THE COURSE / M12 THE FAB AS A SYSTEM / M12.6
Keeping it running
After this lesson you can:
- Say where a tool's hours go, and work out availability from MTBF and MTTR.
- Work out a tool's OEE, and say why a fab never runs every tool flat out.
- Use Little's law to link work in progress, throughput and cycle time.
Loading the lesson…
Sources
- FabTime Cycle Time Management Newsletter, Vol. 2 No. 4, "In-depth guide to OEE resources" (April 2001). After the SEMATECH OEE Guidebook and SEMI E10/E79. OEE = availability × performance efficiency × quality rate; availability = uptime (productive + standby + engineering) ÷ total time; performance efficiency = rate efficiency (ideal ÷ actual process time) × operational efficiency (productive ÷ uptime); quality = good wafers ÷ wafers processed. Example week: 44 h down (30 planned maintenance, 4 shutdown, 10 unplanned) gives 0.738; 4 vs 5 minutes a wafer and 73 of 124 up hours processing give 0.471; 788 good of 876 give 0.900; OEE 0.313, the same as 788 good of 2,520 possible. Typical fab OEE under 50%: unreliable tools, uneven arrivals, test and monitor wafers, setups, scrap, rework, operators, spare capacity; some idle time planned on non-bottlenecks to keep cycle time down.
- C. Keith, M. Anvar and M. Arturo (Applied Materials), "Comparison study to evaluate the relationship between equipment uptime variability metrics and cycle time," Winter Simulation Conference (2024). Leading-edge flows can exceed 1,000 steps, cycle times measured in months; giga-fabs 1,000 or more tools; leading-edge fabs USD 20–40 billion over the first ten years, mostly tools, so fabs run round the clock; tools down 15% of the time or more. Kingman's wait = variability × utilisation × time. Availability from MTTF and MTTR. Simulation at 85% uptime and 80% utilisation: MTTF 45.5 h / MTTR 8 h, 3.9 days / 16 h, 15 days / 2.55 days; cycle time X-factor with 5 tools 6.62, 10.33, 27.2; 10 tools 4.10, 5.03, 14.54; 20 tools 2.54, 2.71, 6.12.
- J. D. C. Little, "Little's Law as viewed on its 50th anniversary," Operations Research 59(3) (2011). L = λW; in operations management TH = WIP ÷ CT (Hopp and Spearman): throughput, work in process, cycle time. Any increase in output needs more work in process or a shorter cycle time, or both, unless the process itself changes; with throughput set, the only way to cut inventory is to cut the wait. Knowing two of the three gives the third.
- S. T. Shikalgar, D. Fronckowiak and E. A. MacNair (IBM), "300mm wafer fabrication line simulation model," Winter Simulation Conference (2002). Products of about 400–600 process steps. An overhead monorail between bays and overhead hoists within them, meeting at stockers at the end of each bay; the interbay vehicle serves only stockers, the intrabay vehicles carry one lot to the tools' load ports. A finished step sends a move request to the material handling, and a move complete comes back. Normal, priority and qualification lots; raising the share of priority wafers lowered throughput; smaller lots raised transport time. Lot release and dispatch rules strongly affect cycle time.