| |
| |
A
Short Segment of an Engineering Career
1951-1958 - John C. Alrich
(© JCA 2005)
I did not realize it at the time but the
early and mid-fifties were the most exciting and productive of my
engineering career of forty years. During my professional history, I
worked at four or five different firms, large and small, but none of the
projects and challenges rose to the extraordinary heights I accidentally
encountered so early on . The factors contributing to this essay were
unique in my experience. In particular, they included two very special
people, a mathematical concept, and a commercial product. Similar products
were so rare that each was given its own acronym--the ENIAC, the EDVAC,
the LARC, and so on. In those days they were ponderously referred to
as , “electronic digital computers.”
It all started in 1951, the development of a
digital computer at the company where I was just starting my employment.
At the time, I was not even aware of the possibility of such a device nor
for what it would be used. The man who promoted the idea of building a
commercial digital computer was Clifford Berry, Ph.d. in physics, a Senior
Scientist at Consolidated Electrodynamics Corporation (CEC), located in
Pasadena. Before WW-II, Cliff had been working on his physics doctorate
under the supervision of Professor John Atanasoff, Iowa State University
who is often credited with being the first person to introduce the concept
of an electronic digital computer.
The hardware part of their design was never
completed due to the intervention of the war . However, Cliff never forgot
his work there as a young student and it was this knowledge and the need
for such a device which propelled CEC into the computer age. At the time,
this moderate sized company did not recognize it had an elephant by the
tail---an elephant which would ultimately dwarf the very company from
which it sprang.
CEC probably no longer exists---I am unable to
find it on the Internet Yellow Pages so it has probably morphed into
another firm---made highly specialized scientific equipment, particularly
mass-spectrometers which were the catalyst for the proposed computer
development. A mass-spectrometer is a large laboratory instrument
for analyzing complex substances at the molecular level. However, to
interpret its results, it was necessary to do a tedious and somewhat
complex mathematical operation called matrix inversion At
that time the inversion was was done at the company by using an analog
computer, invented by Cliff and a co-worker, Sibyl Rock.
The precision of the analog computer was limited
and operation was time-consuming so Cliff suggested to the officers of CEC
that they consider developing an electronic digital computer---at the
time,1951, a highly risky endeavor in an entirely new field for a modest
company without the proverbial “deep pockets.” But it had great
potential if it did succeed.
After the decision was made to proceed, new
hiring was started at CEC for the program and I was the second engineer to
be brought in for the project. My experience was limited to vacuum-tube
circuit design for equipment carried aboard the Aerobee
weather-sounding rocket and some specialty work for Fairchild Camera
Corporation in Pasadena. Fairchild decided to pull up stakes and move
East, so I applied for work at CEC and was accepted into the newly formed
computer design section where I, along with most of the other new hires,
was a complete neophyte.
The reason the company hired such novices was
because in 1951 there probably weren’t more than several hundred
engineers and mathematicians in the world who knew a great deal about the
subject. Essentially, we pulled ourselves up by our bootstraps using the
few texts available and two consultants, Drs. Harry Huskey and Ernst
Selmer. Both worked under part-time contracts to help educate us as we
slowly acquired some understanding about the arcane world of computer
design.
Interestingly, Huskey brought many of his ideas
back to the US from British designers, such as Alan Turing and some of the
other Bletchley Park scientists who broke the German codes during the war,
whereas Selmer took our technology back to his country, Norway, to
introduce it to his scientific community. Huskey gave evening seminars to
us while working on a new machine, called the “SWAC”, at UCLA
for the National Bureau of Standards but he was not directly involved with
our design.
Selmer had been sent to the US from the
University of Bergen, specifically to learn about programming and digital
computer logic design. In the early 50’s he studied with John von
Neumann’s group at Princeton, possibly the preeminent computer design
group of engineers and mathematicians in the world. After completing his
work at Princeton, Selmer came to California and became a visiting
professor at Cal Tech in Pasadena. His area of specialty was Number
Theory, an ideal background for a logic designer. As our program got
under way, our management called Professor von Neumann asking if there
were any candidates here locally and Selmer was recommended.
Selmer did the design for what we called the
Central Timing System, which included the arithmetic section; Operating
System software was unhead of in those days. As his logic design was
evolving, we engineers did the implementation---wiring tables, circuit
design, load calculations, module design and placement, cooling, cabinet
housings, drum memory, console and display, paper tape and printer
input/output; cards and magnetic tape were to come later.
My assignment, along with several others, was the arithmetic processor,
which is now called the “CPU” and today is generally confined to part
of a single semiconductor chip. In addition to performing the four basic
arithmetic instructions, the CTS could perform fifty other instructions
such as shifting and branching operations which are fundamental
requirements for computers even today.
About a year after design work started, what we
called the “bread-board” implementation got under way. This was a
wooden framework roughly the size of five or six home refrigerators placed
side-by-side and built to support 1,600 vacuum-tubes, now in 8-tube
modules, interconnect wiring, and the drum memory. This temporary
structure enabled us to start testing and debugging the newly
created hardware and software prior to completion of the first prototype.
Software compilers and assemblers were unknown in those days, at least by
us, so all testing was done in hand-written machine code. Since the
computer was designed to work internally with binary-coded-decimal (BCD),
coding was relatively straignt-forward.
This method of representing numbers and
characters was unusual, if not unique, even during those early days.
Today, of course, all machines work in pure binary internally, transparent
to the user and slightly more efficient. BCD, while less efficient than
binary, allowed for a certain amount of self-checking, a real advantage
when using unreliable vacuum-tubes. Forty bits in BCD can represent up
to10^10 whereas 40 bits in Binary can represent 10^12.
I remember one of our early tests was to generate
successive primes and check the results against printed tables which
previously had been laboriously calculated using mechanical calculators.
The text listed all primes less than 10,000,000. Our new computer could
have done the same task in about a day.
Testing of the bread-board lasted perhaps several
months and went surprisingly well. While the bread-board was being
checked, manufacturing of the first prototype was already under way. I
believe the prototype was completed in 1953.
ElectroData continued to flourish in this new and
stimulating field. Shortly after shipments began, our company was spun off
from CEC as a separately owned subsidiary so now we had our own financial
responsibilities, our own factory, systems were being individually wired
and assembled manually and shipping commenced at the blazing rate of two
or three systems a month. Prices started at $135,000; with today’s
inflated dollars, the same system would cost $600,000. Field service
personnel were setting up and testing the equipment in our factory and in
the field. Programmers were writing custom and generic programs for
clients. A national marketing team was put in place. Everything was
humming.
Development started almost immediately on the
next model after the 205 prototype was running. We were now part of
Burroughs Corporation and the new machine, the B220, would be at least two
orders of magnitude faster than the Datatron 201 and 205. A blind
physicist, an ex-IBMer named Ted Glaser, was the chief logician for the
new system which would have a faster clock and a core memory but still use
vacuum-tube technology.
By this time, Selmer had returned to his home in
Norway to pass on to others what he had learned in the States. He probably
was the first Norwegian to introduce computer logic design to engineers
and mathematicians in his own country. Today I am told he has been retired
for ten years from teaching and in 1983 “he got a high distinction
from the King, the Order of St. Olaf, for his academic and non-academic
achievements.” An honor richly deserved.
While the popularity and utility of computers improved for both commercial
and scientific users, particularly the latter, limitations with
fixed-point (the 205 had ten decimal resolution plus sign) computing
became apparent and a strategy first used with card equipment became
necessary. This technique involved the replacement of the first two BCD
decimals, stored in the register of ten decimals, with an exponent ranging
from 00 to 99 followed by a fixed decimal point and eight succeeding
decimals, called the “mantissa”. To account for the sign of the
exponent, the number was biased; i.e., the number 50 represented an
exponent of zero, 51 an exponent of 1, 49 an exponent of negative 1, etc.
All exponents were to the base 10. This method had the great advantage of
expanding the number range the computer could handle by 50-orders of
magnitude but at a cost of losing 2-orders of magnitude in number
resolution. Consistent with the nomenclature at the time, it was called
“Floating Point Operation”. Six new commands were added to the 205
vocabulary: the four arithmetic operations plus conversion from fixed to
floating and conversely.
At the time I was given the responsibility of designing the FPC (Floating
Point Control), Selmer had returned to Norway. However, I was able to
discuss the expected function of the FPC with a very fine mathematician
and Senior Programmer on our staff, Stanley Katz, Ph.d. Essentially, my
design emulated what Stan had been doing with sub-programs, a slow and
laborious process at best, using commands already available within the
fixed-point set.
The design took approximately a year to complete. It resulted in a smaller
matching cabinet with plenum and vacuum-tube modules similar to those
employed in the main-frame and was attached to one end so the necessary
wiring could be extended. It could be field installed on existing 205’s
or attached as an add-on at the factory.
JCA 5/16/03
Rev. 2 |
| |
| |
| |
| |
| |
| |
| |
|
