IEEE Solid-States Circuits Magazine - Winter 2022 - 43

the number of comparators. Chuck
Lane of Analog Devices had published
the interpolation concept in
1989 [4] and recognized the benefits
to
differential nonlinearity.
For practical purposes, the two
were contemporaneous in recognizing
the benefit
to DNL. However,
Matsuzawa would stay with
interpolated flash over his career
at Matsushita Electric, migrating
the principle to CMOS. Very likely,
he took the technique farther than
anyone else because video-rate
ADCs being developed elsewhere
were gradually converging to the
pipeline architecture.
While working at Matsushita Electric,
a vertically integrated concern
that designed and fabricated its own
ICs, recording heads, magnetic tape,
packaging, and battery for camcorders
and marketed the product under
its own brand, Matsuzawa must have
learned a great deal about the total
system integration in a complex consumer
product. When the designer
of chips that must fit into small
form factors and operate in powerconstrained
environments participates
in the development of the final
product, he or she gains a much
richer perspective that guides tradeoffs
in circuit design. Engineers
at merchant semiconductor companies,
such as TRW LSI Products and
Analog Devices, could not enjoy
that perspective.
Also, for these reasons, early on,
Matsuzawa understood the need to
migrate flash ADCs to CMOS. There,
he would translate interpolation
with resistors in bipolar flash A/D
into interpolation with capacitors [5].
He adapted offset-cancelled CMOS
inverters as preamplifiers, a classic
CMOS technique from those days.
With all of these methods, a 10-b
A/D clocked at 20 MS/s showed
an 8× lowering of power in a CMOS
realization. This is not an incremental
advance: it was dramatic. It
would enable practical HDTV consumer
equipment.
To tackle the rising complexity
of equipment at declining cost,
Matsuzawa was in a unique position
to confront the design problems
of mixed-signal systems on
chip (SoCs): indeed, he might have
been one of the strongest advocates
at Matsushita for integrating entire
systems. However, Japan was not as
far along in its experience in developing
SoCs as the United States,
where companies like Broadcom
were leading the mixed-signal SoC
revolution for consumer applications,
such as digital cable modems.
Nor was there the access in Japan to
the latest CAD tools and frameworks
suitable for complex systems that
allowed for the simulation of analog
circuits with logic.
Therefore, Matsuzawa and his colleagues
developed their own methodologies
and improvised tools to
enable the development of SoCs for
digital video cameras and, later,
DVD players. For handheld devices,
power consciousness was extremely
important. One can trace this in
early efforts, such as one incorporating
analog circuits-based image
compression on the same chip as
an image sensor [6]. As technology
scaled, later chips would gradually
become mostly digital, integrated
with ADCs. When CMOS imagers
would later supplant CCD imagers,
an ADC would be integrated on the
imager chip at the foot of every column
of pixels.
From these development efforts,
Matsuzawa attempted to abstract
the principles of low-power mixedsignal
system design. While others
in the United States and Korea were
acquiring comparable experience, he
was one of the few to publish it. This
is revealed in a succession of papers
on low-power system design [7]-[9].
Perhaps a little quaint when judged
by today's state of knowledge, these
papers discuss device- and circuitlevel
principles to achieve low power
as well as projections based on the
then-semiconductor-scaling
road
maps of relative power consumption
and chip area of analog, logic,
and memory. Matsuzawa [9] makes
a case for system-in-a-package (SiP)
solutions using different chip technologies
for radio frequency/analog,
logic, and memory. The SoC-versusSiP
debate continues today.
At the culmination of his career
at Matsushita, Matsuzawa led the
development of the most ambitious
mixed-signal SoC undertaken at
that company, possibly in all Japan.
This was the first single-chip DVD
player, announced in 2003 [10]. It
integrated one digital signal processor,
two 32-b CPUs, three dedicated
processing units, a partial-response
maximum likelihood read channel
with an analog front end, and many
other subsystems on the same die.
It contained 24 million field-effect
transistors. I found it remarkable
then how this was designed and put
into production with very little prior
experience or, indeed, even access
to the extremely sophisticated CAD
software that is necessary to manage
a chip of this scale.
Matsuzawa has been an energetic
teacher throughout his career.
Even when employed at Matsushita,
in his spare time he developed
thousands of slides that he used in
short courses offered to working
engineers in evenings or over weekends.
There was, in him, an unusual
impulse to teach others, reaching
well beyond the junior engineers at
his own company.
It is not surprising, then, that, on
taking retirement from Matsushita,
he embarked on a second career in
academia by joining the faculty at
the Tokyo Institute of Technology.
This is one of Japan's top engineering
and science universities with a
very impressive history of influential
IEEE SOLID-STATE CIRCUITS MAGAZINE WINTER 2022
43
Matsuzawa would stay with interpolated flash
over his career at Matsushita Electric, migrating
the principle to CMOS.
IEEE Solid-States Circuits Magazine - Winter 2022

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Winter 2022
