`Imaging Array Technologies:
`Technology Review
`
`Stuart A. Taylor
`
`Technical Report EPC-1998-106
`
`Copyright ª
`
` Xerox Limited 1998
`
`Xerox Research Centre Europe
`Cambridge Laboratory
`61 Regent Street
`Cambridge CB2 1AB
`
`Tel: +44 1223 341500
`Fax: +44 1223 341510
`
`Page 1 of 15
`
`GOOGLE EXHIBIT 1011
`
`
`
`CCD and CMOS Imaging Array Technologies
`- Technology Review -
`
`Stuart A Taylor
`Xerox Research Centre Europe
`Cambridge, UK.
`staylor@xrce.xerox.com
`
`Abstract
`This paper provides an overview of both CCD (charged coupled device) and CMOS
`(complimentary metal oxide semiconductor) imaging array technologies. CCDs have been in
`existence for nearly 30 years and the technology has matured to the point where very large,
`consistent (low numbers of defects) devices can now be produced. However, CCDs suffer
`from a number of drawbacks, including cost, complex power supplies and support
`electronics. CMOS imaging arrays, on the other hand, are still in their infancy, but are set to
`develop rapidly and offer a number of potential benefits over CCDs. This review provides an
`overview of both CCD and CMOS imaging technology, and includes explanations of how
`images are captured and read out from the imaging arrays. Also covered are issues such as
`performance characteristics, cost considerations and the future of imaging arrays. This review
`does not provide details of colour sensors, colour filter arrays and colour interpolation, etc.,
`as these will be the subject of a separate report.
`
`Introduction to CCDs
`The Charged Coupled Device (CCD) was invented in 1970 by Willard Boyle and George
`Smith at Bell Laboratories, USA [Sharma97]. The idea originated from research into
`magnetic bubble memories, and as with many great inventions, Smith is quoted as saying
`“[we] invented charged-coupled devices in an hour” [Lucent96]. In the intervening twenty-
`eight years, CCDs have found their way into a huge range of products including fax
`machines, photocopiers, cameras, scanners and even children’s toys.
`
`CCDs consist of thousands (or millions) of light sensitive cells or pixels that are capable of
`producing an electrical charge proportional to the amount of light they receive. Typically, the
`pixels are arranged in either a single line (linear array CCDs) or in a two-dimensional grid
`(area array CCDs). The particular application will, in general, dictate the type of CCD that is
`used. Flatbed scanners, for example, use linear array CCDs and, in this case, it is necessary to
`progressively move the CCD over the object being imaged (or vice versa) while capturing
`multiple one-dimensional images in order to build up the final two-dimensional image.
`Digital cameras, on the other hand, normally use area array CCDs, thus allowing the full two-
`dimensional image to be captured within a single exposure.
`
`One of the fundamental parameters of a CCD is resolution, which is equal to the total number
`of pixels that makes up the light sensitive area of the device. One of the first area array
`CCDs, manufactured by Fairchild in 1974, had a resolution of 100x100 [Oregon97]. Today,
`the largest commercially available device is approximately 9000x7000 or roughly 63 million
`pixels [Pixelv97]. Other parameters that characterise CCDs will be discussed in more detail
`later.
`
`Page 1 of 14
`
`Page 2 of 15
`
`
`
`CCDs are in essence integrated circuits (ICs) and are hence rather like computer chips.
`However, to allow light to fall on the silicon chip (or die) a small glass window is inserted in
`front of the chip. Conventional ICs are usually encapsulated in a black plastic body to
`primarily provide mechanical strength, but this also shields them from light, which can affect
`their normal operation. CCDs are manufactured using metal-oxide-semiconductor (MOS)
`fabrication techniques, and each pixel can be thought of as a MOS capacitor that converts
`photons (light) into electrical charge, and stores the charge prior to readout.
`
`Examples of CCD Arrays
`Before moving on to explain the operation of CCDs in more detail, it is useful to illustrate
`what CCDs actually look like. The following table shows three different devices for
`comparison, along with their respective pixel counts. The transfer method will be explained
`in a subsequent section.
`
`Manufacturer: Kodak
`Type: area array
`Pixel count: 4096x4096
`Transfer method: full frame transfer
`
`Manufacturer: Dalsa
`Type: area array
`Pixel count: 256x256
`Transfer method: frame transfer
`
`Manufacturer: Kodak
`Type: linear array
`Pixel count: 5000
`Transfer method: full frame transfer
`
`Table 1 – Examples of CCD Arrays [Kodak98, Dalsa98]
`
`CCD Fundamentals
`As mentioned above, each pixel that makes up a CCD is essentially a MOS capacitor, of
`which there are two types: surface channel and buried channel. The two differ only slightly
`in their fabrication, however; buried channel capacitors offer major advantages, and because
`of this, nearly all CCDs manufactured today use this preferred structure.
`
`A schematic cross section of a buried channel capacitor is shown in Figure 1 [Sharma97,
`SITe94]. The device is typically built on a p-type silicon substrate (approx 300m m thick) with
`an n-type layer (approx 1m m thick) formed on the surface. Next, a thin silicon dioxide layer
`(approx 0.1m m thick) is grown followed by a metal electrode (or gate). The application of a
`
`Page 2 of 14
`
`Page 3 of 15
`
`
`
`positive voltage to the electrode reverse biases the p-n junction and this causes a potential
`well to form in the n-type silicon directly below the electrode. Incident light generates
`electron-hole pairs in the depletion region, and due to the applied voltage, the electrons
`migrate upwards into the n-type silicon layer and are trapped in the potential well
`[Muncaster85]. The build up of negative charge is thus directly proportional to the level of
`incident light.
`
`Once the exposure time (also known as the integration time) has elapsed, the charge trapped
`in the potential well is transferred out of the CCD before being converted to an equivalent
`digital value.
`
`+Ve
`
`Metal Electrode
`
`Photo-generated
`electron-hole pair
`
` Oxide Layer
`
`N-Type Silicon
`
`Depletion Region
`
`P-Type Silicon
`
`Earth
`
`Figure 1 - Buried Channel Capacitor CCD Pixel
`
`An illustration of a practical buried channel capacitor is shown in Figure 2 [SITe94]. This
`diagram also shows the channel stops, which are usually created by heavily doping these
`regions to form p-type semiconductor. In addition, a thick layer of oxide, the field oxide, is
`applied over these regions. The purpose of the channel stops is to minimise the diffusion of
`electrons from one pixel to another.
`
`Figure 2 – Practical Buried Channel
`Capacitor (From [SITe94])
`
`The Charge Readout Process
`The charge readout process takes place in two stages. The first involves moving the pixel
`charges across the surface of the array. The second involves reading out the pixel charges into
`a register prior to being digitised.
`
`Page 3 of 14
`
`Page 4 of 15
`
`
`
`The charge transfer process can be explained as follows. Each pixel is divided into a number
`of distinct areas known as phases [Gallagher]. Three-phase sensors tend to be the most
`common form due principally to high yields and high process tolerances, although one, two,
`and four phase formats do exist [SITe94]. Taking a three phase sensor as an example (see
`Figure 3), the integration period, phases 1 and 2 will be in charge holding mode, and phase 3
`will be in charge blocking mode (a). At the end of the integration period, when it is time to
`transfer the captured image out of the array, the following process takes place. Phase 1 is
`placed in charge blocking mode, which has the effect of transferring the total charge of
`phases 1 and 2 into only phase 2 (b). Phase 3 is then placed in charge holding mode, which
`allows the charge in phase 2 to distribute itself evenly between phases 2 and 3 (c). Next,
`phase 2 is placed in charge blocking mode, forcing the charge into phase 3 (d). This process
`repeats, until, as illustrated by (g), the charge from pixel two has been moved into pixel one.
`An alternative representation [Oregon97] of the charge readout process is illustrated in
`Figure 4.
`
`Pixel 1
`
`Pixel 2
`
`Ø 3
`
`Ø 2
`
`Ø 1
`
`Ø 3
`
`Ø 2
`
`Ø 1
`
`a
`
`b
`
`c
`
`d
`
`e
`
`f
`
`g
`
`Figure 3 - Charge Transfer
`
`Figure 4 - Charge Transfer; Alternative Representation
`(From [Oregon97])
`
`The second stage of the readout process occurs after each row of pixels has been transferred
`by one complete row. Located adjacent to the top row of pixels is one additional row of
`pixels called the readout register. As the charge from each row of pixels is moved up one
`row, the charge from the top row will be moved into the readout register. At this point, the
`charge values in the readout register are transferred horizontally into the readout stage, from
`which they are then sent to an analogue-to-digital converter (ADC) before being stored in
`memory. This process is illustrated in Figure 5 .
`
`Page 4 of 14
`
`Page 5 of 15
`
`
`
`Readout
`Register
`
`Readout
`Stage
`
`Readout
`Register Transfer
`
`Row
`Transfer
`
`1
`
`2
`
`3
`
`Figure 5 - Charge Readout
`
`The transfer and readout processes described above apply to both linear array and area array
`CCDs. In the former case, a single transfer followed by the readout stage is all that is
`necessary to transfer the image out of the array. For area array CCDs, the transfer and readout
`stages must, in general, be repeated for each row of pixels, until the complete image has been
`read out. This generic description of the readout process for area arrays will vary, however,
`depending on the overall architecture of the device.
`
`Area Array CCD Architectures
`The architecture of area array CCDs generally falls into one of four categories: full frame,
`frame transfer, split frame transfer and interline transfer. These are illustrated in Figure 6.
`
`Imaging Section
`
`Imaging Section
`
`Storage Section
`
`Readout Register
`
`Readout Register
`
`Full Frame Sensor
`
`Frame Transfer Sensor
`
`Readout Register
`
`Upper Storage
`Section
`
`Imaging Section
`
`Lower Storage
`Section
`
`Readout Register
`
`Readout Register
`
`Readout Register
`
`Light Shielded Vertical CCD
`
`Imaging CCD Elements
`
`Split Frame Transfer
`
`Interline Transfer
`
`Figure 6 - Area Array CCD Architectures
`
`Page 5 of 14
`
`Page 6 of 15
`
`
`
`Full Frame – Here, the image is transferred directly from the imaging region of the sensor to
`the readout register. However, since only a single row can be transferred to the readout
`register at a time (the contents of which must then be transferred to the readout stage) the rest
`of the imaging pixels must wait. During this period, pixels that have not yet been read out of
`the array are still capable of recording image information. The problem with this is that the
`imaged information will now be offset from the original scene recorded, and this can lead to
`smearing and blurring of the final image. Another problem is encountered during high-speed
`applications. In these situations, the integration time will be only a small percentage of the
`total time required to record and read out the resultant image from the array. The effect of this
`is that the image will have a lower contrast since very little time is spent actually recording
`the image. One solution to these problems is to use a mechanical shutter, whereby the array is
`shielded from light once the image has been captured.
`
`Frame Transfer – A frame transfer device utilises a light shielded storage section of at least
`the same size as the imaging section of the array. Following the integration period, the
`captured image is quickly transferred to the adjacent storage section. Whilst the next scene is
`being captured, the previous scene, now held in the storage section, is transferred to the
`readout register as described above. Using this technique effectively allows both the image
`capture and readout processes to run in parallel. This greatly increases the time available for
`integration whilst still maintaining a sufficiently high frame rate. However, since the captured
`image still has to be transferred across the surface of the imaging section of the array, a
`mechanical shutter may still be required to avoid the problems described above.
`
`Split Frame Transfer – This type of device is essentially the same as a frame transfer device,
`except that the storage section is split in half, with each half being located above and below
`the imaging section. The advantage of this is that it allows the image to be transferred out of
`the imaging section in half the time that is required for a frame transfer device.
`
`Interline Transfer – An interline transfer device has columns of photosensitive elements
`separated by columns of light shielded registers. At the end of the integration period, all of
`the photosensitive elements simultaneously transfer their accumulated charge to the adjacent
`storage registers. The light shielded registers then transfer the charge to the readout register
`as previously described, during which time, the imaging elements begin capturing the next
`scene. Although not requiring a mechanical shutter, this type of device has the significant
`drawback that a large proportion (typically 40%) of the imaging section is not sensitive to
`light. To minimise the effects of this, microlenses are often placed directly over the imaging
`section of the array. The lenses cover both the light sensitive and light shielded portions of
`each array element, and have the effect of focusing the incoming light onto just the light
`sensitive areas of each element.
`
`Interlaced and Progressive Scan
`The image collection mode of an area array sensor can be either interlaced or progressive
`scan (non-interlaced). The interlaced technique is used by the major broadcast standards PAL
`and NTSC and is done to reduce the bandwidth of the image for transmission. In this mode,
`the frame is divided into two fields: an odd field consisting of all the odd numbered rows, and
`the even field consisting of all the even numbered rows. Half of the frame is recorded by the
`odd field at time T1, and the other half of the frame is recorded by the even field at time T2.
`This means that it takes two cycles to build up a complete image, so for PAL, which operates
`at 50Hz, images are captured at a rate of 25 per second. A major drawback of the interlaced
`method is encountered when the subject is moving. Since the two fields are separated in time
`
`Page 6 of 14
`
`Page 7 of 15
`
`
`
`by 20ms, the position of a moving object will have changed between the two fields, resulting
`in blur when the two fields are combined to produce the final image. Figure 7 shows a picture
`of a moving hand which illustrates this problem (note, the effect has been emphasised in this
`image to more clearly demonstrate the problem). The progressive scan mode sensor reads out
`the complete frame in one go, thus making it possible to capture images of moving objects
`without the blurring effects associated with interlaced mode [Pulnix97].
`
`Figure 7 - Moving Hand Captured With an
`Interlaced Camera
`
`CCD Performance Characteristics
`In an ideal world, CCDs would exhibit perfect scene-to-digital image conversion. However,
`in reality, CCDs are not perfect. They suffer from a number of problems and limitations
`[Oregon97, SITe94], some of which are discussed in this section.
`
`Fill factor – The fill factor is basically the percentage of each pixel that is sensitive to light.
`Ideally, the fill factor should be 100%; however, in reality, it is often less than this. Features
`to control blooming (see below) and, for CMOS sensors, additional control electronics,
`occupy space within each pixel, and these areas are insensitive to light. The net effect of
`reducing the fill factor is to lower the sensitivity of the array.
`
`Dark current noise – Dark current can be defined as the unwanted charge that accumulates in
`the CCD pixels due to natural thermal processes that occur while the device operates at any
`temperature above absolute zero. At any temperature, electron-hole pairs are randomly
`generated and recombine within the silicon and at the silicon-silicon dioxide interface.
`Depending on where they are generated, some of these electrons will be collected in the CCD
`wells and will appear as unwanted signal charges (i.e. noise) at the output.
`
`The principal sources for dark current in order of importance are: generation at the silicon-
`silicon dioxide interface, electrons generated in the CCD depletion region, and electrons that
`diffuse to the CCD wells from the neutral bulk. The first two sources usually dominate the
`dark current. In addition, the generation rate can vary spatially over the array leading to a
`fixed noise pattern.
`
`Page 7 of 14
`
`Page 8 of 15
`
`
`
`For applications requiring very low noise levels, for example astro-photography, dark current
`sources can be reduced by cooling the CCD since they are strongly temperature dependent.
`The level of cooling is largely dependent on the longest integration time desired and the
`minimum acceptable signal-to-noise ratio.
`
`Quantum efficiency (QE) – Quantum efficiency (QE) is the measure of the efficiency with
`which incident photons are detected. Some incident photons may not be absorbed due to
`reflection or may be absorbed where the electrons cannot be collected. The quantum
`efficiency is the ratio of the number of detected electrons divided by the product of the
`number of incident photons times the number of electrons each photon can be expected to
`generate. Visible wavelength photons generate one electron-hole pair, thus the QE for visible
`light is given by the ratio of the number of detected electrons divided by the number of
`incident photons.
`
`There are a number of techniques used to improve the QE of CCDs, one of which is to
`illuminate the CCD from the back, as opposed to the front. In front-illuminated devices,
`incident photons must pass through the gate structure in order to generate signal electrons.
`Photons will be absorbed in these layers and thus won’t contribute to the final signal. The
`absorption is also wavelength dependent, with short wavelength photons being absorbed
`more than long wavelength photons. This effect results in poor blue and UV spectral
`responses. In order to increase the short wavelength response, a technique of thinning the
`silicon substrate has been developed (typically from 300m m down to 15m m). In this case, the
`CCD is illuminated from the back and thus photons do not have to pass through the front gate
`structure. It has taken almost a decade to perfect the thinning process, the main problem
`being non-uniform thinning, i.e. the corners were thinner than the centre which leads to non-
`uniform response, the ‘potato chip factor’ [Oregon97].
`
`Blooming – Blooming is an effect that occurs when, during the integration period, a potential
`well becomes full of electrons; this is usually caused by the presence of a bright object in the
`scene being imaged (assuming that the overall exposure is correctly set). When a potential
`well overflows, the electrons flow into surrounding potential wells, thus creating an area of
`saturated pixels. If blooming isn’t controlled, the resultant image will suffer from large over-
`exposed regions.
`
`Many techniques have been developed to combat blooming, but one common method is to
`use lateral overflow drains (LODs) [Parulski96], which is illustrated in Figure 8.
`Conceptually, they work in a similar way to an overflow in a sink; when the potential well
`fills to a certain level, any further electrons that accumulate are allowed to drain away
`without affecting surrounding pixels. One of the drawbacks of using anti-blooming systems is
`the fill factor is often reduced, typically from near 100% down to around 70% [Kodak98].
`
`Pixel
`
`Electrons
`
`LOD
`
`Figure 8 – Schematic of LODs
`
`Page 8 of 14
`
`Page 9 of 15
`
`
`
`Charge transfer efficiency (CTE) – The CTE is a measure of the percentage of electrons
`which are lost at each stage during the charge transfer process. Modern buried channel CCDs
`have CTE values in excess of 99.999%. Another aspect of the charge transfer mechanism is
`that the rows of pixels closest to the readout register will undergo fewer transfers than those
`on the opposite side of the array. The net effect is that the image quality will vary across the
`width of the array. CTE effects and the non-linearity of image quality across the array are
`contributing factors to the upper size limit of CCD arrays.
`
`Introduction to CMOS Imaging Arrays
`CMOS, or complimentary metal oxide semiconductor, image sensors have been around for
`nearly as long as CCDs [Dyson97], but it is only recently that commercial sensor chips have
`become available. These devices were made possible through research carried out at the Jet
`Propulsion Laboratory (JPL), and in 1993, they produced a CMOS sensor with a performance
`comparable to scientific-grade CCDs [Photobit97b]. This section provides an overview of
`CMOS image sensors, and highlights some of the advantages and disadvantages they offer
`over CCDs.
`
`CMOS sensors, like CCDs, are formed from a grid of light sensitive elements, each capable
`of producing an electrical signal/charge proportional to the incident light. However, the
`process of achieving this is very different for each of the technologies. As previously
`explained, a CCD pixel is formed from a biased p-n junction that creates a potential well in
`which charge accumulates during the integration period. Each CMOS pixel, on the other
`hand, employs a photodiode, a capacitor and up to three transistors. Prior to the start of the
`integration period, the capacitor will be charged to some known voltage. When the
`integration period begins, the charge on the capacitor is allowed to slowly drain away through
`the photodiode, the rate of drain being directly proportional to the level of incident light. At
`the end of the integration period, the charge remaining in the capacitor is read out and
`digitised. Figure 9 shows an example of an active pixel (see next section) along with graphs
`of voltage vs. time taken from various points within the pixel [Hurwitz97]. It is worth noting
`that other circuit arrangements are possible, for example, the capacitor is charged during the
`integration period, as opposed to being discharged.
`
`Figure 9 - Active CMOS Pixel Structure (From [Hurwitz97])
`
`Page 9 of 14
`
`Page 10 of 15
`
`
`
`CMOS Detector Types
`CMOS image sensors typically come in two forms: passive pixel and active pixel. Passive
`pixel devices have charge amplifiers at the bottom of each column of pixels, with each pixel
`having just a single transistor (in addition to the photodiode and capacitor). This transistor is
`used as a charge gate and switches the contents of each pixel’s capacitor to the charge
`amplifier. Active pixel arrays implement an amplifier in every pixel (as shown in Figure 9).
`The two different detector types are illustrated in Figure 10.
`
`Photo-detector
`
`Photo-detector
`
`Row
`Selectors
`
`Switching
`Transistor
`
`Row
`Selectors
`
`Active Charge
`Amplifier
`
`Charge Amplifiers
`
`Column Outputs
`
`Passive Pixel Array
`
`Column Outputs
`
`Active Pixel Array
`
`Figure 10 - Passive and Active Pixel CMOS Arrays
`
`Although only a schematic, this diagram highlights the fact that even for the passive pixel
`type, a proportion of each pixel is occupied by additional components which are not
`themselves sensitive to light. The effect of this is to greatly reduce each pixel’s fill factor;
`Hurwitz quotes a figure of only 26% [Hurwitz97]. In an attempt to overcome this problem,
`micro-lenses are sometimes mounted directly in front of each pixel in an attempt to focus the
`incoming light onto the sensitive region of the pixels [Pixelv97].
`
`The Readout Process
`Both passive and active pixel arrays use the same technique for reading the image out of the
`array. Each of the row selectors are sequentially clocked. This in turn causes the switching
`transistor or charge amplifier for that row of pixels to activate and thus transfer each pixel’s
`charge to the column outputs. A readout register then serially transfers the column output
`values to an analogue to digital converter in a similar way to a CCD.
`
`Clocking the row selectors sequentially allows the full image to be progressively read out
`from the array. However, it is also possible to clock a limited number of row selectors, which
`has the effect of reading out a small horizontal strip from the array. By discarding pixel
`values from the beginning and end of each of these rows, a specific area of the image can be
`obtained. Sampling a small number of relevant pixels in this way dramatically increases the
`readout speed; using this technique, some manufacturers claim speeds of a million frames per
`second [Graydon97].
`
`Page 10 of 14
`
`Page 11 of 15
`
`
`
`Product Integration
`Another advantage of CMOS arrays over CCDs lies in the potential for high levels of product
`integration. It is possible to include timing logic, exposure control, analogue-to-digital
`conversion and image compression circuitry on-chip with the sensor to make a complete
`single-chip camera [WhatDC97]. It is technically feasible, but not economic, to use the CCD
`process to integrate these functions, and thus most CCD based cameras are formed from
`several chips. This can result in the need for up to 5 different supply voltages which leads to
`high power consumption. By using a single chip, coupled with the inherent low power
`consumption of CMOS devices, power savings of 100 times can be achieved over CCDs
`[Photobit97b]. Another advantage of integrating functionality is that CMOS sensors can be
`tested ‘on the wafer’; that is, rejects can be found before the costly production steps of cutting
`and mounting each individual device [Dyson97].
`
`However, there are a number of drawbacks of producing multi-functional CMOS sensors.
`Firstly, there is the cost of producing the very large dies needed to accommodate the sensors
`and associated components (see the next section for a more detailed examination of cost
`considerations). Also, there is the difficulty of embedding leading edge intellectual property;
`to embed new technology requires a new chip to be produced, therefore multi-chip solutions
`remain popular. Finally, each function of the image sensor technology requires a different
`manufacturing process. For example, companies producing memory chips use a dramatically
`different process to those making analogue-to-digital converters. Consequently, multi-
`function CMOS sensor production must employ a hybrid approach, and this can sometimes
`be difficult to optimise [Pixelv97].
`
`Cost Considerations
`It is estimated that between 90% and 95% of all chips found in computers and other
`electronic goods today are manufactured using CMOS technology [Graydon97, Photobit97a].
`Chip manufacturing plants cost many millions of dollars to set up, but provided they produce
`chips in sufficient quantities, the price per chip is very low, particularly when compared with
`other technologies.
`
`CMOS wafer production lines are already making 8-inch wafers, and heading towards 12-
`inch in the near future. Also, the feature size of CMOS is already 0.4 microns with 0.2 on the
`horizon; wafer sizes of between 4 and 6-inch are typical for CCDs, and their feature sizes are
`typically 0.6 micron. Larger wafers with smaller features means more devices per production
`cycle, and that leads to lower costs [Dunn97].
`
`Performance Issues
`CMOS imaging is still a relatively immature technology and, although there are many
`potential benefits to be gained over CCDs, there remain a number of problems yet to be
`addressed. It has already been mentioned that the fill factor of CMOS devices is typically
`around the 25% mark, and that micro-lenses are used to help improve sensitivity. However,
`there are other more serious problems that have to be addressed.
`
`Photodiodes, which are the light sensing component of every pixel, are easy to make, but
`have an exceedingly non-linear response to light. It has also proved very difficult to control
`the fabrication process well enough to obtain comparable responses from a million individual
`diodes across the chip [Dyson97]. Another related problem is one of pixel defects where
`individual, or groups, of pixels are insensitive to light; it is worth noting that this is also a
`problem found in CCDs. Devices with high defect levels can still be used for non-critical
`
`Page 11 of 14
`
`Page 12 of 15
`
`
`
`applications such as toys, and such devices tend to be very much cheaper than near-perfect
`devices. Currently, the main solution to these problems is to use image processing software,
`i.e., the defects and non-linearities are still present in the sensors, but software is used to
`reduce their effects. For example, by capturing a light shielded image, this can be subtracted
`from the actual image to help reduce the non-linearity effects. Another technique used to
`overcome defects is to use a ‘blemish map’. Using this map, software can be used to look at
`neighbouring pixels and calculate what a likely value should be for defective pixels
`[Dunn97].
`
`On the plus side, CMOS sensors consume very little power. In addition, they do not suffer
`from problems of blooming. Since the pixels are not accumulating charge, unlike a CCD,
`they can not ‘overflow’ and affect neighbouring pixels. Another advantage is image
`consistency (ignoring the effects of defects, etc). Transferring the image directly out of the
`array, as opposed to having to move the image from one pixel to the next, removes problems
`of charge transfer efficiency (CTE) found in CCDs.
`
`The Future of Imaging Arrays
`CCD imaging sensors have matured over the last 30 years to the point where they are now
`available in very high-resolution formats with very low defect rates. Because of this, they
`remain the sensor of choice for nearly all scientific and professional imaging applications.
`However, for the low-end consumer market, where cost is often more important than quality,
`CCDs are beginning to loose out to a new generation of CMOS devices. Over the next 5 to 10
`years, it is likely that CMOS imaging technologies will evolve to the point where they offer
`all of the benefits of CCDs, along with low cost, low power consumption and high levels of
`product integration.
`
`The field of digital photography has seen a huge growth over the last few years, and it is
`believed that this growth will increase once the current problems with CMOS sensors have
`been overcome. In short, the next few years will see CMOS sensors having a dramatic impact
`on the field of digital imaging.
`
`Page 12 of 14
`
`Page 13 of 15
`
`
`
`References
`
`[Dalsa98]
`
`Dalsa web site; CCD Image Capture Technology; http://www.dalsa.com
`
`[Dunn97]
`
`[Dyson97]
`
`Dunn, James F.; A New Digital Camera Startup Busts Price/Performance
`Standards with CMOS Sensor; Advanced Imaging; Jan 1997.
`
`Dyson, P.E, Rossello, R.; CMOS Challenges CCD in Vivitar’s 3000; PMA
`‘97: new technology and applications for digital photography; The Seybold
`Report on Publishing Systems; April 1997, Vol.26, Num. 13.
`
`[Gallagher]
`
`Gallagher, P.; Considerations For Selecting CCD Sensors and Cameras for
`Vision Applications; EG&G Application Note.
`
`[Graydon97]
`
`[Hurwitz97]
`
`[Kodak98]
`
`[Lucent96]
`
`Graydon, O.; Selected pixels deliver pictures by the million;
`http://www.vvl.co.uk/whycmos/text/olejune.htm; Extracted from the full
`text from Opto & Laser Europe, June 1997
`
`Hurwitz, J.E.D, Denyer, P.B, Baxter, D.J; An 800K-Pixel Colour CMOS
`Sensor for Consumer Still Cameras; Presented as part of Electronic
`Imaging at SPIE Photonics West, 1997; Obtained from
`http://www.vvl.co.uk/whycmos/text/spie800.htm
`
`Kodak web site; Full frame sensors, KAF-0400(L);
`http://www.kodak.com/daiHome/genInfo/sensors/kaf0400.shtml
`
`Lucent Technologies web site; Invented Here: Charged-Coupled Devices;
`1996; http://www.lucent.com/ideas2/discoveries/telescope/docs/ccd1.html
`
`[Muncaster85] Muncater, R.; A-Level Physics; Stanley Thornes (Publishers) Ltd; 1985;
`pp.769-778.
`
`[Oregon97]
`
`The University of Oregon Physics Department web site; Evolving Towards
`The Perfect CCD; 1997; http://zebu.uoregon.edu/ccd.html
`
`[Parulski96]
`
`Paraulski, K. and Jamerson, P.; Enabling Technologies for a family of
`digital cameras; SPIE – The International Society for Optical Engineering;
`Solid State Sensor Arrays and CCD Cameras; Volume 2654; 1996.
`
` [Photobit97a]
`
`Photobit web site; Advantages Over the CCD;
`http://www.photobit.com/advant1t.htm
`
`[Photobit97b]
`
`Photobit web site; Technology, CMOS Active Pixel Sensor Technology;
`http://www.photobit.com/techt.htm
`
`[Pixelv97]
`
`Pixelvision web site; Types Of Solid State Imaging Sensors; 1997;
`http://www.pv-inc.com/tutorial-types.htm
`
`[Pulnix97]
`
`Pulnix Web Site; Progressive Scan [Progressive Scan vs. Interlace]; 1997
`
`Page 13 of 14
`
`Page 14 of 15
`
`
`
`http://www.pulnix.com//pstech.htm
`
`[Sharma