Основные понятия (Image Sensor Lexicon)

Temporal noise with a dominantly low frequency content. In this context the origin is typically in MOS or MOSFET interface states. A.k.a. flicker noise.

Light, as all electro-magnetic radiation, is absorbed (and detected!) in silicon over a characteristic depth. Blue light is absorbed in about 0.5 to 1 um; green light in a few um; red light in a layer over 5 to 10 um thick. Infrared light is absorbed in deeper layers. Ultraviolet light is typically absorbed (and not detected!) in the top oxide layers.

Pixel, containing an active (= power dissipating) element. In most cases the active element is a MOSFET in an amplifying or buffering configuration. The presence of only one MOSFET acting as a switch only does not make a pixel 'active'.

Analogue to Digital Converter: An electronic circuit that converts an analogue voltage, as comes out of an imager core, into a digital signal.

Properties of image sensors that are inspired by, or similar to, properties of the human (- eye). Examples: logarithmic response, RGB sensitivity, and log-polar geometry.

Active Pixel Sensor: Image sensor using active pixels.

Anti Reflective Coating: Dielectric layers on top of the receptor with specific thickness and refractive index, so that only a minimal fraction of the light impinging on the receptor is reflected.

Imager with a two dimensional array, or matrix, of pixels.

The average time for a pixel to saturate due to dark current alone, i.e. non-illuminated. The absolute upper limit for useful integration time.

An imager can be illuminated from the backside, if the silicon can be uniformly thinned to a few micrometers. If successful, there is no obstruction of the light by metalization and dielectric layers, and the Fill Factor and UV sensitivity can be very high.

Bipolar + CMOS: CMOS technology extended with bipolar transistors.

Mode of operation that is available only in some types of CCD (FTCCD). The charge of multiple pixels is accumulated and read in a single operation. This type of binning has no noise penalty. Binning is sometimes emulated in passive pixels, active pixels or in the digital domain. But here the noise advantage is less prominent.

A popular type of Red-Green-Blue CFA, invented by Bayer (Kodak).

Background Limited Integrated Performance: Camera operation in a condition where the noise is dominated by the inherent shot noise in the background illumination level.

Substrate. In MOSFETs, the symbol for bulk or substrate is B.

Normally the source-drain current through a MOSFET is regulated by the gate potential. Nevertheless, the bulk potential has a generally unwanted parasitic effect too.

Similar to a MOSFET inversion layer, but by the proper game of implantations the charge is separated from the interface.

The better "buried channel" CCDs store and transfer their charge in a buried channel under the electrodes.

Charge Coupled Device: Originally the name of a structure invented in 1970 of adjacent MOS gates ("electrodes") that allows both confinement (storage) and transport of free charge. Later the name of the modified MOS technology in which the device is made. Even later the name of image sensors based on the CCD principle. And finally the generic name of a solid state image sensor.

Correlated Double Sampling: A method to read the differential magnitude of a charge packet, thereby canceling a source of FPN, as well as the kTC noise that is caused by the reset of the charge packet.

Color Filter Array: The mosaic of colour filters overlaid on the pixel array.

Charge Injection Device: CCD-based technique and image sensor using this technique. Detection based on local charge transfers only and non-destructive read out.

Two-dimensional representation of color space, representing all physical colors, i.e. more than what can be generated with the color triangle.

CMOS Image Sensor: go straight to jail, skip three turns, do not collect $200.

Charge Modulation Device: Image sensor technology based on the detection and charge sensing in buried channel MOSFETs.

Complementary MOS: The mainstream mass manufacturing Silicon technology. Characterized by the fact that both nMOSFETs and pMOSFETs ("complementary" pairs) can be manufactured on the same wafers.

Chip On Board: Mounting the image sensor chip directly on the PCB, without package. Cheap but a bit fragile, unless protected well.

Or the question: Which part of the system (digital or analogue) intelligence is put on the same chip as the analogue imager core?

The raw image from an imager with CFA has only one color component per pixel. Trivial or more clever reconstruction algorithms must restore the missing color components for every pixel.

Charge amplifier or voltage amplifier that is located at the edge of columns op pixels. Serves to buffer or sense the signal of one, selected, pixel in the column. Typically, the signals of the column amplifiers are multiplexed to one imager core output.

As a spot or a mark in the face. Small defects, localized non-uniformities, unwanted patterns, blemishes, dust particles, etc. that degrade the image quality.

Cathode Ray Tube: The classic display device, e.g. TV or computer monitor.

Digital to Analogue Converter: as ADC, but read the explanation back to front.

The parasite of photocurrent. In the absence of light, a non-zero current will flow in the photodiode, spoiling the image and adding its own shot noise.

"Detecting" apparatus. Most people don't tell the difference with "sensor". Sounds more "active" than "sensor" - or does it sound more expensive? The nuclear physics community considers a detector as the bigger apparatus containing many smaller "sensors".

Insulating, transparent layers on top of the IC. Typically SiO2, SiN and organic materials.

Electronic device with two terminals, that conducts the current in only one direction. Typically it consists of a p-n junction, the juxtaposition of an n-type semiconductor and a p-type semiconductor, typically on the same substrate.

Method to extend the dynamic range of a normally linear-transfer imager, by combining the images taken with a long and a short integration time into one image. The resulting electro-optical transfer curve is bi-linear.

Double Sampling: A method to read the magnitude of a charge packet, canceling a main source of FPN.

Digital Signal Processor: software-programmed or hardwired logic device, specialized in performing many calculations at high speed on a signal. The raw image coming from an imager can be/needs to be post-processed in the digital domain.

A pixel array features some rows/columns of real but unused pixels at its perimeter, thus ensuring that the "useful" pixels are not affected by peripheral circuitry.

ratio of highest non-saturating optical flux on the smallest detectable flux for a sensor. Equates to the signal to noise ratio if and when the sensor features a linear transfer curve.

Elementary particle with negative unit charge. Or was it a wave?

A method to operate an electronic image sensor in a way that the effective integration or exposure time (shutter time) is variable. The result is the same as that of the mechanical shutter found on the classical photo camera.

Electro-Magnetic radiation: Visible light is just one small part of the huge EM spectrum.

Electro-Magnetic Interference: System noise.

Epitaxial wafers. Most CMOS image sensors need processing on highly doped wafers with on top a few microns thick "epitaxial layer" of lowly doped silicon.

Field Emission Array: Micro-machined silicon device that combines the light detecting mechanism of a vidicon tube and the circuit possibilities of an integrated circuit.

Full Frame CCD: A FT CCD lacking storage area. Needs an external shutter.

FF: ratio [%] of the light sensitive area of a pixel on its total area. Should be quoted in conjunction with spectral response.

Popular circuit technique for the conversion of (photo) charge to a voltage. Used in most CCDs and in most APSs.

Light power per unit area [W/m2].

Ratio between the opening (diameter) of the lens and its focal length. A high f-number means a large attenuation of the light power.

Fixed Pattern Noise, or spatial noise. Is the unwanted static (DC) variation of the response of all pixels in the image. Expressed as an offset in volts or in ADC bits.

The electronic form of one image.

The frequency at which consecutive images are taken by a camera.

The time it takes for an imager to capture and output an image. Equals 1/frame_rate.

Frame Transfer CCD.

Name of FillFactory's family of random addressable, logarithmic response, image sensors. Mainly intended for industrial and automotive vision. The named was given by Bach fan Bart Dierickx when IMEC management demanded something less arcane than "XYW", the latter being the first FUGA-like CMOS APS IMEC made back in 1986. Incidentally, FUGA means 'flight'. A real premonition if you take into account that the FUGA15 was to become the first CMOS APS in space.

Allowing for more charge capacity, better dynamic range, reduced noise and delivery of optical resolution, Kodak's full-frame sensors allow for the full RGB matrix to be captured instantaneously.

Full-frame CCDs consist of a parallel CCD shift register, a serial CCD shift register, and a signal-sensing output amplifier. In a full-frame CCD, the exposure is controlled by a mechanical shutter or strobe to preserve scene integrity, because a parallel register is used for both scene detection and readout.

Images are optically projected onto the parallel array, which acts as the image plane. The device takes the scene information and partitions it into discrete elements, which are defined by the number of pixels "quantizing" the scene. The resulting rows of information are then shifted in a parallel fashion to the serial register, which shifts the information to the output as a serial stream of data. This process repeats until all rows are transferred off chip. The image is then reconstructed as dictated by the system.

Response correction in cameras originally introduced to counteract the non-linear response of CRTs. The typical gamma correction takes the signal luminance to the power of 0.45

Control terminal of a MOSFET. In a CCD the gates are called "electrodes".

The zero reference voltage in most electronic circuits. In our science it is typically defined as equal to the substrate voltage, sometimes simplistically called substrate potential. At FillFactory ground is found about 12 meters below our floor.

In a semiconductor the local absence of a bound electron, acting to all intents and purposes as a free charged particle with positive unit charge, and thus opposite of a negatively charged free electron. A hole is absolutely not the same as a positron, the anti-electron, which also has a positive unit charge, but which is a real physical (but rare, and thus expensive) particle.

Fear of emptiness (Latin). The natural fear of an image sensor designer for empty silicon, and her/his insuppressible desire to fill this space with circuits, test structures, yield killers, daft logos, etc.

Imager consisting of a CMOS readout array chip and a separate detector chip.

FillFactory's family of integrating APS sensors with on chip FPN correction, intended for photography and related applications. "IBIS" stems from I-BIS, I2, i.e Integrating Imager. Eye-Bis is also a neat explanation.

InterLine CCD.

Digital domain operations in hardware or software intended to convert the raw image from the imager to a "nicer" image. Typical image processing operators are: spatial and temporal filtering, missing pixel correction, offset and gain correction, gamma correction, color (CFA) reconstruction, white balance, saturation and hue correction.

Combination of the words image and sensor.

The essential part of an image sensor chip. In fact nothing more than the pixel array and the immediate peripheral circuits required to operate the pixel array.

Exposure time or electronic shutter time. The time span during which light is being converted to charge and accumulated in a pixel.

Smart sensor. The question is: what is "intelligent"? The first image sensor that can give an answer to that question is really intelligent.

In this science the border between Silicon and whatever is on top. The favorite meeting place for embarrassing things like dark current generation centers and 1/f noise centers.

In the CCIR (PAL) and EIA (NTSC) video formats, the image information is divided in odd and even (sub) frames. Alternating frames contain only the odd or even lines. Imagers for these formats should read their lines in the same sequence.

This type of sensor utilizes separate arrays for image capture and charge transfer, allowing images to be read out while the next image is captured. Interline transfer CCD sensors are commonly used in lower-end digital cameras, video cameras, and broadcast cameras for motion capture.

Undoped Silicon has (in theory) a low concentration of both holes and electrons. To distinguish it from n- or p- type, it is indicated by "i".

Under the influence of an electric field an n-type material can contain a majority of holes, and thus "invert" effectively to p-type, or vice versa. The inversion layer under a MOSFET gate is the conducting path of the MOSFET.

Silicon photodiodes are sensitive for light with wavelengths between 350 and 1100 nm. The human eye sees light with wavelengths between 350 and 750 nm (the "visible spectrum"). An IR cut-off filter is a piece of special glass, or another material, put in front of the image sensor and absorbing or reflecting the light outside the visible spectrum. Quite necessary for color sensitive imagers.

Empirical measure of the light sensitivity or speed of photographic film. Transferring this to an electronic image sensor spec is a real shot in the dark.

Bolzman's constant x absolute temperature x capacitance. The thermodynamic uncertainty on the amount of charge left on a capacitor after disconnecting it from a DC voltage source. kTC noise is also called reset noise.

Liquid Crystal Display: One type of flat panel display, based on the light polarizing capabilities of so called liquid crystals in an electric field.

a light power (P) to voltage (V) conversion that can be approximated by a linear relation as: V = something x P. Transfer curve of our IBIS range.

Image sensor with an (almost) one-dimensional array of pixels, i.e. one or a few rows of pixels. Don't confuse it with a linear response image sensor!

A light power P to signal voltage V conversion that can be approximated by a logarithmic relation as: V = something x log(P). Transfer mode used in our high-dynamic range FUGAs.

Arrangement of pixels, organized in concentric circles, with a constant number of pixels per circle. The pixel size is smallest in the center and grows towards the edges. This mimics the human retina.

Human eye equivalent (i.e. subjective) of the objective light flux measured in W/m2. As the spectral sensitivities of the eye and Silicon are not matched, there is no general conversion factor between lx and W/m2. As a rule of thumb, consider 1W/m2 = 70lx (visible white light) to 180lx (visible + NIR).

Technique to increase the fill factor of pixels by giving each pixel a tiny individual lens that converges all light on the even-tinier photodiode below it.

Metal Oxide Semiconductor: In ancient times MOSFETs were made as a stack of aluminium (metal) on SiO2 (oxide) on Silicon (Semiconductor). Today the aluminium is in most cases replaced with Polysilicon, but the name persisted. Otherwise we would now be making POSFETs, and CPOS Image Sensors. Or perhaps even SOSFETs ...

MOS Field Effect Transistor: transistor type based on the MOS layer structure. The current between source (S) and drain (D) passes through the inversion layer that is controlled by the voltage at the gate (G).

Modulation Transfer Function: Ratio between the amplitudes of a sinusoidal pattern in the optical image and the resulting electronic image. A measure for a sensor's inherent sharpness of focus.

Near InfraRed: EM radiation between 750 and 1100 nm, invisible for the eye, but still visible for Silicon.

n-type MOSFET, a MOSFET in which the conduction between n-type source and drain happens by electrons.

A word with many meanings. In our science it is the unwanted fluctuation of the signal of one pixel over time. To differentiate from (static) FPN it is sometimes called "temporal noise".

Mode of operation of CCDs and other image sensors, where the size of a charge packet is read, without resetting it. This information can thus be read several times, so as to improve the S/N of the reading operation

Under zero electric field and equilibrium an n-type semiconductor contains a majority of electrons and few holes. n-type Silicon is usually obtained by doping with donor ions.

The Nyquist limit equals to the spatial frequency 1/(2 x pixel pitch) of a pixel array. Image information beyond this limit is "aliased" to lower spatial frequencies. In practice this aliasing appears as Moire effects and color artifacts.

Light impinging on a pixel can influence the signal of neighboring pixels through various mechanisms: diffraction of light, diffusion of photo charge towards the neighbor, or plain electrostatic cross talk. Together these make up optical crosstalk.

Pixel not containing an active element, so essentially a photodiode with access switches.

All silicon diodes or p-n junctions are sensitive to light. A diode that explicitly is designed for detecting light is a photo diode.

Light receptor made as a MOS structure. Can be understood as a p-n photo diode, where one of the parts is the inversion layer under the MOS.

Image sensor consisting of an array of passive pixels, where the passive pixels contain only a photodiode and a multiplexing switch.

The basic mechanism of light detection in any solid state imager. A quantum of energy from the impinging EM radiation (a photon) excites a bound electron to an unbound state.

The light quantum. The amount of energy exchanged in a single interaction between EM radiation and a charged particle.

p-n diode with a relatively important lowly doped (intrinsic) layer "i" sandwiched between the p-type and n-type regions.

Photo diode which is covered by an additional junction on the top. The goal is to pin the interface potential to a nearly fixed value, and to shunt the dark current centers at the interface.

From picture + element. Originally the atom of a photograph, of an electronic image, of an electronic display. In our profession the atom of an image sensor. Minimally contains a receptor and elementary means to feed the information to the outside world.

Type of flat panel display, based on the emission of light by the interactions of free electrons in an ionized gas (plasma).

p-type MOSFET, a MOSFET in which the conduction in the inversion layer between p-type source and drain happens by holes.

Mode of operation of a large, slow-scan, imager: It is read at high frame rate but at reduced resolution.

Local property of a position in space. Energy that must be given to or taken from a hypothetical free electron to be stable at that certain position. Is a relative measure, thus it must be referred to some "reference". In absence of any reference the reference is the potential in vacuum at an infinite distance. Ever tried? For most earthlings potential is quite the same as "voltage".

Photo Response Non-Uniformity: non-uniformity in the slope of the photo-electric transfer curve between pixels. Also see FPN for offset non-uniformity.

Position Sensitive Device: A kind of solid state optical sensor that outputs analog signals that are measures of the position of a light spot on the sensor.

Under zero electric field and equilibrium a p-type semiconductor contains a majority of holes and few electrons. p-type Silicon is usually obtained by doping with acceptor ions.

A linear imager scanning a two-dimensional scene. Often used for earth imaging.

QE. Ratio [%] between the number of generated electrons and the number of impinging photons. Closely related to spectral response.

A mode of operation of an image sensor, where pixels are addressed and read individually and randomly, like a RAM or ROM.

The generic name of an elementary radiation sensitive device. Photodiodes, film grains, eye cones and rods are examples of light receptors.

In the imager context, reset means clearing the charge contents of a pixel to the initial dark value.

Red Green Blue: In most color cameras light is separated in 3 components labeled red, green and blue. This matches the 3-color sensitivity of the human eye and the 3 phosphor emission of color TV displays.

Alternative, complementary, colors are sometimes used: Ye (yellow), Cy (cyan) and Ma (magenta). But the choice is not even limited to these 6 colors.

A.k.a. curtain shutter. Moving slit mechanical shutter as used in SLR photo cameras. Also used for the type of electronic shutter which does not expose all of the image's pixels in the same period of time. Contrary of synchronous shutter.

High-Speed Rolling Shutter

In Rolling shutter, each line in the image has the same amount of integration, however the start and end time of integration is shifted in time as the image is scanned (rolled) out of the sensor array, similar to a curtain shutter of a film camera. Although it is a pure electronic operation, the shutter seems to slide over the image.

An object, which moves during the typical 1/30 sec readout time, can display a stretched or skewed perspective, in the direction of motion. For example, a vertical line can appear tilted if the object moves several horizontal pixels during readout.

The faster the object moves the larger the tilt. The crispness or detail of the line is determined by the shutter speed (integration time). The extreme high-speed readout of the camera any such motion artifacts are minimized. The benefit of rolling shutter mode, is that exposure and readout are overlapping, enabling full frame exposures without reducing frame rate.

Random Telegraph Signal: The noise of a single trap or interface state. Is of the same origin as MOSFET 1/f noise

Signal to noise ratio: Ratio between the saturation signal (in volts) and the noise level (in volts RMS).

The light level (the input quantity) or voltage (the output quantity) at which the pixel has received so much light that its differential response drops to zero.

The property of a good sensor. A word with too many meanings to list here. The difference with detectivity and responsivity is not clear.

"sensing" apparatus. In our world a circuit that in some way is able to sense the amount of incoming electromagnetic radiation. Most people don't see the difference with "detector".

Digital circuit that generates the pulse trains that are necessary for the proper operation of an analogue imager core.

The fundamental statistical uncertainty on the amount of photoelectrons that are generated by light falling on a photoreceptor.

Single Lens Reflex: denomination for standard reflex 35 mm photo camera.

Imager combined with logic, on one chip. The chip should do something more than just outputting the image contents.

SR: Response of a photoreceptor over a spectral band. The ratio between the photocurrent and the light energy. Expressed in [A/W]. Depends on the wavelength of the light.

The total time impinging light has impact on the resulting frame. In the case of a rolling blade shutter, stare time = exposure time + frame time. With a synchronous, snap-shot, shutter, stare time = exposure time.

MOSFET gate or CCD electrode that stores charge in the inversion layer or buried channel underneath.

Mode of operation of an image sensor. Pixels are read sequentially, but not contiguously. I.e. some pixels are skipped, to obtain an image with lower resolution, and perhaps at a higher frame rate.

The bulk of the integrated circuit. The "useful" circuits normally lie only on or just below the surface.

A type of electronic shutter, where all pixels in the imager are sensitivity to light during exactly the same time span. Differs from the rolling curtain shutter, where the instant of light sensitivity depends on the pixel's position in the image.

Time Delay Integration: Mode of operation of an area CCD to obtain very high sensitivity.

MOSFET gate or CCD electrode that acts as a switch preventing or allowing the transfer of (photo-) charge.

In a twin well CMOS process, there is of course the "real" well, but also an implant of opposite type to increase the doping level of the substrate. All recent CMOS processes are twin well.

Supply voltage of the polarity associated with "D", the drain of nMOSFETs. The general supply voltage of an integrated circuit. Typically 5 or 3.3 volts referred to GND.

In Complementary MOS (CMOS), nMOSFETs and pMOSFETs cannot be placed in the same substrate. One of them must reside in an area of the opposite type as the substrate. In a p-type wafer the nMOSFETs are in the p-type substrate, and the pMOSFETs are in the n-type (n-)well.

FillFactory's patented method to obtain a very high fill factor active pixel. The full pixel area, except for the metallized areas, is light sensitive.

A mode of operation of an image sensor, where only the pixels within a window of interest are read out.

Sales demonstration of a mediocre quality image sensor, where the quality of the resulting images are "pushed up" by external means like image post-processing, defect hiding, etc.