THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN
Second Interim Report
Clair Brown, Editor
11.2 The Development Process
At JapanTech, development of the business plan, which includes discussions
from users on what will sell as well as production requirements,
is an important process. The plan for memory is sent to the R&D
division annually. The development process is conducted for two
years by the central research lab, which undertakes research for
all the divisions of the company (including semiconductor) and then
by the development team for two years before it is handed off to
the memory division. Central lab engineers produce fundamental research
reports and then develop process architecture and modules (circuit
design) based upon this research. Until last year, they also developed
the engineering sample.
At JapanTech, the development and fabrication process are divided
into three stages: research, development, and fabrication of the
engineering sample. The development lab does the prototype development
(including the study of new technologies, integration, and fabrication
of a prototype.) A cooperative relationship exists between the central
research lab and the development lab. Since the semiconductor division
commissions research on future technology, it is in charge of the
research plan, and funds and receives the research results. The
development lab also does the engineering sample development (i.e.,
establish processes, confirms characteristics, completes and ships
engineering samples to the engineering division). The engineering
sample is then handed off to the business division, which makes
the commercial sample, which is then transferred to the fab. The
hand off to the fab is made at an early stage (e.g., 5 good chips
per wafer). In a recent transfer, they needed over 1000 wafers to
establish reliability. The biggest challenge is transferring technical
know-how and understanding where equipment bottlenecks will be.
The number and complexity of the modifications made decrease as
the development progresses (e.g., in a recent generation, five to
ten mask changes were made in the engineering sample, fewer than
five changes were made in the commercial sample, and three to four
minor mask changes were made at the first receiving fab.)
Daily technical reports during development of the prototype device
total 2000 pages per device; there are far fewer reports during
the engineering sample stage. In contrast, meetings for prompt exchange
of information between engineers becomes more important the further
they are into production. Process engineers communicate with production
engineers weekly in the early stage and then daily in the last stage.
JapanTech estimates that each stage costs approximately the same
amount. Since development resources are limited, the strategic decision
of what device to concentrate on is important. JapanTech is continually
adjusting resources to research, development, and production.
Participation with equipment vendors appears controlled and contingent.
When the product requires specialized nonstandard inputs from the
supplier, or when the equipment is being tailored to the specific
needs of JapanTech, participation with the equipment vendor is extensive.
When these conditions are absent, there is relatively little participation.
Once purchased, equipment performance is continuously monitored.
Since equipment is not restricted to specific devices, one of the
first decisions made is what equipment on hand would work for the
device being developed and what equipment needs to be purchased.
For the existing equipment chosen for continued use, production
engineers make the equipment modifications for the new device. These
modification projects give the equipment engineers the opportunity
to work on a creative project.
JapanTech engineers have discussions with equipment manufacturers
on how to reduce price, on what machine options to order, and on
required safety requirements for machines, which differ by prefecture.
Today they are more frank in discussions with equipment vendors
than previously, since the "cost of hiding information is perceived
to be too high". JapanTechs openness with the equipment
vendors (in contrast to USTech) reflects the widespread use of similar
equipment throughout the memory sector and the decline in the price
of equipment as modifications become standardized.
JapanTech has introduced a "total concept" approach to
improve communications between engineers; design, process, and product
engineers must work together to produce total concept through meetings
and rotation. Discussions with equipment and materials manufacturers
are also integrated into process. This is to structure more of a
focus on innovation in product design, since the days of price competition
and incremental improvements have been replaced by emphasis on innovative
products. The "total concept" approach includes six months
of education on the production line for junior development engineers
before they specialize in engineering tasks. Half of new hires at
the development lab have spent six months on the production line,
and one-fifth of all development engineers have spent six months
on the line.
Work Assignments and Knowledge Creation
The project leader assigns work and makes
schedules according to abilities. Ideas are generated about technologies
by team members. When a new idea is created by an individual, verification
and evaluation of the idea is performed through fabrication by the
test team, which decides the best choice for the new technology.
In the final stage, however, there is no time to try ideas by making
samples, so the leader chooses among competing ideas. The person
whose idea is rejected is usually discouraged for a short time.
The project leader tries to pick an idea as soon as possible and
rejects other ideas right away, even if they are equally good. He
will consider using the rejected idea in the next generation. Usually
only small problems have many ideas for a solution; conflicts among
important ideas do not occur frequently.
At the research and development stages, brainstorming meetings are
held 2-3 times per month. With the chief engineers (kacho)
help, engineers write up technical reports and patent applications,
and the kacho distributes them.
In the development of the device we studied, only two major technologies
were changed. To develop the two technologies, a team was appointed
that consisted of the key person (shunin), plus five engineers for
process development, plus three engineers for circuit design. The
shunin learns about basic information from central lab members and
about customer information from the memory division. The shunins
supervisor (kacho) decides the main job targets for the shunin,
who decides job assignments for members. The shunin also does a
lot of training for the team.
JapanTech is typical of Japanese semiconductor companies by relying
on the team (reported by seven-tenths of the engineers) with the
supervisors input (one-fourth) for assigning job tasks, and
by relying on the team to prioritize tasks and define technical
problems (each reported by nine-tenths of the engineers). (Chart
11.1B.) Project specifications are usually written by the team
(reported by eight-tenths of the engineers) as well as by individuals
working alone (reported by six-tenths). Individuals are also heavily
involved in setting deadlines on their own (half) as well as with
the team (seven-tenths) and the supervisor (two-tens). Compared
to the U.S., Japanese managers appear to be more involved in assigning
tasks and less involved in setting deadlines. Teams appear to be
more involved, and individuals less involved, in prioritizing tasks
in Japan, and individuals are more involved in setting deadlines
On-the-job development of skill level for engineers is intended
to include development of originality and creativity. Job rotation
is a major mechanism for skill development. The goal is to rotate
the jobs of the group members yearly across devices and stage (e.g.,
year 1 in research on two generations out; year 2 in development
of the next generation; year 3 fabrication of the engineering sample
with 80% of time on the fab line of current generation device).
For example, one group had 10-14 team members in development, 2-3
members (with none next year) in fabrication, and 4-5 members (growing
to 5-7 next year) in research.
Japanese engineers would rather do research and development work
than fabrication work, which is hard and hands on ("for an
operator and not suitable for self"), and the breakthroughs
and patents are mostly in research and development. However, the
company believes that fabrication work has important educational
aspects so engineers are required to work on fabrication.
JapanTech also uses the transfer of engineers to transfer knowledge.
This is part of their policy of periodic job rotation between R&D,
the business division, and the fabs. At the time of transfer, development
engineers are sent to the manufacturing fab for one month. However,
one to two engineers (process and design) are sent for two years
from the production fab to the business unit fabricating the commercial
sample. A number of engineers are sent to the volume fab for the
hand-off. In the latest generation, the hand-off to the second high-volume
fab six months later needed only one-tenth as many engineers as
the original fab.
JapanTech recently implemented specific improvements in the transfer
process; a 5% rotation between product and process engineering divisions
and tracking of data for critical points in each process by product
line (matrix of process by product line).
The source of ideas, which reflect the significance of learning
from past experience, include (in order of importance):
1. journals, literature
2. previous JapanTech reports
3. customers: memory division and marketing engineers
1. previous JapanTech reports
2. problem information from product lines (particles, yields,
1. discussion in team meetings; development
engineer in team provides support.
Both Japanese and American engineers
ranked colleagues within their company (by eight out of ten and
nine out of ten, respectively) and journals and books (by six out
of ten and five out of ten, respectively) as important/very important
sources of technical information (Chart 11.2B).
Conference presentations were ranked as important/very important
by seven out of ten Japanese and American engineers. Patents were
ranked important by one half of the Japanese engineers, and engineers
at other companies were ranked important by six out of ten U.S.
engineers. U.S. engineers were more likely than Japanese engineers
to rank outsiders (i.e., engineers at other companies, equipment
vendors and material suppliers) as important sources of technical
information. Japanese engineers were more likely than U.S. engineers
to use their team procedures and findings from previous projects
(two times and .8 times per week, respectively) and both were less
likely to use company documents on previous projects (1.1 times
per week for Japanese engineers and .7 times for U.S. engineers).
For JapanTechs development engineers, an important goal is
to be one of the first companies to present a paper on the next
generation memory device, which marks its public debut at the ISSCC
international conference. The companies already know about individual
processes before the presentations at the ISSCC conference, where
they learn what the other companies are emphasizing and their general
direction of development. However, companies do not present information
about process integration at the ISSCC conference. Overall, the
ISSCC is important for motivating engineers as well as for marketing
Compensation, Evaluation, and Incentive Systems
Like all large Japanese companies, a
complex system is used to determine monthly salary, which is a combination
of age pay plus grade and position pay. Until they reach management,
which typically takes twelve or so years, engineers are in the company
union. During this period, earnings are fairly rigidly set with
only minor variations among engineers with the same education and
tenure. Monthly salary and bonuses are negotiated with the union,
and annual increases are determined by national wage negotiations
in the spring (Shunto, or annual spring offensive.) Twice yearly,
employees are paid a bonus (annual average of 5 months), with performance
accounting for up to 25% of the bonus. Performance pay amounts to
only a small fraction of regular salary with recognition being more
important than money. However, performance affects ones rate
of promotion, especially to management, which begins with the position
of kacho. All group leaders now become a kacho. However, with the
aging of the work force, JapanTech is running into the problem of
too many shunin, so more engineers must become technical specialists
(tanta kacho). Currently, JapanTech is experimenting at the development
lab with paying an annual salary based more on performance and less
on experience and grade.
At JapanTech, an annual goal is 3 to 4 patents per engineer. About
Y5000 is paid at the time of patent application and about Y10,000
is paid if accepted (license). In the infrequent case of a patent
being used in the product line, a much bigger reward is paid (Y1,000,000).
Patents in the development lab are mostly for process (80%, of which
70% is process only and 30% is equipment only) with the remaining
20% for circuit design. All equipment patents are connected to process
Like other large Japanese companies, JapanTech relies on company-based
education for the development of its young engineers, who are hired
after receiving a BS. Unlike their American counterparts in R&D,
Japanese engineers usually do not have masters or doctoral degrees
when they go to work for a company. They will often receive an advanced
degree from their company or from an affiliation with a university
while working for a company. JapanTechs development lab has
only 2 PhDs, one from a university and one from the company.
Go to Next Section
of this Chapter
Go to Table of Contents for this
Go to Table of Contents for the
CSM-HR Interim Report