at the end of the 1960s by scientists at Bell Labs, and were originally conceived not as a method of capturing photographic images, but as a way of storing computer data. Obviously this idea didn't catch on; today we instead have RAM (Random Access Memory) chips in our computers which are, ironically enough, manufactured using the CMOS process.
Where CCDs did catch on, however was recording images by 1975 CCDs were appearing in television cameras and flatbed scanners. The mid 80s saw CCDs appearing in the first "filmless" still cameras CCDs rapidly attained great image quality, but they weren't perfect. Perhaps most significantly, CCDs required a manufacturing process which was different to that used for manufacturing other computer chips such as processors and RAM. This means that specialized CCD fabs have to be constructed, and they cannot be used for making other components, making CCDs inherently more expensive.
Interline Transfer CCDs consist of many MOS (Metal Oxide Semiconductor) capacitors arranged in a pattern, usually in a square grid, which can capture and convert light photons to electrical charge, storing this charge before transferring it for processing by supporting chips. To record color information, colored filters are placed over each individual light receptor making it sensitive to only one light color (generally, Red, Green and Blue filters are used, but this is not always the case). This gives a value for one color at each pixel, and the surrounding pixels can provide eight more values, four each of the two remaining colors from which they may be interpolated for our original pixel.
After the exposure is complete, the charge is transferred row by row into a read-out register, and from there to an output amplifier, analog/digital converters and on for processing. This row-by-row processing of the CCD's light "data" is where the sensor gets the term "Charge-Coupled" in its name. One row of information is transferred to the read-out register, and the rows behind it are each shifted one row closer to the register. After being "read out", the charge is released and the register is empty again for the next charge. Repeat the process a number of times, and eventually you read out the entire contents of the CCD sensor. (Think of a bucket brigade, moving water from point A to point B by pouring it from one bucket into the next...)
A number of disadvantages to this approach to sensor design now become apparent, in addition to the already mentioned cost. For one thing, the entire contents of the CCD must be read out, even if you're only interested in a small part thereof (for example, when using the digital zooms that are all the vogue in digital cameras, you have no interest in a large part of the sensor's data, so why take the time to read it out?) There are also a number of supporting chips required for the CCD sensor, each of which adds to the complexity and size of the camera design, increasing cost and power consumption. CCDs also suffer from blooming (where charge "leaks" from one light receptor into surrounding ones), "fading" (a loss of charge as it is passed along the chain before being read out), and smearing (where the image quality can be adversely affected by light arriving during the read-out process, leaving streaks behind bright scene areas).
There's also the issue of speed. The step by step process used in a CCD is not exactly conducive to very high speed, and for just this reason a second type of CCD exists. The Frame Interline Transfer CCD features a read-out register as large as the light receptor area is, allowing the entire contents of the CCD to be read out in one pass. This, though, adds significantly to the area of silicon required, and hence to the cost of the CCD.
This is where CMOS image sensors step in. CMOS, or Complementary Metal Oxide Semiconductor, is actually a generic term for the process used to create these image sensors, along with numerous other semiconductor items such as computer RAM, processors such as those from Intel and other manufacturers, and much more. CMOS image sensors can be made in the same fabs as these other items, with the same equipment. This technology is, of necessity, very advanced with the amount of competition in processor and other markets contributing to new techniques in CMOS fabrication. Add to this that there is a very significant economy of scale, when your fab can make not only CMOS image sensors, but other devices as well, and you find that CMOS image sensors are much cheaper to make than CCDs.
This cost advantage is even more significant when you consider the way a CMOS sensor works. The Active Pixel CMOS image sensors used in digital imaging are very similar to a CCD sensor, but with one major difference supporting circuitry is actually located alongside each light receptor, allowing noise at each pixel to be canceled out at the site. Further to this, other processes can be integrated right into the CMOS image sensor chip, eliminating the need for extra chips things such as analog/digital conversion, white balancing, and more can be built into the CMOS sensor. This reduces cost of supporting circuitry required, as well as camera complexity, and also power consumption, as does the fact that CMOS sensors require a significantly lower voltage than CCD sensors. CMOS sensors themselves also claim lower power consumption than CCD sensors, with one manufacturer claiming their CMOS sensors draw some 10x less power than equivalent CCD sensors.
CMOS sensors have other advantages, as well. For one thing, they can be addressed randomly. If you're only interested in a certain area of the image, you can access it directly and don't need to deal with the unwanted data. Blooming and smearing are also less of a problem with CMOS sensors. CMOS sensors are capable of much higher speeds than their CCD rivals, with one CMOS chip we've heard of capable of running at over 500 frames per second at megapixel resolution.
With these advantages, you'd think CMOS would be a shoe-in to replace CCD in digital cameras, but thus far it has really only impacted the lower end of the market, with CMOS rapidly becoming dominant in the entry level digital cameras and tethered cameras. Why hasn't CMOS taken over at the high end? Well, up until now, image quality has not been on a par with CCD CMOS sensors, with their many amplifiers at each pixel, suffer from so-called "fixed pattern noise". The amplifiers aren't all equal, and this creates a noise pattern across the image. In the D30's CMOS sensor, Canon has tackled this by first taking the image off the CMOS sensor in 10 milliseconds, and then reading just the fixed-pattern noise from the sensor in the following 10 milliseconds. Subtract the second image from the first, and you neatly remove the noise.
There's also the fact that CMOS sensors are generally less sensitive than their CCD counterparts. High end "Full Frame" CCD image sensors have a "fill factor" of 100%, because the whole CCD sensor area is being used for light capture but in a CMOS sensor the fill factor is lower, because the extra circuitry alongside each pixel takes up space. This space can't be used to capture light, and so you lose some of it Two techniques exist to combat this firstly reducing the size of this support circuitry, and secondly the microlens. Reducing the size of the support circuitry is the less ideal of the two methods the smaller you make it, harder the sensor is to manufacture, and the more expensive it becomes. The microlens is considered to be the better answer, then. Essentially, the support circuitry is covered by an opaque metal layer, and a microscopic lens is placed over the entire area of the light receptor and support circuitry, redirecting the light that would otherwise fall on the support circuitry and focusing it on the light receptor.
Canon's EOS D30 is the first high-end digital camera we've seen using CMOS technology, and it is likely that the projected price advantage the camera has in comparison with its nearest rivals (the Fuji FinePix S1 Pro and Nikon D1) is in large part due to the choice of the CMOS image sensor. The image sensor in the EOS D30 is only ever so slightly smaller than those used in these two cameras, and significantly bigger than the sensors used in consumer cameras, as can be seen in the comparison photo above, which shows the CCD sensor from Canon's PowerShot S20 digital camera alongside the CMOS sensor from the EOS D30. The illustration below shows the difference in sizes (to scale) of a consumer CCD, the EOS D30 sensor, the D1/Fuji S1 Pro sensors, an APS film frame, and a standard 35mm frame.
Canon thus far has been fairly closed-mouthed about their CMOS sensor technology,
but have talked about a few details of it. As with other Active-Pixel CMOS sensors,
theirs does in fact have a signal amplifier located at each pixel site. More
intriguing though, is that they also claim to have an A/D (analog to digital)
converter at each individual pixel site as well. If this last is true, then
it must be a very different sort of A/D than is normally used with CCDs, as
those circuits are quite complex and space-consuming. We suspect we'll hear
more details as Canon's patent position is solidified, but it sounds as though
there's been some genuine innovation in Canon's back labs. It's unusual these
days to see a company moving toward vertical integration, developing component
technology in-house rather than farming it out to specialist companies. As the
digicam market continues to evolve, it will be interesting to see whether Canon's
sensor technology will constitute a competitive advantage for them relative
to other manufacturers.
Reader Comments! --> Visit our discussion forum for the Canon EOS D30 Digital SLR!