IEEE Solid-State Circuits Magazine - Spring 2017 - 20

John had placed the egg in the nest
with vision, technical skill, and
responsiveness, but with little ego
and without an "it's mine" attitude.
hatchling is theirs, they work feverishly to feed it until it leaves the nest
and thrives on its own.
Elsewhere in the issue, James
Boddie gives an excellent explanation of how the single chip digital
signal processor (DSP) work began
in Bell Laboratories in the mid1970s. We had just completed a first
phase of an 8-bit microprocessor
that lent itself to being programmed
in high-level languages (C was ours)
rather than assembly language. In
short, C compilers compiled better
code if there were ample registers
in a CPU, so we found an architectural way to do that in a world
where area on a silicon device was
very tightly budgeted. At the time,
we lived in a world of 4.5-μ metaloxide semiconductors (MOS) (imagine transistors and conductors as
wide as a two-lane highway), not
today's 10- and 14-nm MOS (imagine that highway has now become
the width of a quarter). We then
wondered if, in that same silicon
world and in an increasingly digital
communications network, we could
use a stored program device, not
hardwired logic, to perform realtime signal processing functions
on a single chip. There were several hurdles, and power consumption and realizable processor speed
were among the most significant
ones. Most experts with whom I
spoke were quite sure it would be at
least five years, counting on Moore's
law (which had been suggested by
Gordon Moore about a decade earlier), before a single chip DSP could
be built. There didn't seem to be any
fruitful paths to pursue. Then I met
John Thompson.
John was, and I'm sure still is,
soft-spoken. Yet he spoke with both
clarity of thought and vision as well

S p r i n g 2 0 17

as a sense of practicality. Over a few
cups of coffee, and, as I recall, a
longer session at a conference that
we attended together, he explained
his vision. Most signal processing
functions we wanted to handle had
an algorithmic regularity that lent
themselves to a pipelined process
for the computation we typically
would want to complete in 125 μs.
He was confident that a four-stage,
pipelined architecture would work
for most applications. Based on
John's ideas, we formed what we
called a "study group" and invited a
number of technical people in Bell
Labs (researchers, silicon experts,
and system designers) to participate outside of their regular job
duties. This group had not just an
understanding of signal processing;
many were responsible for designing products, like data sets (going
higher than 1,200 b/s in an inexpensive unit was a tough problem at the
time), tone generators and receivers, echo cancellers, and so on.
In the study group, John ex plained how he thought the processor should operate. Others would
explain how it would have to operate to actually be a part of a useful design in their own work. The
best ideas were incorporated. The
design evolved in a process similar
to ones in which many of you have
participated. John remained in control of the architecture, but he did
so in a thoughtful and responsive
way. It may have been John who
taught me that the most important
communication skill for managing
is listening.
Toward the end, it wasn't solely
John's architecture and approach. It
belonged to many members of the
group. John had placed the egg in
the nest with vision, technical skill,

IEEE SOLID-STATE CIRCUITS MAGAZINE

and responsiveness, but with little
ego and without an "it's mine" attitude. Many of us became adoptive
parents of the hatchling.
John continued to work as part of
the system design group, playing a
major role in its success. The group
had many extraordinarily talented
people. While many problems arose
and many improvements were made,
John's fundamental notion of the
four-stage, pipelined architecture
remained and was the cornerstone
of how the device was built and the
signal processing algorithms were
programmed. His thoughtful, egoless style pervaded the group. Many
more became adoptive parents during the development.
On a personal note, mainly because of the caliber and nature of
people in both the system design
and silicon design, the insight
and thoughtfulness of our management, a nd t he env ir on ment in
which we worked, the DSP effort
was a "Camelot" during my career.
It wouldn't have happened if John
T hompson hadn't brought the
round table.

About the Author
Dan Stanzione (dstanzione@gmail
.com) is president emeritus of Bell
Laboratories. He received the B.S.
degree in electrical engineering in
1967, the M.S. degree in environmental systems engineering in
1968, and the Ph.D. degree in electrical and computer engineering in
1972, all from Clemson University.
He is a Fellow of the IEEE; has published papers on computer simulation, microprocessors, and software
design; and has been granted four
patents. He is on the board of directors for Quest Diagnostics and Internap and has served as the board
chair for both companies. He also
ser ves on the boards of private
companies and on the Clemson
Foundation Board. He and his wife,
Lisa, have five children and live in
Stuart, Florida.

Table of Contents for the Digital Edition of IEEE Solid-State Circuits Magazine - Spring 2017