CCD Imaging: When Every Pixel Counts
Industry experts discuss the progress and future of CCD technology.
Paula M. Powell
With CCDs, production of a quality image remains their selling point, whether the sensors are used with a microscope, the Advanced Camera for Surveys (ACS) on the Hubble Space Telescope or your cell phone. In all cases, though, time is also of the essence.
Consider a 2002 retrofit of the Hubble, for example. After installation of the ACS with its increased sensitivity, the Tadpole Galaxy was imaged in only four hours, instead of the 40 hours required for a similar deep-field exposure with an older onboard camera.
With last year’s addition of new CCD technology for the Advanced Camera for Surveys, the Hubble Space Telescope is able to produce more-dramatic images, such as this one of the Cone Nebula. Courtesy of NASA, STSci and ESA.
Imaging sensitivity and speed also play a role in single-molecule fluorescence imaging simply based on the fact that individual molecules have only a finite lifetime before bleaching occurs.
There’s no question that CMOS-based sensors have a critical place in the imaging market. As the following articles illustrate, though, CCD technology still offers image quality that is hard to beat in many cases, even in commercial applications such as cell phone imaging, once predicted to become the bastion of CMOS technology.
Imaging the Heavens
Possibly the most demanding application of CCDs involves astronomy — especially space-borne instruments such as the cameras on the Hubble Space Telescope. The demands placed on the performance of these devices are extremely high because, for astronomers, every photon is precious. Most of the interesting science is carried out at the 10-e level. Also, in space these devices must perform for years in a harsh radiation environment.
Even here, though, scientists now have access to CCD technology with nearly 100 percent quantum efficiency from 400 nm through the visible spectrum to 950 nm in the near-IR, noise of 2 e that allows reliable detection of as few as eight or nine photons per pixel with a signal-to-noise ratio of 4, and a charge transfer efficiency so close to perfect that a small packet can be transferred from the most remote part of the device to the output amplifier. This can mean transferring through 8000 pixels. With the equipment cooled to 280 °C, dark currents on the order of 3 e per pixel per hour are routine. On the best chips, a dynamic range of 18 bits is possible.
In addition, because telescopes are expensive, users want to employ as much of the available focal plane as possible. The earliest CCD chips that were used in ground-based telescopes were the size of a postage stamp, which translated to a tremendous waste of an expensive resource. Since then, the desire to use more of the focal plane has led to devices with greater than megapixel size and to mosaics of devices.
Demand for larger size also promoted development of the first truly wafer-scale integrated circuits. For many years, Scientific Imaging Technologies offered a 2048 x 2048-pixel device that occupied a full 100-mm wafer, and 4096 x 4096-pixel devices are now routinely available from several sources. CCDs as large as 7000 x 9000 and 9000 x 9000 pixels have been built, and a large focal plane containing more than 40 2000 x 4500-pixel chips is nearing completion at the Canada-France-Hawaii Telescope.
For astronomical applications, the strengths of CCD technology include high quantum efficiency, low noise, a very large dynamic range and very low dark current. Chips also can be optimized to form superpixels. In addition, recent development of the Orthogonal Transfer CCD allows the sensor to follow the motion of a target in two dimensions, thus performing on-chip first-order adaptive optics correction.
CCD technology on the Advanced Camera for Surveys can produce wide-field, IMAX-quality images with more than 16 million pixels in each exposure. Courtesy of Scientific Imaging Technologies and Ball Aerospace & Technologies Corp. CCD technology still has some weaknesses in these applications, including the modest tolerance to a radiation environment. Also, the device readout is still serial; i.e., pixels are not randomly addressable, and a slow readout speed is necessary to drastically limit noise. Therein lies the opportunity for CMOS sensors. If they can achieve the same noise performance and quantum efficiency, they will erode the grip that CCDs have on the astronomy market.
In the near term, end users can expect increases in the readout speed for thinned devices. This may occur through architectural changes or processing and design changes. The L3CCD from E2V of Chelmsford, UK, and the Impactron from Texas Instruments Inc. in Dallas will be developed further. These devices have all the capabilities of high-performance CCDs listed above as well as effectively subelectron read noise, providing the potential for single-photon counting in an all-solid-state device. It would be difficult for CMOS sensors to approach this level of performance.
Contact: Morley Blouke, Scientific Imaging Technologies Inc., Tigard, Ore.; +1 (503) 431-7163.
Examining On-Chip Multiplication
One biological application area that places high demands on detectors is intracellular ion microscopy. Techniques for the study of intracellular ions find wide use; for example, to identify spatial variations in calcium levels within living cells, to measure the concentrations of intracellular ions including pH and, most importantly, to monitor how these concentrations change with time.
Calcium channels are representative of an important class of cell membrane molecules known as ion channels. As these open or close — for example, in response to extracellular messenger molecules — intracellular ion concentrations can also change, altering how the cell behaves. Therapeutic agents that act directly on these ion channels may provide effective treatments for many diseases.
Another demanding CCD imaging application is that of single-molecule microscopy. There is a rapidly growing realization, particularly within the life and materials science fields, that the information content afforded by imaging a single emitting molecule markedly exceeds that offered by bulk measurement, yielding invaluable insight into individual molecular properties and their microenvironment, and enabling characterization of individual molecular interactions.
For these applications, the primary CCD specification requirements filter down to two fundamental parameters: sensitivity and speed. The camera must be sensitive to detect the weak signal of low dye concentrations and to cope with the lower photon fluxes afforded by shorter exposure times (complementing fast frame rates). It also must be able to detect the weaker photon fluxes afforded by reduced excitation powers — which decrease photobleaching of dyes and photodamage to tissues, and lengthen experimental lifetimes — and to overcome the significant readout noise detection limit of a high-speed readout rate.
Scientists used a spinning disc laser confocal microscopy setup, including an Andor iXon camera at 100-fps full frame, to capture this kinetic series (top to bottom) showing calcium flux change as caffeine is administered to muscle cells loaded with fluorescent dye.
High frame rates are necessary to study dynamic interactions between single biomolecules, to track single molecules and to study their fundamental transient blinking effects. End users also need high speed to record fast calcium flux processes, in accordance with the major temporal resolution requirements of intracellular ion signaling studies.
In such applications, traditional CCD detectors, while having high and broad quantum efficiency curves (especially back-thinned devices), low dark current (if effectively cooled) and negligible crosstalk, still suffer from a significant read-noise detection limit, particularly at the fast readout speeds required for dynamic low-light applications. Therefore, longer exposure times are necessary to collect enough photons to overcome this read-noise threshold, leading to a sacrifice in dynamic ability.
Technologies that can amplify signals above the read noise can effectively eliminate this detection limit. Intensified CCD technology is one such approach, but it suffers from many drawbacks, including complexity and cost. Another option first offered by Andor Technology in 2001 is electron-multiplying CCD technology, which has become known as on-chip multiplication. Built around L3 Vision sensor technology from E2V of Chelmsford, UK, the device offers an electron-multiplying feature that allows an image sensor to detect single-photon events without an image intensifier, thus avoiding the quantum efficiency and resolution limitations of intensifier tubes. This means that intensified CCDs will no longer be required solely for sensitivity purposes, but will remain essential for gated time-resolved measurements.
Near term, the most immediate and prolific development in this area will be the launch of back-illuminated electron-multiplying CCDs that promise to combine extremely high photon-collection efficiency with the ability to eliminate the read-noise detection limit. Other developments may entail combining electron-multiplying technology with microlens interline CCDs, multiple readout ports and increasingly faster readout. Other standard CCD developments may result in steadily increasing quantum efficiencies of less expensive front-illuminated devices, to help circumvent the expensive back-thinning process.
Contact: Colin Coates, Senior Scientist, Andor Technology Ltd., Belfast, Northern Ireland; +44 28 9023 7126.
Multidimensional Imaging of Cells
An application where CCD sensors are perhaps the only option involves the multidimensional imaging of cells. This technique refers to the collection of data sets that include multiple wavelengths over time, multiple Z depths over time or both. The net result is a multiwavelength, or three-dimensional, data set that can reveal the intricate movements of cellular structures such as organelles, vesicles or the cytoskeleton over time.
These studies are performed in living cell preparations, so the essential requirements are to reduce phototoxicity to the cells with the cumulative UV illumination. One way is to minimize exposure time per image acquisition, something that demands a highly sensitive sensor.
In addition, the system should collect the multiwavelength image sets as close in time to one another as possible to eliminate any artifacts due to changes in the distribution of the labeled proteins. This can be accomplished through back-to-back image collection, which requires that the imaging system allow tight synchronization between the camera and the wavelength switching or Z-drive hardware. The best way is for the camera to provide instantaneous feedback when an image frame is completed so that the filter wheel or the Z-drive can be activated immediately. This can be done at the hardware or driver level.
An example of a camera system optimized for multidimensional imaging is Roper Scientific’s CoolSnap HQ, which uses a high-resolution interline sensor with good quantum efficiency in the visible region. Combined with extremely low readout noise, this makes the camera sensitive enough for wide-field fluorescence microscopy. The camera also can provide a transistor-transistor logic mirror of the exposure time, as well as driver-generated interrupt at the end of the frame readout. Either option can provide tight timing and synchronization.
Alternatively, some researchers work with dual cameras on different ports of the microscope or on a single port that has a wavelength-splitting optical module. In this configuration, the ability to simultaneously trigger two cameras is essential to the fidelity of the data collection. When done properly, the simultaneous collection of data at two wavelengths increases Z-series speed over time.
In such an application, CCD sensors will outperform intensified cameras by enabling visualization of more details in the multidimensional data set. They also allow deconvolution on the images and, hence, even greater 3-D resolution than is possible with the intensified device. Noise would limit CMOS as a sensor option here. In addition, a CMOS sensor would be throwing away light because the fill factor is not as high in interline CCDs that have dual microlens structures.
CCD-based multidimensional imaging can reveal intricate movements of cellular structures with greater 3-D resolution than possible with intensified devices. Courtesy of Roper Scientific.
Further developments in this field will benefit from high-quantum-efficiency CCDs that can read from multiple ports simultaneously and that do so with low noise. Some very large imagers with multiple readout ports are available, but they are primarily for astronomy and other physical science applications.
Thus, an interline CCD with good quantum efficiency and low-noise readout capability from two to four ports would be useful for multidimensional imaging, and end users should expect to see such devices available within 12 months. These should enable higher-speed data-set collection for samples that are not photon-limited. Also, combining the use of fluorescent probes in the near-infrared with detectors with good quantum efficiency in those wavelengths should increase the viability of live-cell preparations for long-term imaging.
Contact: Mark Christenson, Business Manager of Life Sciences, Roper Scientific Inc., Tucson, Ariz.; +1 (520) 889-9933.
Where CCDs Are Outpacing CMOS Sensors
The higher image quality of CCDs over CMOS imagers continues to hold even in commercial imaging. To the shock of some, CCDs dominate as the image sensors in wireless personal devices. Various reports place current production of wireless handsets with integrated cameras at 4 million to 5 million units per month, driven by the demand of Far East markets.
Consumption is projected to further increase as the application leadership in wireless technology from the Far East migrates to handsets destined for Europe and North America. The CCD producers supplying sensors to the vast majority of imaging handsets are straining to meet demand.
So what fueled the overwhelming success of CCD image sensors in an application once heralded by some as the triumph ground for CMOS imagers? Throughout the last decade, as CMOS technology has matured, CCDs have made continuing progress to reduce power dissipation through lower clocking voltages. Micropackaging technology has evolved so that two-chip solutions could be both cost- and space-competitive with single-chip CMOS proposals. The separation of the imager function in CCDs from the clocking and signal processing delivered overall image quality superior to that of integrated devices, with only minimal trade-offs in price and size. Above all else, consumers value the image quality and sensitivity that CCDs deliver.
When CMOS developers endeavored to improve image quality, they required wafer processing customization and adaptation that moved them away from the economies of scale of mainstream logic and memory integrated circuit production. In parallel, CCD technology derived cost-effectiveness from piggybacking on the production economies of numerous other high-volume and growing CCD applications. It gained additional momentum from an ability to achieve smaller useful pixel sizes than CMOS imagers, further supporting the drive to small size and low cost to propel image sensors into cellular devices. CCDs’ success is a sharp departure from where many thought the image sensor industry would be today.
CMOS technology has been a success, too, but in different ways than some expected. It has achieved greatest marketplace traction in cost-challenged, moderate-performance applications such as optical mouse sensors, PC videoconferencing cameras, low-end scanners and noncontact sensors, certain machine vision sensors, in vivo medical devices, and security and toy cameras. CMOS imagers also have found niches in exotic industrial and scientific applications. They have come to complement CCDs and have expanded the tool kit of imaging technologies available to serve industrial, consumer, commercial, scientific, medical and military/aerospace applications.
The industry is settling down to an era of coexistence for CCD and CMOS imagers. The hype and divisive debate of years past have largely moved into the background. CCDs are now seen as the technology of choice where image quality is of primary importance. CMOS technology is strongest in applications that require aggressive integration along with limited image quality.
We can expect a pace of regular improvement for both imaging technologies. They will compete, but in a more structured manner than was the case in the recent past.
Contact: Dave Litwiller, vice president of corporate marketing and business development, Dalsa Corp., Waterloo, Ontario, Canada; +1 (519) 886-6000.
- An instrument consisting essentially of a tube 160 mm long, with an objective lens at the distant end and an eyepiece at the near end. The objective forms a real aerial image of the object in the focal plane of the eyepiece where it is observed by the eye. The overall magnifying power is equal to the linear magnification of the objective multiplied by the magnifying power of the eyepiece. The eyepiece can be replaced by a film to photograph the primary image, or a positive or negative relay...
MORE FROM PHOTONICS MEDIA