P

P2P
(1)
Peer-to-Peer. A decentralized network, as opposed to the traditional client-server approach. Mainly used for Internet file-sharing. See also darknet.

(2)
Point to Point.
P#
For P54C, P55C, P6, and P7, see Intel.
packet-switched
Network traffic is routed based on addressing information carried in the header of each data packet, as opposed to the old circuit switching, which establishes a dedicated channel.
PAL
(1)
Programmable Array Logic. See PLD.

(2)
Phase Alternating Line system. Obsolete analog color TV format with 625 lines refreshed at 50 Hz, used in Europe & elsewhere. Similar to SECAM. Compare NTSC.
PAM
Pulse Amplitude Modulation. Sampled AM, wherein the amplitude of a train of carrier pulses varies with the amplitude of the modulating signal. Used for electrical signaling over wires, but not for RF modulation. It’s part of the PCM process. PAM-4 is a version with four amplitude levels instead of two, hence encoding two bits per clock cycle rather than one. Three-level PAM-3 is used in some applications to send three bits over two clock cycles.
PAN
Personal Area Network. A short-range, invariably wireless network that serves one person or a small workgroup. See 802.15.
PAO
Process-Architecture-Optimization. Intel’s new product development approach.
PAP
Password Authentication Protocol.
PAPR
Peak-to-Average Power Ratio.
PARC
Palo Alto Research Center. A Xerox facility in Palo Alto, California. It developed many groundbreaking software and hardware innovations that were exploited by other companies.
PA-RISC
Precision Architecture-Reduced Instruction Set Computing. See Intel.
parity bit
An extra bit added to each byte of data for error-checking. If the system uses odd parity, the number of zeros in the modified byte must be odd. Even parity requires that there be an even number of zeros. The transmitter sets the parity bit to 1 or 0 to make this true.
partial-response modulation
A modulation technique that introduces controlled ISI at the transmitter and removes it at the receiver. This allows the use of theoretical minimum bandwidth (half the symbol rate) with no ISI, but, because it has extra signal states (three instead of two for duo-binary, nine instead of four for QPR), it requires more power to achieve a given SNR.
The method involves sampling the waveform midway between two peak values. This gives a value that has contributions in equal part from the two adjacent pulses, and none from all others. The system is not being measured at full response, hence the term “partial response”.
Pascal
See programming language.
passkey
A more secure form of user ID than a password, based on the concept of separate public and private encryption keys. The private key is stored only in the user’s authenticator, which is typically an app on a smartphone or PC. The user activates the authenticator with biometrics (face image from a camera, or a fingerprint from a scanning touchscreen) or a master password. The public key is stored on sites at which the user has an account; when the user logs in, the site uses it to generate a one-time digital challenge, and the authenticator uses the private key to resolve the challenge and send back the solution. Each login session uses a new, pseudo-unique challenge.
If the authenticator is on one device but the user is logging in from another, the remote site can generate a QR code to be scanned by the authenticator. Cross-platform incompatibility can be an obstacle in this scenario.
PATA
Parallel Advanced Technology Attachment. Name applied to the old ATA since the introduction of SATA.
patch cable
Also called a straight-through cable. Originally, any cable used to link two communications ports. It connects the transmit pins of one end to the corresponding receive pins of the other: TX to RX and ground to ground or, for balanced-line signaling, TX+ to RX+ and TX- to RX-.
With the rise of networking, the term has come to mean specifically a cable that connects the communications port of a DCE (router, switch, hub) to that of a DTE client such as a PC. The most common by far is the Ethernet patch cable with 8P8C modular connectors, commonly but inaccurately called RJ-45 connectors. See Cat# for the wiring diagram of an Ethernet patch cable. Contrast crossover cable, rollover cable.
PBCC
Packet Binary Convolutional Coding. A CCK 11 Mb/s waveform with a Viterbi convolutional code overlay, used in the IEEE 802.11b and 802.11g standards.
PBX
Private Branch Exchange. The private telephone system of a company or other organization. The first one was set up in 1879. The dial tone when you pick up the receiver comes from the PBX, and you must dial “9” (usually) to open a line to the telco’s local switching office.
PC
Personal Computer. This usually means any personal computer (formerly microcomputer), but can be used more narrowly to mean an x86-compatible, or just one of the early IBM models with “PC” in the name – although this last usage has faded.
The 1971 Kenbak-1, with 256 bytes of memory and logic that relied on TTL IC chips and used binary I/O, is often cited as the first pre-assembled personal computer. The 1973 Micral-N was the first based on a microprocessor, the Intel 8008 clocked at 500 kHz. The first production PC, in 1975, was the Altair 8800, from a company called Micro Instrumentation and Telemetry Systems (MITS). It had an 8-bit, 2 MHz Intel 8080 CPU and 256 bytes of memory, and ran an interpreted version of BASIC written by Bill Gates & Paul Allen. These early PCs lacked such amenities as a video screen, keyboard, mouse, or disk drive.
The original IBM PC Model 5150 debuted in 1981. It had a 4.77 MHz, 16-bit internal/8-bit external Intel 8088 processor; a maximum 64 kB of memory (later 256 kB); no hard drive; one or two 5¼" floppy disk drives for 160 kB single-sided floppies; an 80-character × 25-line MDA display (CGA optional – see graphics); and IBM’s proprietary PC-DOS 1.0 operating system. It ran the Lotus 1-2-3 spreadsheet program. Most significantly, it was an open system – other hardware and software vendors could manufacture products for it, notably Microsoft’s cheaper MS-DOS, and produce similar computers (which were dubbed PC clones). IBM followed with the PC-XT and the PC-AT.
PC/104
The standard ISA bus (a.k.a. PC bus) repackaged for industrial use. It employs 3.6" × 3.8" stackable card modules with 64- and 40-contact male/female headers. Electrically, the interface is the same as the ISA bus, except for a 4 mA current limit.
PC/104-Plus
A version of the PC/104 bus & form factor that includes PCI. The cards are the same size, and up to four can be stacked together. It has a 104- or 120-pin connector for the 33 MHz PCI bus, in addition to the ISA bus connectors.
PCB
(1)
Printed Circuit Board. Also called a printed wiring assembly (PWA). An insulating substrate (the most common material is FR4, a green epoxy reinforced with glass fiber) inlaid with conductive traces to interconnect mounted components. The term is often understood more narrowly to mean a board that has the circuit components installed, as opposed to a bare board, which is a printed wiring board (PWB). Not everyone makes this distinction.
A new trend is to incorporate passive (R, L, C) components into the board at manufacture.

(2)
Poly-chlorinated Biphenyl. A class of insulating oils once common in power transformers, but banned in 1979 due to their toxicity.
PC Card
Newer name for the PCMCIA card, a credit-card-sized flash memory device also configured for use as a fax/modem, transceiver, network adapter, and even a solid-state hard drive in PDAs. The most recent standard, type III (replacing 2.1), supports 3.3V, 32-bit interfaces and operating speeds up to 132 MB/s. It’s been supplanted by the MMC and the later SD card.
PCD
Printed Circuit Design.
PCE
Power Conversion Efficiency.
PCH
Platform Controller Hub. Introduced in 2008 as the new architecture for the chipset of the PC motherboard, replacing the northbridge/southbridge design of the Hub Architecture.
PCI
Peripheral Component Interconnect. Now obsolete but once the dominant local bus standard for PCs, begun by Intel in 1992. PCI 2.1 specifies 33 MHz, 32-bit, 3.3V or 5V connections. PCI 2.2 (the industry standard as of 12/98) allows 64-bit and 66 MHz, and favors 3.3V but continues to accommodate 5V – although 5V cards are limited to 33 MHz. PCI bus slots are keyed for either 3.3V or 5V implementation, not both. PCI cards can be 3.3V, 5V, or universal. They must have the correct set of notches in the pin module to fit into the corresponding type of slot. Most PCI 2.2-compliant cards are universal. 64-bit cards and slots have an additional 64 pins, making them physically longer (see figure). A 64-bit card can run in a 32-bit slot, although its extra pins hang out over the motherboard. The bus will default to the speed of the slowest device. At 66 MHz, PCI can accommodate only one or two devices.
Three standard PCI slots
Three standard PCI slots
In addition to the bus specifications, there are PCI-to-PCI bridge specifications (1.0 and 1.1). PCI is bridged to the processor bus through a bridge chipset that provides CPU signal abstraction, making PCI processor-independent, unlike its vanquished rival VLBus. PCI supports no more than 4 slots, but bridging permits additional slots on a system.
Alongside its obvious role as an expansion bus, PCI also supports EIDE channels on the motherboard. PCI machines typically don’t give the user a way to control which resources (DMA, IRQ, memory address) their PCI add-on cards claim. See PICMG.
PCIe
PCI Express. Sometimes abbreviated PCI-E, but not PCI-X, which is something else. This serial version of the PCI bus was standardized by the international PCI Special Interest Group (PCI-SIG) to initially supplement and later replace both PCI and AGP. The physical interface is very similar to PCI, but is neither physically nor electrically compatible. However, PCI-dependent applications can use PCIe.
The basic implementation, PCIe 1x, uses 36 contacts to control two LVDS line pairs (one Tx, one Rx) comprising a single lane. Higher-order implementations use more lanes: 4x (64 contacts), 8x (98 contacts), and 16x (164 contacts). Lower-order cards work in the larger slots. The 2x, 12x, and 32x implementations are non-standard. Warning: Motherboard manufacturers sometimes use 16x physical slots for 8x or 4x connections.
Four standard PCIe slots
Four standard PCIe slots
All versions use the same physical interface, and cards are backward-compatible with the slots of older versions. Each new version doubles the transfer rate of the previous version, although changes to the forward error correction (FEC) mean that the actual data throughput doesn’t always quite double:
PCIe 1.0 – (2004) 250 MB/s per lane.
PCIe 2.0 – (2007) 500 MB/s per lane. Uses [8,10] FEC.
PCIe 2.1 – Same speed as 2.0, but adds some pre-3.0 upgrades.
PCIe 3.x – (2010) 985 MB/s per lane. Uses [128,130] FEC.
PCIe 4.0 – (2017) 1.97 GB/s per lane.
PCIe 5.0 – (2019) 3.94 GB/s per lane.
PCIe 6.0 – (2021) 7.56 GB/s per lane.
PCIe 7.0 – (expected 2025)
PCIe 1x expansion slots provide up to 10W of power, 4x and 8x slots up to 25W, and 16x slots up to either 25W or 75W. (Slots have +3.3V and +12V connections, and achieve 75 watts with 5.5A @ 12V plus 3A @ 3.3V.) Graphics cards that use a 16x slot and need more than 75W have to get it from PEG connectors on the motherboard or coming straight from the PSU. The card will have connector cables built into it, or just headers to attach cables. Depending on the combination of PEG inputs it has, a card can draw 525 watts or more, which requires not only a robust PSU but also massive cooling to avoid torching the system.
Mini-PCIe is a smaller form factor introduced in 2005 for laptops and other small devices. It supports both PCIe and USB signaling. The card measures 30 × 50.95 mm, or 30 × 26.8 mm, with thickness 1 mm, and has 52 contacts (26 each side) with 0.8 mm spacing. Mini-PCIe slots commonly support mini-SATA (mSATA) cards as well. See also M.2.
Mini-PCIe card
Mini-PCIe card
PCI-X
Officially known as Addendum 1.0 to the PCI 2.2 specification. This is a design effort led by Compaq, IBM, and HP for a 64-bit, 3.3V version of PCI with speeds of 33, 66, 100, and 133 MHz and peak throughput of 1066 MB/s. Along with PCI itself, it’s now obsolete.
At 133 MHz, a PCI-X slot can accommodate only one device. At 100 MHz, it can handle two. It uses protocol enhancements to improve throughput over PCI at the same clock speeds. A requirement that it be fully backward-compatible with PCI 2.1 and 2.2 complicates its implementation. A universal PCI-X card will work in a 5V PCI slot, although at 33 MHz only. A 3.3V PCI card will work in a PCI-X slot, but might be restricted to 33 MHz (compliance with 66 MHz PCI is optional for PCI-X vendors). A 5V PCI card will not work in a PCI-X slot.
PCM
(1)
Pulse Code Modulation. A misnomer, as this is a source coding (digitizing) technique, not a type of modulation. Basic PCM samples a speech waveform at 8 kHz, and encodes and transmits the samples in a serial bit stream as 8-bit words (8 kHz × 8 bits = 64 kb/s). Starting in the late 1970s, PCM became the standard format for signals on telephone trunk lines connecting US central offices. One copper twisted pair line, instead of carrying a single analog voice signal in one direction, could carry 24 to 32 multiplexed PCM channels (although this requires digital repeaters every 1.8 km). And then came fiber. See companding.

(2)
Phase-Change Material. A substance that cools into either a crystalline or an amorphous state depending on the temperature to which it’s heated, with the states having very different optical and electrical properties. This makes PCM a suitable medium for storing binary data. In rewriteable CDs, the PCM is a layer of chalcogenide glass written by heating with lasers, and the read mechanism – a weaker laser – exploits the crystalline state’s much lower index of refraction. For phase-change memory, entry (3) below, heating is by electric current, and the read mechanism uses the much lower electrical resistance of the crystalline state.

(3)
Phase-Change Memory. Also called phase-change RAM (PRAM). A non-volatile, solid-state memory technology based on (2), much faster than flash memory, as well as being more power-efficient and longer-lived. Each memory cell has a top and a bottom electrode. Applying a particular voltage across the cell heats it to a precise temperature, and it then cools to the desired state. Unlike memory based on charge storage, PCM is largely impervious to radiation, although the resistance of the amorphous state does drift over time.
OUM was the first, and has seen some commercial use, but phase-change memory is mostly still in the lab as of 2011. Some types are able to hold multiple bits per cell by using voltage levels that affect only some of the cell material, meaning it can be reliably set to more than two resistance states.
PCMCIA
Personal Computer Memory Card Industry Association. Creators of the standard for the credit-card-sized flash memory device that was first called the PCMCIA card and later the PC Card. Re-interpreted by some as “People Can’t Memorize Computer Industry Acronyms”.
PCS
(1)
Personal Communications Systems (or Services). Wireless systems operating in the FCC-allocated 1800 MHz band. The idea was that these systems would be simpler and cheaper than cellular, for more specialized and diverse applications, but cell phone companies wound up with most of the spectrum.

(2)
Physical Coding Sublayer.
PCU
Packet Control Unit. A component of a packet-switching network required in the base station controller and the subscriber handset.
PDA
Personal Digital Assistant. 1980s/1990s: A calculator-sized PC that runs personal information manager (PIM) applications and is most often used for E-mail, paging, memos, and scheduling. It uses non-volatile ROM for the OS and software, and a PC Card or USB device for data storage. The user interface can be a touch-screen LCD, a pad and stylus with handwriting recognition, and/or a tiny keyboard. A variety of mostly wireless technologies support connection to a network or dial-up ISP. PDAs were made obsolete by early smartphones, but the concept later returned as the tablet PC.
PDC
Personal Digital Cellular. A 2G cellular system deployed in Japan. Changing to W-CDMA.
PDE
Pulse Detonation Engine. A refinement of the pulsejet (the WWII V-1 engine), using explosive detonations of fuel-air mix for propulsion.
PDF
(1)
Portable Document Format. Adobe’s format for documents containing text & graphics. It became a de facto Internet standard, and then an ISO open standard in 2008, although it’s better suited to document printing than to digital display. The files have a .pdf extension, and are a mixture of text and binary, with metadata.

(2)
Probability Density Function. Given a continuous random variable X, the PDF describes the statistical probability p(x) of its taking on any given value x within its allowed range. (This description is not mathematically accurate, but it gets the idea across.) For any value x, p(x) must be between 0 and 1.0, where 1.0 is certainty. Integrating the PDF across any two values X1 and X2 gives the cumulative distribution function (CDF), which is the probability of finding X between those two points. If X1 and X2 are the bounds of X, then the CDF = 1.0. Some common types of continuous probability distributions:
exponential – The distribution of the interval X in time or space between a series of independent events, where that interval has a known average (mean) value. The PDF declines at an exponential rate with increasing interval.
Gaussian – The continuous equivalent of the Poisson distribution. Also called normal distribution. It’s unbounded, in contrast to the discrete-value Poisson, so the plot of its PDF is always symmetrical, forming a bell curve centered on the mean.
normal – Same thing as Gaussian distribution.
Rayleigh – It’s the distribution for the magnitude of a vector defined by two independent, normal-distributed variables. The PDF is asymmetrical, rising from zero to a peak and then declining more gradually back toward zero.
uniform – Can be discrete or continuous. Every value within the possible range of this distribution is equally likely, so the PDF is a constant.
exponential distribution
Gaussian (normal) distribution
Rayleigh distribution
uniform distribution

Four common continuous distributions


(3)
Probability Distribution Function. The discrete equivalent of (2). To reduce confusion, some texts abbreviate the probability density function in lower case and the distribution function in upper case. Some common types of discrete probability distributions:
Bernoulli – A special case of the binomial distribution in which number of trials N = 1.
binomial – The distribution of the number of successes N in a series of repetitions of a binomial trial – an experiment with two possible outcomes, one defined as success and the other as failure, where the outcome probabilities are constant.
geometric – The distribution of the number of binomial trials N required to achieve one success in a series of sets of trials.
Poisson – The distribution of a series of independent events in either time or space when the events occur at a known average (mean) rate, although if the mean is low the PDF’s curve is less symmetrical, skewing toward the higher values. The PDF is fully defined by the mean and the variance, and usually approximates a bell curve centered around the mean rate.
uniform – Can be discrete or continuous. Every value within the possible range of this distribution is equally likely, so the PDF is a constant.
PDH
Plesiochronous Digital Hierarchy. A term for the European telecom network standard that preceded SDH. Its network nodes used multiple clocks with closely matched but not identical frequencies.
PDIP
Plastic Dual In-line Package. See JEDEC.
PDK
Process Design Kit. A set of IC chip design tools for a specific semiconductor process.
PDP
Plasma Display Panel. A video display using two glass substrates enclosing a neon-xenon mix. A display grid is formed by address electrodes running parallel to display phosphor channels, and pairs of discharge electrodes perpendicular to them. The electrodes ionize the gas, which in turn emits UV radiation that illuminates the phosphors.
The PDP uses much more power than a CRT of comparable size, and has a shorter life. Its advantages over the LCD in cost, color brightness, and (especially) size gave it market share for a while in flat-screen TVs and multi-user displays, but LCD is winning out. See FPD.
PDP-#
Programmed Data Processor. An influential minicomputer family made by DEC from the late 1950s through the 1980s.
PE
Portable Executable. The extension of COFF that Microsoft uses for its Windows executable code files.
PEBCAK
Problem Exists Between Chair And Keyboard. A rude summary employed by technical support personnel throughout the English-speaking world. See PICNIC.
PECL
Positive Emitter-Coupled Logic. See logic family.
PED
Personal Electronic Device. E.g. smartphone, laptop, tablet, PDA.
PEEL
Programmable Electrically Erasable Logic. See PLD.
PEG
PCI Express Graphics. Refers to Molex connectors wired for the auxiliary power required by some PCIe expansion cards, mainly graphics cards. There’s a 75-watt 6-pin and a 150-watt 8-pin version (also called 6+2), as shown in the figure below. Cards exist with as many as three 8-pin PEG inputs. On top of the 75 watts provided by the PCIe 16x slot itself, that’s a maximum power draw of 525W.
PCIe 6-pin auxiliary power header
PCIe 6-pin header
PCIe 8-pin auxiliary power header
PCIe 8-pin header
Power-sense signals tell the card exactly what it has, including a 6-pin output supplying an 8-pin card input, which is permitted. That allows the card to avoid overdrawing its power supply by running at less than its full capabilities, or shutting down if necessary.
PEM
(1)
Proton-Exchange Membrane. See fuel cell.

(2)
Processor Expansion Module. An open standard from Spectrum Signal Processing Corp. that supports a direct connection between each TI ‘C6000 DSP and boot ROM, SDRAM, or asynchronous I/O on a mezzanine board, avoiding the complexity that comes with bridge chips. Each PEM connection also contains routing for the TI ‘C6x serial ports and two configurable DSP interrupts, and is capable of supporting two DSPs. PEM currently supports a 400 MB/s synchronous connection to SDRAM. The physical interface is two 0.8mm pitch, 60-pin SMT connectors.

(3)
Privacy-Enhanced Mail, originally. A base64 ASCII encoding format for SSL and TLS encryption certificates. Although it uses ASCII characters, it is not meant to be human-readable.
Pentium
See Intel.
PER
Packet Error Rate.
PERL
Practical Extraction and Report Language. See programming language.
PET
Positron Emission Tomography. A medical 3-D imaging technology that detects positron (positive electron) radiation from a tracer substance as it passes through the subject’s body. Compare CT scan, MRI.
PFAS
Polyfluoroalkyl Substance, or Perfluoroalkyl Substance. Any of a class of synthetic chemicals used for decades in consumer products, and now known to cause cancer, infertility, and other health issues. There are thousands of variants, some of them toxic, and they’re extremely long-lasting, earning them the name “forever chemicals”.
PFC
Power Factor Correction. The process of reducing the reactive power drawn by a load, thereby increasing the power factor and reducing overall power use. This is a standard feature in modern power supplies and motors. Active PFC (such as a boost regulator plus DC/DC converter) increases power factor to 0.95-0.99. Cheaper & simpler passive PFC achieves power factors around 0.90 with capacitive filters.
PGA
(1)
Pin Grid Array. For the IC chip form factor, see JEDEC. For the Intel CPU socket series, see Socket #.

(2)
Programmable Gain Amplifier.
PGP
Pretty Good Privacy. See encryption.
pH
From “power of hydrogen”. A negative logarithmic scale for characterizing a substance’s concentration of hydrogen ions (H+), from 0 (extremely acidic) to 14 (extremely alkaline). Distilled water is the neutral point because it has about 10-7 moles of H+ per liter, so its pH = -log10(10-7 moles/L) = 7.0, the middle of the scale.
pharming
See phishing.
phased array
A class of beam-forming, multi-antenna systems common in mobile phone systems, radar, and astronomy. A phased array transmitter creates targeted beams and nulls by adjusting the time delay and gain of each antenna’s signal. Similarly, a phased array receiver applies time delays and amplification or attenuation to incoming signals from its antennas before coherently adding them, effectively improving reception in certain directions while shutting out others.
phasor
A complex-number notation created by German-American engineer Charles P. Steinmetz (1865-1923) for representing sinusoids algebraically. Euler’s identity shows that, for any angle θ, e = cos(θ) + j sin(θ) and e-jθ = cos(θ) - j sin(θ). Plugging in values for θ reveals that e describes a vector rotating counterclockwise as a function of θ on the unit circle with the y axis being the imaginary component j, while e-jθ is the same vector with opposite rotation (the negative phasor). From Euler’s equations, it’s easy to derive cos(θ) = (e + e–jθ)/2 and sin(θ) = j (e–jθ – e)/2.
Let θ be the instantaneous phase angle of a sinusoid with cyclic frequency f0. Now θ = 2πf0t, a function of time. Substituting angular frequency ω = 2πf0, this makes the above derivations cos(ωt) = (ejωt + e–jωt)/2 and sin(ωt) = j (e–jωt – ejωt)/2. The term ejωt is the phasor representation, and encapsulates the orthogonal in-phase (I) and quadrature (Q) components of the carrier.
Why bother? Because now, instead of time-domain calculations with the trigonometric sine & cosine values, we face frequency-domain ones using simple algebra. Radio waves, AC power, and some other physical phenomena have traits that fluctuate sinusoidally, so phasors provide a simpler way to describe the technologies that deal with them. As the sinusoid’s angle θ varies with time, its phasor – the complex number e = cos(θ) + j sin(θ) – describes a vector rotating counter-clockwise in the I/Q plane at a frequency ω. The phasor magnitude is the signal’s normalized amplitude, and its angle is the signal’s phase.
I and Q are often referred to as real and imaginary because Q carries the imaginary-number tag j, but they’re simply orthogonal components of the same signal; there’s nothing imaginary about Q. In AC circuits, I is resistive and Q is reactive power, with +j values caused by inductive reactance (purely inductive current lags voltage by 90°) and –j values by capacitive reactance (capacitive current leads voltage by 90°). In RF demodulation, I is what you get when you mix the incoming signal with a sinusoid that matches the carrier’s frequency and phase angle, while Q is what you get if you use a sinusoid that lags the carrier’s phase angle by 90°. This is true even if the carrier is suppressed in transmission, which it usually is.
P-HEMT
Pseudomorphic High Electron Mobility Transistor. See transistor.
phi (φ)
See Golden Mean.
phishing
Impersonating a trusted organization – bank, government agency, major corporation – with forged E-mails and counterfeit Web sites, to persuade people to reveal account information or other valuable data, or to open a file or hyperlink that launches a malware attack. Most often done via spam. The use of counterfeit Web sites is sometimes called pharming, especially if the perpetrators hack servers to re-direct searches to their site rather than simply bringing in victims with spoofed E-mails.
Variants include vishing (voice phishing – getting victims to call a phone number at which an IVR system will try to elicit data), smishing (SMS phishing, i.e., using text messages), spear-phishing (tailoring a phishing attack to a specific person), and whaling (targeting a rich or high-ranking person). Sophisticated spear-phishing attacks appear to come from people the target knows, and refer to subjects those people really would mention. They can even quote text from the target’s own e-mails to appear more genuine. This indicates an account or device belonging to the target’s correspondent has been hacked.
photonic crystal
A crystal that passes radiation of wavelengths roughly corresponding to the spacing of its lattice elements, while blocking others. It’s usually made of insulating or semiconducting material, because metal crystals absorb the visible and IR light that photonic circuits use. Material that absorbs strongly at a particular wavelength when cool will emit strongly at that wavelength when heated. Photonic crystals are used to create switches that convert light to electrical signals and vice-versa (see O-E-O).
PHP
PHP: Hypertext Processor (originally, Personal Home Page). See programming language.
PHY
Physical. Refers to the physical layer (layer 1) of the OSI network model.
pi (π)
A transcendental number: the ratio of a circle’s circumference to its diameter. The value is 3.1415926536… The sequence 113355, as 355/113, approximates it to six places.
π/4-DQPSK
Simply π/4-QPSK with differentially encoded data. With raised cosine filtering, it’s more spectrally efficient than GMSK. See QPSK.
π/4-QPSK
See QPSK.
PIC
Peripheral Interface Controller, originally. A family of (relatively) simple, low-power microcontrollers made by Microchip Technology, Inc.
PICMG
PCI Industrial Computer Manufacturers Group. A consortium of PC product vendors that develops specifications for PCI-based systems and boards.
PICNIC
Problem In Chair, Not In Computer. Newer version of PEBCAK.
piconet
See Bluetooth.
pid
Process Identifier. A number assigned to a running process by the operating system.
PID
(1)
Packet ID. A field in a network data packet that identifies the packet type. Used by DVB, USB, et al.

(2)
Product Identifier. A 16-bit tag embedded in a USB device. The host computer uses the combination of PID and VID to determine which device driver to use. Manufacturers can register PID/VID combinations with the USB Implementers Forum, for a fee. See also GUID, HID.
piezoelectric
A material that changes its physical dimensions in response to an applied electric field, and, conversely, develops an electric field in response to physical stress. Quartz, the best-known piezoelectric, is the basis for most crystal oscillators (see XO). PZT is also a common piezoelectric. As two-way transducers, piezoelectrics have many complementary applications, such as both producing and detecting sound. Piezoelectric generators, which produce current, remain experimental.
pigtail
Sometimes called a dongle. A short adapter cable to connect a communications port to something else. It has some kind of cable connector on one end and either bare wire or a different connector on the other. There are fiber pigtails as well.
PIM
(1)
Processor In Memory. Tightly coupling a processor with memory, usually on the same chip. This greatly increases memory bandwidth while reducing latency and power use.

(2)
Platform-Independent Model. A software system model that avoids tying itself to specific hardware.
PIN
(1)
Positive-Intrinsic-Negative. See diode.

(2)
Personal Identification Number. Most chip-and-PIN credit cards use the EMV standard.
PINE
Program for Internet News and E-mail. A freeware U. Washington client for handling messages using SMTP, POP3, IMAP, NNTP, and MIME. Popular for a while in the mid-1990s.
PING
Packet Internet Groper. A TCP/IP procedure that uses the ICMP echo to check whether a network device is active. From an operating system’s command-line interface, the command ping 192.168.1.64 attempts to send data to the IP address 192.168.1.64 and reports on the reply, if any.
PIO
Programmed Input/Output. See DMA.
pipelines
A RISC technique, adopted by CISC chips starting with the Pentium. It breaks processing into stages like an assembly line, with the CPU operating on several instructions at once, each at a different stage of completion. Branch prediction is used to guess which instructions will come next and perform predictive fetch & execution on them. Superpipelining uses more and shorter stages (14 for the Pentium Pro vice 5 for the Pentium), permitting the chip to run at even higher clock speeds.
PKI
Public Key Infrastructure. A system for providing authenticated encryption keys to users of a network, including the Internet. TLS, and its predecessor SSL, are the best-known technologies to rely on a PKI. Many experts are predicting (2024) that quantum computing will soon be able to crack public-key cryptography. They have been for years, but that doesn’t mean it won’t eventually happen.
PKZIP
Named for Phil Katz, its creator. A popular file compression/decompression utility that started as a DOS command-line tool in 1989, and later became a Windows app.
PLA
(1)
Programmable Logic Array. See PLD.

(2)
Polylactic Acid. A biodegradable thermoplastic with poor heat tolerance, used for 3D printing and medical implants. The name is inaccurate – it’s a polyester, not a polyacid.
Planck’s constant
Named for German physicist Max Planck (1858-1947). The ratio of a photon’s energy to its frequency, represented as h. Its value is approximately 6.626 × 10-34 [J-s]. Per German physicist Werner Heisenberg (1901-1976), the constant is also the lower bound for a subatomic particle’s [mass × uncertainty in position × uncertainty in velocity].
plasmon
An electron wave in the surface of a metal, produced by incident radiation, and considered as a quasiparticle. Important in optics, with potential to be the basis for new types of devices called plasmonics that will probably depend on graphene.
PLC
(1)
Programmable Logic Controller. A specialized computer used to automate monitoring and control of an industrial plant, standing alone or in conjunction with a SCADA system.

(2)
Power Line Communications. Transmitting data through power lines. Introducing a high-frequency signal into a vast network designed for 50 or 60 Hz creates complex, far-reaching EMI risks. This is a major reason why PLC has so far (2015) not been widely used. Typical commercial proposals require servers with high-speed Internet connections at electricity substations, and a conditioning unit (CU) at each customer facility to split the data signal from the power with high-pass and low-pass filters. The HomePlug Alliance, at www.homeplug.org, is the industry group developing technologies and standards for PLC.
PLCC
Plastic Leaded Chip Carrier. See JEDEC.
PLD
Programmable Logic Device. A microprocessor that doesn’t have hard-wired connections between its logic units. Instead, a developer specifies these connections with a program written in a hardware description language (HDL), digitally rewiring the chip. This gives a PLD system flexibility and reconfigurability that an ASIC design lacks, while allowing it to be much faster and use much less power than a conventional processor or even a DSP. The tradeoff is that PLDs require more work to program.
Programming a PLD requires a device programmer unit that holds the chip, and a user interface (typically the chip-maker’s software running on a PC). Some older generations of PLD could be programmed only once, but reprogrammability is now the standard.
Broadly speaking, there are three types of PLD, discussed below. CPLDs are preferred for I/O-intensive applications, because they offer faster but simpler data handling and typically use less power. FPGAs are better for complex manipulations and logic – DSP, FEC, etc.
SPLD – Simple PLD. The smallest devices, containing 4 to 22 interconnected macrocells. Devices that fall under this heading include the one-time programmable PAL (programmable array logic), GAL (generic array logic, an erasable PAL), and PLA (programmable logic array, just another name for a PAL). “PLD” is often used to refer specifically to this device type.
CPLD – Complex PLD. A number of SPLD units (2 to 64) in one package. All macrocells are fully interconnected, which makes this type of chip less efficient as it gets bigger. Data storage is non-volatile. Sub-types include the EPLD (UV erasure), EEPLD or PEEL (bulky and slow), and flash ROM (electrically erasable, but has the compactness of the EPLD). Examples: Altera MAX (Multiple Array Matrix) family, programmed with the MAX+PLUS II tool set.
FPGA – Field Programmable Gate Array. Rather than interconnect all of its I/O blocks as a CPLD does, an FPGA has an array of logic blocks (based on look-up tables, or LUTs) surrounded by a ring of I/O blocks, and employs sophisticated, programmable routing and switching to make connections. This makes it more efficient than a CPLD as higher densities are reached. Data storage is volatile, so it requires an external device to program it at power-up. FPGAs are built using SRAM latches, one-time programmable anti-fuse, or EPROM/EEPROM transistors. Examples: Altera APEX 20K family, programmed with Quartus software, and Xilinx Virtex-E family, programmed with Foundation Express or Alliance IDE.
plenum
The space between a ceiling and the roof (or the floor above), used for air circulation and ducting. Plenum cable (see CMP) has fire-retardant coating that doesn’t produce toxic smoke, and so is approved for use in this space.
plesiochronous
From Greek plesios, near to, and chronos, time. Using more than one clock in a system, and attempting to match their frequencies. The results are generally unsatisfactory. See PDH.
PLL
Phase Locked Loop. A device for matching the frequency of an incoming carrier signal. A phase detector compares the phase of the carrier with the phase of a VCO signal, and outputs a voltage corresponding to the differential. This voltage feedback-tunes the VCO to match the carrier frequency. Phase lock is achieved when the two signals are at the same frequency and in phase.
plug-in
A program that provides an added capability to some other program, particularly to a Web browser. For example, to run a Java applet embedded in a Web page, a browser would need a Java plug-in. Plug-ins are also called add-ons or extensions, although some sources use these terms to mean slightly different things.
PM
Phase Modulation. The phase of the carrier wave varies with the amplitude of the modulating signal. This is the newest of the three types of RF analog modulation. The others are AM and FM.
PMA
Physical Medium Attachment. The part of a network MAU containing the functional circuitry.
PMC
PCI Mezzanine Card. An IEEE P1386 or P1386.1 standard daughtercard that can be mounted on expansion cards (mothercards) for a Multibus II, PCI, cPCI, VME, or other backplane system. It was originally conceived as a way to add I/O and memory to a single-board computer (SBC) or blade server, but is increasingly used for specialized functions such as video processing, DSP, encoding, decoding, etc. There are follow-on specifications for some of these applications: PCI telephony mezzanine card (PTMC, or PICMG 2.15), processor PMC (PPMC, or VITA 32), and conduction cooled PMC (CCPMC, or VITA 20).
PMC board
PMC board
The PMC interface is electrically the same as 32- or 64-bit PCI, supporting 33 MHz at 3.3V or 5V and 66 MHz at 3.3V. Physically, it uses between two and five low-profile, two-row, 64-contact, 1.0mm pitch, tongue-to-groove surface-mount connectors. P1 and P2 carry 32-bit PCI, P3 hosts the 64-bit PCI expansion, and P4 & P5 (P5 requires an extended 249mm-long PMC) are for user-defined I/O. P3-P5 can be omitted if not used. The plugs are on the PMC, and the receptacles are on the mothercard. The PMC mounts parallel to the mothercard rather than at right angles to it, which is why it’s called a mezzanine card. Two 74mm × 149mm PMCs will fit side-by-side on a single Multibus II or VME mothercard.
PMD
Polarization Mode Dispersion. The spreading of short pulses in time due to propagation rates of the pulse components varying with polarization.
PME
Precious Metal Electrode. A technology for making multi-layer ceramic capacitors with precious metals, which have higher conductivity, but are more expensive and commonly alloyed with toxic metals such as lead. Compare BME.
PMIC
Power Management Integrated Circuit. Just what it says.
PMLCD
Passive-Matrix Liquid Crystal Display. See LCD.
Pmod
Peripheral Module. A simple, general-purpose interface for connecting peripherals to circuit boards. It has six lines: power, ground, and four signals. The designer chooses what the signals and power level are. The physical port is a 6-pin, single-row header with 0.1" pin spacing. There are double and quad Pmod connectors.
Single Pmod header
Single Pmod header
PMOS
P-channel Metal-Oxide-Semiconductor. A semiconductor IC based on p-channel MOSFETs, so that the basic conduction mechanism is governed by holes (regions of positive charge created by a dearth of electrons) rather than electrons. Compare NMOS. See also CMOS, transistor.
PMT
Photo-Multiplier Tube. A device that converts photons to electrical signals with high gain, making it a very sensitive photodetector.
PMU
Phasor Measurement Unit. A device that generates time-stamped measurements of AC voltage amplitude and phase in a power grid. Paired with automated control systems, PMU inputs allow grids to respond immediately to developing problems, including downed power lines.
PN
(1)
Pseudo-Noise. Also PRN for pseudo-random noise (or number). A signal generated with a uniform distribution, similar to Gaussian (random) noise. Software alone can’t produce true randomness, so pseudo-random number generator (PRNG) programs base the signal on a pseudo-random binary sequence. PN sequences, such as the Gold, Kasami, and orthogonal Walsh-Hadamard codes, are used as spreading codes for spread-spectrum signals. See AWGN, Gold code, LFSR, RNG.

(2)
Positive-Negative. See diode.
PNA
Phone Networking Alliance. See HomePNA.
PNG
(1)
Portable Network Graphics. An ISO-approved open standard for image files, developed in the 1990s to replace GIFs for Web page images. It uses lossless compression, and supports image transparency and a wide range of color depths. The files typically have the .png extension.

(2)
Pressurized Natural Gas. See LNG.
PNP
Positive Negative Positive. One of two semiconductor doping profiles for bipolar junction transistors. NPN is the other.
POCSAG
Post Office Committee Standardization Advisory Group. An international advisory group on pager standards. Also, a digital pager standard introduced in the early 1980s for vendors to the British Post Office, and later popular in Europe & Japan.
podcast
Named (2004) for Apple’s iPod, the first device to support it. It means an audio-video or other file distributed via syndicated feed (push) for playback, rather than as streaming media or downloadables the user has to go get (pull). The feed is an originator-updated link to podcasts. User software detects podcasts autonomously and makes them available.
POE
Power Over Ethernet. A technology that supplies power to devices through a network connection.
POF
Plastic Optical Fiber. Cheaper and more durable than glass fiber, but with much poorer performance. Losses of 0.15-0.2 dB per meter at 650 nm are typical.
polarization
For an electromagnetic (EM) wave, polarization refers to the way the electric field vector E of the wave changes with time.
Imagine looking directly into an approaching EM wave, and seeing the E-field vectors of its individual photons as they pass through an X-Y coordinate plane perpendicular to the direction Z of the wave’s propagation. (The orientation of the X and Y axes is arbitrary.) No matter how many individual E-field vectors there are, they’re being considered in just two dimensions, so their sum can have just two components: an X and a Y.
Even at a single frequency, most natural EM sources are incoherent, which is to say unpolarized. Their aggregate field vector has negligible amplitude and randomly shifting direction. Filtered or man-made waves, however, can exhibit polarization – a strong bias in a particular direction. And since the fields of a simple harmonic EM wave fluctuate sinusoidally, the X and Y of a polarized wave are sinusoidal functions: they define a vector rotating at a constant angular rate, or frequency, tracing an elliptical path on the incident plane.
There are two conventions for a rotating polarization – the classical and the IEEE definition – and, for entirely logical reasons, they’re mirror opposites. The classical optics view imagines the spiral path around the wave that the E-field vector traces as the wave moves through space. Moving in the wave’s direction of propagation, this path has either a right- or a left-handed twist, and that’s the polarization of the wave.
The IEEE view instead confines itself to a 2-D plane through which the wave passes. The E-field vector rotates in time on this surface, and its direction of rotation relative to the direction of propagation establishes its handedness by the right-hand rule. So what the IEEE calls right-handed polarization, the classical view calls left-handed. In engineering and communications, the IEEE definition is standard.
All RF polarization is fundamentally elliptical, i.e., the rotating field vector can be said to define an ellipse. However, there are two special cases: circular and, less obviously, linear polarization.
Linear polarization occurs when the phase differential of the X and Y components is either 0° (in phase) or 180° (orthogonal). This creates an ellipse with eccentricity of 1, which collapses to a line segment; rather than rotating, the vector simply oscillates between positive and negative maxima along a straight line. If the X axis is defined as parallel to Earth’s surface, a wave is said to be vertically polarized when the X component remains zero, horizontally polarized when the Y component remains zero, or slant polarized when both are non-zero.
Vertical polarization is typical for omni-directional transmission over short to medium distances, e.g., most radio stations. Horizontal polarization is used for television and other signals that traverse longer distances, to reduce interference from the more common vertically polarized signals. Receiving antennas must be aligned accordingly.
Circular polarization (CP) is the other special case, and it occurs when the X and Y components have equal magnitude and a 90° phase separation. This creates an ellipse with eccentricity 0 – a circle. CP is necessary for satellites and other systems where movement, environmental anomalies (Faraday rotation passing through the atmosphere), etc. would disturb the alignment of a linear-polarity signal and compromise reception. CP reception requires a physically symmetrical antenna, commonly a circular dish. RHCP and LHCP refer to right-handed and left-handed circular polarization, respectively.
Using the IEEE convention for describing handedness of circular/elliptical polarization, the JavaScript animation below takes a viewpoint looking into an approaching wave (with artificially low frequency). The red line shows the electric field vector, and the pale blue lines are the vector’s X and Y components. Remember that X-Y alignment is arbitrary.




Your browser does not support the HTML5 <canvas> element used for this animation



polymorphism
In object-oriented programming, this is the capability of an object to vary its behavior depending on the type of data passed to it. In the world of malware, it’s the capability of newer viruses and worms to change their structure when they self-replicate, foiling signature-based detection.
PON
Passive Optical Network. An optical network with no active (powered) components, just passive splitters and couplers. It’s the basis for virtually all fiber-to-the-curb (FTTC) concepts: an optical line terminator (OLT) at the central office, optical network terminals (ONTs) at customer facilities, and a PON connecting it all. Types include ATM-based PON (APON), later called Broadband PON (BPON), Ethernet PON (EPON), and Gigabit PON (GPON).
POP
(1)
Post Office Protocol. A member of the TCP/IP protocol suite at the Application layer, this is the original Internet e-mail retrieval protocol. The newest version is POP3. Using TCP port 110, or port 995 for encrypted connections, it links to a SMTP server and downloads any stored messages to the user’s e-mail client application. IMAP is a more capable protocol for the same job, but also more complex, so POP is more widely supported.

(2)
Point of Presence. A location where inter-exchange carrier (i.e. long distance) facilities meet up with the local access facilities of telephone companies or other service providers.
POR
Power-On Reset. A feature of IC chips that verifies supply power has stabilized and the circuit is initialized before starting operation. It also resets the circuit to recover from power drops that could otherwise leave it in an unknown state.
POSIX
Portable Operating System Interface for Unix. Pronounced “PAUSE-ix”, not “POSE-ix”. IEEE and ISO standard for Unix system-level (i.e. kernel) APIs, meaning interfaces between programs and the operating system. Windows NT and most flavors of Unix are POSIX-compliant. Linux supports it, but has its own, more frequently updated APIs as well, so many GNU/Linux applications use those instead.
POST
Power-On Self Test. A set of routines that test the primary system components when a computer or embedded device boots up. In a PC, it’s stored in motherboard ROM. Similar to BIST.
PostScript
Adobe’s 1984 page-description language, which produces ASCII-format text files with a .ps extension. Displaying or printing them with the intended layout requires a PostScript interpreter such as Ghostscript, either on the PC or running in a processor on the printer itself. In desktop publishing, Adobe’s own PDF has largely replaced PostScript.
potentiometer (or pot)
A tunable resistor, used for adjusting circuit gain or attenuation. Also called a voltage divider, because it uses two contacts at either end of a resisting element and a third, sliding contact on the element, in parallel with the output contact. Compare with rheostat.
POTS
Plain Old Telephone Service (or System). Also called PSTN.
power
A measure of energy over over time, which SI (the metric system) expresses in watts (W), or joules/second. Horsepower (h.p.) is used for mechanical work, and BTU/hour for heat. 1 HP = 746 W, and 1 BTU/hr = 0.293 W.
Just as impedance is the complex sum of resistance and reactance, power in an AC system is both resistive and reactive. AC power is not the simple algebraic product of the RMS voltage and RMS current; that’s the VA (volt-ampere) value. Instead, it’s a scalar quantity resulting from the dot product (vector product) of two vector quantities: volts and amps.
For example, given vac = V cos[ωt + φ1] and iac = I cos[ωt + φ2], power P = I V cos[φ1 - φ2]. So power is at maximum when AC current and voltage are in phase (when φ1 = φ2), and declines with increasing phase shift. Reactive power, in other words, does no useful work. See power factor, crest factor.
Power#
See PowerPC.
power factor
The ratio of the watts a device or system consumes to the volt-ampere (VA) value required to deliver that wattage. Power factor equals 1.0 only for purely resistive loads or DC systems. The more reactive the load, the greater the VA required to deliver the same power, which is why it’s critical for electricity grids to minimize the phase difference beteen voltage and current. Harmonic distortion, a mismatch between the frequencies of voltage and current, also reduces power factor.
Reactive loads are either capacitive or inductive. In a capacitive load, current leads changing voltage because of the stored charge. In an inductive load, current lags voltage because of impedance from the inductor’s EM field.
Old AC-DC power supplies had a power factor of about 0.7, but see PFC.
PowerPC
Performance-Optimized With Enhanced RISC PC. Originally a 1990s RISC microprocessor architecture developed by IBM with Apple and Motorola. Early family members include the PowerPC 601, 603, 604, 615, and 620. Later came the G# series (G = Generation), and the Power# series. Some of these were not sold separately, only as part of a system from IBM, Apple, or a partner.
G3 – (1997) 32- and 64-bit, using 0.35-0.25 µm process and probably an integrated (on-chip) L2 cache.
G4 – (1999) 32- and 64-bit, using 0.25-0.18 µm process.
G5 – (2003) 64-bit, up to 2 GHz, supporting up to 500 GB of disk space, and 8GB of RAM linked to the processor by a 1 GHz system bus. Apple used it in the PowerMac G5 PC.
Power4 – (2001) 64-bit implementation of PowerPC architecture. 180nm process, CMOS with copper interconnects, dual-core, 1.1 to 1.3 GHz clock. There’s a multi-chip module (MCM) hosting four Power4 dies with up to 128 MB shared L3 cache, and eight-way symmetric multiprocessing (SMP). The Power4+ reached 1.9 GHz.
Power5 – (2004) 130 to 90 nm process, dual-core, 1.5 to 2.2 GHz clock. Each core has 32 kB L1 cache and 1.875 MB L2 cache; they share 36 MB off-die L3 cache. Power5+ added a quad-core model.
Power6 – (2007) 65nm process, dual-core, 3.5 to 5.0 GHz clock. Each core has 128 kB L1 cache, and L2 cache. Cores share an off-die L3 cache via an 80 GB/s bus. Power6+ reached 5 GHz.
Power7 – (2010) 45nm process; 64-bit; 4, 6, or 8 cores; 3.0 to 4.14 GHz clock. Includes Power 750, 755, 770, and 780.
PowerShell
A Microsoft command-line application and scripting language introduced in 2006. Starting with Windows 10, Microsoft is pushing PowerShell as a more powerful, .NET-capable replacement for the old Windows Command Prompt. PowerShell scripts have the .ps filename extension.
PPI
Pixels Per Inch.
PPM
Pulse Position Modulation. The same as PWM, except that position is indicated by two very short pulses rather than one long pulse, greatly reducing power use.
PPP
Point-to-Point Protocol. A member of the TCP/IP protocol suite at the Link layer (the Data link layer of the OSI model), PPP is in effect a protocol suite unto itself. It’s based in part on IBM’s HDLC data protocol, and provides remote LAN access with more capability than the older SLIP. Two nodes with a non-TCP/IP connection can use PPP sub-protocols to exchange IP packets encapsulated by HDLC-formatted data frames, creating a virtual network link. Other PPP sub-protocols provide authentication, data compression, data encryption, and support for multiple connections.
PPPoE (PPP over Ethernet) is the standard way for DSL customers to exchange IP packets with their Internet service providers.
PPS
Precise Positioning Service. See GPS.
PPTP
Point-to-Point Tunneling Protocol. A Microsoft extension to PPP to create multi-protocol VPNs over the Internet. It provides encryption with 40- or 128-bit keys.
PQC
Post-Quantum Crytpography. Same as PQE.
PQE
Post-Quantum Encryption. Starting in 2015, a NIST initiative to select new encryption standards capable of standing up to cracking attempts backed by quantum computing.
PQFP
Plastic Quad Flat Pack. See JEDEC.
pragma
Pragmatic. In some programming languages, including ANSI standard C or C++, this is a pre-processing directive inserted into code to give special instructions to the compiler. The instruction set that pragma can use varies with the compiler.
PRC
Primary Reference Clock. The master clock in a synchronous system such as SDH or SONET. It can drive a number of slave clocks. See SEC.
PRI
Primary Rate Interface. See ISDN.
PRM
(1)
Partial Response Modulation.

(2)
Programmer’s Reference Manual. Explains the assembly programming and hardware of a microprocessor.

(3)
Probabilistic Road Map. A 3D point graph created by the motion-control systems of industrial robots. It determines the optimal path through physical space for the system to follow from one point to another without hitting anything.
PRML
Partial-Response Maximum-Likelihood. A technique for reading the digital sequence represented by a weak analog signal. Rather than simple threshold detection, it relies on digital sampling and interpretation to determine the most likely binary sequence. PRML and its successor, the even more efficient Extended PRML (EPRML), are standard features in modern hard drive controllers.
PRN
Pseudo-Random Noise (or Number). See PN.
PRNG
Pseudo-Random Number Generator. See PN.
probability distribution
A function or data set defining the probability p(x) that a random variable will take on the value x, for some range of x. It can be discrete (taking on any of a limited set of exact values, e.g., x can be any integer from 1 to 10), or it can be continuous (taking on any allowed value within some range, e.g., x can be any number from 1.0 to 10.0, which includes an infinite number of possibilities). See also probability density function.
process
Referring to computers, this means an instance of a running program.
Referring to microcircuit manufacturing technology, the process or, more accurately, process node is traditionally the distance in microns (µm) or nanometers (nm) between the gate and drain in each field-effect transistor (FET) on the chip. It’s no longer an accurate descriptor of modern microcircuits, but persists for lack of a broadly accepted replacement.
process management
A method for specifying how an organization will carry out its tasks. This includes general-purpose methods such as Total Quality Management (TQM), Six Sigma, CMMI, and the Peter Principle.
The oldest systematic approach specifically for developing software is called the waterfall or plan-driven method. It divides the work into a series of stages, each of which must be complete before the next begins. It’s meant for large, complex systems where safety and reliability are crucial, and it’s still in wide use. However, waterfall projects frequently run over schedule and over budget.
The software engineering community has come up with a number of alternatives that apply the principles of Agile programming. The 2001 Agile manifesto lists these 12 principles, which include close collaboration with customers and end users, early and continual testing, frequent incremental releases of the software, and an emphasis on working software over documentation. Some of the better-known Agile methods:
DevOps – Development Operations. An extension of Agile principles to include deployment and operation of systems in addition to development. It focuses on rapid delivery, and requires ongoing collaboration and feedback between developers and operators to help both improve their process.
DevSecOps – Development Security Operations. An attempt to balance DevOps and SecOps, maintaining the rapid development of the former while applying practices to minimize risk.
GitOps – A spinoff of DevOps for automating IT infrastructure, using configuration files to generate the infrastructure in a consistent, repeatable way, and using Git (see GitHub) for source control.
Scrum – A structured process that breaks work into sprints lasting 2 to 4 weeks. A sprint starts with a short sprint-planning event, and has daily meetings (Scrums) not to exceed 15 minutes to report task progress and any obstacles. At the end of each sprint, there’s a review with all the stakeholders to go over what was accomplished and revise the task backlog, and a separate retrospective with just the developers to consider what went well, what didn’t, and possible changes in approach.
SecOps – Security Operations. A method that prioritizes information security.
XP – Extreme Programming. One of the first Agile methods, it focuses on the most valuable features from a business standpoint, and identifies 12 practices to follow. This invites confusion with the 12 principles of the Agile manifesto, which undoubtedly inspired them but aren’t the same thing.
Despite the Agile de-emphasis on documentation, having clear, correct, documented requirements prior to development remains crucial to project success.
programming language
A set of rules for writing computer instructions in human-readable form. This broad description embraces several approaches, which have given rise to thousands of individual languages. A scripting language is just a programming language that uses an interpreter and does certain kinds of jobs. See also Turing machine.
Low-level languages, including all assembly languages, require the programmer to deal with details of processor hardware: registers, I/O ports, timers, memory addresses, etc. An assembler program turns the assembly source code into a binary (machine-language) executable. Low-level code can produce very small, fast, efficient programs, but they work on just the one type of processor for which they were written.
High-level languages treat the hardware as an abstraction, so the programmer doesn’t need to know how the processor will handle instructions. This spares the programmer from having to learn a separate language for each processor family, and permits use of the same code on many different platforms. High level vs. low level is a subjective evaluation; some high-level languages are less high-level than others, leaving room for a medium-level category to further muddy the waters.
High-level languages fall broadly into two types:
A couple of other ways to categorize or describe languages, both having to do with handling of data types:
Strong typing vs. weak typing – Weakly typed languages such as C make it easy to treat a value as though it were another type, e.g., reading a floating-point number as an integer, or adding a character and an integer. This gives programmers a lot of flexibility, but can encourage carelessness and create security vulnerabilities. Strongly typed languages such as Ada generate errors when implicit type conversion is attempted.
Static typing vs. dynamic typing – Statically typed languages verify correct type usage at compile time, and dynamically checked languages do so at runtime. Dynamic typing incurs some overhead, and can result in program failures or crashes. It’s needed because some things can’t be checked statically, and it might be better to have a program crash than quietly perform an unintended action. Many languages use both methods.
There are two recognized methods for turning high-level code into machine-language instructions: compilation and interpretation. Each has benefits and drawbacks. (It’s inaccurate, though common, to refer to languages as being compiled or interpreted. The distinction isn’t in the language itself, but in how it’s handled. There’s no technical reason why a given piece of source code can’t be compiled for one use and interpreted for another, and some implementations use both methods.)
Compiled programs are generated with a compiler (and, commonly, a linker as well) as a single, monolithic executable. This step need be done only once, and the executable can then be run any number of times under a given operating system or directly on the processor. The same source code can, in principle, be compiled for different operating systems and CPUs by using different compilers.
Interpreted programs are read and executed by an interpreter at runtime. Rather than creating an executable file, the interpreter generates machine-language instructions on the fly as it reads the source code. This means the program is easier to create and modify, but has to run under the interpreter every time it’s used. To reclaim some of this lost efficiency, many interpreters are actually hybrids – they compile source code into an intermediate, platform-specific stage called byte code, which they can cache for use at runtime. An interpreter can be much faster working from byte code rather than source code, though this approach, just-in-time (JIT) compilation, is still not as fast as fully compiled code.
Functional programming is a set of practices that have been found to reduce errors, including immutable data (variables don’t change once they’re set), pure functions (functions can’t change data that exists outside them), and disallowing null references. Some newer programming languages enforce these practices.
A few of the better-known programming and/or scripting languages:
Ada – A general-purpose, object-oriented language named for Augusta Ada Byron King (1815-1852), daughter of Lord Byron. She’s considered the first computer programmer because of her 1842-1843 work for Charles Babbage’s analytical engine, a mechanical computer that was never built. Ada was created for the US Department of Defense based on Pascal, and registered as an ANSI standard in 1983. Ada 2005 supports real-time applications and many different OS environments, even a compiler for producing Java byte code. The source code files have the filename extensions .ads (specification, i.e., header) and .adb (body).
BASIC – Beginner’s All-purpose Symbolic Instruction Code. (Yes, it is an acronym.) A language developed at Dartmouth College in 1964 by math professors John Kemeny and Thomas Kurtz. It’s simple to use, but has a reputation for producing slow programs and teaching bad programming habits. Some of the reason for this is that early implementations of BASIC relied on interpreters, and interpreter overhead on computer systems of the 1960s and 1970s was considerable. It has many descendants, including Microsoft’s Visual Basic (VB) and VBA languages. The source files often use the .bas filename extension.
C – The name is sequential; less successful predecessors were called A and B. C was developed by Dennis Ritchie at Bell Laboratories in the early 1970s, made an ANSI standard in 1989 and an ISO standard in 1990. While extremely flexible, it has many low-level features and requirements, hence is often considered a mid-level language. In addition, C’s flexibility permits unsafe memory accesses and type conversions, making its programs inherently vulnerable to certain kinds of malware attacks. Nonetheless, it’s probably the most widely used of all languages, and remains very influential. A typical C project compiles source code files (*.c) described by header files (*.h) into machine-language object files (*.obj), which it then links with library files (*.dll) to build the executable (*.exe). There’s no rule that the program files must use these filename extensions.
The ISO and IEC have issued a series of standards for the language, starting with C89 in 1989. As of 2022, the latest is ISO/IEC 9899:2018, or C18. C compilers are built for backward compatibility, so they can compile most source code written to older standards.
C# – Pronounced “C sharp”. Object-oriented Microsoft language based on C++, created as Microsoft’s answer to Java. It’s for developing Windows applications only. All C# classes are subclasses of System.Object. C# source code files usually have the .cs filename extension, and compile into an intermediate language (IL) that runs in Microsoft’s CLR.
C++ – An object-oriented extension of C, created in the early 1980s by Bjarne Sjoustroup at Bell Laboratories. Powerful, flexible, and very widely used, but it inherited the security weaknesses of its parent. C++ projects can use both C and C++ source code files (*.c and *.cpp).
COBOL – Common Business Oriented Language. One of the first (1960) high-level languages, later standardized by the ANSI and the ISO. It was meant for file-oriented business applications, and so was designed to be simple and inflexible. COBOL 2002 climbed on the object-oriented bandwagon.
Fortran – Formula Translation. The first high-level language, developed in the 1950s primarily for mathematical applications.
Go – Also called “golang”, to distinguish it from the older Go! language (note exclamation point). Created by Google researchers and released in 2009, with open-source compilers. It’s intended to provide the power of C++ without the complexity. The developers baked their code style and design preferences into the language, and created tools to check some of them.
Java – A network-savvy, object-oriented programming language from Sun Microsystems, based on C++ syntax, and widely used to add interactive capability to Web pages and mobile wireless interfaces. A Java program runs over a network, with part of the code executing on the server and part of it passed to the client device. Full-featured, stand-alone Java programs require an installation of the Java runtime environment (JRE). Java applets are embedded in some other program – typically a Web page – and require only that the browser have a Java plug-in, a kind of JRE-lite.
A Java compiler turns the source code into an intermediary form called byte code, which takes the form of .class files, or a .jar archive of .class files. (And it’s possible to compile byte code from languages other than Java.) Here’s the trick: Rather than addressing a specific platform (Windows, Mac OS, Unix, etc.), the compiler’s system calls are addressing the API for an abstraction of the hardware called the Java virtual machine (JVM).
When a computer executes byte code, the local version of the JRE launches a JVM to host the process. The JVM’s platform-specific Java interpreter fields the platform-agnostic API calls, converting them into machine-language instructions native to the host.
Just-in-time (JIT) compilation, introduced for Java around 1997, greatly improves running speed in part by caching and re-using byte code and machine code, saving the interpreter some time.
Some applications require machine-specific code that bypasses the interpreter. Sun provided a mechanism in Java to do that. Microsoft issued its own version of Java with a different mechanism, which wouldn’t work with any OS but Windows. Sun won the resulting court battle, so Microsoft created C# instead.
JavaScript – An object-oriented scripting language for creating text-based apps (scripts) that a JavaScript-enabled browser can run within HTML documents. Its syntax is based on the C language. It has a much smaller, simpler command set than its distant cousin Java, and runs much more slowly. A common use is to link Java applets, or applets and Web pages, primarily for client-side scripting. Stand-alone JavaScript files use the filename extension .js.
A 2013 specification for a subset of JavaScript called asm.js achieves higher running speed than manually coded JavaScript by restricting language features to those compatible with optimization and ahead-of-time compilation, including a requirement to use static typing. It requires a special compiler, which creates asm.js source code from code written in another language, usually C or C++. Older browsers that don’t apply the asm.js optimizations will still run the code, because it’s valid JavaScript, but they won’t get the speed boost. A newer standard called WebAssembly, or wasm, in development as of 2017, will run programs in browsers at about 80% of the speed of the native code from which they were converted.
Lisp – List Processing. A venerable language with several versions.
Objective-C – An object-oriented superset of C that has inter-object messaging. It was the primary language for Mac OS and iOS applications before the introduction of Swift. Header files still use the .h filename extension, while source files are .m, or .mm for Objective-C++.
Pascal – Named for French mathematician Blaise Pascal. Niklaus Wirth created it in the late 1960s, largely as a tool to teach structured programming, which primarily means dividing programs into logically related sections with a single point of entry and data structures that relate to the sections. Like Java, modern Pascal employs compilation into an intermediate stage (P-code), which can then be either compiled into an executable or interpreted at run-time.
PERL – Practical Extraction and Report Language. A flexible Unix scripting language created by Larry Wall for high-level system control, and often used to manage Web servers.
PHP – PHP: Hypertext Processor (that’s a recursive initialism, kids). Originally stood for Personal Home Page. A FSF-supported, non-GPL, server-side Web scripting language that can be embedded in HTML. PHP variable and class property names are case-sensitive, but function and class names are not. Variable names must begin with $ (dollar sign). Web servers expect an HTML document containing PHP code to have the .php extension. Vulnerabilities in PHP code are one of the most popular targets for malware.
Python – Created in the late 1980s by Guido van Rossum, and named for Monty Python’s Flying Circus. It uses interpreters, and is popular for server-side Web scripts as well as desktop apps. Python uses the hash character # for single-line comments, and three double-quote characters """ to start and end block comments. The most surprising thing about the language is that indentation with whitespace characters defines the extent of functions. Python 3.0 (2008) is not backwards-compatible. The source code files usually have the .py filename extension, while .pyc and .pyo are used for cached byte code.
R – An interpreted, open-source language created specifically for data mining, data processing, and data analysis. The basic R environment uses a command-line interface, but there are third-party tools that provide a GUI.
Ruby – Open-source, object-oriented language developed in the early 1990s by Yukihiro “Matz” Matsumoto. Ruby on Rails, a mid-2000s Web development platform written in Ruby, popularized the language.
Rust – (2015) Backed by Mozilla.org for Web development, Rust is designed to greatly reduce unsafe uses of memory relative to C, on which it’s based, while producing software that runs almost as fast. Unlike C, it employs static typing and rejects implicit type conversions.
Swift – Apple’s 2014 security-conscious open-source language for Mac OS and iOS applications, supplanting Objective-C and Objective-C++, which, with their inherently less safe uses of memory, are more vulnerable to hacks.
VB – Visual Basic. A visual programming language from Microsoft, strictly for creating Windows applications. As the name suggests, it’s an extension of BASIC. The later VB .NET is not compatible with the original VB.
VBA – Visual Basic for Applications. A pared-down VB included with Microsoft Office, for creating macros to run within Office applications.
VBScript – Visual Basic Script. A Microsoft scripting language derived from Visual Basic. Its primary use is generating HTML content dynamically for Web pages – see ASP. The files usually have the .vbs filename extension.
PROM
Programmable Read-Only Memory. An old type of IC chip that can be programmed once, and only once, by the user. (Strictly speaking, the re-programmable EPROM and EEPROM are also PROMs, but “PROM” is understood to mean one-time programmable ROM.) This requires a device programmer, which essentially blows fuses in the PROM chip to hard-code the data (a process called burning the chip), and a user interface – usually a PC running device-programmer software.
A PROM can’t perform logical operations. It’s just storage for programs, repeating functions, lookup tables, etc. used by other devices.
PROM chips have mostly gone away, replaced by flash memory. They retain niche roles in aerospace applications because they’re highly resistant to radiation, and in consumer electronics DRM or critical systems because they can’t be hacked.
propane
A gaseous alkane hydrocarbon, C3H8, widely used as a fuel. Stored under pressure as a liquid, it’s also called liquefied petroleum gas (LPG). It vaporizes when released for combustion. A propane-air mix must be between 2.2% and 9.6% propane to burn. See LNG.
propanol
Either CH3CH2CH2OH (1-propanol) or its isomer (CH3)2CHOH (2-propanol). The latter is also called isopropanol or isopropyl alcohol, and is used as medical alcohol. Both compounds are toxic and highly flammable. “Propanal” is a different compound. Compare ethanol, methanol.
protected mode
The native running mode of the old AT CPUs, allowing them to address extended memory.
proxy
See server.
PRS
Partial Response Signaling. See partial response modulation.
PS/2
Personal System/2 (originally). This was the designation for a series of IBM PCs introduced in the 1980s. It has come to refer to the mini-DIN mouse interface used for the machines. The chief alternative at the time was the serial mouse, which uses a DE-9P serial port. Both have been supplanted by USB or wireless connections. The PS/2 interface was common for keyboards as well.
PSD
Power Spectral Density. A function describing how the power of a signal is distributed with respect to frequency, normalized to 1 Hz.
psi
Pounds per Square Inch. A standard US unit of pressure.
psig
Pounds per Square Inch in Gauge. A standard US unit of pressure relative to ambient atmospheric pressure. That is, ambient pressure is always 0 psig.
PSDN
Packet-Switched Data Network.
PSK
Phase Shift Keying. A family of digital modulation techniques, derived from analog PM, for shifting the phase of a carrier wave to convey information. There’s BPSK (binary PSK, 1 bit/symbol), QPSK (quaternary PSK, 2 bits/symbol), and 8-PSK (ainary PSK, 3 bits/symbol), depending on the number of phase positions the signal uses. 16-PSK is rare. These are also called M-ary PSK, where M is the number of symbol states – 2, 4, 8, or 16. Phase imbalance refers to the amount by which the phase shift varies from the ideal. Amplitude imbalance is the power differential between the waveforms.
Because of AM-to-PM conversion from its 180° phase discontinuities, M-ary PSK is not a constant envelope modulation, although variants such as OQPSK and MSK are. See also DPSK.
PSN
Packet-Switched Network.
PSOP
Plastic Small Outline Package. See JEDEC.
PSpice
See SPICE.
PSRR
Power Supply Rejection Ratio.
PSTN
Public Switched Telephone Network. Also called POTS.
PSU
Power Supply Unit. An incorporated AC/DC power converter, as in a PC; see regulator. A switching PSU won’t run without a load. Most have protection circuitry that will shut them down if no load is present, and some have artificial resistive loads to keep them running when not connected. Most also have the capability to use either 110V/60Hz (American) or 220V/50Hz (European) AC power, although the user might have to manually set a switch to the appropriate type.
Power supply unit, ATX form factor, with 80mm rear exhaust fan
PSU with 80mm rear exhaust fan
Many PSU designs have an incorporated cooling fan – either one that pushes warm air out as in the image above, or one that pulls cooler air in. The former is probably more effective, although the latter approach is often able to use a larger, slower fan that runs quieter.
PTC
Positive Temperature Coefficient. A thermistor with resistance that increases with temperature. The opposite type is NTC.
PTH
Pin Through Hole. The original circuit-soldering approach, using component form factors such as DIP and TO-92 (see JEDEC) with long wire leads that are inserted into holes in a circuit board and soldered into place. Excess lead is normally snipped off afterward. Modern electronics instead use SMT.
PTN
Pager Telephone Number. A phone number that connects to a paging system, and corresponds to a particular CAP code. A single pager can have multiple PTNs.
PTT
Push-To-Talk. A half-duplex (i.e., transmits and receives but not at the same time) voice radio that transmits only while a key is pressed.
pulse generator
A laboratory instrument that generates electrical signals containing square-wave pulses with properties defined by the operator. Compare AFG, AWG, signal generator.
push-pull
A logic circuit output that consists of two open circuits – one current sink and one current source – so that the output can push or pull. Also called a source-sink or totem pole output, although the latter term is usually specific to TTL. Multiple push-pull outputs cannot be tied together. Compare open collector.
PV
Photovoltaic. The technology that generates electricity from light – and not only sunlight, as PV-powered calculators that work indoors demonstrate. As of 2013, research and aerospace PV systems can achieve conversion efficiencies above 40%. However, they rely on unusual semiconductors, and more complex construction: compound cells with multiple junctions, each one optimized for a different part of the light spectrum, and concentrators that focus more light on the cells. This makes them much too expensive for general use.
As of 2022, most commercial PV cells are made with polycrystalline silicon, and have 15% to 20% efficiency. (Moncrystalline silicon and gallium arsenide achieve higher efficiency, but cost more.) Commercial thin-film PV cells using amorphous silicon, carbon-based dyes, polymers (plastic), or nanostructures deposited on various non-semiconductor substrates are just a few percentage points less efficient, and are lighter, cheaper, and flexible. However, they’re a much smaller part of the PV market, partly because most rely on toxic metals such as cadmium (see RoHS), and partly because deploying thin-film PV as rigid panels requires two panes of glass rather than one, making them heavier than crystalline silicon panels. Perovskite – meaning not only calcium titanate (CaTiO3), the mineral named for Russian scientist Lev Perovski (1792-1856), but any compound with its crystalline structure – is a promising material to replace silicon in commercial panels.
PV panels produce DC power. To supply AC power compatible with the electric grid, a PV panel installation requires both a large rechargeable battery and an inverter.
Thermophotovoltaics (TPV) systems first convert heat into light, and then use that light to generate electricity. Researchers are trying to improve their efficiency, which has exceeded 40% as of 2022, by designing them to generate only those wavelengths that the PV element can use.
PVC
(1)
Permanent (or Private) Virtual Circuit. An always-available path through a packet-switched network. It’s intended for hosts that communicate frequently. Compare SVC.

(2)
Polyvinyl Chloride. A family of tough plastics used for water pipes, cable jackets, junction boxes, and many other things.
PVCS
Polytron Version Control System, originally. See version control.
PWA
Printed Wiring Assembly. Same as PCB.
PWB
Printed Wiring Board. Supposedly means a bare circuit board, i.e., a PCB or PWB before it has been populated or, in industry slang, stuffed with circuit components. It’s often used more loosely to refer to populated boards as well.
PWM
Pulse Width Modulation. Also called pulse duration modulation. The signal carrier is a constant-frequency train of amplitude pulses, and the width (duration) of each pulse corresponds to instantaneous signal amplitude.
PWM is a poor RF modulation scheme, but an efficient way to regulate power supplied to a load, because it dispenses with the need for the load to have timing circuitry and power-wasting resistors. It’s therefore a common control signal format for analog electronics. For example, many PC motherboards control cooling fan speed with PWM. A motherboard with this capability has a four-pin rather than three-pin header for the fan power cable.
Motherboard PWM fan header
Motherboard PWM fan header
PWR
(1)
Power.

(2)
Pressurized Water Reactor. See LWR.
PXI
PCI Extensions for Instrumentation. Originally (1997) from National Instruments, PXI is an open-standard but compatible variant of the CompactPCI data bus, and is meant to support PC-based testing, measurement, instrumentation, and automation. The most common use is a PXI backplane bus in a rack-mount chassis that includes a PXI controller running an operating system. Card modules with the needed capabilities are inserted into slots in the chassis, and the controller makes them available to other installed software such as NI’s LabVIEW.
The PXI standard is maintained by the PXISA. They issued PXI v2.0 in July 2000. In 2005, they released the PXIe (PXI Express) standard, based on the newer PCIe local-bus standard used in PCs. It runs at 6 GB/s full-duplex. Unlike PCIe, PXIe retains backward compatibility at an electrical if not a physical level with its ancestor PXI, so a PXIe chassis can have hybrid slots that accept either a PXI or a PXIe card. PXIe v3.0 was released in 2015.
PXIe chassis
PXIe chassis
PXISA
PXI Systems Alliance. See PXI.
Python
See programming language.
PZT
Lead Zirconate Titanate. (Lead’s elemental symbol is Pb.) A perovskite ceramic with high piezoelectric and pyroelectric sensitivity, commonly used for sensors and actuators.