This is an OCRed version of OralHistory_JCLogue_TC-68.pdf
The OCRing was complicated by an obvious fault in reproduction
of the original text - almost vertical stripes of missing print,
especially on the left edge of some pages.
Some places I made some reasonable sounding guesses
Others, I inserted question marks
And I used a spell checker to detect other blunders
memory is not spelled rnemory
The spelling in the original was not perfect,
I corrected some that might be disturbing
transitors to transistors, for example
This has been somewhat proof read, but there are errors.
by Ed Thelen July 6, 2009
March 30, 1970
Memorandum to: Mr. L. M. Saphire
Subject: Interview TC-68 in the IBM Oral
History of Computer Technology
I have finally found time to edit the transcript of the interview we recorded
on April 2, 1968. After rereading the transcription, I concluded that it
would be more productive if I summarized what appear to me to be the
key events in computer technology over the past 15 to 20 years. I think
you will find this will make a much more coherent story than what would
result from editing the transcript of a question and answer session.
Since you are interviewing many other people, I will limit my comments to
those events in which I was personally involved or in which people that
reported to me were involved.
Defense Calculator
I left the academic life of Cornell University as Professor of Electrical
Engineering to join what I considered at that time to be a coming field,
computers. In May 1951 I joined IBM to work on the Defense Calculator
which later became the 701. Two interesting problems arose, that had to
be solved to make the machine a success. Naturally, there were many
other problems that demanded solutions. As I have said, I will limit my
remarks to activities in which I was personally involved.
The first of the problems we were faced with was improving the read-
around-ratio of the electrostatic memory. Very briefly, this involved
how many times memory addresses surrounding a given address could
be interrogated by the electron beam of the cathode ray tube before the
information at the address in question would be destroyed. We were
striving for, as I recall, a read-around ratio of 350 or more. This
meant that the addresses surrounding a particular address could be
interrogated a total of 350 times before the information at the given address
had to be regenerated or it would be destroyed. By experimentation I
was able to show that the read-around-ratio was directly dependent on the
stability of the deflection circuits which were in turn dependent on the
stability of certain power supplies that powered the memory. After re-
designing these power supplies I was able to achieve the required read-
around-ratio.
Mr. L. M. Saphire -2- March 30, 1970
The second problem became known as the indian blanket problem because
of the characteristic pattern of ones and zeros displayed on the face of all
36 cathode ray tubes. The problem was intermittent and frustratingly
impredictable. The memory would work fine for hours and instantaneously
lose its information causing the 701 to come to a halt until the memory was
reloaded. This problem persisted, as I recall, for two weeks. One
Friday afternoon Jerry Haddad told me he had heard that I thought I knew
what was causing the problem and what had to be done to fix it. 1 told him
I felt sure I knew what was wrong but had not been able to try my idea
because I could not get the use of the 701. He suggested that we come
in Saturday morning and try out my idea. I was able to show Jerry that a
noninductively wound feedback resistor only had to change value by .05%
suddenly and it would cause the indian blanket problem. The following
Monday I replaced the guilty resistors and the problem was cured. Jerry
asked me to redesign the deflection circuits to eliminate that type of
resistor, which I did.
Point Contact Versus Junction Transistors
After completing my work on the electrostatic memory for the 703 I was
placed in charge of a small group of engineers that had been developing
transistor switching circuits using point contact transistors. This was
about December of 1952. Since I knew nothing about transistors I pro-
ceeded to learn all I could and build circuits at the bench. Within one or
two weeks, two important problems became obvious to me:
1] The point contact transistor had very serious limitations.
Their current gain was respectable only when transformers
could be used which limited them to pulse logic. Finally I felt
they would always be unreliable because of their construction.
2) Minority carrier storage could seriously limit the turn off
speed of a transistor.
To correct the first problem, I took Bob Henle and Carter Dorrell off of
the Small Accounting Machine model that they had been trying to build for
some time using point contact transistors. After one false start I assigned
Bob Henle the task of developing a family of logic circuits using junction
transistors. By June of 1953 we had such a family of logic circuits working.
Mr. L. M. Saphire -3- March 30, 1970
I solved the second problem by inventing a circuit that clamped the collector
of the transistor out of saturation. This was accomplished within a few
weeks after taking over the direction of this circuit development group. I
make this comment because we were at that time still working with point
contact transistors. Several years later Warnock of Philco addressed
the minority carrier storage problem for junction transistors and arrived
at a solution that was dependent on some of the claims in my patent when it
was properly written.
New Technology Education
In the summer of 1953 Ralph Palmer asked me to structure and direct a
course to train 28 engineers in transistors and cores. The course I put
together consisted of three weeks of classroom work and six weeks of
laboratory work. During the six week period I split the class into six
groups and had them lay out on paper six machines using the knowledge they
gained during the three weeks of formal classroom work. The course was
known as Logue's College of Digital Knowledge.
At the termination of this course I recommended to Ralph Palmer that three
of the machines be built. He agreed and the three machines built were the
Small Accounting Machine, the Data Transceiver, and a transistorized 604.
My group supplied the circuit designs for these machines. Since we had been
working with Ed Lorenz and his people we had demonstrated that printed
circuit cards could effectively be used to package the circuits.
With the exception of wire wrap, current switch circuits, and transistor
driven core memories the machines that were built as a result of this
course demonstrated all of the concepts contained in the SMS program out
of which was built the 7000 series of machines. It must be remembered
that this was all accomplished in 1953 and two thirds of 1954.
Blue Sky Activities
In the Spring and Summer of 1954 Ralph Palmer asked me to have my group
look at blue sky ideas. These were far out concepts that would hopefully
anticipate things to come. We engaged in this activity plus assisting the
three machine groups described above. To keep my group productive I was
forced to develop ideas that were far out at that time. Several that come to
Mr. L. M. Saphire -4- March 30, 1974
mind are:
1) Semiconductor equivalent of the gas tube counter. The idea
I had was to use the double base diode structure in a bar of
semiconductor material in such a way as to permit it to
have 10 stable states. Input pulses would shift the position of
an ON -state to an adjacent position and reset the starting position
to the OFF state. I assigned a summer student whose name
was Rick Dill to this idea and before the summer was over he had
demonstrated a device with 4 stable states. Rick went on to get
his PhD. and is now working in Research.
2) A one dimensional image disector consisting of a bar of germanium
with ohmic contacts at each end and an alloy junction collector at
one end near one of the ohmic contacts. In operation, an image
of a character was projected onto the germanium bar for a
short time. A voltage pulse was applied across the ohmic
contacts to sweep out the carriers generated by the image of the
character. The carriers were collected by the collector and the
collector current was a time function that represented the spacial
distribution of light falling on the germanium bar. I had Kenny
Sih work on this idea and he demonstrated a working model.
While all of this was quite interesting it was at the same time very frustrating
for three reasons:
1) We did not have our own semiconductor laboratory. John
Little naturally objected to our using his furnaces for diffusion.
2) It became clear that silicon might have properties that would
open up other possibilities. Since there was no silicon technology
within IBM at that time there was no equipment that would handle
silicon.
3) The work we were doing was very far out and it was difficult to
get the type of person that can be productive in that type of
environment.
Mr. L. M. Saphire -5- March 30, 1970
Silicon versus Germanium
In retrospect I must say that while the blue sky work was very frustrating
it convinced me that silicon would be the semiconductor of the future.
Another point that may or may not be valid is that if we had been able to
get the equipment to work with silicon we may have been led to the
invention of integrated circuits. We were certainly nibbling at the edges
of the concept. It is a fact that in 1955 I tried to get Lloyd Hunter and his
people to work on silicon. One of the reasons I gave for my interest in
silicon was the possibility that the surface could be oxidized thus forming
SiO2 which could passivate the surface. This concept is very fundamental
to integrated circuits as is the idea of performing complex operations
within a single piece of semiconductor.
MAC Program
In January of 1955 my group was asked to design transistor circuits for
the Modular Accounting Calculator program. We were given three
months in which to come up with a circuit manual. At the end of one month
Max Paley said he had to have the manual within one month. As I recall
we met the schedual with a crude manual. During 1955 we were giving the
people on the MAC program circuit designs before the ink was dry on the
paper. It must be mentioned that in 1953 and 1954 my circuit group con-
sisted of George Bruce, Ray Emery, Carter Dorrell, Jim Walsh, Bob
Henle and myself. I transferred George Bruce to the 604 group in early
1954 because I had had him developing transistor circuits for use with
core memories and I wanted that machine to incorporate a core memory
into its design. Unfortunately, this did not happen. Sometime in 1954 I
transferred Kenny Sih into my group and Rick Dill worked as a summer
student.
In January of 1955 my group increased from about 6 people to 18 with the
12 new people having no knowledge of transistor circuits. To get a circuit
manual out in such short time with new people was very difficult.
In addition, we had only a handful of samples of the Raytheon junction
transistors that were to be used in the MAC program. Our semiconductor
people could not help us with specifications because they were still working
with point contact transistors.
Mr. L. M. Saphire -6- March 30, 1970
With all the problems I had at that time, I can't understand why I
also took on a contract with our military group in Endicott to develop six
servo amplifiers using silicon transistors. These amplifiers had very
tight design specifications. My reasons at the time were two fold:
l. My conviction was that silicon would be the semi-
conductor of the future and we had to get some
experience with it to see if that conviction was
well founded.
2. I was also convinced that my group had to be
knowledgeable in technologies that would be
fundamental to the application of computers to the
process control field. I felt that by selecting the
proper contracts, my group could learn what was
needed for future process control activities.
In late 1955, Max Paley reversed the assignments of George Hawkins
and Dick Richards. Previous to this reversal, I had been dealing
with George Hawkins who had responsibility for the hardware of the
608 machine. Dick Richards had responsibility for the system design.
When their jobs were reversed, Dick Richards took the position that
the circuits my group had been designing were no good because they were
unreliable due to poor noise tolerance. Needless to say this created a
large flap. A number of circuit groups in Poughkeepsie and Endicott
proposed 10 and 1 Z volt circuits as an answer to this apparent problem.
I took the position that Dick Richards was all wet. There was so much
emotion and so little fact that I pulled my people off the program and
assigned them to developing circuits with lower voltage swings than the
5 volts used in the 608. My reason for going to lower voltage was that
this was the direction in which to proceed if higher circuit speeds were
to be achieved.
To the best of my knowledge, the 608 machines were delivered with the
same circuits we designed. I do know that one or two years later Lloyd
Hunter accused me of overdesigning the circuits because the 608 machines
were too reliable in the field.
STRETCH
In March 1956, I was given responsibility for the hardware design of
Stretch. I reported to Steve Dunwell who had overall responsibility
for the program. I recall I had my circuit designer's work during the
regular two week summer vacation that year. We measured the
Mr. L. M. Saphire -7- March 30, 1970
characteristics of they transistor that our semiconductor people were
proposing to use to determine the best operating point for maximum
performance. To no one's surprise, it turned out that best performance,
in terms of cut-off frequency, would be obtained if the device were
operated in the conduction region. I instructed my people to come up
with a design for a logic circuit that would keep the transistor in its active
region and at the same time perform logic. This was a new concept and
it departed significantly from the approach I took in the end of 1952 to
keep a transistor out of saturation. Bob Henle gave structure to the con-
cept and Hannon Yourke worked out the details. It involved using comple-
mentary symmetry devices, namely PNP and NPN transistors. This was
not a new concept for us since we had worked with it earlier.
As a result of this analysis and invention, I posed the following alternative
to John Little and Bob Swanson who were to develop the high speed tran-
sistor for Stretch:
l. If we were to use one type of transistor then we would
require a transistor that had a low saturation drop and
come out of saturation fast.
2. If saturation specifications were too difficult to meet,
then we would need both an NPN and a PNP transistor
with no saturation specification. They decided in favor
of providing complementary types of transistors.
Before we had fully developed the current switch circuit and certainly
before we had a firm transistor design, I told the systems designers
that I would commit to a performance of 20 nanoseconds delay per stage
with reasonable fan in and fan out. I had to develop this definition since
Gene Amdahl and Steve Dunwell were talking in terms of a ten megahertz
clock in the machine but were then being very loose in specifying through
how many stages a signal would have to propagate in a machine cycle. I
felt that it would be possible in the development time we had available to
achieve the performance I had defined. I fully understood, and so stated,
that my predictions were based on anticipated invention. I felt we could
not build a superior machine if we did not count on inventing our way out
of certain problems. It takes a finely tuned intuition to know what is
doable and worth doing and what is not doable within a certain time frame,
???n one is dependent on new technologies and new inventions. It is
on this point more than any other that Steve Dunwell and I disagreed.
Mr. L. M. Saphire -8- March 30, 1970
The SMS Program
In March of 1957, I was asked by Bud Beattie to head up a circuit
group to develop the circuits and package for what later became the
7000 series of machine, including Stretch. There was a great deal of
motivation for standardization with the intent of keeping the part numbers
to a minimum. Ed Wyma and Newton Noell were developing a package
called the "Bird Cage". This consisted of placing a circuit, composed
of discrete components, on a small phenolic card with sheet metal pins
crimped around its edge. These pins were in turn wire wrapped to
mating pins mounted in a large board that would support hundreds of
the small phenolic cards. In crude concept, this was the forerunner
oŁ the SLT module.
During the period from mid-1957 until about late 1958, Ralph Palmer had.
me reporting into him. He explained to me that Mr. Watson, Jr. , was
determined to achieve standardization and minimize part numbers. If
a low part number count was not achieved, Ralph would be in trouble
and one microsecond later I would be in trouble since he had me re-
porting into him. I got the message.
Ralph suggested that I would be a good idea to make a movie, to be shown
at the laboratories, that would extoll the virtues of standardization. After
giving this some thought, I had a movie made which included a pan shot
of a shelf of eight tube pluggable units. When the camera dolly backed off,
the viewer saw that it was not a shelf these pluggable units were on, but
was in fact the gym floor in the Poughkeepsie IBM Country Club. This
sea of pluggable units represented all of the pluggable unit part numbers
needed to service the 700 series machines. The movie was obviously too
well done because Bud' Beattie saw it and put it under lock and key.
Ralph's comment was "If anybody saw that movie, I would get fired."
That ended my short career as a movie director.
Ralph decided that a one-sided printed circuit card, with 16 contacts that
would hold one circuit, would be an ideal package. He conveyed his
idea to Ed Garvey in Endicott and that was how the Standard Modular
System -- SMS was born. I invited Ed down to Poughkeepsie to see what
Ed Wyma and Newton Noell had been doing with wire wrap and he adopted
this idea that had been developed by Bell Labs. My people and I worked
Mr. L. M. Saphire -9- March 30, 1970
with the Endicott people, and jointly set up a release procedure
for circuits and cards. All logic circuits were approved by my
group before they could be adopted as a standard circuit and published
in the standard circuit manual.
During this period I set Bob Domenico up to do circuit simulation on
digital computers and to develop and prove out circuit design concepts.
We developed an end of life circuit design using a statistical analysis
of component-parameter drift. Somewhat earlier, I had set up a group
under Herman Wolff to develop techniques for measuring the electrical
characteristics oŁ transistors so that we would be in a position of being
able to predict the performance of a transistor in a circuit prior to
building the circuit. The reason I got this work started was twofold:
1. A sound scientific base for circuit design is very
important when large production runs of computers
using these circuits are involved. It is equally
important when the limit of performance is to be
attained.
2. I felt that as devices, circuits, and packages get smaller
and smaller the day will arrive when it is not practical
to model a circuit on the bench.
In mid to late 1958 Ralph Palmer brought Nate Edwards on board to handle
standards work in White Plains and asked me to turn over to Phil Stoughton
the Product Engineering functions associated with the SMS program and staff
Phil's group with competent in my organization. This I did and the transi-
tion was very smooth.
Preparing for Batch Fabrication Techniques
During 1957 and 1958 while most of my people were working on the SMS
program, I had a small group composed of Gcrry Maley, Paul Low and
others investigating how machines should be designed so as to minimize
the part number problem when many circuits could be fabricated on one
substrate. At that time, the technologies that tended to indicate what was
coming in the future, with respect to batch fabrication techniques, were
cryogenics, electroluminescence/photo-conductors, and to a lesser degree,
integrated circuits. I knew that T.I. was experimenting with putting one
Mr. L. M. Saphire -10- March 30, 1970
or more circuits down on a piece of silicon because Willis Adcock
of T.I. asked me for and I gave him sample circuit configurations
with which they could experiment. Since it was impractical for my
group to work with cryogenics, we used electroluminescence/photo-
conductors as a vehicle with which to explore the impact that batch
fabrication capability would have on machine organization.
It was at this time Rex Rice carried the concept from Poughkeepsie to
Yorktown. He gave it the name Pattern Logic. Dan Slotnick was in
his group for awhile and left IBM to join Westinghouse. Slotnick has
continued to explore a branch of this concept ever since. While my
group did demonstrate some "gee whiz" embodiments, these never
found practical application.
ASDD Formation
In the summer of 1959, I was promoted to Manager of Technical
Development. ASDD had just been formed and our first job was to
tructure a meaningful set of programs. I inherited several programs
from Research when they were transferred, with people, to ASDD. I
phased some of these programs out and continued to support others that
ooked worthwhile. Since ASDD's charter was to open up new market
and product possibilities for the Corporation, I felt we had to initiate new
techrnology programs with long-termpayoffs.
There were two areas that I felt had significant possibilities for growth,
if the proper technologies could be developed to support them:
1. Unconstrained handprinted character recognition.
2. I coined the term Image Processing to describe the
other area.
To further develop an idea I had pursued in my spare time during the
Summer and Fall of 1956, 1 got Evon Greenias to join my group in
Ossining. I wanted to set up a small but effective character recognition
group to develop a practical means for identifying handprinted numerals
and characters. The concept of using a flying spot scanner as a curve
follower, which I had developed on paper earlier, made a good starting
point. After I got Evon set up and the program started, it turned out I
had to put my job on the line to keep the program going. My management
Mr. L. M. Saphire -11- March 30, 1970
took the position that ASDD should not be doing this type of work since
Endicott had the responsibility for character recognition. My position
was that Endicott was not working on unconstrained handprinting recogni-
tion and I could envision future product opportunities if IBM had this
capability. I was specifically thinking of reading credit slips, stock
inventory cards, etc. Subsequent developments have confirmed the
convictions I had. The 1287 machine, in a sense, got started back
in 1956 with the idea I had for pattern recognition. Certainly Evon's
significant contributions to this program made the program a success
and contributed towards his being named an IBM Fellow.
The image processing group that I established under Bob LeHane un-
fortunately did not really get off the ground. My idea was that while
we had gotten fairly sophisticated at processing data we had done very
little toward processing images. It was my feeling that a lot of man
hours were spent in drafting and doing styling design in the auto industry.
The reason this effort got no place was due to a lack of a focus. I
later found out that at about the same time Elmer r Sharp who was in
DP was negotiating with General Motors for a Graphic Expression
Machine - GEM. This was exactly in line with the concept I had in
mind in connection with image processing. As it happened the GEM
program reported into me when I returned to Poughkeepsie.
There were at least two other technologies that I anticipated but did not
achieve success with. One was hydraulic logic. I tried to get Pat Panassidi
to stop thinking in terms of pistons and valves as components with which
to assemble a hydraulic system but instead think in terms of jets of air
that can be deflected by other jets of air. Unfortunately, Pat did not have
the background to develop this concept and I did not take the time to do so.
I had Jack Morton investigating the direct generation of microwave energy
in a single block of semiconductor. We were using silicon and a magnetic
field and were attempting to understand and improve on a mechanism, as
I recall, RCA first discovered. As we now know, the concept was sound
but it took Gunn and a different semiconductor to implement it.
As can be seen from this sample, the technical development work in ASDD,
at that time, was quite broad, highly speculative, but quite anticipatory
of future developments.
Mr. L. M. Saphire -12- March 30, 1970
Back to Poughkeepsie
At the end of 1960 I was promoted to Manager of Technical
Development of the Poughkeepsie Laboratory. In this position, I
had responsibility for memory development and product engineering,
machine technology, advanced technology, tape devices development
including Hyper and Tractor, communications products development,
and the GEM program. During this period a number of significant
products were announced or delivered as a result of the efforts of
these groups. Some that come to mind are: the 7302A air cooled
memory, Hyper, Tractor, an incremental tape drive, and GEM. Some
of these programs required my getting involved in technical details in
order to make the programs a success. Hyper and GEM are notable
example.
Hyper was a tape drive system that was far superior to anything
available at the time with the possible exception of Tractor. It took
me several months to convince Jim Weidenhammer that the drive could
not meet its start-stop specs. When he made the correct experiments
and was convinced I was right, I suggested a capstan was needed that
would absorb the energy stored in the tape. This is the fix that was
adopted by Jim and Ray Barbeau. I was instrumental in getting Jim to
adopt the phase encoding method of recording on the tape. In order
to reduce the cost of the drive, I spent a month going over the design
and was able to eliminate $l,000.00 of product cost without impacting
the reliability of the drive. Last but not least, I had to get personally
involved in a tape quality problem that held up a product test.
The GEM program was another difficult technical program. It was
further complicated by SDD having subcontracted the job to FSD Kingston.
In the Summer of 1961, I called an audit on the program and found my
worst fears to be correct. I took over the program from FSD and put
Ernest Newman in charge. To staff the group, I assigned many of the
people in the communications area and the incremental tape activity to
the program. It was a difficult technical program because it involved
exposing microfilm from a CRT and developing, fixing, and drying the
film in about 10 seconds. The film could then be viewed in a projector built
into the machine. The machine also had a CRT display console with a light
pencil. The system was checked out in Kingston prior to shipment to
General Motors. A system checkout in Kingston had not been planned in
Mr. L. M. Saphire -13- March 30, 1970
the original schedule. This caused a further slip in delivery date.
The machine was shipped early in 1963 and has been doing useful
work in G. M.
The 91 Program
In January 1963 Bob Evans and Carl Reynolds asked me to join the
systems areas as manager of Advanced System_ s Development. I had
responsibility for the Stretch machines in the field and made sure the
customers that had these machines were happy. A group led by Dick
May that was doing programming for NSA reported into me plus other
programmers such as John Carter who was improving the performance
of a FORTRAN I V for Stretch. Bob Meade had a small group of very
competent engineers designing the 608 that later became the 91. This
group worked closely with Bob Henle's group in CD who were developing
ACP.
A basic problem with the 608 program was that the systems effort in my
shop did not match the components effort in CD. I did not agree with the
late delivery schedule for the machine since I felt competition would force
an earlier schedule and if this did not happen then the component approach
would be obsolete by the time the machine was delivered. Toward the end
of 1963 the situation deteriorated rapidly when the technical approach that
CD had been taking was challenged by other engineers in SDD. CD had
designed their circuits around a tunnel diode which is a fast device but
which, when placed at the end of a transmission line, caused a "ping
pong" problem.
In retrospect, I would say my position at the time was somewhat ambiguous.
On the. one hand I had pushed very hard to get more funds into the systems
effort to accelerate the program. On the other hand, I permitted CD to
delay too long the dropping of the tunnel diode approach. In support of
my position to let CD continue to explore the tunnel diode I felt at that
time me that there would be no great difficulty to revert to resistive loads if
the tunnel diode approach did not buy anything. The rapidity with which
the resistive load approach was implemented, once the program was
properly supported, attests to the correctness of my feeling.
By the time the program was being given the support it required the
???otionall and political problems had gotten out of hand. As an example,
I found one of my recently acquired managers had made an organizational
Mr. L. M. Saphire -14- March 30, 1970
change without complying to my written request to review all such
changes with me prior to their implementation. This created two
personnel problems which is what I wanted to avoid by my request
to review proposed organizational changes. Since I found myself
in the position of not being able to control problems that I had antici-
pated, I asked Max Paley to relieve me of responsibility for the 608
program. The request was granted and I was transferred to CD.
Integrated Circuits
When I joined CD in January of 1964, I was very surprised to learn
that there was no work directed at developing integrated circuits.
Furthermore, there was an antagonism toward integrated circuits.
They were considered a threat to the SLT program rather than
complementary to it. With the exception of Leon Atwood, all of the
people in the Advanced Technology group that reported to me were
negative toward integrated circuits. I later learned that Bob Henle
made a presentation to the R&D Board in September 1963 in which
he took the position that integrated circuits could never compete with
a hybrid technology. All of this was very strange to me since these
people had always been at the forefront of technical development. I
found my conviction about integrated circuits was as welcome as a case
of the bubonic plague. Dr. Pietenpol, to whom I reported, asked me
as soon as I came on board to draw up a five year plan for Ad Tech.
Naturally, I gave heavy emphasis to developing integrated circuits.
My five year plan was not well received. I found myself in the position
of having to lay out a plan for the next generation technology NGT and
my management and my people completely negative to integrated circuits.
Bob Doxtator and then John Haanstra were asked to give an assessment
of the integrated circuit technology and its impact on SLT. The results
of these task forces are well known and I won't belabor them here.
The problem I faced was converting my people to the idea that integrated
circuits was the way to go and defining a packaging system that would
enhance their capabilities. By August of 1964, I had defined a four phase
program that would enable IBM to ship machines with integrated circuits
by 1969 or 1970. Phase one was essentially, the MST program as it stands
today. Phase two was to further exploit the SLT package but with a
tighter grid spacing. Phase three was what later was known as the NGT
program. It consisted of integrated circuit chips mounted on a multilayer
ceramic module and the module mounted to a board through an areal
connection of pins on the board and solder buckets on the module. The
board was to be wired either through the use of the printed circuit
technology or by discrete wires that were imbeded in a low melting point
alloy such as Wood's metal. Phase three was aimed at cost performance
Mr. L. M. Saphire -15- March 30, 1970
with Phase four directed toward high performance.
In addition to the above, I had another effort called MINT. This was
a benchmark program aimed at investigating the advantages and dis-
advantages of large scale integration LSI for small low cost machines
by using the ECHO keypunch as a test vehicle. This program was
directed by Bill McAnney and there was a total of about eight people in
his group. We had a contract with T.1. to supply two sets of nine
modules with about 120 circuits on one wafer on each module. A model
of the ECHO keypunch was completed and working by April of 1966. It
consisted of nine modules each containing 1.20 circuits on one wafer and
the nine modules were mounted on a printed circuit board along with
some special. circuits.
During the early part of 1964 T.I. people described to us their status
with integrated circuits. I asked them about monolithic memories and
they indicated that they could indeed build many memory cells on a chip.
I talked to Bob Meade about the use of monolithic registers in the 91
and conveyed to him the information that T.I. had given me. This inter-
action started what later became the SP-95 memory.
At the same time, I had Fred Buclow and Leon Atwood in Fred's group
set to work investigating the advantages of monolithic memories in con-
figurations other than shift registers. Leon's work led to the medium
speed local store, MSLS, which was used in the Mod 44.
In the Fall of 1964, Bob Henle decided on monolithic memories as his
direction of effort for his IBM Fellow program. Upon learing this, I
told Fred Buelow to discontinue further work on memories since to do so
would place his group in competition with Bob Henle and create unnecessary
duplication. I felt very strongly that monolithic memories would. be very
important in the future and in fact, informed Bob Evans of this. I happened
to see Bob Evans at the Poughkeepsie Airport at this time, and told him
that I felt monolithic memories would obsolete thin film memories. I
also told him that as a responsible manager, I was having my people dis-
continue the work I had gotten started on monolithic memories. My
explanation was that Bob Henle was a competent engineer and I could not
justify a duplicate effort even though we had started work in February or
March of 1969 before Henle became an IBM Fellow.
Mr. L. M. Saphire -16- March 30, 1970
During 1964 and 1965, my group worked on developing the capability
to do the NGT Program. It must be remembered that the Ad Tech
effort in CD had been aimed at ACP which later became ASLT. There
had been no effort on integrated circuits prior to the NGT Program.
The ACP Program did pioneer the multilayer board technology, and a
very small effort under Bernie Schwartz was exploring complex ceramic
structures. To a very large degree, therefore, the NGT Program had to
pick a direction and then set about developing support tools as well as the
technologies to be used without the advantage of being able to draw on
previous efforts.
The direction chosen was mentioned earlier and 1 would say the concept
is still sound. Let me outline it again:
1. Phase one of NGT became the MST Program.
2. Phase two of NOT was not done in the NGT Program
but was explored by Kurt Trampel's group in Poughkeepsie
and they call it NEXT.
3. Many ideas and technologies of phases three and four
of NGT were adopted by NLT-HP and the SNS Programs;
a) Multilayer ceramics
b) Capillary Refill
c) Multi-chip module
d) Areal connections between module and board
e) Solder bucket connections
f) Heaters imbedded in module
g) Discrete wires on a board imbedded in a
low melting point conductor.
Many support tools were developed in parallel with technology development.
I had a printed circuit generator modified to improve its precision so
that it could be used to generate masks two hundred times actual size.
These masks were then photo reduced for use on a step and repeat
camera. I had a group under Gary Gied develop and build integrated
circuit chip testers. There were other people developing design automa-
tion techniques for mask layout and developing test techniques for testing
complex integrated circuits. All this activity, while peripheral to the
main effort, was needed for a successful program.
Mr. L. M. Saphire -17- March 30, 1970
While the people in the semiconductor shop under Dave DeWitt,
Don Seraphim, and Ernie Van Derveer developed the integrated
circuit technology -- Rico DePietro who was in Owego, supplied
our first integrated circuit chips. These consisted of 3 emitter
coupled logic circuits per chip and the circuit delay was 5 nano-
seconds. Jerry West's group in Endicott mounted the chips on
multilayer ceramic modules made by Bernie Schwartz's group in Fish-
kill. The modules were designed and checked out by Brian Malbec
and his people.
One of the major problems we faced in 1964 and 1965 was getting experienced
people on the program. The cryogenic effort reported to me in early 1964
and was headed by Holly Caswell. I gave Holly the challenge, which he
accepted, of completing a working model of an eight by eight by eight
memory by July 1964 or if unsuccessful, terminating the program. They
were not successful, and I terminated the cryogenic program. The people in
this group were given choices of joining my other groups or joining the
semiconductor groups and most of them joined my other groups. These
were experienced people and they were badly needed. There was a great
risk involved in terminating the cryogenic program since GE, T.I. and
RCA still continued their work which might have been successful.
The people from the cryogenic program enabled me to staff a packaging
;coup, a test equipment group, and contribute to the staffing of a multi-
layer ceramic effort. By the end of 1964 the basic technology work was
vell underway. The semiconductor and. module work was still being done
or the program on a subcontract basis. The multilayer ceramic group was
nadequately staffed and I therefore had to depend on the subcontract
module group. This was very unsatisfactory because the subcontract group
ould only think in terms of variations on SLT. The semiconductor sub-
ontract efforts was somewhat better, but not much. As an example of
hat the NGT program was faced with, Ed Davis had Casper Barscia do
?? analysis which showed that the maximum level of integration desirable,
from an economic viewpoint, was 5 circuits per chip. This was in
February and March of 1955. Ed Davis publicized his findings without
checking with me. I found myself in the position of targeting my cost
performance program toward 10 to ZO circuits per chip and my MINT
program toward 120 circuits per chip. It must be remembered that I
was depending on Ed Davis and his people for my semiconductor and
module designs.
Mr. L. M. Saphire -18- March 30, 1970
I was forced to take the position that I had to have my own fully
staffed module effort if the NGT program was ever to develop and
demonstrate the route I felt we had to take. In the Summer of 1965,
we redistributed the manpower in Ed Davis' group, and I was able to set
up a multilayer ceramic pilot-line. We were then able to make progress
and turn out multilayer, multichip modules.
In December of 1965, Jerry Haddad headed a task force to review the
NGT program and decide what technology would be used to support the
next family of machines. The task force recommended that the corporation
go with what was first proposed as phase one of NGT. At this time, it
was called SLX. It later became MLT and then was called MST which is
its present name.
In January of 1966, I placed Jim Walsh in charge of the MLT program
and staffed his group with people from the NGT program. By June of
1 ?66 it became clear the NGT program could not survive in the environ-
ment in Fishkill and I joined Jerry Haddad's staff in Armonk.
Staff Director of the CTC
Alter assisting Jerry Haddad for about six months, I was promoted to
Staff Director of the Corporate Technical Committee and reported to
Mannie Piore. This was January 1967. At this time the Memory
Strategy was submitted for review by the CTC. I felt the strategy put
improper emphasis on ferrite cores and thin magnetic films and inadequate
emphasis on monolithic memories. To prove my contention, I went back
to Bob Henle to see where he stood after a little over two years study of
monolithic memories. By chance, he had just written a memo to file with
no carbon copy list in which he identified advantages of monolithic
memories. I used this to support my position that the Memory Strategy
was wrong. Mannie Piore concurred and we made it an issue with DP
Group. It took DP Group one year to study the situation and conclude
that the CTC was correct.
During 1967 1 drafted a procedures manual for the CTC. It defined the
content of the strategy documents and the procedures that were to be
followed for their approval. It also became clear that the CTC could not
limit itself to addressing only technical issues as it was charted to do.
I felt the CTC had to understand the product direction and the relationship
of the technologies to this direction. SDD was very emphatic in wanting the
CTC to stay out anything smacking of a product issue. During 1967, SDD also
adopted the concept of written strategies. Some of these strategies were
Mr. L. M. Saphire -19- March 30, 1970
???iquely SDD's, others were initially CTC strategies that SDD incorporated
?? part of their family of strategies. This created a certain amount of
confusion and tended to water down the effectiveness of the CTC strategies
cause of the dilution caused by the SDD strategies.
??? had a beneficial effect with respect to the CTC being exposed. to SDD's
product strategy. SDD was obviously more interested in a product strategy
than a technical strategy. They therefore spent more effort on defining
??? product strategy. At the same time, they did not want the CTC to
formally review the product strategy. After several meetings with SDD
people, I got them to agree to write their strategies in two parts; a
technical strategy and a product strategy. The CTC could review the
finical strategy with the benefit of having a product strategy to review
in parallel. It was agreed that the CTC would not raise issues with the
product strategies. This arrangement worked fairly well.
By my staff and myself reviewing the strategies and having a first hand
familiarity with the technical work in the laboratory that supported the
strategy, I was able to inform Dr. Piore of the weaknesses in the strategy
and issues that should be raised that may not have been raised by the CTC
members. I felt, and Dr. Piore agreed, that one main measure of my
success in the job of Staff Director of the CTC was how infrequently the
Corporation would experience technological surprises. There were no
technological surprises during my three years in this assignment. Mono-
lithic memories could have been a disasterous surprise but as I des-
cribed earlier, this was very effectively avoided.
The above briefly but accurately summarizes the highlights of my 19 years
with IBM and in the computer field. I would say my major early contribu-
tions involved pioneering in solid state logic circuit development which per-
mitted IBM to switch from tubes to transistors without a major disruption.
While I personally contributed in terms of patents, my main contribution
was in selecting the right people, training them, leading them down the
right technical paths, and challenging them to do more and better professional
work. As can be seen, there have been many times when I have selected,
at a very early slate, the right technology that ultimately survived. Con-
versely, I have been instrumental in terminating technologies that were
wrong to pursue further.
Mr. L. M. Saphire -20- March 30, 1970
During the last ten years, my contributions have been at a somewhat
higher management level. In addition, my assignments have been more
varied. They have involved peripheral equipment, systems, and to a
much lesser degree programmer supervision. My last three years
were spent looking after the technical health of the Corporation.
I hope my next fifteen years will be as interesting as my last nineteen
years. Lastly, I hope this will be of use to you, Larry.
[a signature]
J, C. Logue
JCL.mc:cpl .