کشاورزی ماشینی

CCD and CMOS sensors

Charge-coupled device (CCD) and complementary metal oxide semiconductor (CMOS) sensors are similar in their architecture and function. They both utilize semiconductor technology to collect an image focused by a lens on to them. Some of the differences between the technologies

will be detailed as we progress through this chapter, but the majority of concepts may be applied to both technologies.

Trying to conceive of a new type of memory for computers, CCDs were invented, almost by accident, by Willard Boyle and George Smith at Bell Laboratories in 1969. CCDs therefore precede CMOS devices and are a more mature technology. CMOS devices initially suffered from high noise and therefore low image quality, though these arguments are not as valid in the present day. CCD sensors are generally used for higher quality imaging and scientific applications, such as astronomy, whereas CMOS technology is used in many devices from cellphones and webcams' to prosumer digital single-lens reflex (ProDSLR) cameras and security devices. The pricing of the devices has fallen considerably as the technologies have matured and digital devices have become extremely common. The sales of digital cameras have surpassed those of film cameras since the last edition of this book.

Both CMOS and CCD imaging sensors are constructed from the same material, silicon, and this gives them fundamentally similar properties of sensitivity in the visible and near-lR spectrum. Sensors are formed on silicon wafers and cut apart, generally with a diamond saw, to produce individual sensors or die. CCD and CMOS sensors can support a variety of photoelements, though the fundamentals are primarily concentrated upon here.



MOS capacitor

At the heart of the CCD is the metal oxide semiconductor (MOS) (Figure 9.5). It is comprised of silicon doped with impurities, such as aluminium, to produce a P-type semi-conductor, formed on a silicon substrate. Above the P-type semiconductor is formed a gate from polysilicon, which is held at a positive bias and produces a depletion region in the semiconductor. Generally, three or four such photogates are used in a single pixel to enable charge transfer (see later).

Silicon has a band gap of 1.1 eV; this is the energy required to promote one of its valence electrons to its conduction band (Figure 9.6). Using the relationship expressed in Eqn 2.35 it is therefore found that it has an inherent sensitivity up to approximately 1100 nm. While this may be an advantage for recording in the IR region of the spectrum for security and other applications, it is a disadvantage for imaging within the visible region of the spectrum and results in incorrect tone and colour balance. Image sensors to be used in the visible region of the spectrum have an IR filter to restrict their response to the visible region below approximately 670 nm.

The positive voltage to the gate (Vgate) causes mobile positive holes to migrate to the ground electrode, creating a depletion region. When light of sufficient energy (greater than 1.1 eV) is absorbed in the depletion region an electron— hole pair is produced. The electron remains in the depletion region whilst the positive holes migrate to the ground electrode. If the positive voltage was not applied.

A fundamental difference between CCD and CMOS technologies is the manner in which the signal is read from the pixels on the sensor. The CCD gets its name from the method of moving the stored image across the sensor (charge coupling) and reading it out at one corner, whereas each pixel in a CMOS device contains several components and may be read individually in most architectures.



CCD readout

Exact architectures may vary, though generally each pixel in a CCD is subdivided with three photogates. The charge may be thought of as gathering under the photogate to which the highest potential is applied. Adjustments of the gate voltages then allow transfer of electrons from well to well as a function of time.

The architecture of CMOS sensors differs significantly from that of CCDs in terms of the manner in which the signal is read. Dependent upon the architecture, each pixel in a CMOS sensor has its own set of transistors to perform amplification and to assist with readout and reset. By including circuitry to select columns and rows, entire columns or rows can be read out simultaneously. Figure 9.11 shows an early passive pixel architecture. The readout transistor for each pixel in combination with the horizontal and vertical readout circuits allows each pixel to be individually addressed. Adding amplification to each pixel allows readout noise to be overcome, creating the active pixel sensor.

As separate amplifier and reset transistors are used for each pixel, CMOS sensors tend to have increased levels of noise over CCDs, as matching these components is a challenge and the photodiode resets to a slightly different value each time. The sophistication of a photodiode-based pixel is related to the number of transistors that can be incorporated into each pixel. The predominant architecture as of writing is three transistor designs (3T) consisting of the photodiode, reset, addressing and source-following components. A number of 4T, 5T and 6T designs exist, reducing read out and reset noise.

Advances in technology over the past decade have significantly reduced the gap with CMOS as the predominant sensor in the consumer market.

Because of the extra area needed to accommodate the circuitry for each pixel, the fill factor and the sensitivity of the CMOS device suffer. CMOS sensors may be made using similar silicon foundry technology to computer memory and other integrated circuits and, hence, they may accommodate image processing and other circuitry that traditionally gets put into supporting chips for CCD devices. This allows for the production of sophisticated devices and full 'camera-on-chip' imagers are regularly seen at the time of writing. However, this extra complexity can increase design time and updates to the sensor's on-chip functions may often require revision of the chip design.

Though the CMOS sensor requires less off-chip support it uses a larger area of silicon for the supporting processing, and the cost benefits of using the same or older depreciated foundries for production have not been as large as expected. CCDs, by comparison, require dedicated foundries and little, if anything else at all, can be produced. Because a CCD uses a single amplifier and ADC, the uniformity of these processes is very good.

The method of reading the CMOS sensor allows for windowing of the image, selecting a reduced portion of the device from which to read an image. This, in turn, allows frame rates to increase as only part of the available data is read. In addition, control of the imaging parameters can be performed on a frame-by-frame basis, making the device highly configurable.

CMOS sensors generally consume less power than CCDs, which need accurate timing circuitry. CMOS sensors are therefore a preferred choice for consumer goods where battery life is important, e.g. small cellphones and cameras. As with many commercial products, the choice between the sensors is a compromise between cost, performance, efficiency and suitability for purpose.