More than 30 years have elapsed since Bell Systems USA first announced the use of CCDs (charge-coupled devices) for a solid state camera in 1972. Essentially, a CCD is a silicon integrated circuit of the metal oxide semiconductor type with a basic structure such as that shown in Fig. 1. It comprises an oxide (SiO2) covered silicon substrate upon which is formed an array of closely spaced electrodes.
The term ‘charge coupling’ refers to the method by which signal charges (which can be photoelectrons created from an image falling on the CCD array) can be transferred from under one electrode to the next. This is accomplished by taking the voltage on the second electrode also to a high level, and reducing the voltage on the first electrode, as shown in Fig. 2. By sequentially pulsing electrode voltages between high and low levels, charge signals can be transported down an array of very many electrodes with hardly any loss and very little noise. By taking three electrodes we get one pixel which, typically, has the dimension of 0.009 mm or 9 micron (9 μm) square. The medium-format Kodak Pro Back (2001) had 16 million of these pixels in an area 37 mm square.
By connecting up three closely spaced MOS (metal oxide semiconductor) capacitors into storage elements, and applying a set of three-phase drive pulses φ1, φ2, and φ3 (typically from +5 to +15 volts), charge signals can be stored under every third electrode in each line of the array, as shown in Fig. 2. Under this system each storage element (or pixel if the input signal is optically generated) has every third electrode connected to the same clock voltage; consequently, three separate clock generators are required, as shown in Fig. 2a.
Fig. 1
Fig. 2
Full-frame transfer arrays
It would seem that current developments favour a full-frame transfer array where all the pixels are employed in the recorded image (earlier types of CCD camera used the field-frame array, where half the array was used for viewing). As may be seen from Fig. 3, a mechanical or electronic shutter is used to expose the pixel array to the image formed by an otherwise conventional camera, the CCD array replacing the film, and a battery installed for the power source. As a three-phase system, there are three MOS capacitors to each (vertical) element, and so a square-shaped pixel is formed by appropriately spaced channel stops. When the array is illuminated by an image, photons can pass through the semi-transparent polysilicon electrodes to generate electron-hole pairs within the silicon substrate. In the example shown in Fig. 3, there is a matrix of 5 pixels shown, and the quantity of charge collected is proportional to the local radiation intensity and time allowed for collection, that is, the exposure time. The exposed pixels are then transferred directly to the shift register at the bottom of the array; from here the rest of the story takes place inside the camera, within the computer and the software.
Once a picture has been taken with a monochrome CCD imager, each image point is then clocked out, pixel by pixel, along the shift register to the output amplifier, and then to an eight-bit (28 = 256) analogue to digital converter (ADC) where each pixel-charge value (along with its x, y location) is registered in the form of a binary code. From here the image data are temporarily stored in an eight-bit buffer memory before being sent to a programmable digital signal processor (DSP) where the image is stored with other data in the dynamic random access memory (DRAM). Once the data are in the single-frame DRAM, the camera's hard disk is started and the image data transferred to it. This process takes a few seconds, depending on the number of pixels in the image.
After the image data have been stored on the camera's hard disk, smart-media, compact flash, or PCMCIA card, the digital data can then be downloaded to the computer (usually by USB or Firewire connection), where they are received by the manufacturer's image acquisition software. Finally, the image-processed data are then sent to the computer monitor's digital to analogue converter (DAC), which does the reverse job of the camera's ADC. This conversion is necessary since the monitor only accepts analogue signals (as a stream of rapidly varying voltages).
Fig. 3
The colour image route
Colour is generally introduced by the spatial multiplex system. This is a popular system as it allows for single exposures and virtually makes the camera little different from conventional film types. The technique is to provide an individual red, green, or blue filter over each pixel, as shown by Fig. 4. Known as the Bayer colour filter array (CFA), it exemplifies the typical repeating 2 × 2 matrix that uses two green cells for every single red and blue cell. The reason for the 2:1 ratio of green cells is to provide for improved green sensitivity and so match the human eye.
Known as the ‘adjacent pixel colour interpolation technique’, area array systems ensure that each pixel filter is always adjacent to the other two. If the subject had a uniform red, or blue, colour, the Bayer filter would only pass 25 per cent to the pixel array, if it were green, then 50 per cent. There would be gaps, but camera software examines the strength of the missing colours recorded by adjacent pixels, and makes a qualified evaluation of the two missing colours on a pixel by pixel basis. If, for example, the camera were to record a uniform red object, it would not be just the red-filtered pixels that created a charge signal, but all of the pixels—red, green, and blue. Were this not so, there would be open gaps in the resolution.
All the pixel filters are broadband types, allowing certain amounts of other colours through, i.e. red-green and blue-green; consequently the software algorithm has some signal information on each pixel regardless of the image colour at that point. In the example of the uniform red image, the interpolation algorithm finds little response in the green-, and even less in the blue-filtered pixels, and since the entire matrix in the immediate area has the same response the algorithm calculates a totally red image at this point.
This method of colour interpolation degrades spatial resolution by a factor of √3 and if three pixels are approximately equal to one line pair, we can put spatial resolution of a monochrome CCD imager as:
rmono ≈ 3 px (μm)and, for a colour CCD imager,
rcolour ≈ 3 px √3 (μm)In terms of line pairs per millimetre this becomes:
Rcolour ≈ 1, 000/(3 px √3) (line pairs/mm)Spatial multiplex systems are not without their disadvantages, however, the main problem being that because of the averaging or resampling of the pixel intensities, artefacts can occur in the image of an object that has sharp edges or strong lines. Known as aliasing, such artefacts manifest themselves as stair-stepped jagged edges (‘jaggies’), particularly when the display or printed resolution is too coarse to hide the effect. However, a suitable anti-aliasing technique known as dithering exists in Adobe Photoshop image-processing software.
Fig. 4
CMOS: the image of the future
CMOS(complementary metal oxide semiconductor) image sensors are cheaper and more integrated, with much lower power consumption than CCD technology. They have considerable advantages over CCDs in that CMOS devices' structure and comparative ease of fabrication lend themselves to low-cost enhancements. The main difference between the two sensors is that while the CCD groups the output of the pixels together, through charge coupling, and then row by row presents them to the read-out section, a CMOS sensor can address each pixel individually. Problems associated with CMOS sensors include higher noise levels and limited ISO sensitivity. However, their potential advantages are increasingly being realized through intensive research and development.
The Californian company Foveon developed an alternative sensor, the Foveon X3. Using CMOS, the X3 does without the usual Bayer CFA and so increases colour efficiency by a factor of 2 or 3. It is the first chip to capture red, green, and blue image data at every pixel site in a single exposure. To make this possible, the silicon is prepared in three layers, similar to the dye layers in colour film technology. As white light is formed by red, green, and blue, with approximate wavelengths centred on 0.4, 0.55, and 0.7 μm respectively, the different wavelengths are absorbed at different depths by the doped silicon substrates. The blue is read by the top substrate, located just 0.2 μm below the surface, the green sensors are below this at 0.6 μm, and the red even further at 2 μm. The first production cameras to incorporate the X3 sensor, the Sigma SD9 and SD10 SLRs, had a mixed reception, and the future of the technology remains uncertain.
Conclusion and outlook
The expansion of the digital market between the early 1990s and the turn of the 21st century was spectacular. In both the high-end and consumer sectors it was characterized by rapidly advancing capability and falling prices; high-end innovations soon trickled down to the mass market. Camera models were superseded at an unprecedented pace: Canon's acclaimed EOS D60 professional SLR, for example, by the even better specified, and cheaper, 10D in less than twelve months. The same year, 2003, saw the launch of the 11MPx (megapixel) Canon EOS-1 DS, and of a promising new (‘Four Thirds’) high-end system by Olympus improving the interaction of sensor and lenses. In the medium-format field the Kodak Pro Back, housed in a Hasselblad 555ELD, found applications in studio photography, aerial survey, and remote sensing. Significantly, its price fell from £16, 500 to c. £12, 500 in about a year.
In the meantime, solutions were found for many intractable-seeming early problems, such as poor-quality LCD screens, high power consumption, high noise levels, storage limitations, and sluggish operation. The pace of innovation seems set to continue. By c.2010 we are likely to see up to 20MPx sensors in 35 mm-equivalent cameras, increasingly sophisticated in-camera image manipulation facilities, cableless links to computers and printers, and more extensive overlaps with other technologies (e.g. mobile communication devices). Digital will probably also conquer most of the global camera market.
— Ron Graham
Bibliography
- Graham, R., Digital Imaging (1998).
- Edwards, S. H. (ed.), ‘Electronic and Digital Photography’,
History of Photography ,22 (1998). - Ang, T., Silver Pixels: An Introduction to the Digital Darkroom (1999).
- Tarrant, J., Digital Camera Techniques (2003)
