Charge-coupled_device

# Charge-coupled device A **charge-coupled device** (CCD) is an [[Integrated_circuit|integrated circuit]] built from a row or grid of closely spaced [[Capacitor|capacitors]] that can pass an electric charge from one to the next under the control of a sequence of clock voltages, the way a bucket brigade passes water hand to hand. Because the charge a capacitor holds can be set by the number of [[Electron|electrons]] a [[Photon|photon]] frees inside it, the same structure that shifts charge along a line also makes an efficient way to record and read out an image: expose an array of these charge wells to light, then walk each row's accumulated charge out to a single output amplifier one clock cycle at a time. For three decades the CCD was the dominant image sensor in cameras generally; ordinary photography has since moved to active-pixel CMOS sensors that read every pixel directly rather than shifting charge across the chip, but the CCD remains preferred wherever the cost of extra circuitry buys a cleaner, more uniform signal, most often in astronomy, spectroscopy and other scientific and medical imaging, and in specialised variants built for very low light. The primary microsim on this page lets the reader drive a small pixel array through the same four stages a real sensor cycles every frame — exposure, parallel shift, serial shift and readout — one keystroke at a time, or set it running automatically and watch the whole bucket brigade play out. Exposure fills each capacitor with charge as photons arrive; parallel shift moves a whole row down into a serial register; serial shift then drains that register one pixel at a time through an output amplifier, filling a growing readout buffer with the finished row. ## History The charge-coupled device originated at [[Bell_Labs|Bell Telephone Laboratories]] in September 1969, when Willard S. Boyle and George E. Smith sketched the idea on a blackboard while thinking about a new kind of semiconductor memory built from a line of coupled capacitors rather than individual transistor cells.[^ccd-bstj] They and their colleagues quickly showed that a chain of metal–oxide–semiconductor capacitors could not only store a bit as the presence or absence of charge but shift that charge, undisturbed, from one capacitor to the next under clock control, and that the same structure would fill its wells with charge simply by being exposed to light, without any separate photodetector. Bell Labs built working charge-transfer imagers within a few years, and by the mid-1970s the CCD had become the sensor behind the first solid-state television cameras. Boyle and Smith shared half of the 2009 Nobel Prize in Physics, the other half going to Charles K. Kao for work on optical fibre, for the invention of an imaging semiconductor circuit, the CCD sensor.[^ccd-nobel] Engineer Michael Tompsett, also at Bell Labs, is widely credited with turning the charge-transfer concept into a working imaging device and building some of the earliest solid-state cameras, though the exact division of credit for that step has been debated since the Nobel announcement.[^ccd-tompsett] ## Basics of operation At the heart of a CCD is the metal–oxide–semiconductor [[Capacitor|capacitor]]: a conductive gate sits above a thin insulating layer of [[Silicon_dioxide|silicon dioxide]], which in turn sits above a doped [[Silicon|silicon]] substrate. A positive [[Voltage|voltage]] on the gate repels the substrate's majority carriers away from the interface and draws minority carriers, electrons in the p-doped substrate common to imaging CCDs, toward it, opening a small potential well that can hold a packet of charge. A photon absorbed nearby frees an electron that drifts into this well and stays there, so the well fills at a rate set by the local light level; over an exposure the collected charge is `Q = integral of I_ph(t) dt`, the integral of that [[Electric_current|photocurrent]] over time, which is exactly the quantity the microsim's own equation display tracks. The device becomes "coupled" because neighbouring wells sit close enough, separated only by a thin barrier under an adjacent gate, that cycling a sequence of clock voltages on the gates in the right order tilts the potential profile and walks the whole charge packet sideways into the next well without spilling it, a bucket-brigade transfer repeated, gate by gate, until a full row has been shifted into a special output row called the serial (or horizontal) register. From there the same clocking shifts each pixel's charge, one at a time, to a single output node, which converts the arriving charge into a voltage that a following amplifier and, eventually, an [[Analog-to-digital_converter|analog-to-digital converter]], turn into a number. ## Detailed physics of operation Two things beyond this outline matter in practice: what happens physically at the instant a photon is absorbed, and what the manufacturing choices behind a real device look like. ### Charge generation A photon reaching the silicon is absorbed only if its energy exceeds the roughly 1.1 electronvolt band gap between silicon's valence and conduction bands; absorption promotes one [[Electron|electron]] across that gap, leaving behind a mobile hole, and the CCD's electric field sweeps the electron toward the gate interface while the hole drains away into the substrate. How deep a photon penetrates before it is absorbed depends on its wavelength: blue and ultraviolet light are absorbed within a fraction of a micrometre of the surface, while red and near-infrared light penetrate several micrometres deeper, so a conventional CCD illuminated through its patterned gate structure loses some blue response to absorption and reflection in the gates themselves. Thinning the silicon and illuminating it from the back, through bare substrate rather than through the gates, recovers much of that lost blue and ultraviolet sensitivity and is the standard way to build the most sensitive scientific and astronomical sensors, at the cost of a harder, more expensive fabrication process. ### Design and manufacturing A CCD begins as a wafer of high-purity [[Silicon|silicon]], selectively [[Doping_(semiconductor)|doped]] with impurities such as [[Boron|boron]] or [[Phosphorus|phosphorus]] to set the conductivity type of the substrate, the channel in which charge is stored and transferred, and the channel stops that isolate one column from its neighbours; wherever a p-type region meets an n-type region the result is a [[P–n_junction|p–n junction]], several of which are used deliberately around the imaging area to define its boundaries and to drain excess charge. A thin, carefully grown layer of silicon dioxide serves as the gate insulator, and polycrystalline silicon gate electrodes, typically two, three or four separate, interleaved sets of them per pixel, each driven by its own clock phase, are patterned on top using the same photolithographic steps as other [[Semiconductor_device_fabrication|semiconductor device fabrication]]. The number of separate gates per pixel is a real design trade-off: more phases give finer control over the potential profile and cleaner charge transfer, but cost silicon area that would otherwise be light-sensitive, so a design that must maximise the fraction of each pixel able to collect light will favour fewer phases even at some cost in transfer efficiency. ## Architecture Beyond this generic design, several specialised architectures modify the basic pixel-plus-shift-register plan to trade extra silicon or extra circuitry for an advantage in speed or sensitivity. ### Frame transfer CCD A frame transfer CCD doubles the imaging array with an identical, light-shielded array of the same size immediately alongside it. Once an exposure ends, the whole image is shifted rapidly, far faster than it could be read out, into the shielded store, where it waits to be read a row at a time while the next exposure is already accumulating in the now-empty imaging area. Because the transfer into the store is so much quicker than a full serial readout, the smear that a slow, unshuttered readout would paint across a moving scene is greatly reduced without needing a mechanical shutter, at the price of roughly doubling the chip area for a given resolution. ### Intensified charge-coupled device An intensified CCD (ICCD) places a microchannel-plate image intensifier ahead of an ordinary CCD, coupled to it through a fibre-optic faceplate or a lens. The intensifier converts each incoming photon into a burst of electrons at a photocathode, multiplies that burst through the microchannel plate, and reconverts it to light on a phosphor screen bright enough for the CCD behind it to record easily; the intensifier's own gain, not the CCD's sensitivity, sets how faint a signal the combination can see. Because the intensifier can be switched on and off in nanoseconds, an ICCD can also act as a very fast optical shutter, a property used in gated spectroscopy and range-gated imaging. ### Electron-multiplying CCD An electron-multiplying CCD (EMCCD) instead multiplies the charge itself, inside an extended serial register held at a much higher clock voltage than an ordinary transfer needs. At each stage of this extended register the higher field gives a small, fixed probability that a stored electron will trigger impact ionisation and add a second electron to the packet; the probability per stage is tiny, but multiplied over several hundred stages in series it compounds into a large overall gain, enough to lift a single photoelectron's signal above the [[Noise_(electronics)|noise]] of the output amplifier that would otherwise have buried it. The gain is bought at the cost of extra statistical noise in the multiplication process itself, so an EMCCD trades away some of its [[Dynamic_range|dynamic range]] for the ability to see single photons at all. ## Use in astronomy Astronomy adopted the CCD earlier and more completely than most other fields, because the qualities that matter for a scientific measurement are exactly the ones a CCD offers and a photographic plate does not. A CCD's response is linear over most of its range, twice the light gives twice the recorded charge, where a plate's density responds to light through a nonlinear characteristic curve that must be calibrated and that fails badly for both very faint and very bright sources; a CCD is also far more efficient, converting a much larger fraction of incident photons into a recorded signal than a plate's few per cent. Because the goal in astronomy is usually to detect the faintest possible source against the sky background, astronomical CCDs are commonly cooled, thermoelectrically, or with liquid nitrogen or other [[Cryogenics|cryogenic]] cooling for the most demanding instruments, to suppress dark current, the thermally generated charge that would otherwise accumulate in a well and erode the [[Signal-to-noise_ratio|signal-to-noise ratio]] of a long exposure. Large survey telescopes and space observatories, including instruments flown by [[NASA]], have tiled dozens of individual CCDs into a single focal plane to cover a wide field at high resolution, an approach that would have been impractical with photographic plates of comparable uniformity.[^ccd-astro] ## Color cameras A CCD's stored charge records only how much light reached each well, not what colour it was, so a single sensor is monochrome unless something upstream of it discriminates by wavelength. Cinema and broadcast cameras built for the best possible colour image often use three separate sensors: a dichroic prism block splits the incoming light into red, green and blue components and directs each to its own CCD, so every pixel position gets a full-strength measurement in all three colours at the cost of three sensors, three sets of readout electronics and precise optical alignment between them. Far more common, because it needs only one sensor, is to cover the array with a mosaic of tiny colour filters, typically red, green and blue arranged so that green, to which the eye is most sensitive, is sampled at twice the rate of the other two; each pixel then reports only the one colour its filter passed, and a full-colour image is reconstructed afterward by interpolating the missing colours from each pixel's neighbours, a [[Digital_image_processing|digital image processing]] step usually called demosaicing. The trade is resolution and a small loss of true per-pixel colour accuracy for a camera that costs a fraction of a three-sensor design. ### Sensor sizes Sensor sizes are still commonly quoted in fractions of an inch, 1/2.3, 1/1.8, two-thirds and so on, a convention inherited from the vacuum-tube cameras that CCDs replaced, where the number named the outer diameter of the round tube rather than the diagonal of the rectangular light-sensitive image it produced; a nominal one-inch format therefore has an actual diagonal considerably smaller than one inch.[^ccd-sensorsize] The consequence that matters for image quality carries over regardless of the naming: for a given pixel count, a physically larger sensor gives each pixel more area to collect light, which generally means less noise and a wider usable range of brightness in one exposure, at the cost of a larger, heavier and more expensive lens to cover it. ## Blooming Every potential well has a finite full-well capacity, the largest charge it can hold before the barrier that is supposed to confine it stops working. Point too much light at one part of a scene, a reflected glint of sun, a light fixture in frame, and the wells under that highlight fill completely and then overflow, most easily along the column in the direction charge is normally shifted, since the barriers between wells in that direction are the ones designed to be lowered on command rather than to hold indefinitely. The result is blooming: a bright vertical streak radiating from the overexposed highlight into pixels that received no such light themselves, an artefact with no analogue in film. Many sensors add a dedicated overflow channel, an anti-blooming drain running beside each column, that carries away charge once a well nears capacity before it can spill into a neighbour; the drain buys resistance to blooming at the cost of some of the pixel area, and therefore some of the full-well capacity, that would otherwise be available to hold genuine signal. ## Microsims The microsim on this page carries the pixel array itself through the pipeline described above. An eight-column, four-row grid of wells (the array can be widened or narrowed from three to sixteen columns with the +/- keys) fills during an EXPOSURE phase as small photon particles arrive and each well's charge climbs toward the fixed maximum the sketch uses to stand in for a real full-well capacity, the same saturation point that, in a real sensor, triggers the blooming described above, though the sketch does not go on to animate charge spilling sideways into a neighbouring well (an illustrative simplification). Pressing the space bar advances the pipeline one step at a time: a PARALLEL SHIFT moves the bottom row down into the serial register, with every other row shifting down behind it and a new, empty row appearing at the top exactly as continued exposure would refill it, and a SERIAL READOUT then drains that register rightmost-pixel-first through the output amplifier, adding each value to a growing readout buffer drawn as a strip of grey tones below. The `a` key runs the whole sequence automatically; `r` resets the array to a fresh exposure. The bottom-right of the display carries the sketch's own equation for what a well is doing during exposure, `Q = integral of I_ph(t) dt`, charge as the running integral of photocurrent, the same relationship described in Basics of operation above. *Try:* Press `a` to auto-run the sequence and watch the HUD's `readout=` count climb to 32px as all four rows of the default eight-column array finish draining through the amplifier. ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Charge-coupled_device) : [Wikitube](https://en.wikitube.io/wiki/Charge-coupled_device) Skeleton mirrored at revision 1361592451. Prose, emphasis and the microsims are Wikitube's own. ## See also - [[Digital_image_processing]] - [[Semiconductor_device_fabrication]] - [[Analog-to-digital_converter]] - [[Signal-to-noise_ratio]] - [[Photon]] - [[Capacitor]] ## References Standard semiconductor-device physics described here, the metal–oxide–semiconductor capacitor, the photoelectric generation of carriers in silicon, and the band-gap figure used in Charge generation, is textbook material and is not separately footnoted, per Wikitube style guide §6.1. [^ccd-bstj]: Boyle, W. S.; Smith, G. E. "Charge Coupled Semiconductor Devices." *Bell System Technical Journal*, vol. 49, no. 4, 1970, pp. 587-593. [^ccd-nobel]: The Royal Swedish Academy of Sciences. "The Nobel Prize in Physics 2009." Nobel Prize outreach, 2009. https://www.nobelprize.org/prizes/physics/2009/summary/ . [^ccd-tompsett]: Citation needed: a primary paper or patent record documenting Michael Tompsett's specific contribution to the first CCD imaging demonstration at Bell Labs, and a secondary account of the post-2009 credit dispute, would settle the exact wording used here. [^ccd-astro]: Citation needed: a NASA or observatory record for a specific tiled-CCD focal-plane instrument (chip count, mission and year) would let this sentence name one rather than describe the practice generically. [^ccd-sensorsize]: Citation needed: a manufacturer or standards-body reference for the historical vidicon-diagonal convention behind fractional-inch sensor-size names would settle the precise ratio. ## External links - [Charge-coupled device — live microsim](https://editor.p5js.org/sciencenibber/full/-qz3TPybl) - [Charge-coupled device — sketch source, p5.js editor](https://editor.p5js.org/sciencenibber/sketches/-qz3TPybl)