THE COMPETITIVE SEMICONDUCTOR MANUFACTURING HUMAN
Second Interim Report
Clair Brown, Editor
11.2 The Development Process
The development team has to simultaneously fulfill several management
goals, including the development of a new product that satisfies
customers and is designed for low-cost, high-quality manufacturing.
Often management will use specific rules to help guide the development
process, such as a ceiling on the number of new fabrication steps
and new types of equipment (or conversely, the process must use
certain equipment). These types of specific rules governing constraints
on process steps or equipment are much easier to identify and implement
(as well as negotiate) than management of the creative process itself.
Here we will focus on the tension between encouraging individual
autonomy and creativity and controlling the direction of the development
process, including the use of teamwork for encouraging, evaluating,
and controlling individual ideas in the development process.
Planning for the Future while Building on the Past.
A companys technology strategy
is encompassed in what is commonly called a technology "roadmap,"
which identifies the key decision points of the design, development,
and production of new technologies over time. In general, the Japanese
companies have been more aware than American companies of the need
for this type of long range planning in the use of equipment and
technology over generations. However, USTech plans long-term use
of their fabs with three generations of technology per wafer size
before converting the wafer.
Strategy decisions proceed from the process chosen to the products
that can be manufactured with the process available. A single process
has to serve a broad range of products over the life of the process,
and this determines the depreciation period and the rate of return
to the investment made in the process. For example, a logic chip
may need low voltage for a computer and high voltage for an automobile,
and the engineers must be able early in the process to create design
rules that will accommodate both types of products. A new process
is developed every two to three years.
The roadmap covers the three major steps of the decision-making
process; new technology choice, development of the new technology,
and use of the new technology in high volume manufacturing of a
product. However, these steps are interrelated and must be thought
of in terms of function rather than organization. For example, a
product such as 16-meg DRAM is first defined, and then the engineers
ask what are the details of the process flow available in order
to decide on the general features and the design rules for the DRAM.
At both JapanTech and USTech, step one (technology choice) is made
at the highest management level, and the required research is done
by the central research lab. The development team is then given
the definition of the new process characteristics and the extent
to which the new process must adhere to the design rules of existing
technology. In step two, the key development changes that need to
be made to the existing technology, both to minimize costs and risks
of the new technology and to maximize the overlap of the new technology
with the subsequent technology, must be identified. The new process
flow will be comprised of modules that are modified to varying degrees--some
already exist with little modification; others are new but use existing
equipment; and others are new and require new equipment and materials.
The development stage focuses on the characterization, optimization,
and integration of new modules. The new or modified modules are
developed and are integrated into a single process flow. In steps
one and two, equipment is chosen. In step three, the new process
flow is transferred to manufacturing, and high volume production
is begun. The level of characterization of the process and the volume
of manufacturing achieved before transfer to manufacturing can vary
a great deal. A critical decision is to determine when the transfer
from development to manufacturing begins. This boils down to a decision
of where it is most effective in terms of time and costs to do many
of the development activities. Sufficient lead time is necessary
to allow purchase and installation of the equipment and provide
the fab space needed for production. Later in the fabrication process,
the decision must also be made when to transfer the ownership of
the technology to manufacturing. During steps two and three, exchange
of engineers occurs between the manufacturing and development facilities.
Encouraging, Evaluating, and Controlling Ideas
Issues of creativity and control arise
throughout the development process. Some of the relevant decisions
how well-defined the characteristics
of the process are and how well-defined the constraints are
(e.g., equipment that can be used) in the assignment made to
the development team;
how the "winning" technologies
are chosen and assigned to team members;
how the ideas generated by the team
members are trimmed down to a manageable number and the members
rewarded for their contributions;
how work assignments are made over
time as development proceeds, and how much autonomy the members
and the team have in making decisions;
what sources the engineers use for
generating and developing ideas, such as referring back to previous
processes, reading theoretical literature, brainstorming in
what sources the engineers use for
information and feedback, including scientists at the central
research labs, team members, engineers at the receiving fab,
marketing experts or customers, engineers at other companies,
or publicly-available technical information.
The team leader has to foster ideas that conform to the development
guidelines and has to choose among the many ideas put forth by the
team members, while ensuring that the members feel their contributions
are valued (even if not adopted) and rewarded.
We now turn to how USTech and JapanTech manage innovation in the
11.3 The Development Process
At USTech a strategic planning committee defines new lead products
by marketing segment. A strategic capabilities committee ensures
that capabilities cut across business segments and lays out process
capability over time, which defines the technology roadmap. The
goals of the development team are set by the strategic planning
committee, which annually updates the technology roadmap and the
lead product for each new technology. Since any process change involves
a lot of time and resources, a process change must be big enough
to make a significant difference in the products manufactured. The
planning committee makes the selection of architecture and gives
only one option with design rules to the development team. The design
of circuit and block architecture with layout is done at the research
Since the new technology is not compatible with the available process,
the development teams goal is to build a prototype of the
product that is commercially feasible. In order to differentiate
between problems caused by modifications of the process versus the
product, a new design is brought up on the old process in small
volume, and then the first shrink is usually made on the new process.
This minimizes debugging problems since it decouples the new design
from new technology (process). The development manager issues design
rules based on algorithms (e.g., 70% of former size), and the design
of the first shrink of the product is done in the development fab.
Next, the product is brought up on a new process at high volume.
The development rules issued provide very simple "rules of
thumb"; for example, 30% of the steps are new (70% are old),
and the equipment associated with any given module lasts two generations
(i.e., through two different recipes).
Learning across development teams is facilitated by keeping archival
records of project decisions throughout a program, and performing
a "postmortem" after a project is completed to see what
can be done better next time.
This structure of development activities reflects the companys
strategy of remaining the product market leader. Time to market
is an important consideration and short-run financial considerations
are not dominant, so much of the work traditionally done at the
manufacturing fab, including the ramping to volume production, is
done at the development fab at USTech. The transfer rule is to freeze
the technology and transfer it to the manufacturing fab when the
development fab has no fundamental yield or manufacturing problems
at 2000 wafers per week. This level of ramping to volume is undertaken
in the development fab since it reinforces USTechs rule that
manufacturing engineers must copy the process exactly and not make
changes. If the development fab can produce 2000 wafers per week
rather than the more traditional 300 wafers, then volume-related
problems will have already been solved and fab engineers become
"believers" in the process. At the time of transfer, the
yield goal is to be within 1% of the development fab yield as the
manufacturing fab ramps up to 6000 to 7000 wafer starts per week.
By ramping to volume at the development fab, the development engineer
is forced to learn and use manufacturing skills. This encourages
"developing for manufacturing" (i.e., developing in a
way that minimizes manufacturing problems and costs) since the development
engineer experiences the problems at high volume and does volume-related
In order for the manufacturing engineers to have ownership of the
new process and to make use of their manufacturing skills, they
are transferred to the development fab up to eighteen months before
the transfer. Their role is to set manufacturing goals, define optimized
recipes, set up maintenance procedures, transfer deliverables (documentation,
knowledge), undergo training, and help characterize the process.
Although they help in the qualification of the product and the transfer
process, their primary tasks are to be trained to run the new process
and to manufacture the chips at the development fab.
Work Assignments and Knowledge Creation
The management style at USTech is distinguished
by the cultivation of peer pressure, aggressive discussion of ideas,
delegation of responsibility to young engineers, and job assignment
along with stock options as rewards. In general, job assignment
reflects the degree of complexity of the task and the experience
of the person. A junior engineer is assigned an open-ended but less
complex problem. A senior engineer has a more detailed assignment
with a complex problem. USTech believes that lack of experience
is beneficial in research and development, since an inexperienced
engineer does not have preconceived notions of what works or doesnt
work. For this reason, young engineers are given major responsibilities
in developing new processes or technologies. Managers described
new graduates as providing a role model for older workers, since
they are hard working and have are creative in their problem solving.
Assignment of the correct people to a project is critical. Engineers
are rewarded for excellent performance by their next job assignment,
and this provides a major incentive to do well. Job assignments
determine compensation, which is strongly performance based. Those
who do not do well are assigned tasks with a lower level of responsibility.
Only in rare instances do engineers ask for more challenging work
or projects, and, in response, they are told they need to perform
well on simple tasks first. Engineers reported enormous peer pressure;
"sitting back is not condoned." Some engineers, who cannot
make the required breakthroughs, burn out and quit. However, USTech
restructures assignments when projects are not going well. One example
was given of a key engineer who "blew it" on a huge project,
so the project was split into two parts, core and support, and he
Conflicts about ideas are resolved through confrontation, or a process
of "disagree and commit". If a technology being developed
is dropped or not integrated, younger module engineers may be discouraged.
However, in a conflict about which idea to use, the boss may ask
the dissident to run the project. Engineers typically do not withhold
their ideas in the development process, which is problem driven.
This is more of a problem in research, which is idea driven.
USTech relies on decision-making through committees, and ones
influence in the company is determined by how many standing committees
one is on. Everyone is part of a group, and engineers have a bi-weekly
one-on-one meeting with their group leader at which they can bring
up problems. In addition, focus teams meet weekly, and a coordination
team of 20 people from integration and modules sets assignments
and goals for focus teams.
USTech is typical of U.S. semiconductor companies in its reliance
on teams making the decisions that structure work. The team decisions
are supplemented by individual decisions and have limited managerial
input (Chart 11.1A). Seven-tenths of the
engineers reported their teams assign tasks, prioritize tasks and
goals, set deadlines and schedules, and define technical problems
or objectives. Six-tenths reported their teams write project specifications,
and three in ten reported individuals write project specifications
alone. All of the engineers reported that individuals by themselves
or with the team prioritize tasks, and nine-tenths reported that
individuals by themselves or with the team define technical problems.
Managers greatest involvement, reported by two-tenths of the
engineers, was setting deadlines and schedules.
USTech has a formal methodology for learning new knowledge:
1. Search the literature. Engineers
think that papers from universities are better than company papers.
However, Sematech has increased the quality and number of industry
2. Compare methods used across equipment sets (e.g., read a report
on defect mechanisms on thin film and see if it works on etch.)
However, USTech engineers reported that
the technical literature is not very helpful "because if anyone
is doing (a solution to the problem under consideration), they are
not sharing it; if they are sharing it, they are doing it poorly."
The engineers often find industry papers misleading since a paper
may show only the best picture and not report other cases. The partial
reporting is hard to evaluate, and the process usually doesnt
work as well as reported. Normally, USTech engineers do not publish
anything outside the company, since publishing is not important
for career advancement within the company or for professional reputation.
USTech engineers are very proud to be working for a technology leader,
and they share in the cachet of the companys name.
The engineers at USTech are very clear that any information sharing
with outsiders has to be done on a quid pro quo basis. Employees
are aware of the risk of information transfer because they have
taken classes on internal document control and are familiar with
past cases where intellectual property was transferred illegally.
Even within USTech, engineers need to have a reason to request company
papers. In general, they do not rely upon the exchange of knowledge
with outsiders because "outsiders have nothing to share."
The company policy is not to answer questions from outsiders, since
USTech receives nothing in return. However, some engineers indicated
that customers and colleagues at other semiconductor and equipment
companies are important sources of technical information.
USTech reported that the biggest challenge in knowledge creation
relates to equipment since the machines needed for the new process
do not exist. USTech engineers do not normally share knowledge with
vendors, who want to know USTechs recipes, because there is
not the possibility of reciprocity. The need to protect their technology
requires USTech to use their own resources to develop equipment
specifications and modifications. On selected machine modifications
that improve the process or die yields, USTech will protect the
knowledge by signing non-disclosure agreements with the vendor and
excluding the vendor from the development process. Other equipment
improvements are not protected, however, since USTech wants them
to become part of the standard machine.
Control over ideas is further ensured by using the patent process
defensively. There has been a renewed emphasis on patenting at USTech
over the past two years, especially in the process architecture
area. Engineers are paid $2000 for filing a patent. The rule of
thumb is to keep knowledge a trade secret, but to patent whatever
can be reverse engineered. For this reason, process integration
does not need to patented.
USTech is typical of U.S. semiconductor companies in its engineers
relying on company colleagues and on journals and not relying on
patents for technical information (Chart 11.2A).
Unlike at USTech, however, engineers at other U.S. semiconductor
companies report knowledge sharing with outsiders through conferences
or personal contacts. Seven-tenths of the engineers reported conferences,
six-tenths reported colleagues at other companies, and one out of
two reported equipment vendors or material suppliers as very important
information sources (i.e., 6 or 7 on a scale of 1-7). USTech engineers
seem to feel rewarded for their creative achievements through their
national reputation of working for USTech without engaging in the
process of presenting papers at conferences.
Evaluation and Incentive Systems
USTech pays their engineers based on
individual performance. Usefulness of ideas is the main factor in
the annual performance review, which includes a qualitative summary
of accomplishments and the employees strengths and weaknesses.
Performance is evaluated on a relative basis by ranking engineers
on an equity curve by quartiles. An engineer is ranked within a
group of 10 to 20 engineers in his or her own grade range. They
are told their relative ranking (e.g., top one-third) and given
a rating of superior/outstanding (15%), successful (80%), or needs
improvement (0-5%). Each person is also told if their performance
trend is greater or less than their peers. The evaluation can affect
pay within the rank group up to 10% (e.g., those with superior receiving
a 6% pay increase, with good receiving 4%, etc.) plus promotions
and responsibility on the next job. Everyone is considered well
paid, and they do not know each others rankings.
Compensation can include profit-sharing and stock options as well
as salary. Exempt and nonexempt employees have the same cash bonus
system, which is based on six month company profits (e.g., 5 to
7 days of pay). Higher grade levels (i.e., beginning with the grade
level at which PhDs enter) receive an executive bonus (percent of
base pay), which varies by position and payoff rate. Stock options
are paid for future potential and are an important incentive mechanism.
Some engineers earn more from their stock options than from their
salary. All grades are eligible for stock options but with different
participation percentages. For engineers, the bottom 10-15% do not
receive stock options; the middle 70-75% receive some options; the
top 15% are identified as key players and receive more options.
USTech is typical of U.S. semiconductor companies by including profit-sharing
(reported by seven-tenths of the engineers in the larger sample),
and stock options (reported by six-tenths). However, performance-based
pay was less prevalent in other companies; only half of the engineers
reported individual performance pay and two-fifths reported team
Promotion is also used to reward outstanding performance at USTech.
Like the large Japanese electronics companies, USTech develops its
engineers and many rise within the company. Engineers can be promoted
to group leader, whose job is still highly technical with supervision
of approximately six engineers and a few techs. The group leader
must have been at USTech at least four years. Competition for group
leader is keen, and those passed over are miffed and ask "Why
him and not me?" The group leader reports to an area manager,
who has been at USTech 10 to 19 years and supervises 50 to 150 people.
The area manager is the big step into management, since the job
is less technical and requires more leadership and administration.
Senior engineers who do not go into management have a parallel path
and report directly to an area manager. Less than 20% of engineers
are on this track. So far, USTechs growth rate has allowed
a satisfactory rate of promotion, but this might become more difficult
to achieve as the company matures.
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