Before Canon could launch its latest range of digital EOS cameras, the Canon EOS-1Ds Mark ll, EOS-1D Mark ll, EOS 20D and Canon 350D, its engineers had to solve the big problem with small pixels, how to deal with the noise that smaller pixels would inevitably create. Canon's latest EOS range features four entirely new imaging sensors, each with a greater number of pixels than its respective predecessor. To squeeze more pixels onto the same sized sensor, the company's engineers had to make the pixels smaller. The problem with smaller pixels is that they are less sensitive to light and greater signal amplification is required to make it readable. This amplification can generate unwanted noise. To control this noise, Canon's engineers came up with an entirely new imaging sensor for each camera. |
350D CMOS sensor - APS-C sized
The APS-C sized sensor on the original Canon EOS 300D already possessed 6.3 million effective pixels. Increasing the pixel count to 8.0 million without increasing the sensor size required a reduction in pixel size that would normally result in an increase in unwanted noise.
350D CMOS sensor - Reduce the signal noise
To combat the problem, Canon devised and developed a new generation of on-pixel, noise-control amplifiers which were first proven on the Canon EOS-1Ds Mark ll, EOS 1D Mark ll and Canon 20D. These new amps are designed not only to drain each pixel of residual charge before exposure, but also to reduce the signal noise that occurs when working with long exposures and exposures in low-light conditions.
350D CMOS sensor - Improving efficiency
The pixels housing each sensitive photodiode had to become smaller in the new high-density sensors, so Canon’s team had no option but to look at making them work more efficiently. Using a process of miniaturisation and electronic-component efficiency, the company's sensor designers reduced both the size of the parts that make up each pixel, as well as their number. By increasing the sensitive area of the photodiode, the team were left with a sensor that was much more sensitive to light, and thus less in need of signal amplification. So even though each pixel site is smaller, its receptive area is proportionally larger than before.
350D CMOS sensor - Sensitivity
Long exposures are read as they happen, with a dark frame subtraction method used to cancel native interference for true colours and smooth tonal rendering. Fixed-pattern and random noise reduction systems have also been improved to produce the cleanest images possible. However, Canon has not restricted its users to short exposures or sensor ISO values of moderate speeds. The new cameras allow fixed time exposures of 30 seconds, as well as an unlimited bulb mode that lets users keep the shutter open for as long as required. Chip sensitivity, too, can be expanded to the equivalent of ISO 1600 on the Canon 350D Digital and ISO 3200 on Canon's EOS-1 series DLSRs.
350D CMOS sensor - Micro lenses
While the micro technologies described above play an enormous part in the image quality these new sensors provide, perhaps the greatest achievement in the conditioning of these highly populated chips is the efficiency with which they use ambient illumination. For smaller pixels to react more effectively to light, they need either more light or the light that exists to be channelled to them in a more direct manner. Canon's engineers have developed a new type of micro lens for each pixel. Larger and more effective than models used in the past, these lenses have a greater wide-angle element, so are better equipped to gather light and direct it straight to the sensitive part of the pixel. As the lenses have less space between them, they are better equipped to gather more light, and allow less light to fall between receptive areas and be wasted.
In effect, these lenses can gather more light in the same way a wide-aperture camera lens can. A larger front element catches a greater section of the information passed in its direction, while the domed aspect of the lenses draws and focuses the light where it matters most. As a result, the micro lenses pick up more of the light channelled to the sensor by the camera’s lens, using it to illuminate and inform the pixels of the sensor. Improving the effective brightness of the received image means the sensor is able to pass information to the processing engine with a lower degree of amplification than would otherwise be the case.
350D CMOS sensor - In the detail
Paradoxically, higher resolutions lead almost inevitably to the confusion of fine detail and edge colour, because the sensor’s resolution exceeds the boundaries where Bayer four-pixel layouts can deal efficiently with the level of detail the sensor can record. Without sensor designer intervention, this results in a distracting misrepresentation of extremely fine detail, 'moiré', as well as a colour fringing on edges where subjects of one colour meet the background colour of another, 'false colour'.
Canon's solution is to fit its cameras with a three-layer, low-pass filter that combats moiré and false colour noise patterns by ordering the incoming light to suit exactly its sampling abilities. The first part of this new filter is an infrared cut-off filter that stops the sensor recording wavelengths beyond the normal visible spectrum. This controls the sensor’s response to reds in general, ensuring they don't 'overexpose' and burn out.
The next layers are a sandwich of a polarizer and single crystal substrates that combine to provide an accurate vertical and horizontal light wave layout to fit the characteristics of the sensor's resolution. This filter organises the light that passes through the lens into a form the high-resolution sensor can take most advantage from, creating the potential for sharper, more detailed images, while at the same time providing clear and defined subject edges that are not influenced by surrounding colours.
350D CMOS sensor - Micro becomes macro
If you were to study any of these cameras' sensors, you would almost certainly be unable to perceive any visible change. You would not be able to see the new micro lenses, the larger photodiodes, the smaller circuitry or the construction of the low-pass filter. To see the difference, you have to compare the images these sensors create with those captured by their immediate predecessors.
Canon's principle aim with the introduction of these new cameras is to increase the amount of information saved to file. The purity of this information depends on the balance of signal and noise, and with these new sensors, Canon has broken the relationship between pitch and interference. This was achieved by developing methods to control exactly how the pixels react to light, and to determine how those pixels receive their light in the first place.
350D CMOS sensor - The gap widens
The pixel race will continue to drive increases in resolution. As this happens, the importance of noise avoidance technologies also increases to ensure that increases in resolution improve, rather than degrade, image quality. Manufacturers relying solely on software and firmware to deal with the greater noise generated by smaller pixels in higher density sensors will struggle, while brands purchasing off-the-shelf sensors have little say in how those components are designed.
As a CMOS designer and manufacturer, Canon is uniquely placed to tailor pixel size, photosite design and sensor componentry to suit each new camera design and respective target market. The inherent barriers to entry of such a position means Canon's leadership in D-SLR is likely to continue for several years. "As a photography and imaging company, it has always been Canon's intention to control the entire light path, from lens through to printed image," says Mr Fukuchi, Senior General Manager, Camera Products Management Division, Canon Consumer Imaging Europe "Taming the power of CMOS has been critical to our success".