The terminology used in a high-tech field like semiconductors is necessarily complex to adequately distinguish advancements and peculiarities of the technologies involved.  However, the jumble of words and acronyms, and their usage, can quickly befuddle the unenlightened.  Yet it is important that high-tech terminology is used properly by speakers and writers, and especially marketers, to assure good communication. 

In this glossary, Strategy Sanity documents industry-accepted use of terms and acronyms, mostly dealing with the technology surrounding digital processors, semiconductors, and electronic equipment.  Beyond just definitions, discussions expand on important issues and clarity is given to implications of the technologies.  In fact, a careful reading of the glossary provides special insight to how the semiconductor world runs.  Diagrams and additional information may be referenced, where helpful.  

Be aware that terminology is often mis-used, either mistakenly, to add perceived qualities to a product, or because “we always say it that way.” 

Over time some technologies may be viewed as moving from good to old or "less good", usually because a newer technology is positioned as better by its promoters.  But this is rarely reality.  Engineering is a trade-off of multi-dimensional pros and cons.  However, this can cause terminology to be stretched to fit inappropriately.  When CISC became old-style, the term "variable-length RISC" was coined to borrow goodness from the newer term. 

Note that capitalization may be used herein to illustrate the source of the acronym but not be appropriate in normal writing. 

Note that words in this glossary are described in the context of high-tech electronics, electronic applications, semiconductors, and embedded processors before other or more-general usage, with hardware connotations taking precedence over software.  Readers taking exception to information presented in the glossary should address concerns to the primary Strategy Sanity analyst. 

Acronym/

Symbol/Term          Description / Explanation / Implications 

µ                  micro-, the Greek mu, refers to one-millionth, 10^-6  For convenience, the lower-case “u” is often substituted.  See also: numbers

Ω                 An ohm is a measure of electrical resistance.

101              101 is often used to indicate very fundamental information about the topic. The source of the term/number 101 comes from college life where courses are numbered according to their level of difficulty (and often credit hours), thus, the ’01 indicates the first, a beginner course.  The author of this glossary actually believes it should be more appropriately named Semis 378, since it is based on decades of experience in semiconductors, electronics, and engineering.  

3D                Three-dimensional is used loosely in many cases to simply refer to object lighting and shading effects which give realism to created objects, but at other times refers to true stereo-imaging allowing each eye to see a slightly different view of a scene which the brain interprets as natural distance and depth.  3D can also apply to audio where the listener can determine a depth or third-dimension to the source of a sound, often through interpretation in the brain of how the sound is altered by near objects like the head, hair, and ears. 

3G, 4G          Third-Generation, Fourth-Generation, is often applied to wireless cellular phones, but appropriate for nearly any system

Ĺ                 Angstrom: 10^-10 meters; perhaps the smallest unit of measure, originally to measure the wavelength of light (electromagnetic radiation), but also used to measure atoms, molecules, and other ridiculously small items.

A-to-D          Analog-to-Digital Converter   See:  ADC, D-to-A

AAC              High-efficiency Advanced Audio Coding version 2 is a lossy data compression scheme for digital audio. It is an extension of Low Complexity AAC (AAC LC) optimized for low bit-rate applications such as streaming audio. HE-AAC version 1 (HE-AAC v1) uses Spectral Band Replication (SBR) to enhance the compression efficiency in the frequency domain. HE-AAC version 2 (HE-AAC v2) couples SBR with Parametric Stereo (PS) to enhance the compression efficiency of stereo signals.  (description originally from www.worlddab.org/technology) 

ABI               Application Binary Interface, a software term

ABS              An Anti-skid Braking System uses a microcontroller to optimize braking on vehicles by rapidly pulsing the brakes when the wheels are skidding.  ABS even allows steering to be effective under lock-up conditions.

ADC, A/D       Analog-to-Digital Converters (sometimes called A-to-D converters) take a voltage reading on their input, which may come from a temperature sensor, pressure gauge, or other real-world continuous signal, and translates it into a digital representation for use by a processor.  ADCs are judged by the speed (1 Msps –million samples per second) and the resolution and accuracy (8-, 10-, 12-bit with ±1/2 bit) of their conversion.  Methods of conversion include successive approximation and delta-sigma (sometimes sigma-delta.)  

Because analog circuitry is best manufactured in a different manner than digital circuitry, the best A-to-D converters are typically designed on an analog process technology (and referred to as mixed-signal – mixed: analog mixed with digital).  To run fast and be built cheaply, most digital processors are built on a digital technology, so it is very difficult to integrate the best A-to-D converters on the same chip with a digital processor.  A D-to-A converter (or DAC) performs the reverse function by producing an analog output voltage that is proportional to its digital data input.  The output voltage can drive a motor, display or control device. 

ADPCM          Adaptive Differential Pulse Code Modulation

ADSL            Asymmetric Digital Subscriber Line

AES              Advanced Encryption Standard    Others:  SHA, DES

AFE              Analog Front-End

ALU              Arithmetic and Logic Unit.  The width of the primary integer ALU in a processor is the truest measure of the “size” (bit-size, or word-width) of a processor.  If two 16-bit pieces of data can be added in a single pass through the ALU, then the device is a 16-bit processor.  See also:  FPU, CPU, processor

AMI              Advanced Metering Infrastructure

AMIS            Advanced Metering Information System

AMP              Asymmetric MultiProcessing.  See also:  SMP (symmetric), heterogeneous, and discussion at homogeneous

AMPS            Advanced Mobile Phone Service

AMR    *        Automated Meter Reading

AMR    *        Advanced MultiRate encoder

AMR-NB  *    Adaptive Multi Rate - Narrow Band, a vocoding compression technique

analog           (see discussion at A-to-D, digital)

ANSI             American National Standards Institute

API               Application Program Interface, a software term

APON            ATM Passive Optical Network  (see Network)

Architecture   Sometimes short for Instruction Set Architecture

ARM    ^*         1) A processor architecture:  ARM refers to the RISC processor cores that execute the ARMv1, v2, … Thumb, or Thumb2 instruction set which were developed by the company that became ARM Ltd.  Common ARM processor cores include the ARM7TDMI, ARM9, ARM11, Cortex A8, Cortex R4, and Cortex M3.  [Other popular multi-vendor architectures include: ARM, MIPS, Power/PowerPC, SuperH/SH]

ARM    ^*         2) A company:  ARM also refers to the company ARM Holdings which designs, maintains, markets, and ultimately licenses the popular ARM processor architecture as intellectual property (IP) to customers who build the processor into a semiconductor chip.  ARM is short for Advanced RISC Machines.  www.arm.com

ARPU            The Average Revenue Per User is frequently used by cellular service providers to measure the viability of a new service. 

ASCII            American Standard Code for Information Interchange – a jumble of words that signifies the coding for the American-English alphabet, numbers and some special characters into a binary code that computers use.  Still the basis for most character-based computer codings, since expanded to include many other languages.  Fonts overlay ASCII.  A sample of ASCII coding into hexadecimal:
   0 = 30     1 = 31     2 = 32     3 = 33
              A = 41     B = 42     C = 43 
              a = 61     b = 62     c = 63 . . .

ASIC             Application-Specific Integrated Circuit.  The acronym is a bit of an inadequate description because there are many ICs dedicated to a specific application.  To fully distinguish an ASIC properly, think of it as a customer-specific or OEM-specific IC.  An ASIC is really designed for only one OEM, and is not sold to any other customer (unless they are explicit licensees of the chip).  The ASIC is generally designed using customer-owned tools (COTS).  However, there are numerous examples where a chip vendor lays out the chip, either because the ASIC is really a special alteration of a standard commercially-sold circuit, or because the chip is designed from the ground up for that one customer.  ASSPs are similar to ASICs except they are sold on the merchant market and are fully described with a traditional publically-available data sheet.  Further, important explanation is given at ASSP and SoC. 

ASP              The Average Selling Price is a common, but broad pricing measure of a category of products; total revenue generated in a product category divided by the number of units shipped.  50,000 $100 processors and 250,000 $20 processors might sell in a year resulting in a $33 ASP.  Note no individual chips need to sell for $33.   See also: KU

ASSP            The Application-Specific Standard Product is sold to the merchant market to address the needs of perform a specified function needed in a stated application.  The chip is described in a traditional publically-available data sheet.  However, many MCUs, MPUs, DSPs can be mistaken for ASICs and ASSPs.  The only real way to distinguish these chips is to apply a rigorous definition of “application-specific”.  A little history also may determine the proper designation. 

                   Whether a chips is a standard processor or an ASIC or ASSP really should be based on three things. 
1) How much of the on-chip circuitry must be specific to the application?  If 20% of the circuitry by area is unique to the application, is that A-S?  If it’s only 5%, it is likely that other applications can make good use of the 95% non-specific circuits and buy the chip for completely different end-equipment. 
2) How narrow must the use of the term “application” be?  If the “application” is as broad as “wired communications” is that “application-specific” or would it have to be as narrow as “VoIP-specific”?  
3) An ASIC differs from an ASSP because an ASIC has no public data sheet and is sold to only one customer. 

                   This analyst deems a chip to be application-specific if “only a fool” would use the chip for an application other than the one specified, and requires the application to be very narrowly defined, such as VoIP or laser-printer.  More than 20% of the chip should be “specific” lest it will attract many other users.  “Industrial” and “Wired Communication” are application segments.  Also see: SoC and MCU. 

                   Note that many MCUs are customer-unique when programs are coded into mask ROM on-chip. However, this is understood in the definition of an MCU and thus an insufficient qualifier to designate the chip an ASIC.  Also, many MCUs are designed targeting an application or application segment yet are not considered to be ASICs or ASSPs. 

ATM             Data transmissions in Asynchronous Transfer Mode utilize a 53-byte cell consisting of a 5 byte header in front of 48 bytes of payload data.  The header consists of information used for switching, packet priority, flow control, and header error control.

ATSC            Advanced Television Systems Committee standards for high-definition TV (HDTV). 

AUTOSAR       Automotive Open Systems Architecture

A/V               Audio and Video

underside of BGA

b                  Lower-case “b” stands for bit

B                  Upper-case “B” stands for byte.  B can also stand for billion (10⁹)

B2B              Business to Business services are not intended for consumers to use

BCD              Binary Coded Decimal.  See also:  ASCII

BGA              A fairly dense package for semiconductors uses a regular pattern of spheres of conducting material.  Advantages of the Ball Grid Array are that it utilizes nearly all of the package area, including the space directly underneath the die (chip) in the package.  Heat from the die can be transferred to the underlying circuit board to which the package is attached, benefiting today’s larger high-power (10+ watt) chips.  BGA packages take special abilities to solder to the PCB and solder joints are more difficult to repair.  The BGA was one of the earliest packages that took advantage of all of the under side of the package, whereas before the BGA, mostly the perimeter of the package had leads connected. 

binning          Binning is a casual term for a test procedure used to select the best semiconductor chips from many in a manner similar to choosing players on a sports team in grade school.  See discussion at Temperature Range.

BIOS             Basic Input/Output System

bit                The smallest representation of a binary state, 1 (on) or 0 (off) used by a computer.  Note that bit is usually abbreviated as a lower-case “b”.  See also: byte

black box       A black box is the physical expression of a function, that is, it produces outputs based on a given specification of a sequence of inputs and time.  The term “black box” is used to denote that it doesn’t matter how the outputs are determined - they might be created by logic or by a processor, some combination thereof, or it may be a simple or complex formulation.  For instance, the braking of an automobile could be the function of a black box:  it doesn’t matter whether hydraulics press on pads or electric motors drive calipers or if a traction-control system intervenes, when the driver steps on the pedal then the car must slow down.  See also: block diagram

block diagram The block diagram illustrates the major components of a system and how they are generally connected.  Many details are left out and orientation is not intended to be accurate. 

blue wires     In the day when printed circuit boards were fairly easily modified to correct design errors by soldering a wire between two points, blue-colored wire was often used.  A quick glance at the number of these blue wires that were on the PCB would show how many small changes were needed on this version of the board.  Ultimately the term “blue wire” became synonymous with correction and a board with “no blue wires” is a sign of a solid design.  As multilayer PCBs, very small traces, and traces hidden below large chips, practical use of soldered wires to correct design flaws has become far more difficult.

BOM             The Bill Of Materials lists all of the components and materials needed to build the product.  Assembly labor, software, testing, royalties, shipping, tariffs, overhead, and cost of sales must be added to the cost of the BOM to arrive at the final product’s total cost.  In speech, BOM is often pronounced the same as bomb rather than spelling it out.  See also: TCO.

bps               bits per second, typically in a serial communication channel

BRAS            Broadband Remote Access Server

BTB              Branch Target Buffer

BTS              Base Transceiver Station.  the cellular “basestation"

byte             A convenient, most-popular size of a number of bits, 8 bits make one byte, representing 0 through 255, or -128 through +127. 
Note that byte is usually abbreviated as an upper-case “B”. 
See also: bit, numbers

cache           Cache memory is intended to provide instant memory access to the processor when most of the memory system is slower or more distant. Caches keep a copy of data that is in more-distant memory. Management of cache is a science to maximize efficient use of the no-wait memory accesses. 

Cache memory benefits from the observation that subsequent memory accesses have a high probability of being very near, often in the very next location, to the most recent memory access.  This “locality of reference” is especially true of program instructions which are normally sequential with occasional branches back to recently-used instructions. 

Data is more problematic in a cache.  Since only 10% of memory accesses are seeking data rather than instructions, any data caching will have a much smaller benefit.  While some data can come from memory near a recent data access, it is almost as likely to come from a very different location.  Some data accesses actually come from peripheral and I/O locations that are expected to change dramatically with each access, and must be “live” rather than an old, saved copy.  Data accesses can also intend to write results out into memory, and management of the changes in the cache versus the original or “stale” data in farther memory can be complex. 

“Icache” and “Dcache”, sometimes greatly abbreviated as I$ and D$ by playing off the dollar sign to indicate cash (sic), refer to instruction and data cache, respectively.  A unified cache leaves it to a proper memory management scheme to partition instructions and data in a large memory array.  There can be multiple levels of cache, but increasingly less-effective benefits of speed, cost, and management tradeoffs usually limit practicality to two, and maybe three levels of semiconductor cache. 

CAD              Computer-Aided Design, although the word “aided” is quite an understatement

CAGR            Compound Annual Growth Rate is determined by the formula: 
( ( [final year value] / [start year value] ) ^ ( 1 / ( [end year] - [start year] ) ) - 1
. . . and usually expressed as a percentage and often considered for a 5-year period.  So if revenue in 2010 is $1M and expected to grow to $2M in 2015, then the CAGR would round to  14.87%    (CAGR is often pronounced Kay’-Ger)

CAN              The Controller Area Network is a standard driven initially by the Robert Bosch Gmbh as a low-level network suitable for the harsh environment of an automobile.  It has evolved in the marketplace to be used in industrial applications as well.

CDMA            Code-Division Multiple Access (IS-96), a cellular technology

CDRS            Common Digital Radio System

CELP             Code Excited Linear Predictive

CISC             Complex Instruction Set Computer:  processing operation, addressing mode, register, and data size designations are encoded into very compact instructions that might find their data in external memory. The 68000 (somewhat morphed into today’s ColdFire), x86, and most microcontroller cores are typical CISC architectures.   See also: RISC

CMOS           Complementary Metal-Oxide Semiconductors pair a P-channel transistor with an N-channel transistor which performs a sort of push-pull effect that very quickly switches the silicon-based transistor from ON to OFF and back, saving power dramatically.  Thus, CMOS became known as the low-power low-cost semiconductor technology.  In the mid-1970’s PMOS was largely left behind in favor of faster NMOS.  By 1980, CMOS had become the technology of choice for most semiconductors. 

CO               The Central Office is the “telephone company” facility where cabling to the subscriber connects to line cards which handle the interfaces, now including data transmissions over the circuit switched network.  See also description at CPE.  CO also stands for Carbon monoxide.

CODEC          coder/decoder

compatible     Compatible does not mean precisely identical.  It means that one technology can work with another.  “Compatible” is one of those words that marketers (and engineers) apply liberally.  Even if I’m compatible with my wife, we may have a lot of issues to work out – even after 20 years. 

Upward compatible generally means the latter or greater product will operate the same as the earlier or lesser product.  Backward-compatible is the view from the lesser side.  For a hardware interface like a serial or parallel bus it means the updated bus will accept signals from a product designed to run on the original bus.  USB 2.0 hosts are backward compatible with USB 1.1 signals because the 2.0 systems can handle the slower 1.1 transmissions.  Some semiconductor chips can be pin-compatible which means the two parts can be interchanged and still operate.  A 100-pin part can even be compatible with a 64-pin part if all the signals on the 64-pin part are in similar positions on the 100-pin part.  See the illustration.

Some semiconductors may have I/O lines that are 5 volt tolerant.  These pins may normally drive (generate) 3.3 volt levels, but can read inputs and, importantly, are not harmed by the 5 volt inputs that other parts generate. 

For processors compatibility can mean that programs written for the old processor will run on the new, upward-compatible processor.  But there are different levels of compatibility. 

Binary or machine code is the 1’s and 0’s a processor reads to get its instructions.  Machine code compatible processors are the most compatible. with one being able to execute the 1’s and 0’s programs written for the other.  For Processor B to be upward-machine-code-compatible with Processor A, all machine-code instructions that can be executed by Processor A must run the same on Processor B.  The memory patterns found in ROM and DRAM should perform the same whether run on Processor A or B.  However, Processor B will usually have some new, improved instructions that Processor A does not recognize – thus the use of the word “upward”. 

Slightly less compatible are assembly-code-compatible processors.  Assembly language is abbreviated but generally human readable instructions that may look like LDA #35 which means load register A with the number 35.  This may “assemble” into machine code resulting in 86 35 or 10010110 00110101.  On Processor B this assembly code might turn into 4086 0035 and so not be compatible at the machine level while being compatible at the assembly level. 

To say processors are compatible in high-level language is marketing nonsense.  High-level languages can be compiled to any processor instructions including DSPs, so all are HLL compatible. 

connectivity   Connectivity is the ability to communicate over a serial link with another device, rather loosely and at close range.  Connectivity is an increasingly important characteristic of many electronic products today, making it easier for them to work in unison, often without express human intervention.  Personal Area Networks (PAN) like Bluetooth and Zigbee are often utilized in “connected” devices.  There is no hard definition distinguishing connected devices from those which otherwise communicate over a serial link, but characteristics of connected devices usually include: 

                   • simple communications protocols
• low bandwidth and/or brief transmissions
• wireless rather than wired links
• mesh rather than point-to-point networks with possibly hundreds of direct contacts
• communication being only one, perhaps minor, function of the device

core             A Core processor, usually licensed, that is a part of a highly-integrated ASIC, ASSP, or OEM-designed custom chip, sometimes combined with other processors  [Discussion at “processor”] 

COTS *         1) Customer Owned Tools generally refers to computer-aided design (CAD) tools that an OEM might use to design its own semiconductor circuits, often leading to a chip in the ASIC or purely custom category.  

COTS *         2) Commercial-Off-The-Shelf products are sometimes specified in aerospace, defense, military, or other high-reliability applications where it is believed that the product will be in a less severe environment such that traditional high-reliability (hi-rel) products can not be justified against the lower cost expectation of COTS products.  

CPE              Customer Premise Equipment is the equipment at the terminating end of a communications network, and typically located on the end-user’s premises.  In some ways, CPE is to communications what “client” is to the distributed computer.  CPE is distinguished from the head end of the system, the central office (CO).

CPR              Chip Rate Processing.  Used in cellular phone technology.  Also consider: Symbol Rate Processing

CPU              Central Processing Unit – the main processor in a system, often the MPU in a PC or mainframe.  [Discussion at “processor”]  See also: MPU, RISC, CISC, processor 
[the “U” of CPU stands for unit, a remnant of the days when many individual components on a board, and later when a single chip, comprised this function.  In speech, “unit” is rarely used now]

CRC              Cyclical Redundancy Checking uses algorithms to verify that received data is uncorrupted.  CRC does not facilitate correction, only detection of errors.  Most often used in serial transmissions.  See also: ECC, EDAC

customer       To the semiconductor vendor, the customer is designer of a piece of electronic equipment, often an OEM.  The engineer at the OEM decides which components to use in putting together the cell phone, television, or network switch being designed based on the needs of the end-user of the equipment and within criteria for the equipment like features, upgradeability, quality, cost, size, power consumption, etc.

Since year 2000, more semiconductor customers have been a middleman, the contracted ODM who does the design of the end-equipment but then turns the design over to the OEM to either manufacture the product or have it manufactured. 

The customer might also be the designer of a board-level product that may not be a fully-finished product but will end up being sold to an assembler who finishes the product, or perhaps will be combined with other boards and software to run a piece of mechanical equipment or shop floor.  See also: design-win, FAE.

D-cache        Cache memory set up to contain data.  Data Cache memory.  Sometimes D$ is used as a short-hand.  [Discussion at “cache”] See also: cache, I-cache

DAB              Digital Audio Broadcasting  (see www.worlddab.org)

DAC, D/A, D-to-A       Digital-to-Analog Converter.  See: ADC

DECT            Digitally Enhanced Cordless Telephone is generally a technology used in Europe

DES              Data Encryption Standard.    Others: SHA, AES

design-win     The effort to sell a semiconductor to a customer may start months before any chips are sold.  The vendor often must provide technical documentation to demonstrate how its product will function in the customer’s application.  When the customer decides to design the vendor’s product into the application, it is considered a “design win” or “design-in”.  This is also referred to as “winning the socket”, because parts may be mounted in a socket. 

Unseating the competition by convincing the customer to replace a competitor’s part with yours can be especially sweet, and referred to as “stealing the socket.” It may still be 6 to 12 months before any real volume of chips is shipped, but the design win is the point at which the vendor gains confidence there will be dependable volume shipping soon.  See also: customer, OEM, FAE

DHCP            Dynamic Host Configuration Protocol

die               One of the last steps in semiconductor wafer fabrication is scribing and breaking the wafer in a grid pattern to isolate each circuit into an individual die.  This is mounted into a special protective package and its electrical connections are brought to the extremities for easy solderability to a printed circuit board (PCB).  At least one vendor uses the term “bar” rather than die.  Engineers find microphotographs of die (a “die photo”) to be fascinating, especially if major circuit functions are marked on it.  Most people might think they look like a satellite photograph of New York City, with Donald Trump’s properties circled. 

digital           Discrete quantities, such as either 0 or 1 but nothing in between, are considered digital.  Digital is often contrasted with “analog” which has infinitely variable quantities, such as 74.6 degrees.  “Fuzzy Logic” was, in some ways, an attempt to bridge analog and digital by using less defined concepts like “more” or “hot” which are both relative and mean different things to different users.  Digital computers and, by extension, digital equipment built from them are more precise in that from a given set of inputs and time sequences, the results can be predictably identical for all similar machines.  Analog electronics is more sensitive to temperature, drift (changes in component specifications over time), and part-to-part variation.  Digital equipment has been marketed as “better than” analog.  The “real world” is actually very analog – a tree branch bends in an infinite variety of ways, sound can be at any frequency, and blue is a band of light frequencies – and the analog of real life must be changed to digital with analog-to-digital converters so computers can analyze it [see discussion at ADC]. 

DIMM            Dual In-line Memory Module

disc, disk       Disc (with a “c”) is often used to designate an optical disc like a CD or DVD while disk (with a “k”) is used for a magnetic disk media like a floppy or hard drive, but the two spellings are also used less distinctively.

DMA             Direct Memory Access controllers transfer large blocks of data from a peripheral or memory to another peripheral or memory location without requiring attention from the main processor after it’s initial setup.

DNA              Deoxyribonucleic Acid is a biological term

DNS              Domain Name System (of the Internet)

DOD              United States Department of Defense

DOE              United States Department of Energy

DOJ              United States Department of Justice

DPD              Digital Pre-Distortion

DRAM            Dynamic RAM requires frequent refreshing to retain data.  See also: Memory

DSL              Digital Subscriber Line modem.  ADSL is an Asynchronous DSL that takes advantage of the fact that most users of Internet service bring data and files down to them (Web pages and downloads) and only infrequently push data and files up to the other people or machines (send and upload).  Most DSLs today are actually asynchronous (unbalanced) so they can download 10 times as fast as they can upload so they maximize the most-noticeable download transfer rate. 

DSLAM          The DSL Access Multiplexer is on the other end of the wire that connects to a DSL and is in the telephone company central office (CO) or possibly in a remote neighborhood cabinet. The DSLAM concentrates and controls many DSL lines onto one very high-speed network cable (a “fat pipe”) which then connects to the greater Internet. 

DSO              Device Software Optimization

DSP              The Digital Signal Processor excels at running algorithms such as FFTs, FIR filters, and others that are often used for processing data representing signals, often after being sampled and converted to digital.  While the MAC operation is the most common sign of a DSP, it alone does not define one.  DSPs utilize separate instruction and (sometimes two) data buses (a “Harvard bus architecture”) to maximize throughput of the constant stream of signal data and constants that typify signal processing.  Early standalone DSPs were dependent upon a host processor, a GPP, to set them up.  More recent DSPs are more self-sufficient, or are on-chip with a RISC processor that works with it.  Signal processing also makes use of extremely tight instruction loops that require no extra clock cycles be used.  DSP data size is typically 16-bits for voice and lower quality audio with 20 and 24 bits used for the best audio.  These data are referred to as "fixed-point".  32 bits are used for floating-point data in a DSP. A DSP data word is organized uniquely for maximum mathematical effectiveness so special normalization and error accumulation operations are important.  [Discussion at “processor”]  See also: Processor, MPU, MCU

DSSS            Direct-Sequence Spread Spectrum

DVB              Digital Video Broadcasting with Cable, Handheld, Satellite, and Terrestrial subsegments

DVI               Digital Visual Interface; alternative to HDMI

DVR              TiVo and ReplayTV simultaneously introduced Digital Video Recorders to the world but TiVo became so successful that its name became synonymous with the product, even being turned into a verb describing the act of recording a TV program onto a hard disc.  DVRs, also known as personal video recorders, completely overran the VCR market.  A hard drive as the recording media eliminated the shuffling of video tapes and an integrated program guide simplified set-up and playback of recorded TV programs.  Eventually, STB’s grew to include hard drives for video recording functionality.E1, E2, E3      2.048, 8.448, and 34.368 Mbps data transmissions for Europe and outside North America  (see also T1)

EAROM/ EEPROM       Electrically Alterable/Erasable Programmable ROM

ECC              Error-Correcting Code uses algorithms to verify that data is uncorrupted.  “Check bits” are appended to the actual data, which allow minor errors in the data to be detected when read or received.  Single-bit error correction and double-bit detection is common.  There is a “cost” associated with ECC: some logic and time is needed to perform the checking/correcting and additional memory is needed to maintain the check bits.  The original Hamming codes require 4 extra bits to correct errors in 8 bits of data, 5 bits in 16, 6 bits in 32, and 7 bits in 64.  One additional bit is required to detect double-bit errors (without correction).       See also: CRC, EDAC

ECU              ECU is used in automotive applications, sometimes to refer to an Engine Control Unit and more loosely as an Electronic Control Unit – in either case it consists of a microcontroller or microprocessor that reads various sensors as inputs and determines the proper action to control its assigned subsystem, from the engine to the climate control to the window lift.

EDAC            Error Detection and Correction (also EDC)   See also: CRC, ECC

EDGE            Enhanced Data rates for Global Evolution, a cellular technology

Embedded    An embedded processor is a processor which is embedded in a piece of equipment that is not considered to be a computer.  This entails most uses of processor chips.  On the other hand, when a PC, server, workstation, or mainframe is purchased, it is bought as a general-purpose computer with the expectation that added software will direct the machine to perform computational tasks.  These are computers and not considered embedded systems. 

EMF, EMI       ElectroMagnetic Field radiation or Interference is good if you’re making use of the radiation or bad if it interferes with the EMF signals you’re trying to use.  Antennae radiate EMF but cables must be shielded from EMI.  EMF can also refer to ElectroMotive Force which describes the electrical energy potential across the terminals of a battery or similar source. 

Endian          The sequence in which data is placed in memory, sometimes referred to as “byte sex”.  Big Endian data organization places the most significant part of the data in the lowest memory address; so the “big end” comes first.  Little endian format matches the least significant byte with the lowest memory address.  (Memory addresses are always assumed to “start” at zero.)  “Endianness” is an acceptable term within the computing community. 

EPA               United States Environmental Protection Agency

EPROM          Erasable Programmable ROM is bulk-erased by ultraviolet light.  See also: Memory

ESD              Electro-Static Discharge

Ethernet        Today’s baseline LAN protocol, the IEEE 802.3 standard, originally operating at 10 Mbit/s using 10baseT PHYs, now extended to 100 Mbit/s (called Fast Ethernet) using 100baseT PHYs, 1000 Mbit/s (Gigabit Ethernet, or GigE), and 10 Gbit/s (10 Gig Ethernet).  CAT5 cables are the physical medium over which Ethernet is transmitted. 

EU ETS          European Union Emissions Trading Scheme relates to the tradeoff of pollution emissions as a result of energy generation or consumption against tax or other financial credits or penalties. EV-DO           Evolution-Data Optimized, a cellular technology

fab                The fabrication facility is the building, equipment, and personnel that manufacture a semiconductor chip from raw materials.  See also: foundry, IDM

fabless          A fabless semiconductor company designs and markets integrated circuits, but relies on a contracted foundry for the fab that implements the design and manufactures the chips.   The fabless semiconductor company avoids the extreme costs of building, equipping, and maintaining a fab (easily $1 billion) while taking advantage of nearly the latest manufacturing technology, but has less control over parameters in the manufacturing facility and its products’ priority within the contracted fab.  See also: IDM     

FAE               The Field Applications Engineer has a critical role in the process of selling complex semiconductors.  Located in “the field” with the salesperson near the customer, the FAE must understand many technical aspects of the chip to help show the customer how to fit the chip into the customer’s application.  See also: design-win, customer

FCC              The United States Federal Communications Commission

FFT               Fast Fourier Transform is a fundamental signal processing algorithm often executed on a DSP

FHSS            Frequency-Hopping Spread-Spectrum

FIB               A Focused Ion Beam system can repair and somewhat “edit” circuits already etched into a semiconductor chip.  Generally, the physical circuits on chips can not be changed at all once they are fabricated.  If there is an error in the circuit design then the entire wafer, lot, and production run has the error and is worthless except for verifying circuits on the rest of the chip.  This is extremely wasteful of time and money and can occur late in the chip development stages or just after an alteration is made to the circuits to enhance a feature.  Sometimes individual chips have a flaw due to a contaminant in the manufacturing materials. 

FIB machines can deposit and remove insulating and conductor material in a very precise fashion, to recover otherwise-wasted chips.  FIB machines are expensive and rather slow to repair a chip, but can be very valuable in saving expensive mask layers and weeks of turnaround-time to run a new set of masks through the fab; all while engineers sit pensively wondering if the part of the chip that can’t be tested will perform as expected. 

FIFO             First In-First Out is a sequence description where the first data to arrive is the data first to be driven out; a formal description of a memory buffer.  See also: LIFO

FIR               Finite Impulse Response filter is a fundamental signal processing algorithm often executed on a DSP

Firewire         IEEE 1394

firmware       The physical components that make up the electrical part of a system is considered hardware: the processor, memory, display drivers, ASICs, analog and discrete components, among many others.  The printed circuit board, chassis, keyboard, and enclosure that make up a complete end-equipment may be considered hardware as well.  Hardware is essentially unchangeable. 

Software is the program that instructs the processor what to do.  Software can be changed infinitely depending on the task it is assigned.  Software is not something a person can touch but a printout of the instructions can be read and, for an electronic system, this software must be stored in the memory system so the processor can quickly read and execute the instructions. 

Firmware is also software but is in memory that can not be changed every second.  Firmware is saved in factory-programmed mask ROM or one of the non-volatile (NV) forms of memory like Flash or EEPROM.  The contents of ROM, Flash, and EEPROM are often copied into RAM (which operates faster) so the processor can read its programs quickly.  Flash and EEPROM can be updated while still mounted on the printed circuit board (PCB), but changing a mask ROM requires chip replacement. 

A simple way to think of software is like the recipe in a cookbook.  A recipe consists of a set of ingredients and instructions for preparing an edible dish.  If the right ingredients are not available or the instructions are not followed in the prescribed order then the results may be very different than expected.  Some substitution is possible (raspberries for blackberries) and chopping the carrots before the apples may not affect the outcome - similarly some programmers or processors can rearrange the order without affecting results.  In this analogy the oven, stovetop, blender, and knives are hardware, the cookbook is like the memory, and the cook is the processor following instructions. 

fixed-function          The beauty of a programmable processor is that a readily-changeable program instructs the processor exactly what function to perform.  However, it takes a while to sequence through the instruction strings and complete the task.  A fixed-function accelerator uses dedicated hardware to scream through the specified function 10 to 100 times faster than a programmable processor, but has none of the flexibility of the processor.  When the system doesn’t need the function, this circuitry is completely idle.  However, that has advantages because it can be turned off to save power, even though it will take some time to be turned back on.  These are sometimes called FASICs – fixed-function ASICs. 

Flash             Flash Memory is a non-volatile, electrically reprogrammable memory.  See also: Memory

FLOPS           FLoating point Operations Per Second is a measure of a processor’s floating-point performance, often too simplistic for adequate comparisons

FPGA             Field-Programmable Gate Array

FPU               A Floating Point Unit is a special processor that has unique logic to perform mathematical operations on data that represents floating point numbers, comprised of a mantissa and an exponent, such as 2.468 x 10¹³.  An FPU often has a general-purpose processor as a host, feeding it data.    See also: CPU, ALU

FSK              Frequency Shift Keying

FTP               File Transfer Protocol

FUD              Fear, Uncertainty, and Doubt are often spread by one vendor disparaging the products or services of his competitor.  The information may be true or may be twisted but is highlighted to raise concerns about the viability of the competitor, often unfairly. 

G.hn             an ITU specification G.9960 being developed for the home network relying on existing home wiring like mains and coaxial cable

GHz              a Gigahertz is one billion cycles per second  (10⁹) .  See also: frequency

GigE             Gigabit Ethernet

GOPS            Giga-Operations Per Second, that is, billion operations per second.  See: FLOPS, MOPS, MIPS

GPIO             General-Purpose Input/Output.   See: I/O

GPP              General-Purpose Processor is a good universal term for all RISC and CISC microprocessors, microcontrollers, and core processors.  GPPs are designed to perform a wide variety of operations, typically based on integer arithmetic and Boolean logic.  Specialty processors like DSPs, FPUs, GPUs, and network processors typically require a GPP as a host: setting them up and performing overall system management and maintenance.

GPR              General-Purpose Register

GPRS            General Packet Radio Service, a cellular technology

GPS              Global Positioning System

GPU              Graphics Processor

GSM             Global System for Mobile Communications, a cellular technology

GUIs             The Macintosh computer was the first commercially successful example of a Graphical User Interface which presented more natural graphical components like icons and used actions such as drag-and-drop in the human-machine interface, replacing the far simpler text-based screens that had existed before.  This required more complex interaction between the display (monitor) and the input device, now including a two-degrees-of-freedom mouse as well as a keyboard.  GUIs have advanced significantly since 1984 in terms of resolution, color, size, and the variety of input devices available. 
See also: WYSIWYG

H.264           H.264 compresses video streams utilizing block-oriented motion-compensating techniques resulting in about half the bits required represent the same video as MPEG-2, MPEG-4 part 2, or H.263 video.  This high efficiency comes at a cost, requiring nearly double the processing performance to handle the more complex decoding involved.  H.264 is used for Blu-ray discs and iTunes video.

HAN              The Home Area Network is really just a LAN but identifying it as being located in the home or a residence may imply ease-of-use, ownership, and standards characteristics that ease discussion of home applicance and home energy control, consumer electronics products, and audio/video data stream transmission (see Networks, G.hn)

HDLC            High-level Data Link Control along with SDLC (Synchronous) established the most basic synchronous (closely tied to a clock signal) serial point-to-point communications protocol while ADLC describes the Asynchronous method.

HDMI            High Definition Multimedia Interface - a trademarked standard

hexadecimal   Base-16 numbers are a handy way to write the binary machine code patterns so that humans can recognize them more easily, since two hexadecimal characters can represent a byte.  After 0-9, characters A, B, C, D, E, and F are used to represent decimal 10, 11, 12, 13, 14, and 15.  The term is often shortened to just “hex”. For clarity, hex numbers are sometimes appended with an “h” (or may be preceded by “$”), and appear in pairs representing bytes, such as 04h.  Binary looks like this in hexadecimal:
   0000 = 0h   0001 = 1h   0010 = 2h   0011 = 3h
   0100 = 4h   0101 = 5h   0110 = 6h   0111 = 7h
   1000 = 8h   1001 = 9h   1010 = Ah   1011 = Bh
   1100 = Ch   1101 = Dh   1110 = Eh   1111 = Fh

heterogeneous         Heterogeneous processors perform a system’s primary function using two or more different processors, fairly tightly coupled.  A good example is a RISC processor coupled with a DSP as is traditional in a cellular phone design.  Since they are not designed to run on their own, any system utilizing a DSP, network processor, or graphics processor is typically heterogeneous since there is a host GPP that sets up the specialty processor, runs the OS, and performs numerous applications functions.  See also: AMP, SMP, and discussion at homogeneous 

HID              Human Interface Device is the electro-mechanical system through which a person inputs data, receives information, or manages the electronics.  Common HIDs include the keyboard, keypad, mouse, touch-screen.  The term is commonly used in USB environments.  See also: HMI

HKMG            high-k/metal gate (generally for very aggressive semiconductor manufacturing technologies like 32nm)

HLL               Writing a program in a High-Level Language like “C” and “C++”allows the programmer to structure algorithms and tasks in the most useful manner without regard to the peculiarities of the processor on which it will run.  HLL programs are considered portable because they can be compiled into machine code for a number of processors.  Assembly language is cryptic but still human-readable and might be considered a low-level language, though the term is rarely used.  Machine code or binary code is just patterns of 1s and 0s and what resides in memory chips.  The processor or “machine” performs its operations based on binary code, and any human that can read binary needs to get out and play soccer a lot more.

HMI              The Human-Machine Interface is the means by which the flesh-and-blood user manages, monitors, and controls the electronic or mechanical system.  While this may be seen as the physical controls and display hardware, aided by significant underlying software to present conditions and interpret inputs, the most important feature of the HMI is how obvious, easy, natural, and intuitive its procedural operations are.  Typing “raise the temperature 3 degrees” may not be as obvious or natural as raising a lever slightly, and searching through chained menus may not be as clear as an icon.  A comprehensive understanding of many factors regarding the user at the controls is required for the design of a good human-machine interface.  See GUI, HID, WYSIWYG

homogeneous Homogeneous processors perform a system’s primary function using two or more identical processors, fairly tightly coupled.  This may be two or four cores on one piece of silicon, or even arrays of 256 standalone processors.  The most obvious examples of late are dual-core MIPS, PowerPC, or x86 processors, though most of these (certainly the x86) were doubled up to contain power consumption rather than the real goal of multiprocessing which is to quickly multiply the performance of a system. Homogeneous processors can be used in a symmetric manner (SMP) where a task runs on any available processor, or in an asymmetric manner (AMP) where the operating system and associated system functions run on one processor and applications programs run on the other processors.  See also: AMP, SMP, and discussion at heterogeneous

HSPA            High Speed Packet Access ;
also: Dual-carrier High-Speed Packet Access Plus (DC-HSPA+)

HSUPA          High-Speed Uplink Packet Access

HTML            HyperText Mark-up Language

https             HyperText Transfer Protocol over Secure Socket Layer (SSL) or Transport Layer Security (TLS)   

HVAC            Heating, Ventilation, and Air Conditioning systems

Hz                Hertz is a measure of periodic frequency, cycles per second

I/O               Input and Output, or I/O, are the signals that interface the real-world to a computer-based electronic system.  Inputs can come from switches, keyboards, sensors, and microphones and outputs can drive LEDs, buzzers, speakers, and actuators.  I/O signals are usually digital, although they may be converted from/to an analog form (see A-to-D).  The I/O line usually isolates the sometimes-harsh outside environment or shift the signal levels to be compatible with the processor.  While the simplest form of I/O is a single on or off indication, I/O is often more useful when grouped in sets of 8 for convenience to be read and written as bytes.  General-Purpose I/O (GPIO) implies a wide range of electrical compatibility is accommodated.  Related: peripherals

IC                Integrated Circuit.   Originally, electronic circuits were built using discreet components: transistors, resistors, and capacitors.  Jack Kilby at Texas Instruments is credited with being the first to successfully integrate a number of circuits onto a single semiconductor substrate.  This integrated circuit has all the good characteristics of a semiconductor - consistent mass production, low cost, low power, and long life.  Almost the only real negative quality of an IC is its lack of flexibility - it is impossible to change its function once it is produced.  But ICs were designed to serve the greatest number of applications, so the first digital ICs formed the most basic gate functions, and later flip-flops, soon adders and shift registers, and ultimately enough transistors were available to design the exotic graphics processors of today on single pieces of silicon.

The level of integration of electronic circuits onto a silicon substrate increases with each passing year.  An approximation of the integration level used to be indicated by a stepping up of the terms used.  Much of that has been abandoned for the present – until marketers see the need for adding a new word to the list.  Following are terms from a more naďve time when VLSI had more than 1000 equivalent gates – in increasing order of complexity.  See also: die, PCB, SoC, SIP
     SSI   Small Scale Integration
     MSI   Medium Scale Integration
     LSI   Large Scale Integration
     VLSI  Very Large Scale Integration
     SoC   System-on-a-Chip

I-cache         An Instruction cache is cache memory set up to hold processor instructions.  Sometimes I$ is used as a short-hand.  [Discussion at “cache”]  See also: D-cache

ICE               An In-Circuit Emulator is a hardware development tool that helps system designers replicate the operation of a processor in a target system (“in-circuit”) but with much better control to monitor and manipulate the system, looking to iron out problems

IDE    ^*          1) Integrated Development Environment for electronic system design

IDE    ^*          2) Integrated Drive Electronics for a PC disk drive

IDM              Traditional semiconductor companies are Integrated Device Manufacturers because they design the electronic circuits, manufacture the circuits in their own fabs, and market the chips all themselves (packaging the silicon may have been done by outside companies).  A very different form of company is the fabless semiconductor company that mostly designs and markets the integrated circuit, but contracts with an external foundry to implement the design or at least manufacture the chips.  

IEEE             Institute of Electrical and Electronic Engineers – a primarily-US-centric professional organization and a significant global standards body  www.ieee.org

IIC, I²C          Inter-Integrated Circuit serial communications standard, spoken as “eye-squared-sea”  See also: SPI, SCI

IIR                Infinite Impulse Response Filter is a fundamental signal processing algorithm often executed on a DSP

instantiation    Instance, copy.  A term often used by chip designers to indicate an individual copy of a circuit which may appear multiple times.  “Each instantiation of the core will consume 120 milliwatts” indicates that if four copies (or instances) of the core were in a chip, they would consume 480 milliwatts of power.  Other versions of this word are used, such as “instantiated”.

IP    *           1) Intellectual Property  (subset: Silicon IP (see: SIP))

IP    *           2) Internet Protocol

IR, IrDA        InfraRed is electromagnetic radiation at the wavelengths just longer than visible red light (frequencies lower than…).  IR is often used to transmit data without wires using LEDs (light-emitting diodes) in a serial form across a short distance (10 meters) but is generally limited to line-of-sight.  IrDA (Infrared Data Association) is a standard method that computing equipment can transmit data.  Most television remote controls use IR to beam to the TV. 

IRQ               Interrupt ReQuest

ISA               Instruction Set Architecture (see Architecture)

ISM              While not a standardized acronym, Industrial, Scientific, and Medical markets are sometimes grouped together.

ISO               International Standards Organization – what more need be said

IT                 Information Technology is a vast modern-day term encompassing all things computer

ITU               International Telecommunication Union - yet another standards body (see also ISO, IEEE)

IVR               Interactive Voice Response

JPEG, .jpg      Joint Photographic Experts Group for still images.   See also: MPEG

JTAG             Joint Test Automation Group, IEEE Standard 1149.1

k-                kilo-, one thousand

KU                KU is a thousand Units.  In market statistics, unit volume shipments are often stated in thousands while revenue is stated in millions of US dollars, so:  150KU   $20.00 ASP   $3M annual market.  Related: $M, ASP

LAN              Local Area Network.  See also: Network

LCD              Liquid Crystal Display

LED               Light-Emitting Diode

LIFO             Last In-First Out is a sequence description where the last data to arrive is the data first to be driven out; a formal description of a typical memory stack.  See also: FIFO

LNA              Low Noise Amplifier

logic             NOT, OR, AND, and XOR (exclusive OR) are the basis of logical operations and, in great combinations and chains, can represent virtually any decision-making process or mathematical function. 

LSI               Large Scale Integration.  [Discussion at “IC”]

LTE               Long Term Evolution, sometimes considered the distinguishing technology of 4G cellular

LUT              A Look-Up Table can be a quick and dirty shortcut to determining the correct output according to an input.  A color look-up table (CLUT) can allow 256 symbols to represent a variety of 65,000 non-uniform colors for a display. 

m-                milli-  one-thousandth.  See also: micro, nano, number

mA               A milliamp is one-thousandth of an ampere, an electrical current measurement.  See also: micro-

MAC    ^*        1) Multiply ACcumulate is the most fundamental of digital signal processing operations.  The MAC multiplies two pieces of data and sums the result to a running total in 1 or 2 clocks. Coupled with a conditional-branch operation that takes no more than 1 additional clock (zero is better), the MAC is ideal for the frequent sum-of-products operation of many DSP algorithms.  Every DSP will have a MAC, but many standard MPUs and MCUs add in a MAC instruction to ease the execution of signal processing functions in those GPPs - it’s a good start, but no substitute for a real DSP.  MAC can refer to either a processor instruction or to the hardware logic that implements the operation.  See also: DSP

MAC    ^*        2) Medium Access Controller

MCU              A microcontroller (MCU) contains a processor, I/O peripherals, and the memory system all on the chip.  An MCU is primarily designed to perform control operations and has quick interrupt capability.  On-chip program memory (mostly ROM and Flash) distinguishes the MCU from an MPU.  While some MPUs have peripherals and cache memory on-chip, an MCU is expected to execute instructions entirely from its on-chip memory (although with 32-bit, this requirement is relaxing).  MCUs run at much lower frequencies and are generally less-powerful than an MPU.  On-chip peripheral circuits may be the most critical part of an MCU, off-loading the processor from being bothered with too many rudimentary I/O operations.  [Discussion at “processor”]  See also: MPU, DSP, processor  [the “U” of MCU stands for unit, rarely used now]

MEMS            Micro-Electro-Mechanical System

MHz              One MegaHertz is one million cycles per second (clock frequency).  See also: frequency

micro- (µ)     One-millionth (10^-6) [the lower-case “u” is often substituted for the more-correct “mu”, µ]  See also: milli, nano, Numbers

micro  *        Sometimes an offhand, shortened version of “microprocessor”.  Twenty years ago, ”Micro-“ prefixed many words such as processor to indicate the item was a semiconductor chip rather than a board-level product made up of many chips.  A microcomputer would be a chip but a minicomputer was a board or box-level computer. 

micron          One-millionth (10^-6) of a meter.  Micron is a common way of measuring feature size (line widths) on semiconductor die.  Around year 1980 most semiconductors used technology larger than 1 micron.  By the year 2000, most advanced semiconductors had shrunk below 0.13 micron and 0.1 micron, ushering in the misnomer “sub-micron” as terminology shifted to sizes called “90 nanometer” and below.  See also: nanometer (nm), Numbers

MIDI             Musical Instrument Digital Interface

mil                One-thousandth of a meter (0.001 m).  The lead spacing, or pitch, on the outside of the chip package is usually described in mils.  These leads align with the spacing of signal traces on the PCB on which the chip is mounted.  1.0 mil spacing of the ’80’s has given way to 0.8 mil and even half-mil center-to-center lead spacing. 

milli-             One-thousandth (10^-3) .  See also: micro-, nano-, Numbers

MIMO            Multiple Input/Multiple Output (antenna technology, often associated with IEEE 802.11n and high-speed 3G cellular)

MIPS    *       1) As a pure acronym, MIPS is short for “Million Instructions Per Second”, a measure of a processor’s performance, often too simplistic for adequate comparisons.  DMIPS refers to measurements made using the Dhrystone benchmarks.  See also: FLOPS, GOPS, MOPS

MIPS    *       2) A processor architecture -  MIPS also refers to the RISC processor cores that execute the MIPS-I, MIPS16, MIPS32, or MIPS64 instruction set which were developed by the company now known as MTI.  Common MIPS processor cores include the MIPS 4K, 5K, 24K, 34K, and 74K.  Originators pretend that MIPS stood for “microprocessor without interlocked pipeline stages”.  [Other popular multi-vendor architectures include: ARM, MIPS, Power/PowerPC, SuperH/SH]

MIPS    *       3) A company - MIPS also refers to the company MIPS Technologies Inc (MTI) which designs, maintains, markets, and ultimately licenses the popular MIPS processor architecture as intellectual property (IP) to customers who build the processor into a semiconductor chip.  Previously the company was MIPS Computer Systems, and once a part of Silicon Graphics (SGI).  (Nasdaq: MIPS) www.mips.com

MMACS, MegaMACS    Million multiply-accumulate operations per second as a measure of a DSP’s potential performance is sometimes too simplistic for adequate comparisons.

MMS             Multimedia Messaging Service transfers images, animation, video as well as text on a cell phone.  See also: SMS

MMU             The Memory Management Unit sits between the processor and the memory and maps virtual memory addresses to physical memory locations while keeping programs from accessing improper sections of memory and I/O.  MMUs must be designed carefully to avoid introducing delay during accesses to memory and I/O.  See also: MPU memory protection.

MOPS            Million Operations Per Second, a measure of a processor’s performance, often too simplistic for adequate comparisons.  “Operations” can be difficult to define consistently between vendors.  This term is often associated with DSPs.  See also: MIPS, GOPS

MOS             Metal-Oxide Semiconductor

MOSFET        Metal-Oxide Semiconductor Field-Effect Transistor

MOST            Media-Oriented Systems Transport - a network for automotive multimedia

MP3              Motion Picture Experts Group Audio Layer 3

MPEG .mpg    Motion Picture Experts Group.  See also: JPEG

MPU *           A microprocessor (MPU) is a processor designed to perform data processing, operating primarily on numeric data.  MPUs usually dedicate the most transistors to instruction execution, resulting in the highest performing processor.  32-bit and 64-bit MPUs are typical, running at frequencies from 100 MHz to over 1 GHz.  Cache memory on-chip is necessary to keep external memory, which is relatively slow, from impairing performance.  An MPU is distinguished by its focus on data processing and its lack of significant memory on-chip.  An MCU is usually self-contained, distinguished by its focus on control operations, its peripheral support, and execution of instructions primarily out of on-chip memory.  [Discussion at “processor”]  See also: processor, CPU, RISC, CISC, core, MCU, DSP

MPU *           Many ARM-architecture processor cores make use of a Memory Protection Unit, which unfortunately abbreviates to the same letters normally used for a microprocessor.  This confusion could have easily been avoided, but then the same boys in Cambridge also thought referring to their ARM Architecture Reference Manual document as the ARM ARM (sic) to be too cute to pass up.  Memory protection in this sense is the first stage of what some Memory Management Units (MMU) perform.  It allows special handling of apparently-improper access to memory such as when less-trusted User programs try to access memory assigned to operating systems (or “Supervisors”), attempts to write data to sections of memory intended to be read-only, or erasure of certain memory areas.  See also: MMU. 

MPW             The Multi-Project Wafer has become practical as high-density process technology met 300mm wafers.  While most semiconductor manufacturing replicates the pattern for just one chip across the entire silicon wafer, in fact a set of wafers, MPW allows patterns for a number of chips onto a single wafer.  Sharing a wafer this way reduces the quantity of die of any one chip pattern that are produced so test chips and low quantity runs are not overwhelmed with potentially un-sellable chips.  Realize that a 300mm (approx 12” diameter) wafer might produce 2,500 die that are 5mm on a side. 

ms                A millisecond is one-thousandth of a second

MultiCore       Multiple processing cores on a chip, usually homogeneous cores

Multiprocessor         Multiprocessor refers to more than one processor; sometimes homogeneous such as a pair of MIPS 4K processors, other times heterogeneous such as an ARM7TDMI with a DSP.  Multiprocessor can refer to processors on individual chips or, in growing use starting about 2005, the processors all being on the same chip, in many cases to increase overall performance while not burning the chip up with internally-generated heat.  Multiprocessing often refers to the logistics and management of computing by using multiple processors, usually homogeneous, and has been around for a long time.  See also: cores, AMP, SMP

MUX              A multiplexer selects which of many input signals to use while a demultiplexer (DEMUX) selects the output line on which to put an individual signal. 

mV               A millivolt is one-thousandth of a volt, an electrical voltage measurement.  See also: micro-, numbers

mW              A milliwatt is one-thousandth of a watt, an electrical power measurement.  See also: micro-, numbers

n-                nano-   one-billionth (10^-9) .  See also: milli-, micro-, numbers

NAN              Not A Number.  Sometimes, Neighborhood Area Network (see: Network)

NFC              Near Field Communication

NRE              A Non-Recurring Engineering charge is a fee charged to cover substantial up-front costs associated with starting a project with a vendor.  NRE may cover equipment, labor, mask charges, or other one-off expenses, and are agreed upon ahead of time.  NRE is in contrast to piece-prices that primarily cover the manufacturing costs of individual parts.  See also discussion at: OTP

NTSC            National Television System Committee, sometimes jokingly referred to as Never Twice the Same Color, describes the 525-line standard by which the United States and some other countries base (the old analog) color television.  Other major color TV standards include PAL (Phase Alternating Line - used in Australia and much of Europe) and SECAM ((long French phrase) - used in France and former Soviet countries).  NTSC radio transmissions in the USA will be turned off on February 17, 2009 forcing broadcasters and consumers to digital TV on the ATSC (Advanced Television Systems Committee) standard. 

numbers        Note: Special attention is given to numbers at the bottom of this glossary

NVRAM          Non-Volatile RAM

ODM             Original Design Manufacturer.  See also: OEM

OEM             The Original Equipment Manufacturer is the company that makes and markets the end-equipment; Ford may be the OEM of a car, though Bosch may make some of the subassemblies that go into the car, and a Taiwanese company designed the circuits (the ODM), and an Israeli company wrote the software that included a communications stack licensed from a British company.  See also: ODM

OFDMA          Orthogonal Frequency Division Multiple Access, often associated with wireless technology.

order of magnitude    10x.  Mathematically, each order of magnitude (size) is ten times.  Thus, 10,000 is an order of magnitude larger than 1,000.  100,000 is two orders of magnitude larger than 1,000.  Order of magnitude is usually used to “wave a hand” at the actual size difference to give a general idea.  For instance, to agree that “a horse is an order of magnitude bigger than a dog” doesn’t require a set of scales.  It is uncommon to use fractions in mention of orders of magnitude, so one does not often hear “two and a half orders of magnitude,” partly because precision is not trying to be conveyed.  Can be hyphenated for clarity

OSGi             Open Services Gateway Initiative

OTP              One Time Programmable Read-Only Memory is non-volatile memory designed to be programmed only once; essentially an EPROM packaged with no window, preventing ultraviolet light from ever erasing it.  Packaging EPROM die so it can be erased noticeably raises costs of the memory or microcontroller.  OTP solved the cost problem and in turn opened a new behavior in the MCU market.  While OTP was more vision than invention, exploiting the extremely simple concept turned the microcontroller industry on its ear and put the company Microchip Technology (MCHP) on the map.  The implication to MCUs of various memory technologies is the backbone of MCU evolution.

Mask-ROM-based MCUs are the least expensive MCUs because mask-ROM has very small die implications and because vendors require very high shipping volumes before they’ll accept an order for a mask MCU.  High volume means low cost.  Offsetting the low unit cost is a mask charge (~$100,000) that is assessed each time the pattern (the program) is changed to correct errors or add features.  The customer would amortize this non-recurring engineering (NRE) charge over the total production run of the MCU.  The ideal situation is to have a 100% solid program committed to a mask-ROM-based MCU and run millions of the chips. 

In the 1990’s OTP MCUs took over as production MCUs for all but the highest-running end-equipment.  Since 2000, Flash-memory-based MCUs have replaced OTP MCUs in all but the most cost-sensitive and most stable applications. 

PCB

P 2P               Peer-to-Peer communication goes from endpoint to endpoint with no superior/inferior, server/client, master/slave relationship nor expectation of a hierarchy of intervening mediators.

PA                Power Amplifier

PAN              Bluetooth is an example of a wireless Personal Area Network designed to provide digital connectivity between devices like mobile PCs, printers, PDAs, cellular phones, and cameras over a range of about 3-10 yards.  See also:  Network

PCB              The familiar green multilayer fiberglass Printed Circuit Board is a mainstay of the electronics industry, usurped only by the integrated circuit that replicates part of the PCB’s function of bringing together a myriad of circuits onto a piece of silicon rather than the PCB.  See also: SBC, IC, SoC, die

PCI               Peripheral Component Interconnect parallel bus for PCs

PCM              Pulse Code Modulation

PDA              Personal Digital Assistant

PHY              Physical layer or interface, often to a serial communications cable like CAT-5 for Ethernet

PIN               Personal Identification Numberersonal Digital Assistant

PIM              Generally, a Personal Information Manager is a very simple PDA, often made up of a calendar, address book, and notepad.  These devices were overtaken by the more sophisticated PDA in the early '90's.

pixel             The picture element (pixel) is the smallest element of an image and may have a color and intensity value associated with it.  The resolution of an image depends on the number and quality of pixels that comprise it.  The quality is often referred to as “depth” which may determine which of 64K (16-bit), 16M (24-bit), or 4G (32-bit) colors is assigned to that pixel. 

PLC   *          1) Programmable Logic Controller - a board- or box-level product often used in industrial control

PLC   *          2) Power Line Communications - where the “mains” power line is the physical medium that carries the data

PLD               Programmable Logic Device - a semiconductor category

PLL               Phase-Locked Loop oscillator (clock generator)

PMP              Personal Media Player

PND              Personal Navigation Device - a handheld GPS-based system

POE              Power Over Ethernet

POP3             Post Office Protocol-3

POS *           1) The Point of Sale is the location at the retailer where the customer exchanges money for goods or services and often the terminating end of an electronic system to track and inventory goods and services.  

POS *           2) Packet-over-SONET

POS *           3) Piece of Stink – loving description of a beat-up, ugly, barely-running old car - or other system

POTS            Plain Old Telephone Service - think: 1970 telephony; circuit-switched telephony

PowerPC, Power       A popular processor architecture managed by IBM and Freescale, with limited availability in IP form.  PowerPC was a commercial version of the more robust POWER architecture IBM had established internally, and was developed and nurtured by a joint venture of IBM Microelectronics, Motorola Semiconductor Products, and Apple Computer.  In 2004, much of IBM’s moderate performance, merchant Power-based MPU business was sold to AMCC, which has further developed the product line.  Around 2006, after Apple abandoned the architecture in its Macintosh computers, the “PC” portion of the name was suppressed.  Advancements of the original IBM-only POWER processors are impressive and found in IBM’s most-powerful servers.  [Other popular multi-vendor architectures include: ARM, MIPS, Power/PowerPC, SuperH/SH]

PPP               Point-to-Point Protocol for networking

PRML            Partial Response, Maximum Likelihood 

Processor      A digital processor executes a sequence of instructions (the program) to perform arithmetic and logical operations on data.  The ability to branch to different parts of the program based on test conditions of the results gives particular intelligence to processors.  A synchronizing clock determines how fast instructions are executed.  The bit-size (or word width) of the instructions determine the variety and intricacy of the operations and the bit-size of the data determines the resolution or range that the data can represent.  Data might represent numeric information (typical for an MPU), I/O (input and output) representing switch, sensor, or control information (typical for an MCU), signal information like audio, video, or vibration (typical in a DSP), and other information.  The infinite programmability of the digital processor has made it an invaluable component of all modern electronics.  This flexibility lets vendors to supply one processor to hundreds of customers.  Vendors can then spread the cost of the design (which is very high) and manufacturing of the processor across the cumulative volumes of all these customers.  Numerous varieties of processors are available including MPU, MCU, DSP, graphical, network, vector, and core processors.  The Arithmetic and Logic Unit (ALU) is the most basic component of a processor.  See also: MPU, MCU, DSP, ALU, core

PTP               The IEEE 1588 Precision Time Protocol is a mechanism for permitting timing and synchronization of events within a microsecond across the normally-non-deterministic Ethernet.

PV                Photovoltaic electricity generation usually involves conversion of photon energy (light) directly to electricity, excluding schemes of heating water to drive turbines

PWM             Pulse Width Modulation/Modulator

QED              The corporation Quantum Effects Design, later Quantum Effects Devices, developed designs and chips using the MIPS processor with favorite markets of printers and networking.  Acquired by PMC-Sierra in 2000.

RAM              Random Access Memory is a fundamental memory type.  See also: Memory

RCH              An extremely small distance barely visible to the naked eye; used in phrases such as “we’re only off by an RCH.”  Popular term among mechanics and mechanical engineers even though few have ever seen the mythical RCH and no metric or Imperial standard has ever been established for the measure.

Read-Modify-Write   Processors need the ability to read a memory location, modify the contents, and write updated contents back to the location without delay or interruption caused especially from other bus-masters.  This allows proper management of semaphores that are used for synchronization with the system and assuring only one master has access to a shared resource.  This read-modify-write sequence must be atomic, that is, uninterruptible.

real-time    A simple description of real-time is imperceptible delay. 

Real-time operation is “fast enough that no delay is noticed.”  This is critical in most situations.  On a factory floor, stopping the drill must be done in real-time, before it cuts the material deeper than the set depth.  However, reporting operating hours or even motor temperature to central maintenance is not time-critical.  In an MP3 player it might not matter whether the audio program starts immediately when the play button is pushed, it will be unacceptable if the audio stutters or skips during playback.  The audio must be processed “in real time”.  If hitting a drum requires so much processing time that the drum sound during playback is delayed in relation to the rest of the instruments – or the original – then the processing is not “real-time” – and unacceptable.  The following typically require real-time operation:
    Audio processing
    Video processing
    Machine control
    Motor control
    Automotive safety – ABS, airbag, traction / stability control

Real-time is nice but not always required in some applications:
    Network traffic (data can be re-sent)
    Voice response (people take time to think)
    User interface (humans are relatively slow)

A processor-based system might not be capable of real-time operation due to the slowness of many activities:  interrupt response, instruction length, memory speed, register save and restore, processor clock frequency, loop or branch efficiency, instruction set effectiveness, operating system efficiency, algorithm complexity, and compiler effectiveness.  The accumulation of all of these activities can prevent a system from attaining real-time operation. 

The PC does not operate in real time there are delays after hitting “enter” and the application reacting.  The Internet is not real-time – it may take seconds for a Web page to download.  Even the fluorescent light bulbs are not real-time – they may take a minute to come up to full brightness.  Waiting for your printer is not real-time.

RTC              Real Time Clock or Time of Day clock (TOD) is a timer that keeps track of the time and date, usually in a form that is easily read or displayed, such as mm/dd/yyyy hh:mm:ss ap.  It is important that such clocks keep time very accurately so that the time is still correct after a number of years.  Most RTCs run from a 32768 Hz “watch” crystal which is true within 20 ppm (parts per million beats). 

The original IBM PC used a real-time clock which was an independent circuit and it was kept alive by the “CMOS battery” that had to be changed every few years.  The circuit became integrated into many other chips.  The power an RTC consumes is the subject of much attention because it may be the only circuit kept alive during “power off” of most electronic systems, and so can be a main source of battery drain.

RF                Radio Frequency

RFID             Radio Frequency Identification

RISC             Reduced Instruction Set Computer - instructions that can execute in a single cycle with functionality limited to a 1) math or logical operation, 2) address pointer manipulation, 3) memory access (two cycle execution) or 4) conditional branch.  RISC processors allow much higher clocking to be attained by keeping the operations simple compared to CISC processors and came into popularity in the mid-1980’s.  ARM, MIPS, SPARC, and PowerPC are well-known RISC architectures.  See also:  CISC

RNC              Radio Network Control

RoHS             Restriction of Hazardous Substances

ROM              Read-Only Memory is a fundamental memory type.  Once programmed, the data in ROM can be read out reliably time-after-time.  ROM is used to hold programs (sequences of instructions) for microprocessors and microcontrollers and because it is unchangeable (for the most part), those programs are referred to as “firmware” as opposed to “software” (which implies readily-changed.)  Sometimes the term ROM is used where the term “program” would be more correct, often in the sense of ROM (which holds the program) and RAM (which holds the variable data). 

ROM might also contain tables of fixed data which might be quicker for the processor to parse than executing a series of mathematical operations to arrive at the same result.  Look-up tables (LUT, mentioned elsewhere) might also translate the color green to the computer value 33FF00h for use on a Web page, or the letter “A” pushed on a keyboard that comes in as 41h to a series of dots in multiple rows to form the character A on a screen. 

“True” ROM is programmed at the vendor’s factory in a metal mask layer, and thus is called “mask ROM.”  Mask ROM is the most-dense form of non-volatile memory.  The special mask layer is costly to make and takes time to run through the factory, and the resulting product can usually only be sold to one customer.  The customer is usually charged an NRE (non-recurring engineering) charge for each mask, and may have to wait 4 weeks to see the final parts. 

The patterns in mask ROM can not be changed, and thus mistakes in the data are costly, requiring “work arounds” that use programming tricks to accommodate the errors or completely throwing away the chip and programming another (with NRE charges, etc).  Over time, other forms of memory solved some ROM difficulties.  See: OTP, Memory.

RSA              A public-key cryptographic system, the letters come from the last names of Ron Rivest, Adi Shamir, and Leonard Adleman

RTC              A Real-Time Clock is a digital time-of-day and calendar circuit.  Some people tried to paralyze the IT industry in the year 2000 blaming improper readings from RTCs.  The sky was not falling, but FUD did sell a lot of computers and “research”. 

RTOS            Most embedded systems rely on Real-Time Operating Systems to respond quickly to handle sporadic events that occur in the equipment. that require attention.  Since an operating system (OS) allocates hardware and software resources and manages the starting and completion of numerous software tasks, very fast interrupt response with a lean switch overhead is critical for initiating the service routines to handle real-world events.  (See Real-Time)

Real-Time      A simple description of real-time is “imperceptible delay.” 

Real-time operation is “fast enough that no delay is noticed.”  This is critical in most situations.  On a factory floor, stopping the drill must be done in real-time, before it cuts the material deeper than the set depth.  However, reporting operating hours or even motor temperature to central maintenance is not time-critical.  In an MP3 player it might not matter whether the audio program starts immediately when the play button is pushed, it will be unacceptable if the audio stutters or skips during playback.  The audio must be processed “in real time”.  If hitting a drum requires so much processing time that the drum sound during playback is delayed in relation to the rest of the instruments – or the original – then the processing is not “real-time” – and unacceptable.  The following typically require real-time operation:
    Audio processing
    Video processing
    Machine control
    Motor control
    Automotive safety – ABS, airbag, traction / stability control

Real-time is nice but not always required in some applications:
    Network traffic (data can be re-sent)
    Voice response (people take time to think)
    User interface (humans are relatively slow)

A processor-based system might not be capable of real-time operation due to the slowness of many activities:  interrupt response, instruction length, memory speed, register save and restore, processor clock frequency, loop or branch efficiency, instruction set effectiveness, operating system efficiency, algorithm complexity, and compiler effectiveness.  The accumulation of all of these activities can prevent a system from attaining real-time operation. 

The PC does not operate in real time there are delays after hitting “enter” and the application reacting.  The Internet is not real-time – it may take seconds for a Web page to download.  Even the fluorescent light bulbs are not real-time – they may take a minute to come up to full brightness.  Waiting for your printer is not real-time.

RF                Radio Frequency “radio”

RFID             Radio Frequency Identification

RISC             Reduced Instruction Set Computer - instructions that can execute in a single cycle with functionality limited to a 1) math or logical operation, 2) address pointer manipulation, 3) memory access (two cycle execution) or 4) conditional branch.  RISC processors allow much higher clocking to be attained by keeping the operations simple compared to CISC processors and came into popularity in the mid-1980’s.  ARM, MIPS, SPARC, and PowerPC are well-known RISC architectures.  See also:  CISC

RKE              Remote Keyless Entry allows the driver to lock or unlock the car doors without inserting a key into the lock.

RNC              Radio Network Control

RoHS            Restriction of Hazardous Substances

ROM              Read-Only Memory is a fundamental memory type.  Once programmed, the data in ROM can be read out reliably time-after-time.  ROM is used to hold programs (sequences of instructions) for microprocessors and microcontrollers and because it is unchangeable (for the most part), those programs are referred to as “firmware” as opposed to “software” (which implies readily-changed.)  Sometimes the term ROM is used where the term “program” would be more correct, often in the sense of ROM (which holds the program) and RAM (which holds the variable data). 

ROM might also contain tables of fixed data which might be quicker for the processor to parse than executing a series of mathematical operations to arrive at the same result.  Look-up tables (LUT, mentioned elsewhere) might also translate the color green to the computer value 33FF00h for use on a Web page, or the letter “A” pushed on a keyboard that comes in as 41h to a series of dots in multiple rows to form the character A on a screen. 

“True” ROM is programmed at the vendor’s factory in a metal mask layer, and thus is called “mask ROM.”  Mask ROM is the most-dense form of non-volatile memory.  The special mask layer is costly to make and takes time to run through the factory, and the resulting product can usually only be sold to one customer.  The customer is usually charged an NRE (non-recurring engineering) charge for each mask, and may have to wait 4 weeks to see the final parts. 

The patterns in mask ROM can not be changed, and thus mistakes in the data are costly, requiring “work arounds” that use programming tricks to accommodate the errors or completely throwing away the chip and programming another (with NRE charges, etc).  Over time, other forms of memory solved some ROM difficulties.  See: OTP, Memory.

RSA              A public-key cryptographic system, the letters come from the last names of Ron Rivest, Adi Shamir, and Leonard Adleman

RTC              Real Time Clock or Time of Day clock (TOD) is a timer that keeps track of the time and date, usually in a form that is easily read or displayed, such as mm/dd/yyyy hh:mm:ss a/p.  It is important that such clocks keep time very accurately so that the time is still correct after a number of years.  Most RTCs run from a 32768 Hz “watch” crystal which is true within 20 ppm (parts per million) beats. 

The original IBM PC used a real-time clock which was an independent circuit and it was kept alive by the “CMOS battery” that had to be changed every few years.  The circuit became integrated into many other chips.  The power an RTC consumes is the subject of much attention because it may be the only circuit kept alive during “power off” of most electronic systems, and so can be a main source of battery drain. 

Some people tried to paralyze the IT industry in the year 2000 blaming improper readings from RTCs.  The sky was not falling, but FUD did sell a lot of computers and “research”. 

RTOS            Most embedded systems rely on Real-Time Operating Systems to respond quickly to handle sporadic events that occur in the equipment. that require attention.  Since an operating system (OS) allocates hardware and software resources and manages the starting and completion of numerous software tasks, very fast interrupt response with a lean switch overhead is critical for initiating the service routines to handle real-world events.  (See Real-Time)

SAM             The Served Available Market (sometimes: Addressable) is the market for which the product / service is intended to be sold, and is less than the TAM.  If the product is an automobile, the TAM might be all passenger vehicles, but the SAM might be cars, excluding light trucks.  However, when considering the TAM, this particular car might draw a few people out of their pickup trucks.  Alternatively, proportionately, truck sales may increase, squeezing the car SAM smaller even though vehicle TAM is constant.  SAM must be realistically specified to be useful.  “Black 2-door Studebakers” is a market that other vendors can not serve, but “sports cars” may be the right SAM for a Mustang.  Pay attention to exactly what SAM is being referenced when market share is discussed.  (see also TAM, SOM, share)

SBC              The Single-Board Computer is a board-level computer that today might be called a “blade” and would include a semiconductor processor on the board and likely the memory as well.  Some standard board configurations include VME or PC104 which entail physical, electrical, and power specifications.  See PCB

SDRAM          Synchronous Dynamic RAM; spoken: ess-dee’-ram

semiconductor     Semiconductors are materials based on chemical elements that are neither good electrical conductors (like gold and silver) nor good electrical insulators.  Silicon (Si) is by far the most useful basis of semiconductors, although in the early days Germanium (Ge) was used, and for some very high-speed circuits today use a molecular combination of Gallium (Ga) and Arsenic (As) – GaAs. 

SERDES         The 10-bit Serialize and Deserializer physical-medium attachment sub-layer for Gigabit Ethernet 

share            The Market Share (or SOM) of a product is the percentage of the market which purchases or uses the product / service.  Competitors serve the balance of that market and market share is one important measure of a product/service success and growth.  Suppliers fight vigorously to gain market share because amortizing costs across the greatest volume usually gives the lowest per-item cost.  Describing the market properly is important and should be agreed among industry specialists.   See also: TAM, SAM, SOM

SH, SuperH    The SuperH processor architecture (sometimes just SH) family includes MCUs (SH-1 and SH-2) and MPUs. Most of Renesas’ (Hitachi) 32-bit MCUs are based on the SuperH family. The original SuperH was designed at Hitachi. Early in 2001, SuperH Inc. was organized to license the architecture, but within five years and after a cooperative venture with STMicro yielded little commercial benefit, the licensing was largely abandoned.  [Other popular multi-vendor architectures include: ARM, MIPS, Power/PowerPC, SuperH/SH]

SHA-1           Secure Hash Algorithm is an encryption technique.   Others: DES, AES

SIM              Subscriber Identity Module

SIMD            Single Instruction, Multiple Data refers to one instruction that causes the same operation to be effected on numerous pieces of data in parallel, for instance adding eight pairs of data to eight individual totals.  Can be useful for media processing, like video.

SiP               System-in-a-Package (multi-chip, stacked…) usually refers to a number of chips connected together in a package (not all need to be semiconductors.  A good example is a processor chip connected to a memory chip or two all in one package, allowing for quite a mix of memory configurations without the difficulty of laying out a chip for each.

SMP              Symmetric MultiProcessing.  See also: AMP (asymmetric), Heterogeneous, and discussion at Homogeneous

SMS             The Short Messaging Service allows text messages up to 160 characters to be transmitted between cell phones.  Originally a low bandwidth channel expected to be used for system maintenance, when made available to users it quickly became the most successful revenue generator for cellular carriers after voice calls. 

SNMP            Simple Network Management Protocol

SoC              A System-on-a-Chip is an easy concept and seemingly self-explanatory, but in reality difficult to pin down indisputably.  The SoC is the most highly integrated circuit today, typically with a variety of functions blended together on a single piece of silicon.  The chip should be comprised of all of the major functions required to operate the system, including the digital processor.  The entire semiconductor memory system is not expected to be on the SoC, mostly because it could be huge – four or five chips in itself.  An SoC should contain at least a quarter million transistors.  

The difficulty with defining an SOC is pinning it down.  Is a processor required or can the chip simply contain 250,000 gates of logic?  What if a massive function is implemented but it takes four similar-sized SoCs to build the entire system?  Does it matter if the SoC requires a few support chips like A-to-D converters or PHYs?  What if the system is relatively small – something that a microcontroller can operate, like an automobile climate control?  Isn’t a microcontroller the original SoC?  Is a highly-integrated embedded processor or 32-bit microcontroller an SoC?  Do all ASICs and ASSPs fit the classification of SoC?  Is a classification of SoC mutually exclusive of MPUs, ASICs, and ASSPs, or an independent label?

Does it matter?  Generally some of these issues are simply areas for discussion, but when companies and marketers declare “the market for SoCs is $150B,” then it is important to understand what is included in the category. 

When spoken the letters are spelled out, with the plural form adding an “s” to the end, thus spoken: ess-oh-seas.   See also: IC

SOHO           Small Office-Home Office

SOI              Silicon on Insulator

SOM             Share Of Market or market share.  See share, SAM, TAM.

SONET          Synchronous Optical Network

SPARC          The RISC processor architecture that executes the instruction set which was originally developed by Sun Microsystems, Inc.  Common SPARC processor cores include UltraSPARC and SPARClite.  SPARC is short for Scalable Processor Architecture.  “Sun” was originally inspired by the alma mater of its founders, Stanford University.  MIPS and Sun’s SPARC were the two architectures that exemplified the commercial push for RISC processors in the mid 1980’s.  MIPS was licensed widely but as the basis of Sun’s products, a lot of software was created for SPARC   [Other popular multi-vendor architectures include: ARM, MIPS, Power/PowerPC, SuperH/SH]

SRAM            Static RAM.  See also: Memory

SSID             Service Set Identifier

STB              A Set-Top Box converts digitally-encoded streams into an analog or different digital signal that is suitable as the input to television sets.  STBs often presents the program guide to the user for channel selection and, in more advanced systems, arranges recording for DVRs.  STBs are often categorized into Analog, Digital, Terrestrial, Cable, Satellite incoming media.

System         A collection of interacting, interdependent parts.  Any of those parts might be referred to as a subsystem or component.  See also: Black Box, Block Diagram

T1, T3          1.544 and 44.736 Mbps data transmissions for North America (see also E1)

TCP/IP Transmission Control Protocol of the Internet Protocol

TAM             The Total Available Market includes the entire market that can be reasonably served by the product.  The TAM for automobiles will normally exclude light trucks and especially motorcycles, even though all might be found in a residential garage.  (see also SAM, SOM, share)

TCP/IP          Transmission Control Protocol of the Internet Protocol

TCO              The Total Cost of Ownership of a product encompasses not only the cost to buy the equipment and its attentive software, but also the costs of installation, training, maintenance, upgrades, and repairs - all over the lifetime of the product.  The cost of consumables such as printer ink, paper, and electricity should also be included.  Consumables might be the largest part of some products from razors to printers.  Service costs like monthly service charges must be included in TCO as well.  Cell phones might be given away because the service provider is underwriting the cost of the handset on the expectation of collecting 24 months of service charges.  The first piece of equipment that requires metric tools when all other equipment makes use of English tools will require purchase of a metric tool set – and should be considered as part of the TCO. 

TDMA            Time Division Multiple Access, a cellular technology

temperature range    Semiconductors are usually tested to operate within a selected temperature range (room temperature is around 25°C):
   Commercial              0°C to +  70°C
   Industrial              -40°C to +  85°C
                               -40°C to +105°C
   Automotive           -40°C to +125°C
   Military                 -55°C to +125°C
At lower temperatures, semiconductors run faster but consume more power.  There is a science and a little art to testing the chip at the extreme temperature points and specifying (“spec’ing”) electrical parameters over the full range.  The problem is that high performance is tied to the fast transistor switching that occurs at low temperature, but a 20% gain in performance may require doubling the power consumption – usually a poor trade-off.  A chip operating at room temperature may run 750 MHz and only draw 3 watts even though the part is spec’d for worst-case which is 1000 MHz and 10 watts at -40°C.

Usually, in a sequence known as “binning” (putting into bins), semiconductors are tested first at the widest temperature range (say, automotive) and chips that pass are categorized and sold at that temperature range.  Next the parts that failed the widest range are tested at the next-narrower (less stringent) temperature range (say, industrial) and chips passing that range are similarly categorized and ultimately sold at a lower price.  After repeating this process enough times any parts that fail to pass the 0-to-70°C range are considered un-usable and either disposed of properly or sent for failure analysis (FA) to determine why. 

Interspersed with temperature tests are speed tests for 1.5 GHz, 1.2 GHz, 1.0 GHz or whatever maximum frequencies the chip has been specified to run.  Market demand may determine the sequence of this testing and whether an industrial 1.2 GHz part is more desirable than a commercial grade 1.5 GHz part.  

thou             This means you.  Ignor it at your own peril, especially when followed by “shalt not . . .”  Similar terms in Olde English and King James’ are thy, thine, and thee, with ye being similar to the modern day y’all, commonly heard in the South and Southwestern United States.  Distinctions between these are beyond the scope of this text.  More help may be available at http://av1611.com/kjbp/articles/bacon-theethou.html

Thumb          The 32-bit ARM instruction set has a subset defined as Thumb that compacts frequently-used instructions into a 16-bit format to conserve memory space and reduce memory accesses.  A Thumb-enabled ARM processor must be switched into the Thumb mode to execute Thumb instructions.  The “T” in the popular ARM7TDMI processor indicates Thumb ability.  Availability of compact Thumb instructions is one of the reasons ARM was chosen to run early digital cellular phones, and ARM continues to dominate that market space.

Thumb2 instructions are a much newer form of compact ARM instructions.  No mode switch is necessary to run Thumb2 instructions so they can be interspersed with normal ARM instructions, and they execute faster than the prior Thumb instructions. 

Other instruction set architectures like MIPS (Tiny MIPS), PowerPC, and SuperH also have compact instruction capability though it may be implemented differently.

TLB              Translation Lookaside Buffer, often associated with memory management

TTL              Transistor-Transistor Logic technology essentially made integrated circuits practical and is the basis of most electronics today.  See also CMOS.

U                 The “U” in a many processor circuit acronyms (ALU, CPU, MPU, MCU, FPU...) stands for unit, a remnant of the days when many individual components on a board, and later when a single chip, comprised this function.  In speech, the “unit” is rarely spoken now. 

UMTS            Universal Mobile Telecommunications Service, a cellular technology

UPCS            Unlicensed Personal Communications Services

URL              Uniform Resource Locater (was Universal Resource Locater)     http://

us                A microsecond is one-millionth of a second (for convenience, the lower-case “u” is often substituted for the more-correct “mu”, µ, for micro-)

USART          Universal Synchronous Asynchronous Receiver/Transmitter (UART is asynchronous only)

USB              Universal Serial Bus.  High speed serial data transfers are possible over this truly universal standard physical and protocol description.  The ubiquitous cable allows increasingly faster speeds for ever-greater types of equipment.  Many devices even recharge their batteries or completely rely on the power (900ma) available down the cable.  The larger “A” end defines the host, hubs can replicate and extend the fan-out of the hierarchical bus, and peripherals are the downstream end-point devices or “functions”.  An “On-the-go” OTG device might act as a host in one circumstance and as a peripheral in another.  For instance a camera might act as a host when printing pictures directly to a printer, but act as a peripheral when uploading pictures to a host PC; all using the same physical port. 

All are backward compatible. 

UWB             Ultra Wide-Band enhanced Bluetooth

VGA              One of many monitor display standards often used to describe a screen size of 640x480.  Other sizes (most with 4:3 aspect ratio):
    320 x  240  QVGA  Quarter
    352 x  288  CIF   Common Intermediate Format (1.22..:1) 
    640 x  480  VGA   Video Graphics Array
    800 x  600  SVGA  Super
   1280 x 1024  SXGA  Super Extended  (1.25:1 aspect ratio) 

VLIW            Very Long Instruction Word

VLSI             Very Large-Scale Integration  [Discussion at “IC”]

VLSI Technology    The corporation VLSI Technology was an early pioneer of ASIC technology and embedding processors (like the earliest ARM processor) into ASICs.  It was acquired by Philips (now NXP) in 1999.

VPN              A Virtual Private Network provides a layer of security allowing a client on a public network like the Internet to safely communicate with an enterprise server without fear of data transmissions being compromised. 

VoIP             Voice over Internet Protocol

word             In describing a processor, “word” usually refers to the primary data size the processor is designed to handle, often the same size as the machine’s ALU.  “Double-word” data, which is twice a word’s width, can often also be handled though may make two passes through the ALU to accomplish the operation.  Half-word and smaller data sizes might be accommodated by sign-extending the smaller data to make it full word size or might even be packed so that two half-word data can be operated on in a single pass (see SIMD).  To a 32-bit processor, the word size is considered 32-bits, double-word is 64-bits, and half-word is 16-bits. 

WAN             The cellular phone network is a good example of a Wide Area Network where digital communications is available wirelessly over an area that may be miles.  See also:  Network

WAP             Wireless Application Protocol

W-CDMA        Wideband Code-Division Multiple Access, a cellular technology

WEP             Wired-Equivalent Privacy

Wi-Fi            Wireless Fidelity; Wireless networking based on IEEE 802.11

WiMAX          Worldwide Interoperability for Microwave Access, wireless networking based on IEEE 802.16

WLAN, Wireless LAN  A Wireless Local Area Network like IEEE 802.11a, b, or g provides digital communications in buildings or within about 100 yards.  See also:  Network

WYSIWYG      What You See Is What You Get describes a good graphical user interface where the size, color, and orientation shown on the monitor are the same as on a printed output.  Pronounced: wiz’-zee-wig

X                 In some logic and engineering uses, X is a shorthand for “trans” as in transistor (Xistor), transceiver (XCVR), transfer (Xfer).  Also X can indicate a “don’t care” or irrelevant condition, one that has no affect on the state.  In a more general sense, X can simply stand for an unknown or “any number of substitutions”.  The origins of this use is a little vague though may stem from the letter’s similarity to two arrows pointing at each other, since no particular direction is implied by “trans”, though movement is suggested.  As a Roman numeral, X counts as ten.   

Zigbee          A wireless communications specification, IEEE 802.15.4

8051             The 8051 is a well-known 8-bit microcontroller architecture that Intel extended from its earlier 8080 microprocessor.  8051-compatible MCUs are still offered by many other companies.  Also: 80C51; simpler: 8031.

   *  Asterisk indicates that multiple definitions are given for this acronym/term

Networks       LAN, MAN, PAN, WAN, WLAN = x Area Network . . . 
Local, Metropolitan, Personal, Wide-Area, Wireless… Area Network   
Bluetooth, Wi-Fi, WiMAX, UWB (UltraWide Band), Zigbee wireless networks.                                                                          

Numbers        Traditional scientific conventions in notation of numbers and measures are utilized in electronics and semiconductors with some slight variations when dealing with the powers-of-2 inherent in binary counting.  Marketers have been known to revert back to traditional convention when it makes their products appear superior, but this should be discouraged.  Any numbers associated with bytes and address ranges should refer to Base-2 numbers.  Certain official standards organizations alter the prefix slightly for Base-2 to terms with “bi” in them like kibi, mebi, gibi, tebi; although these are not used in common practice.  Clock rates and pixel counts are usually tied to the Base-10 numbers, even if they are often doubled or ultimately dependent upon bytes for storage.  If in doubt, check specs because the 2% differences do multiply up to become nearly 10% at the tera- level.  For additional information on numerical representations see: byte, word, hexadecimal.      Tables follow.                                                  

      1 inch = 2.54 centimeter = 0.0254 meter

             Power of 10                       Power of 2

          Base-10 (decimal)                   Base-2 (binary)

m  milli-  10^-3    0.001

µ  micro-  10^-6    0.000001

n  nano-   10^-9    0.000000001

p  pico-   10^-12   0.000000000001         bits

                                            8   2⁸ =                256

k  kilo-   10³               1,000  kilo-  10   2¹⁰ =             1,024

                                           16   2¹⁶ =            65,536

M  mega-   10⁶           1,000,000  mega-  20   2²⁰ =         1,048,576

                                           24   2²⁴ =        16,777,216

G  giga-   10⁹       1,000,000,000  giga-  30   2³⁰ =     1,073,741,824

                                           32   2³² =     4,294,967,296

T  tera-   10¹²  1,000,000,000,000  tera-  40   2⁴⁰ = 1,099,511,627,776

                             2⁶⁴ = 1.84467441 × 10¹⁹ = (reportedly)

                             2⁶⁴ = 17179869184 gigabytes