Computer-aided engineering

Shared by: mercy2beans119

Tags

computer aided engineering, computer-aided engineering, finite element analysis, engineering software, mechanical engineering, computer aided design, college of engineering, product development, computer-aided design, software engineering, case tools, computer aided, computer aided software engineering, engineering design, engineering technology

Stats

views:: 157
posted:: 1/11/2010
language:: English
pages:: 6

Public Domain

Document Sample

Computer-aided engineering 549

globally distributed manufacturing is possible. Vir- Program structure. A CAE program usually consists
tual corporations can be established involving partic- of a series of mathematical models and a data struc-
ipants from around the world. Design, production, ture. Figure 1 illustrates a simpliﬁed view of a typical
and assembly can occur wherever is most conve- CAE program operating in an engineering worksta-
nient and cost-effective for the enterprise. Develop- tion environment. First, a mathematical description
ments in CAD/CAM, including advances in product of the physical phenomena being analyzed is writ-
data models and international standards, have made ten. This engineering model may consist of equa-
this possible. See COMPUTER PROGRAMMING; DIGITAL tions such as Newton’s second law to describe the
COMPUTER; ENGINEERING DESIGN; INTERNET; WORLD dynamics of a system or the Navier-Stokes equations
WIDE WEB. K. Preston White, Jr.; Larry G. Richards to analyze a ﬂuid ﬂow ﬁeld. Next, a model of the phys-
Bibliography. C. R. Asfahl, Robots and Manufactur- ical conﬁguration is created. This geometric model
ing Automation, 2d ed., 1992; D. D. Bedworth, M. R. may consist of two- or three-dimensional (2D or 3D)
Henderson, and P. M. Wolfe, Computer-Integrated curves, surfaces, faceted approximations to surfaces,
Design and Manufacturing, 1991; T.-C. Chang, R. A. or solid elements. The results of the engineering anal-
Wisk, and H.-P. Wang, Computer Aided Manufac- ysis are frequently displayed on the geometric model
turing, 2d ed., 1998; C. Machover, The CAD/CAM by color fringing to show the variation of a scalar pa-
Handbook, 1996; C. McMahon and J. Browne, rameter. Large amounts of data are created during
CAD/CAM: Principles, Practice and Manufacturing this modeling phase, and the need for a data struc-
Management, 2d ed., 1998; P. K. Wright and D. A. ture to store and retrieve them is greater than for the
Bourne, Manufacturing Intelligence, 1988; I. Zeid, engineering model. See DATA STRUCTURE.
CAD/CAM: Theory and Practice, 1991. Although the engineering and the geometry are
fully described, they cannot be viewed on the dis-
play until a model for rendering has been formu-
lated and coded. A mathematical description of the
Computer-aided engineering lighting conditions, the approximate intensities of
Any use of computer software to solve engineering
problems. With the improvement of graphics dis-
plays, engineering workstations, and graphics stan-
dards, computer-aided engineering (CAE) has come
to mean the computer solution of engineering prob-
lems with the assistance of interactive computer
graphics. See COMPUTER GRAPHICS.
CAE software is used on various types of comput-
ers, such as mainframes and superminis, grid-based
computers, engineering workstations, and even per- analysis
results
sonal computers. The choice of a computer system
is frequently dictated by the computing power re-
quired for the CAE application or the desired level
and speed of graphics interaction. The trend is to- graphical disk
ward more use of engineering workstations. See DIG- input
ITAL COMPUTER; MICROCOMPUTER.
Design engineers use a variety of CAE tools, in-
cluding large, general-purpose commercial programs
engineering workstation
and many specialized programs written in-house or
printer
elsewhere in the industry. Solution of a single engi-
operating system and
neering problem frequently requires the application graphics support software
of several CAE tools. See ENGINEERING DESIGN.
A typical CAE program is made up of a number of
mathematical models encoded by algorithms written
algorithmic description
in a programming language. The natural phenomena
of engineering problem
being analyzed are represented by an engineering plotter
model. The physical conﬁguration is described by engineering model
a geometric model. The results, together with the
geometry, are made visible via a user interface on
the display device and a rendering model (graphics geometric model
image). See ALGORITHM; COMPUTER PROGRAMMING; image copy
MODEL THEORY; PROGRAMMING LANGUAGES. rendering model
CAE allows for many more iterations of the
analysis-design cycle than was possible by hand com-
putation, especially when the CAE is coupled with
optimization systems that drive this cycle automati-
cally. The beneﬁts are translated into improved pro-
ductivity and quality of design. Fig. 1. CAE program in an engineering workstation environment.
550 Computer-aided engineering

light reﬂected by the nodes of the geometric model, The result is a hierarchical graphics data structure
and the corresponding shades of color provide the well suited for animation and representation of en-
graphics data necessary for the model to be real- tities with multiple components. A number of ex-
istically shaded. Before viewing, the graphics data tensions to PHIGS, called PHIGS PLUS (PHIGS+),
are transformed from geometric model coordinates were adopted as a separate international standard in
to a normalized coordinate system. Parameters are 1990. PHIGS+ provides support for most of the ren-
set for an orthographic or perspective projection dering model and some of the geometric model in
from 3D model coordinates to 2D coordinates for a CAE program. PHIGS+ routines address lighting,
ﬁnal transformation to the display screen in de- shading, hidden surface elimination, transparency,
vice coordinates. The CAE applications programmer nonuniform rational B-spline (NURBS) curves and
accomplishes most of these tasks with the help of surfaces, and improved user interaction and control.
precoded algorithms provided by graphics-support PHIGS and PHIGS+ were merged into a single inter-
software. Interactive communication with the graph- national standard, PHIGS, in 1997.
ics image and the geometric and engineering models The emergence of PHIGS+ was in response to dra-
occurs through a user interface written with the as- matically improved computer graphics hardware ca-
sistance of the graphics software or windowing soft- pabilities in the late 1980s and early 1990s, and that
ware provided with the workstation. were rapidly being incorporated into the new breed
In the past, CAE programs have been predom- of CAE workstations. Vendors did not wait for in-
inantly coded in the FORTRAN programming lan- ternational committees to address these new capa-
guage. The present trend is toward the C, C++, and bilities in their standards, but deployed their own
Java programming languages together with the UNIX proposed extensions in the interim. While this made
and Windows operating systems. See OPERATING SYS- these new capabilities available to the CAE software
TEM. developers, it also prevented the CAE software from
Graphics standards. Portability, the ability to move easily being ported from one hardware platform to
programs easily from one computer to another, is another, which is the main purpose of a standard.
important for CAE software. Although it has long Taking advantage of this noncompliance to the
been possible to make a computational program adopted international standard and the slow rate at
code portable by using standard programming lan- which new technologies and capabilities were ad-
guages, this was not previously possible for CAE soft- dressed by these standards, an industry consortium
ware because of the lack of graphics standards. A was established in 1992 to support a new graphics
proposed 3D graphics standard (CORE), introduced standard known as OpenGL. The aim of OpenGL is to
to the American National Standards Institute (ANSI), provide a timely integration of new technologies and
was superseded by the adoption in 1985 of the 2D capabilities as they emerge and to avoid the problem
Graphical Kernel System (GKS) as an international of partial conformance to the standard. GKS, PHIGS,
standard by the International Organization for Stan- and other standards have been plagued by partially
dardization (ISO). conforming deployments, which makes portability
A new 3D device-independent graphics standard, between platforms difﬁcult and defeats a key pur-
the Programmer’s Hierarchical Interactive Graphics pose of standards. The OpenGL Architecture Review
System (PHIGS), was proposed by ANSI in 1985 and Board, which draws its membership from nine lead-
adopted as an international standard by ISO in 1988. ing computer graphics hardware vendors, avoids this
Where possible, the concepts and nomenclature of problem by permitting only the term OpenGL to be
GKS have been used in PHIGS. applied to products that conform completely to the
The principal limitation to CAE software portabil- OpenGL standard. OpenGL is today the most widely
ity has been the wide variety of graphics hardware used and supported 2D and 3D graphics application
and the direct dependence of CAE software on this programming interface (API).
hardware. Both GKS and PHIGS give the program- Hardware. Computer systems ranging from work-
mer device-independent graphics primitives and co- stations to mainframe computers are used for CAE
ordinate systems as well as a set of logical graphics software. For example, the aerospace and automo-
input devices to replace the wide variety of input tive industries make wide use of very large main-
hardware. For example, the pick action (selection of frame computers and large grid-based computer sys-
a graphics entity on the screen) could be physically tems to support computer-intensive CAE software
accomplished by a light pen, a cursor and tablet, such as ﬁnite element analysis and computational
or a mouse. Using graphics standards, the CAE pro- ﬂuid dynamics. These large computational resources
grammer will always specify a logical pick device are connected by high-speed data transmission lines
regardless of the physical device used to achieve the to ﬁle servers holding the model and resulting data.
pick. The precomputation model description and the
PHIGS has several important advantages over GKS postcomputation interrogation of the results, and
for CAE software. It is a full 3D system for view- frequently the computations themselves, are usu-
ing and modeling transformations, and allows both ally performed on ultra-high-end workstations capa-
graphics and nongraphics data to be stored in its ble of managing these vast amounts of data. These
data structure. The data structure can invoke other workstations are typically 64-bit multiprocessor sys-
structures and store transformations as attributes. tems, with 12 to 24 gigabytes (GB) of shared memory
Computer-aided engineering 551

(RAM) and a UNIX operating system, which are con-
nected to the ﬁle servers via one or more 1- or
10-gigabit/second network connections. Ultra-high-
end workstations are necessary when working with
large, complex models and datasets, such as the solid
modeling of a complete automobile or aircraft.
For less complex models and datasets, or for sub-
sets of the larger models and datasets, it is typical to
use high-end engineering workstations. These work-
stations are typically 32-bit single- or dual-processor
systems with up to 4 GB RAM running a Windows
or UNIX operating system, the former being more
common. These engineering workstations distin- Fig. 2. Solid modeling and simulation on a mainframe via a
CAE workstation.
guish themselves from personal computers by hav-
ing more RAM and faster central processing units
(CPU), graphical processing units (GPU), hard disks, Though many small CAE applications programs are
network connections, and internal interconnects written for personal computers, of which most engi-
and data transfer rates. These high-end engineering neering companies have a wide array, most large,
workstations cost two or three times more than a typ- commercial CAE programs do not perform effec-
ical personal computer and are commonly used by tively or at all on personal computers. Engineering
engineers. In contrast, ultra-high-end workstations workstations, on the other hand, have the necessary
typically cost three to ﬁve times more than a high- computer power and local resources, and usually run
end engineering workstation, and consequently are only one CAE program at a time and serve only one
less used. user. As a result of this independence, a demand-
Computer-intensive engineering problems are fre- ing CAE code at one workstation node in a com-
quently distributed across multiple CPUs. These puter network will not affect the response time for
CPUs can be connected in a grid structure where other users. See DISTRIBUTED SYSTEMS (COMPUTERS);
each CPU has its own RAM and where communica- LOCAL-AREA NETWORKS.
tions between these CPUs and their RAM takes place The personal-computer gaming industry has been
over a computer network. Occasionally these grids instrumental to the recent widespread availability
are composed of regular personal computers and en- of affordable, high-quality graphical processing units
gineering workstations that are accessed whenever that are capable of supporting the heavy demands of
they are idle, such as during nighttime hours. The CAE graphics processing. These graphical process-
bottlenecks of such systems are the limited band- ing units contain a number of specialized high-speed,
width between the CPUs and the personnel time re- parallel graphics engines, display list memory, image
quired to conﬁgure and manage these grids. Grid memory, buffer memory, and alpha memory to per-
systems tend to be most successful when the com- mit the real-time generation of color-shaded images
putations can be easily parallelized and when there desired for CAE applications. The graphics engines
is minimal need to communicate and move data be- perform graphics modeling and viewing transforma-
tween the CPUs. Computational problems that do tions, lighting and shading computations, and scan
not fall into this category are best performed on conversion of models for raster-scan display. Display
large multiprocessor, shared-memory systems with list memory retains the rendering model structure.
high-speed, high-bandwidth interconnects between The z-buffer provides fast hardware support for hid-
the CPUs. Examples of such systems include main- den surface elimination, whereas alpha memory al-
frames and the ultra-high-end workstations described lows fast hardware-assisted transparency computa-
earlier. The advantages of these systems include the tions. The ﬁnal rasterized image is stored in one of
ﬂexibility of how data can be accessed simultane- two image memories for fast transmission to the dis-
ously by multiple CPUs without having to be moved play screen. Most graphical processing units support
across a network, and the reduced need for manu- the OpenGL graphics standard by performing its op-
ally managing the distribution of the computational erations directly in hardware as opposed to comput-
resources within the system. See MICROPROCESSOR; ing them in software. See COMPUTER ARCHITECTURE.
MULTIACCESS COMPUTER; MULTIPROCESSING; SUPER- The graphical processing unit used in a typical
COMPUTER. personal computer accounts for approximately 2–
These large centralized computational resources 4% of the personal computer cost. In contrast, in a
are often connected to networks of from 10 to 100 high-end engineering workstation, the graphical pro-
client workstations that act as display terminals with cessing unit typically accounts for 20% of the hard-
local controllers (Fig. 2). When a large number ware cost. Where ultra-high-end workstations are
of CAE users invoke computer-intensive programs, also used for graphics processing and visualization,
a signiﬁcant delay in response usually occurs. For it is increasingly common to use multiple high-end
this reason, as well as initial cost considerations, graphical processing units working in parallel, with
most companies today deploy high-end engineering each unit capable of computing 750 million triangles
workstations on a two- or three-year upgrade cycle. per second or 8 billion pixels per second.
552 Computer-aided engineering

known as ﬁnite-element preprocessing, and the va-
lidity of the solution is largely dependent on the skill
of the modeling engineer. Although very advanced
methods of interactive graphics are used in ﬁnite-
element preprocessing, it is still a very tedious and
expensive operation. The solution of the equations is
highly computer-intensive. Finite-element methods
are applicable for a wide variety of physical phe-
nomena, including mechanical stress and strain, ﬂuid
ﬂow, acoustics, heat transfer, and electrical ﬁelds.
Figure 3 illustrates the solution of a ﬁnite-element
stress analysis problem with a commercial CAE code
running on a mainframe computer. See FINITE ELE-
MENT METHOD.
Computer-aided design and manufacturing. CAD/
CAM systems were created by the aerospace industry
in the early 1960s to assist with the massive design
and documentation tasks associated with producing
airplanes. By the late 1970s, these codes were being
distributed to other industries. CAD/CAM systems
Fig. 3. Finite-element stress analysis using commercial CAE software. have been used primarily for detail design and draft-
ing along with the generation of numerical control
A wide variety of output devices are available for instructions for manufacturing. Gradually, more CAE
CAE images. The most popular are ink-jet and laser functions are being added to CAD/CAM systems. A
printers for color images and line drawings on paper trend toward open architecture with ﬂexible geom-
and Mylar. Others include wall projection and stere- etry interfaces is stimulating the addition of more
ovision display systems, often combined with posi- analysis and manufacturing functions. Modeling with
tional and orientation input from the observer. Many CAD/CAM systems has become fairly sophisticated.
types of graphics input devices are available for work- Most popular commercial systems support 2D and
stations and mainframe displays. Cursor and tablet, 3D wireframe, surface models, and solid models.
mouse, joystick, ball, glove, function key box, and Rendered surface models differ from solid models
dial box are the most widely used. Light pens are in that the latter have full information about the in-
generally in use on older-technology displays. See terior of the object. For solid models, a combination
COMPUTER PERIPHERAL DEVICES. of three types of representation is commonly used:
Commercial CAE software. Most companies use a constructive solid geometry, boundary representa-
combination of a few commercial CAE programs and tion, and sweep representation. Complex systems
a number of smaller, specialized CAE applications require signiﬁcant amounts of processing power, pri-
programs typically written in-house or elsewhere in mary memory space, and disk space. See COMPUTER-
the industry. There are several classes of commer- AIDED DESIGN AND MANUFACTURING.
cial CAE software. The ﬁrst large commercial pro- Integration of CAE software. Frequently, all of these
grams used for CAE were dynamic simulation sys- CAE tools are needed together with specialized CAE
tems, which usually had no graphical output or at applications programs to solve a single engineering
best a printer-plotter output of graphs. They required problem. The integration of these tools or the com-
the engineer to give input in the form of ordinary dif- munication of data between them is a challenging
ferential equations or building-block elements to de- problem. To successfully integrate these programs,
scribe the dynamic behavior of a real system, in many a centralized database is necessary. Data from the
ways similar to the patching (programming) of ana- various CAE tools are processed into and retrieved
log computers. These systems numerically integrate from this database through a database management
the set of coupled differential equations, thereby sim- system. If the CAE software is proprietary and the
ulating the dynamic behavior of the particular sys- data structure not easily accessible, the proprietary
tem. Simulation systems are still widely used, and programming interfaces may be used or, sacriﬁcing
although the graphics output has improved consid- of functionality and interactivity, a data interchange
erably, the description of the input has not changed standard may be used. See DATABASE MANAGEMENT
greatly. See ANALOG COMPUTER; NUMERICAL ANALY- SYSTEM.
SIS; SIMULATION. The Initial Graphics Exchange Speciﬁcation (IGES)
A need to solve detailed problems, such as stress was developed under the leadership of the National
or deformation analysis, for components of systems Bureau of Standards (today known as the National In-
led to the development of the ﬁnite-element method stitute of Science and Technology) and was accepted
or ﬁnite-element analysis. In this method, a complex as a standard by ANSI in 1981 (Y14.26M-1981). The
object is broken down into simpler elements. With goal of IGES is to allow the transfer of geometric
these, a set of equations is formulated which, when data between dissimilar CAD systems. IGES is widely
solved, predict the behavior of the object as modeled available and is generally capable of providing the
by the set of elements. The modeling of the object is translation of a snapshot of the model in one CAD
Computer-aided engineering 553

system into a static model that cannot be edited in Applications. The CAE methods for electrical and
another CAD system. electronics engineering are well developed. The ge-
The standard for exchange of product model data ometry is generally two-dimensional, and the prob-
(STEP) has been adopted as an international standard lems are primarily linear or can be linearized with
by the International Organization for Standardization sufﬁcient accuracy. Chemical engineering makes ex-
(ISO 10303). The aim of STEP is to provide a neutral, tensive use of CAE with process simulation and con-
computer-interpretable representation and descrip- trol software. The ﬁelds of civil, architectural, and
tion of product data throughout the life cycle of a construction engineering have CAE interests simi-
product that is independent from any particular sys- lar to mechanical CAE with emphasis on structures.
tem. It can be used for both data exchange and for Aerospace, mechanical, industrial, and manufactur-
archiving data over time. The latter is an important is- ing engineering all make use of mechanical CAE soft-
sue when the product life cycle exceeds the life span ware together with specialized software.
of the software and hardware that created the prod- An example of CAE is the design of an aircraft
uct data. Given this extremely broad objective, ISO landing-gear mechanism. The ﬁrst step is deﬁnition
10303 is not a single standard but a collection of in- of the problem and creation of a set of performance
terrelated documents that form a multipart standard. speciﬁcations. Next, the conceptual design phase
A number of these documents have been adopted as may be aided by specialized programs to determine
an international standard, while many others are still a size estimate of the landing gear based on spec-
in development. The table lists of some of the docu- iﬁed loads and deﬂections. Commercial CAE pro-
ments that have been adopted as part of ISO 10303. grams are available for kinematic synthesis of mech-
The software industry’s support for adopted ISO anisms based on speciﬁed motion requirements, but
10303 documents has been signiﬁcantly slower than this landing-gear mechanism will be designed with
was the case for PHIGS (graphics standards). In the an in-house program written for the purpose. See
case of STEP, the software vendors have been hesitant LANDING GEAR.
to support the standard, both because of the high up- The next phase is preliminary design. An applica-
front cost associated with complying with the com- tions program is used to analyze the deﬂection and
plexity of the standard, and because of their desire to response of a shock-absorbing, energy-dissipating
provide a comprehensive yet proprietary solution for strut. Dynamic analysis of the guiding mechanism
the customer in which the customer’s data is locked and complete assembly is determined by a commer-
into. Thus far, it is primarily the large mechanical cial code. When the dynamic loads are determined, a
CAD/CAM/CAE vendors serving the automotive in- ﬁnite-element stress analysis of each link of the mech-
dustry that have given into customer pressure and anism is done by using commercial ﬁnite-element-
competition to provide support for STEP (AP 203, method software. Following the stress analysis, some
AP 214). Recently it appears that AP 210 and AP 212 links are changed in size and the dynamic analysis re-
have also started to receive some initial vendor sup- peated to determine new loads. Using the new load-
port. ing, another iteration of the ﬁnite-element-method
software is made to verify that the stresses fall below
the strength limits.
Adopted ISO 10303 documents The next phase is the ﬁnal design. All components
of the assembly are drawn in 2D on a CAD/CAM
Document
number Title system and detailed, giving dimensions, material
speciﬁcations, and other instructions. An assembly
STEP AP 201 Explicit Draughting drawing is created from the components, and mat-
STEP AP 202 Associated Draughting
STEP AP 203 Conﬁguration Controlled Design ing of components are veriﬁed. An alternative ap-
STEP AP 204 Mechanical Design using proach is to create a solid model of each compo-
Boundary Representation
STEP AP 207 Sheet Metal Die Planning &
nent, assemble the solid components, and run an
Design automatic interference-clearance check; 2D drop-
STEP AP 209 Composite and Metallic offs are then automatically made of each compo-
Structural Analysis and
Related Design nent and manually detailed (at the workstation) by
STEP AP 210 Electronic Assembly, a drafter. From the ﬁnal part geometry, instructions
Interconnection and for numerically controlled machine tools are gener-
Packaging Design
STEP AP 212 Electrotechnical Design and ated to produce the part. Some systems may sup-
Installation port tooling design and process planning. Finally, a
STEP AP 214 Core Data For Automotive
Mechanical Design
design release is made to the manufacturing depart-
Processes ment. Jan Helge Bøhn; Arvid Myklebust
STEP AP 215 Ship Arrangement Bibliography. J. D. Foley et al., Introduction
STEP AP 216 Ship Moulded Forms
STEP AP 218 Ship Structures to Computer Graphics, 1993; T. Gaskins, PHIGS
STEP AP 224 Process Planning Using Programming Manual, 1992; M. K. Gillenson,
Machining Features Database, Step-by-Step, 2d ed., 1990; F. R. A.
STEP AP 225 Building Elements Using Shape
Rep Hopgood et al., Introduction to the Graphical Ker-
STEP AP 227 Plant Spatial Conﬁguration nel System (GKS), 2d ed., 1987; M. E. Mortenson,
STEP AP 232 Technical Data Packaging: Core
Info & Exch.
Geometric Modeling, 2d ed., 1997; D. F. Rogers, Pro-
cedural Elements for Computer Graphics, 2d ed.,
554 Computer architecture

1997; D. Shreiner et al., OpenGL Programming processor—has been the core issue in deﬁning the ar-
Guide, 5th ed., 2005; A. Tizzard, An Introduction chitecture of a computer system. For example, how
to Computer-Aided Engineering, 1994. many machine instructions will there be, will the in-
structions be of uniform length or variable in length,
and how many ﬁelds will the machine instruction
contain. See MICROPROCESSOR.
Computer architecture Registers. The typical central processor contains
The art or practice of designing computer systems. several special-purpose registers (storage units), in-
Just as the architecture of a building describes its cluding an instruction register, a program counter or
overall structural concept and principal features, the instruction pointer, a processor status register, per-
architecture of a computer system consists of a de- haps one or more separate index registers, segment
scription of the overall layout and major features of (base) registers, one or more general-purpose regis-
a computer system. In both cases, we are concerned ters to hold integer or ﬁxed-point data (integer reg-
with how the object, whose architecture we are de- isters), and either the same or a distinct set of reg-
scribing, appears to the user. In computer architec- isters to hold ﬂoating-point data. Although integer
ture, the user is not the ﬁnal user of the system, but registers, a single instruction register, and the pro-
the person who writes computer programs to be run gram counter or instruction pointer have tradition-
on the system. Compared to the architecture of build- ally been of the same width (number of bits), there is
ings, which is largely an issue of esthetics, the archi- no law requiring such. Floating-point registers may
tecture of a computer system is principally about or may not be of different width from integer regis-
efﬁciency and economy of operation. There may be ters. See COMPUTER STORAGE TECHNOLOGY.
an esthetic dimension to computer architecture, but Native data types. These are the data types that will
when present it is distinctly lower in importance. be separately and distinctly represented and recog-
At present, computer architecture is the highest nized by the computer hardware. Some computers
level of the closely related ﬁeld of computer design. may natively represent and carry out arithmetic on
The ﬁeld has evolved from being the top level of the decimal integers via an underlying binary represen-
design process used by the early developers of com- tation. Floating-point representation may or may not
puters into a separate area of specialization within be present. See NUMBERING SYSTEMS.
the overall hierarchy of computer designers and en- Word size. The most commonly encountered num-
gineers. Underneath computer architecture are the ber of bits contained in a single word of memory
ﬁelds of computer organization and implementation, (word width) are 32 and 64. See SEMICONDUCTOR
and computer engineering. See COMPUTER; DIGITAL MEMORIES.
COMPUTER. Memory addressing scheme. There are two principal is-
Computer organization and implementation. Closely sues under the topic of memory addressing scheme.
related to the ﬁeld of computer architecture is a The ﬁrst is the smallest unit of addressable mem-
subsidiary discipline known as computer organiza- ory, which is usually the byte, although there are
tion and implementation. It is difﬁcult to deﬁne pre- several machines of major signiﬁcance that are not
cisely where computer architecture stops and com- byte-addressable but word-addressable. The second
puter organization and implementation begins. In issue is the order in which the bytes are stored for
general, computer architecture is concerned with the several bytes that constitute a single word. In the
the principal design parameters, while computer or- scheme known as big-endian, the leftmost byte (that
ganization and implementation is concerned with is, the byte with the numerically smallest address)
the subsidiary points of design of the system, includ- contains the most-signiﬁcant bits of the word con-
ing the composition of the logic circuits that make tent, while the byte containing the bits of least sig-
up each structural component of the system and the niﬁcance possesses the numerically highest address.
implementation of each of those logic circuits as an The opposite arrangement is known as little-endian.
arrangement of speciﬁc electrical circuits of some Maximum size of memory. A major issue for the mod-
particular type or types. Computer organization fo- ern computer architect is how large the memory ad-
cuses on issues such as whether an adder circuit dress space should be, since memory has been get-
will be implemented as a ripple-carry adder or as a ting cheaper at a prodigious rate. For every doubling
carry-lookahead adder and whether the under- of size of the memory address space, the width of an
lying electrical circuits will be implemented in address increases by one bit. Therefore, the designer
TTL (transistor-transistor logic), bipolar transistors, is not free to increase the memory address space
CMOS (complementary metal-oxide semiconduc- by an arbitrary amount, as every increase in address
tors), or gallium arsenide (GaAs). See INTEGRATED width has much wider implications for instruction
CIRCUITS; LOGIC CIRCUITS. set architecture, as well as imposing requirements
Major issues. There are several major issues in the affecting lower levels of computer organization
design of a computer system covering all or nearly all and implementation. On the other hand, insufﬁ-
of the architectural considerations. See MICROCOM- ciency of memory address space has been one of the
PUTER. principal factors causing an architecture to become
Instruction-set architecture. Traditionally, the instruc- outmoded.
tion-set architecture—the detailed design of the set Number of buses and their relationship. A modern com-
of machine instructions implemented in the puter typically has a frontside bus and a backside bus,