CCD turns 40
Invented by Willard Boyle and George Smith at Bell Labs, the charge-coupled device (CCD), used in imagers from high-end cameras to space probes, is 40 this October.
The charge-coupled device (CCD) had a quiet start: the result of a post-lunch brainstorming session between Willard Boyle and George Smith one sunny October afternoon at AT&T Bell Labs in New Jersey. Boyle had already produced the first continuous ruby laser, and had a nice big office to show for it, but his boss, semiconductor pioneer Jack Morton, was not about to allow him to relax. Telling Boyle to leave transistors alone for a while, he urged him to "come up with something different".
Borrowing from the then popular notion of using magnetic bubbles to store data, Boyle and Smith considered how pockets of charge could be moved in a silicon matrix. They proposed a device that would temporarily hold charge within a given cell and then shift that charge to an adjacent cell by applying a suitable pulse. Whole rows of charges could be moved together to a collecting row, virtually without loss, and then individually stepped towards a device that would measure the charge in each cell. Given that differing amounts of charge could represent data values, it was clear that the CCD would make a capable data storage device.
An hour or so later, they were deciding what to call it. The online histories make it seem like a film script: "It's got charge. And we're moving the charge around by coupling potential wells," said Smith; "Let's call it a charge coupled device," said Boyle.
The truth was probably somewhat less succinct, but when asked about the course of events those 40 years past, George Smith confirmed: "In a discussion lasting not more than an hour, the basic structure of the CCD was sketched out on the blackboard, the principles of operation defined, and some preliminary ideas concerning applications were developed." Indeed, subsequent progress was also rapid. "In less than a week, masks were made and devices were fabricated and tested", says Smith.
Although the research was initially aimed at producing an electronic analogue to the existing 'bubble memory', the CCD found its true vocation as a solid-state imager. By focusing an image on an array of cells, light values could be converted to charge and stored temporarily in the CCD. Individual values could then be shifted out of the CCD, stored in another medium or transmitted to another device.
Its co-inventors had already foreseen this application. "The use of the CCD as an imaging device was discussed in my very first notebook entry describing the structure of the CCD and its use as a shift register", says Smith. "Bell Labs required all important ideas and experimental results to be put in the official notebook because lawyers insisted that entries had to be made as soon as possible after anything occurred which might result in a patent. This was used to establish priority and legally puts the inception of the imaging application at the same date as the shift register application."
Once the CCD had been identified as an imager, the cells could be relabelled 'pixels', or picture elements, a term coined in the mid-1960s to describe the elements of pictures returned from spacecraft sent to the Moon and Mars (though the word's detailed etymology is unclear).
Of course, the resolution of today's digital imagers is measured not in pixels, but in megapixels - an indication of how the field has developed in the past four decades. It is interesting, however, that this came as no surprise to the inventors: "I predicted that a one megapixel area device was possible," says Smith, proudly referring to his original notebook.
From the first experimental imaging CCD with its eight pixels, and the publication of the first paper in the Bell System Technical Journal of 29 January 1970, it was clear that the CCD had great potential. Unfortunately for those engaged in predicting the development of the semiconductor market in the 1960s, the way the CCD would be used was not so obvious.
Gordon Moore, co-founder of Intel Corporation but better known for his eponymous law of semiconductor development, was asked to write an article for the 35th anniversary issue of Electronics magazine in 1965, predicting developments in the semiconductor components industry over the next ten years. In essence, his prediction was that chip complexity would double each year.
Moore revised his projection when asked, ten years later, to deliver a speech at the International Electron Device Meeting and cut the pace of development of a doubling in density every two years. But it was the CCD that Moore saw as the pacemaker. In his 1995 paper 'Lithography and the Future of Moore's Law' he admits that what he couldn't know was that "CCD memories would not be introduced".
In effect, explained Moore, the CCDs were too sensitive to incoming alpha particles to form a reliable memory: "As memories, they became just that...memories". Luckily, the phenomenon that made the CCD a poor memory would make it an excellent imaging device.
The first commercial imaging CCD was produced by Fairchild Electronics, the former employer of Moore and colleagues at Intel, in 1974 with a format of 100 × 100 pixels, but the charge transfer efficiency of the device was so low that some users of existing imaging technologies were less than enthusiastic. Professional astronomers preferred to stick with their high-definition glass plates and space technologists with their vidicon-based TV cameras.
However, the very next year, arguably less critical users began to adopt the first CCD-based TV cameras for commercial broadcasting and use flatbed scanners incorporating the first CCD integrated chip, a 500-sensor linear array from Fairchild.
Kitt Peak CCD
By 1979, even the astronomers were convinced. RCA had introduced a 320 × 512 pixel liquid nitrogen-cooled CCD system that would see first light on a 1m telescope at Kitt Peak National Observatory. Cooling reduced the troublesome noise of dark current, facilitating the long exposures required by astronomers. According to Gregory Bothun of the University of Oregon, "early observations with this CCD quickly showed its superiority over photographic plates". The quantum efficiency was at least a factor of 50 higher (in the red), said Bothun, and the device itself was highly linear (in contrast to photographic plates which are highly non-linear), making it easy to calibrate.
Long before CCDs found their way into mass-market cameras, the most common source of images derived from CCD sensors was, perhaps surprisingly, the spaceborne imager.
In fact, the concept of producing still pictures in a digital domain arose from a proposed space application. In 1961, at the American Rocket Society's annual convention, Jet Propulsion Laboratory engineer Eugene Lally presented a paper on 'Mosaic Guidance for Interplanetary Travel'. He described the design of a camera employing a mosaic array of photodetectors, discrete optical sensors whose output was digitised to produce a map of star and planet positions that could be used for the navigation of manned spacecraft to Mars. It was, in essence, the first presentation of the concept of digital photography. The only thing missing was the availability of something like a CCD.
Following the development of the first practical CCD arrays by Fairchild in the mid-1970s, there was the inevitable period of qualification before such hardware could be specified for satellites. Initially, for operational missions, this was restricted to linear arrays used as so-called 'pushbroom scanners' on Earth observation satellites.
A pushbroom scanner produces a line of data at a time and uses the orbital motion of the satellite to build a two-dimensional image, rather like a fax machine or photocopier which moves the paper instead. And like a broom, the scanner sweeps across the Earth in a so-called "swath", with a swath-width governed by the width of the sensor array and the camera optics. The first spaceborne pushbroom imager, MSU-E, was flown on the Russian Meteor-Priroda 5 (Meteor I-30) spacecraft launched in June 1980.
The pushbroom scanner has advantages over the optical-mechanical line scanner in that it has no moving parts, lower mass, lower power requirements, longer operational lifetime and higher geometric resolution - all of which are a good fit for space missions where mass and power are severely limited and mechanical breakdowns are unrepairable.
By the time the first commercial imaging satellite, SPOT-1, was launched in February 1986, the scanners had become quite advanced. Thanks to a brace of CCD line scanners, it was the first non-military satellite to deliver image resolutions of 10m in panchromatic (black-and-white) mode and 20m in multispectral (colour) mode. This involved sampling 6,000 and 3,000 detector elements, respectively, from a set of four CCDs, each with 1,500 pixels.
Not surprisingly, the debut of such a discerning instrument caused concern among military authorities, as it had the resolution to spot aircraft on runways and missile silos. Since then, the commercial space imaging sector has grown accustomed to image resolutions of less than 1m.
Although several space telescopes had been launched prior to the Hubble Space Telescope, which was deployed in 1990, none of them had galvanised the public imagination and gained the attention of newspaper and magazine editors to anything like the same extent. This is due largely to the unprecedented detail and uncompromised beauty of the images it has produced.
The telescope originally carried five main observing instruments, including cameras, spectrographs and a photometer, but it was the Wide-Field/Planetary Camera that was fitted a CCD detector. Although its array of only 640,000 pixels - a square with 800 elements on a side - seems small by today's standards, this was because the instrument was designed in the early-1980s. The telescope's launch had been planned for 1986, but the Space Shuttle Challenger accident delayed it until 1990.
The total imaging area was increased, however, using a mosaic of four separate CCDs to give a total 2.56 megapixels, which was good for the time. Bolting on a 'telephoto' with a 2.4m-diameter primary mirror and deploying it beyond the murky turbulence of the Earth's atmosphere did much to improve the resultant images.
Hubble to Kepler
CCDs have moved on considerably since the Hubble telescope was built, as shown by the array on Nasa's Kepler Space Observatory, launched this March. Designed to look for Earth-sized planets in other solar systems, Kepler carries a Schmidt telescope with a 1.4m primary that focuses light onto an array of 42 CCDs, each with 2200 × 1024 pixels, giving a grand total of about 95 megapixels.
For earthbound applications, increasing the pixel count on a single CCD has been a key goal, especially with the rise of digital photography. And the CDD has come under pressure from competing CMOS sensors. Many of the recent launches by makers of professional-grade single lens reflex cameras have included CMOS rather than CCD imagers for cost reasons.
For specialist use where budgets are higher, the CCD has offered better performance. Canadian supplier Dalsa Semiconductor delivered the world's first 35mm full-frame imager, a six-megapixel device, as long ago as 1998. Although, as the company points out, "image quality depends on more than just pixel count" - for example, high throughput, low noise and high dynamic range - it was inevitable that the 100 megapixel barrier would be broken, which is just what DALSA achieved in 2006 with its 111 megapixel CCD. The device itself measures 10cm on a side and incorporates an array of 10,560 × 10,560 pixels, each 9µm across.
The record-breaking device was developed for the Astrometry Department of the US Naval Observatory for the accurate measurement of star positions, but as one waggish blogger put it: "Don't hold your breath for the consumer version".
But the CCD has survived 40 years without staying for long in any consumer application. And it keeps finding applications.