Vous êtes sur la page 1sur 8

Computer-Aided Engineering:

Seven-League Boots to the Gates


of Progress

By John Michael Williams


jmmwill@comcast.net
2015-09-18

A brief history of computer-aided design and an analysis of the


relevance to it of the now-defunct Daisy Systems Logician design
tool.

Copyright (c) 2015, by John Michael Williams.


Reprinted With Permission of ETC -- All Rights Reserved.

Abstract
The semantics of computer-based electronic design have become
more and more accurate over the past years from 1960 to 1985; the
label typically applied to the field was changed correspondingly
from Computer-Aided Drafting (CAD1) to Computer-Aided Design
(CAD2) to Computer-Aided Engineering (CAE). In CAE, semantics
are assigned (intensionally) to electrical component elements, and
the designer then extensionally tests the resulting, often huge,
circuits with reference to those elements.
Consistent design syntax is prerequisite to usable design
semantics: A similar requirement has been applied to natural
language in the form of the Whorf Hypothesis.

Historical Notes:
The original version of this paper was presented at the Tenth International Conference
on General Semantics, San Diego, California, in August of 1985. It later was published
in the journal, ETC: A Review Of General Semantics. I found a copy of the
manuscript among some old documents belonging to my now-deceased father, John Max
Williams, and have somewhat rewritten it here.
An historical note:
Daisy Systems, my employer in the 1980's, went defunct in about 1990. It attempted
to manufacture tools for all aspects of computer-aided engineering -- applied engineering
methodology, workstations, and software -- and consequently suffered fatal quality
errors. One of its competitors, Mentor Graphics, restricted its attention to the software
and has prospered with great success to this day.
The original Daisy facility was located in Sunnyvale, in the San Francisco Bay area,
in 1981. In about 1983, Daisy moved to a large, newly-built facility located at
Middlefield Road, Mountainview, California. This facility now is occupied by the morerecent Synopsys corporation, another presently successful software provider.
The terms intensional and extensional are derived from General Semantics and are
used here technically:
Briefly, an intensional orientation is one in which understanding and planning is
done strictly based on human thinking and logical assumptions and can lead to
misevaluation, prejudice, and failed actions. An exclusively intensional orientation
describes a form of mental illness.
By contrast, extensional orientation is based on observation of physical fact and can
lead to an insistence upon routinely repetitive details and device-driven, unusual
behavior.
The greatest advances in science have resulted from simultaneously successful
intensional and extensional orientations, with emphasis upon the latter.
---------------------------------Note: Mentor Graphics and Synopsys are trademarks of their respective companies.
Additional information on Daisy Systems may be found online. Two background articles
of use are located (as of 2015-09) at:
http://product-wisdom.com/2013/03/23/daisy-systems-an-anatomy-of-a-blunder-anecdote /
and https://en.wikipedia.org/wiki/Daisy_Systems .
Some more detail on Daisy Systems is in the Footnotes section below.

Introduction
In recent years, the exponential rate of increase of knowledge in modern society has
not been unrelated to a corresponding rate of increase in the complexity of integrated
electronic circuits. The vacuum tubes of the 1950's, components with no more than a
half-dozen distinct functions each, have been supplanted by transistorized chips. These
chips have increased in complexity from a few functions each in the early 1960's, to
hundreds of functions each in the 1970's; and to hundreds of thousands of functions each
in the early 1980's. Current chips can exceed tens of millions of functions each, with no
clear limit in sight.
How is it possible to design such devices? How can a design engineer, or a designteam of several engineers, keep track of the development of an integrated circuit
containing, say, half of a million independent logic gates?
The answer seems to be, by the process of abstraction.

The Need for Extension in Design


In the past, structural and mechanical engineering was done by planning in paper and
ink, by engineers trained as skilled draftspersons. Electrical engineering and circuit
design followed suit, so that, by the mid-1950's, electrical circuits such as those for radio
or television receivers were being captured in schematic form by being plotted
painstakingly by hand on graph paper. In those days, electronic circuits hardly were
complex enough to require more than a day or so for correction of a design error or for
incorporation of a new design feature.
Before the 1950's, engineers had adopted "breadboarding" as a means of extensionally
testing their ideas. In breadboarding, a prototype circuit was assembled on an extended
flat surface: Such a breadboard, covered with sockets, clips, wires, and components,
resembled nothing so much as a set of recently cut animal cookies, all laid out -- and soon
to be collected up and baked.
Sometimes, when power was applied to a breadboard, an error might cause a
component actually to bake; this would require a review of the breadboard wiring and
possibly of the schematic itself. An important value of breadboarding was that it saved
the multiplication of errors which otherwise might have occurred if an inaccurate
schematic had been used directly in a full-scale production run. Often, design errors
which would cause incorrect circuit performance were not detected until the
breadboarding stage in the design process.
Breadboarding was expensive and time-consuming; and, so was any revision of a
paper-and-ink schematic: To make changes, the most effective approach was to cut away
parts of the breadboarded schematic in order to move them to new locations; then, errors
might be corrected by pasting revisions on top of the schematic. This "cut-and-paste"
approach to design held sway through the 1960's; but, the increasing size of circuits soon
created a need for more efficient methods of capturing and extensionally testing
schematics. The advent of the electronic computer provided that more efficient
approach.

The Need for Intension in Design


A design must start with an idea, and ideas must be abstract. The level of abstraction
may vary considerably, depending upon the problem to be solved. In any event, given a
problem, the circuit solution first must be captured, as it were, in a schematic. Given a
schematic, a physical layout (plan) then may be created so that the circuit actually might
be assembled, thereby realizing the idea. The present discussion does not address itself
to any of the special problems involved in the creation of a circuit layout.
A schematic may be viewed as a linguistic expression: Specifically, the schematic
expression provides an abstract representation which defines the function and
connectivity of the circuit elements. A schematic usually will not represent the relative
size, shape, location, or orientation of the objects making up a working circuit; these last
will be defined by the circuit layout plan. However, schematics primarily are graphical
entities in that they are two-dimensional representations in which the shape and
arrangement of the circuit symbols have been assigned conventional meanings.
It is not easy to construct, order, erase, or connect two-dimensional symbols with pen
and ink; so, in the 1960's, the cathode-ray tube, controlled by a computer, was adopted as
a more convenient means of manipulating schematic symbols. Following the lead of
structural and mechanical engineering usage, the new cathode-ray-tube-based schematic
design workstations were called CAD systems, "CAD" standing for Computer-Aided
Drafting.
The meaning of "CAD" later became Computer-Aided Design as design engineers
began to replace draftspersons at the workstation. The idea of a CAD workstation,
entirely and directly operated by a design engineer, became established in the 1960's and
1970's. Stand-alone, microcomputer-based workstations began appearing in the 1980's.
A CAD workstation, in the narrow sense, would be an expressional device only: Such a
workstation would allow the capturing (perception and expression) of a schematic. Such
a capture would be an extremely important feature in three-dimensional mechanical
design. However, CAD workstations, strictly defined, can not perform any test of
correctness of the schematic, either against the ideas of the designer or against the
(extensional) reality eventually to be produced from the schematic: The accuracy of a
CAD-captured schematic depends solely upon the care and the sharp eye for error of the
designer.
After the early CAD workstations, the next step, not at all exclusive of the progress of
previous steps, was to incorporate automatic checking of the syntax and the semantics of
the captured schematic. This step was taken in the late 1970's for systems based on
mainframes or large-scale minicomputers; it was adopted for stand-alone workstations in
the early 1980's. The new discipline was called CAE, for Computer-Aided-Engineering.
The Logician workstation, made by Daisy Systems Corp., in those days defined the
CAE state-of-the-art. Such CAE workstations, implemented on IBM and other personal
computers, soon became as much a part of a design engineer's set of tools as once were
slide rules or drafting tables.

Semantics of the Logician


The primary purpose of a CAE workstation is to assist its user in abstracting workable
designs. By workable, I mean (a) consistent with known electronic principles; and (b)
constructible from components which might be assembled according to a schematic
representation of the intended design.
The Logician workstation facilitated circuit design in three main ways:
First of all, the Logician allowed the designer to create a schematic which might be
drawn, and altered interactively, while viewing it on a cathode-ray tube screen.
Corrections, changes, or improvements might be made with no time lost, as in the past,
by the need to redraw the schematic on graph paper. Nor was there any cutting or
pasting of intermediate results. These features greatly lessen the need to attend to
irrelevant detail: They permitted better focussing of effort on the more abstract concepts
of the purpose and the function of the circuit elements which the Logician's display
represented. The display, in CAE, also might be used to abstract functional groups of
wires, components, etc., into block-schematic format, the further design process
containing detail represented only at a block-diagram level. Thus, the "schematic
capture" defined above became the more efficient and more abstract creation of a
computer-generated schematic.
Secondly, the Logician allowed a captured schematic to be modelled. I am using the
word model here to refer to a consistency check: This is a standard usage of the word in
science or philosophy. A successfully modelled design might not perform functionally as
intended, but it would be guaranteed consistent.
A computerized modelling system is a special set of programs. These programs may
be run to test automatically the design and to report errors resulting from such tests.
The modelling system may be considered an extensional device in that it allows
correction and validation of the structure of the schematic against its built-in rules
(=observations) of combination of elements. Such elements of a design might be
represented by gates, integrated circuits, wires, blocks, inputs, outputs, etc.
Considering the schematic as though it were a linguistic expression, we may apply, of
course, the Whorf hypothesis.
In its original context of language, the Whorf hypothesis is the hypothesis that the
format (syntax) of a language determines the understandings (cognitive effects) which it
might have upon the speaker or listener.
In the context of use of the Logician's modelling system, the syntax may be viewed as
determining how the design engineer should be allowed to perceive a schematic. Should
any perceived (captured) schematic not suit the designer's need, the schematic may be
changed or rearranged -- but only provided that it comply both with modelling-system
syntax rules and with the designer's functional requirements. Such requirements may be
viewed as the intended semantics. In practice, should a design entail an intractably
unmodellable schematic, it would be possible, at relatively great expense of time and
effort, to introduce new design elements or new design rules into the modelling system
itself.

Examples of typical design inconsistencies which would be reported by the modelling


system are: Input or output wires which led nowhere, more than one wire connected to
the same output pin (a situation which, in a breadboard test, might risk a destructive
short circuit), and components which themselves had been designed incorrectly.
Thirdly, the Logician allowed simulated operation of a successfully modelled design.
This is the step which required semantics.
In order to simulate a design, for each major part, the schematic of the circuit first
must be transformed into a network of (symbolic) electrical nodes, each pair of nodes
being separated by a functional representation of some circuit element. This usually is
done automatically, after modelling, by running a program as described in the next
paragraph. The functional representations of the transformed circuit elements in the
resulting node-net might include truth tables defining logic gates, time delays caused by
components or long wires, constant voltage differences, or sequences of logic-levels caused
by the running of a computer program. The electrical node-net in effect defines a set of
electrical functions; it also places boundary conditions upon the circuit operation during
simulation.
After creation of the electrical node-net, the designer could simulate the Logician
circuit by entering a set of initial conditions, running the simulator program, and then
studying the simulated output as a graphical or tabulated display. A circuit not
performing as designed might be changed, remodelled, and resimulated until the
designer was satisfied. It was during simulation that the designer's preconceptions were
put to an extensional test at the relatively low level of abstraction of simple logical
assertions and simple electrical connections.

Conclusion
The technological impact of computer-aided engineering has been greatest on the
design of the large logical circuits which have become commonplace. Simulation has
replaced breadboarding as an intermediate step between schematic and chip. In
practice, no one would consider wiring, say, a million components together just to test a
design. Rather, after modelling and simulation, an attempt would be made to
synthesize, lay out, and print the chip all at once. I do not discuss logic synthesis here.
The semantic importance of computer-aided engineering is that it has made possible
electronic design at a very abstract level, the design always being kept consistent with a
syntax which permits it to be tested against the reality of the known laws of electronics.
The Logician, in its day, allowed the ever-increasing demands of larger- and larger-scale
circuit integration to be met, simply by guaranteeing that the product of the designer's
ever-increasing level of abstraction would be consistent with the physical reality which a
workable circuit always must represent.

Footnotes
The standard Daisy Systems logo was

The early Logicians (~1982) were sold as a unit: A metal desk allowed a seated
engineer to operate a monochrome terminal using a keyboard and I think mouse. The
desk included an 8-inch floppy drive with a capacity of a little over 1 MB, very generous
at the time. The hardware for the proprietary operating system, maybe a half-dozen 24'
boards, was contained in a metal chassis near the floor of the desk. I have not been able
to find an internet-posted picture of this desk.
Later Daisy software versions used a Sun Microsystems chassis and unix operating
systems.