TechArchive

Original article date: January 1999

CCD CMOS or CID? NEIL STEWART of Sira explains the selection factors between different types of imaging sensors.

With the advent of much lower cost semiconductor imaging devices digital imaging technology has come on by leaps and bounds in the last five years; so much so that many more business areas should be able to take advantage of the techniques.

Digital imaging devices incorporate an array of many (up to tens of millions) separate semiconductor detectors. This array is situated at the focal plane of a lens system. Each individual device measures the light intensity (in either monochrome or colour) at one point in the array. These individual intensity measurements are read out by an electronic circuit which forms an image. Over the last few years technological advancements in electronics and semiconductor technology have greatly advanced the capabilities of focal plane array sensors used in imaging.

Two types of semiconductor detector are predominantly used in focal plane sensing: photodiodes and MOS (Metal Oxide Semiconductor) capacitors. In both cases changes to individual pixels are detected by a readout circuit which arranges the information into an image.Visible light sensors are based on miniature silicon detectors. The range of detectable wavelengths is limited because wavelengths longer than 1100nm do not possess sufficient energy to excite electrons and wavelengths shorter than 400nm are quickly absorbed in surface layers so that they do not reach the detector.

The data can be read out from silicon focal plane arrays using different methods. By far the most common is the Charge Coupled Device (CCD) which has been around for 25 years and has undergone huge development in that time. The CCD device is a charge storage and transfer device. The detector signal is representative of the total light falling on the sensor during exposure. CCD devices account for the majority of imaging sensor sales. They produce the best quality images currently available from electronic imaging sensors. Although there are competing technologies it will be many years before they can compete with high quality CCD in fields such as digital still photography and image reproduction. For high quality digital imaging applications they are unrivalled.

One example of a CCD imaging sensor is the Frame Transfer (FT) architecture. The arrangement on a FT imaging chip is shown in Fig 1. The device has a high sensitivity to light and it is possible to build up very large high resolution FT CCD sensing arrays (one commercially available camera uses a 4096×4096 pixel array). In general readout from these devices is slow particularly with very large pixel arrays. To achieve faster frame rates with a reduction in quality of the image other architectures such as Interline Transfer or Frame Interline Transfer can be used.

CCD devices can produce extremely high quality images but they do also have some disadvantages. The device depends on accumulation of charge. If one pixel is over exposed the charge from that pixel can leak onto neighbouring pixels and create distorted values in the surrounding area. This effect is known as blooming and is present in all CCD devices. In addition the readout electronics require a non-standard power supply and have complex clock requirements in order to sequentially compose an image from the sensor.

Alternative devices are now available in which a fully functional camera can be placed on a single sensor chip. These sensing arrays are manufactured using CMOS (Complementary Metal Oxide Semiconductor) technology which should ultimately produce very cheap imaging sensors in large quantities. At present CMOS image quality is not as good as CCD sensors and it will be many years before they can match high quality CCD. The best current CMOS sensors produce images which are comparable to the lower end of CCD quality but this should improve.

Unlike CCD CMOS arrays are not charge transfer devices. They are charge detection devices which means that charge is not stored and moved around. It is created in the pixel element and detected by a readout circuit. It is not an integration device the charge represents the light intensity falling on the pixel at the instant of readout. Although if two pixels are readout they will not represent the same instant in time the readout rates are typically in the region of 10MHz so normally the time delay is not significant.

Charge does not leak between pixels and blooming does not occur as in CCD sensors. Pixels in CMOS arrays can also be randomly accessed either individually or as a group of pixels. Typically CMOS arrays can produce images at video rates but if the number of pixels of interest is small only these pixels need to be accessed. This means that the effective frame rate can be thousands of frames per second.

In some of these devices each pixel contains a photodiode and other electronics “on-chip”. Additional electronics can be integrated into each individual pixel to perform processing. This means that operations such as analog to digital conversion and arithmetic logic operations can be performed at the pixel level prior to readout introducing the concept of the “smart pixel”. All of the operations which would required to build up a camera can be performed on a single chip so that the sensing chip is the entire camera.

CMOS sensors also have standard power requirements and low power consumption and they are capable of producing high dynamic range sensors. Pixels can be manufactured with either linear or logarithmic response. In the latter case dynamic range can be as high as 1:1 000 000 comparing favourably with the typical CCD dynamic range of 1:3000. This enables sensors to distinguish areas with varying illumination.

Because the chip incorporates additional electronics which occupies some of the surface area not all of the chip is light sensitive. Quoted fill factor values lie between 25 and 60 In addition the presence of electronics particularly amplifiers on the pixel make it difficult to exactly match the pixel outputs. Each individual pixel has a noise value associated with it so any image has fixed pattern noise. Although this can be overcome by sampling in a particular manner it adds complexity to the system.

The third type of imaging sensor which is used for visible light detection is the Charge Injection Device (CID). The idea behind this device dates back to 1972 but almost all the development has been concentrated into the last five years. The device operates by injecting charge from a detector into a substrate and readout can be performed by individually accessing pixels as in CMOS.

CID devices have several interesting characteristics. Pixels accumulate charge which is injected into the substrate. Pixels can be individually accessed however in doing so the charge is sensed and then replaced so that the integration process is not disturbed. Light continues to be collected while the pixels are monitored. This is called Non Destructive Readout and it makes it possible to carry out real-time exposure monitoring and real-time noise detection and cancellation. These devices are also typically 100 times less sensitive to potentially harmful radiation than CCD sensors. In addition the surface layers on the detector tend to be very thin so lower wavelengths are not absorbed. As a result CID devices can detect wavelengths down to 185nm (UV).

CID devices do have advantages over the CCD and CMOS devices but the big drawback is the unit cost which is much higher than for a CCD with equivalent resolution. These factors restrict them to specialised applications such as in space or in high radiation environments.

January 1999