| |
 |
|
Cycle Time
Management Styles
There are many factors that affect cycle time in wafer
fabs: utilization, lot size, hot lots, tool dedication,
variability, and so on. Understanding these issues is key
to improving wafer fab cycle time. FabTime’s cycle
time management newsletter has discussed many of these
issues in detail. Some of these articles have left people
thinking “OK, I understand that utilization and
variability and tool dedication and all these things
influence fab cycle times, but what do I do about
it?” In Issue 3.06 of FabTime’s newsletter, we propose three distinct cycle
time management styles, and describe how each can be used
to improve cycle time under certain conditions. We have
named these three styles: The Traffic Cop; The Shepherd;
and The Relay Coach.
The Traffic Cop approach involves managing starts to
control utilization and thus keep cycle times lower for
all lots. The Shepherd approach involves managing inactive
lots, to reduce variability in the fab, and thus reduce
both cycle times and cycle time variability. The Relay
Coach approach involves focusing close attention on a
small number of hot lots. This gets those particular lots
through with very low cycle times, though at the expense
of other lots. These approaches all require access to
accurate, real-time data about fab behavior, along with
management authority to make recommendations or changes.
These approaches can be used together (with some
trade-offs required), and can be used in conjunction with
other cycle time improvement efforts such as variability
reduction programs. To read the full article, please subscribe
to FabTime’s cycle time management newsletter.
What
follows here is a series of graphical examples, using standard charts from FabTime’s
cycle time management software. These are
examples of the kinds of analysis we have seen people go
through when applying the cycle time management styles. We
have added many of these charts to our FabTime software
specifically to facilitate this type of analysis.
All of the sample charts are taken from the same set of
demo data. The demo was derived from simulated output
using an adapted version of one of the SEMATECH testbed
datasets. The SEMATECH datasets were derived from actual
wafer fab data, scrubbed to maintain confidentiality. Move
transactions for the demo were generated using the Factory
Explorer® simulation engine, distributed by Wright Williams &
Kelly. |
