Our article this month is a continuation of last month’s discussion on including cycle time in the capacity planning process. Last month we talked about how people do this implicitly, though the use of capacity loading factors. This month, we talk about a more explicit method of including cycle time in the capacity planning process, through the use of simulation models. This article is based on a project that we did for Seagate Technology several years ago. The method involves using simulation to estimate the cycle time of candidate models, and adding tools on the basis of greatest cycle time reduction per dollar of fixed cost. The main point from this study is that other factors besides equipment loading have an influence on the cycle time contribution of individual toolgroups. Considering those other factors can allow you to plan for more cost-effective toolsets. Navi Grewal, one of the original authors, collaborated with us on this article.
Discussion topics in this issue include: several responses to the 300mm lot size question; a proposal for calculating the cost of cycle time; a statement of the continued need for moves as a daily fab performance metric; a modification to the cycle time calculations in the characteristic curve generator; a case study comparing actual performance to short-term goals; and questions about the implications of 300mm factory size, relating OEE to cost per wafer, modeling operator impact, modeling cycle time and WIP during a volume ramp, the industry definition of "loading", calculation of product and factory line yield values, and benchmarking cycle time for wafer production.
Cycle time is always considered in the capacity planning process for wafer fabs. In most cases, however, cycle time is considered implicitly, rather than explicitly. If your capacity planning team was not considering cycle time, they would plan for the minimum toolset to meet throughput requirements, with perhaps some additional tools to account for potential product mix changes. Instead, they include planned idle time for essentially all tool groups. They also try to avoid one-of-a-kind tools, frequently recommending duplicates of even very lightly loaded tools. In this article, we will talk about these traditional methods of implicitly accounting for cycle time in the capacity planning process. Next month we will look at ways to be more explicit, and shoot for specific cycle time targets.
Discussion topics in this issue include: a question about the standard for 300mm lot size; a question about quantifying cost savings from cycle time reduction; an inquiry about the availability of published productivity report indices for fabs; a request for references on literature regarding new product introductions; and a practical best-case X-factor for cycle time goals taking human performance into account.
We are surrounded by performance measures. Goals help us to convert these absolute numbers into relative “good or bad” indicators. At higher levels of an organization, you deal with aggregated goals. More detailed goals must be set, however, at lower levels of the organization. These detailed goals must be consistent with the higher-level goals, and must be useful for day-to-day operations. The closer you look at the process, the more you see the proliferation of goals. If you can address this proliferation, you can generate appropriate goals for a wide variety of intermediate performance measures. It’s important to remember the implicit assumptions behind long-term goals, however, and to mix long-term goals with appropriate short-term targets.
Discussion topics in this issue include: a question about generating operating curves for the wafer test area; a description of experiences in measuring process time variability; and a request for the logic behind the variability parameters in the FabTime characteristic curve generator.
The FabTime Cycle Time Characteristic Curve Generator is an Excel-based tool for exploring cycle time and utilization trade-offs for single tools with failures. You can enter parameters for process time, mean time between failures, downtime percentage, and system coefficients of variation for up to three scenarios. The calculator then displays characteristic curves for the scenarios, allowing you to get a quick visual impression of the impact of both downtime and variability attributes. The curves are based on a queueing approximation that we received several years ago from Ottmar Gihr of IBM Germany. You can download the characteristic curve generator from FabTime’s website (here).
Discussion topics in this issue include: the method for ordering the SEMI E-79 Standard document; a description of where to find abstracts to INFORMS articles; a request for fab cycle time benchmark data; and a request for tool cycle time benchmark data.
The article was written by Frank Chance, with assistance from Stuart Carr (consultant and FabTime affiliate), and Ken Beller. Frank started thinking about this question because, as President of a cycle time management software company, he is frequently asked about the dollar benefit of cycle time reduction. This article outlines several potential ways to quantify this benefit, and focuses in particular on the timely issue of inventory write-off during an industry downturn. The article references an Excel-based cycle time payback calculator that was formerly available from FabTime’s website. The calculator has since been replaced by a more comprehensive calculator described in Issue 3.5.
Discussion topics in this issue include: the SEMI E-79 Standard definition of ideal process time; and a clarification of the OEE calculations for quality rate.
This issue contains the abstracts to all previous issues. It also contains additional discussion on OEE, and several industry announcements.
Most of our readers are familiar with the general concept of Overall Equipment Efficiency (OEE). OEE is a tool-level measure reflecting how much good product the tool produced relative to some theoretical amount that it could have produced. Typical OEE values in a wafer fab are less than 50%. Given the high cost of equipment, there is a clear incentive to make OEEs as high as possible. OEE is the measurement that’s used in TPM (Total Productive Maintenance), a methodology for improving the entire manufacturing process.
In this article, we review the formulas for calculating OEE (both the full formula and a short-cut version), as well as some of the reasons for low OEE in wafer fabs. We also include a series of links to OEE resources on the Internet (including primary resources from SEMI and SEMATECH), as well as some additional published OEE references.
The power of OEE is that it provides a clearly defined metric by which equipment performance improvement projects can be measured. SEMI and SEMATECH have gone to great lengths to define OEE, and also the necessary supporting metrics like the SEMI E-10 equipment states. The nice thing about this is that it means that you can compare OEE values across factories, and even across companies, and get a true picture of your factory’s performance. Another nice thing about OEE is that it drives you to do good things, like reduce setup and rework and scrap and starvations due to WIP or operator shortages. By focusing on the six types of losses highlighted by OEE, you can design a strong equipment improvement program, and monitor your progress through trends in the overall metric.
Downturns are a fact of life in the cyclic semiconductor industry. Various factors contribute to their existence - capacity buildup (and the long lead-time required in capacity purchases), decline in selling prices, inventory build-up, and the general state of the economy. This one seems to have been triggered mainly by the last two factors, but explanations and predictions also seem to change every day.
The quickest way to reduce cycle time in a wafer fab is to significantly decrease start rates. This moves your factory to the left on the cycle time vs. factory loading curve, to a region of lower cycle times. The irony is that just when customers aren’t clamoring for product, your fab can delivery product with record cycle time and on-time-delivery performance. It’s very easy under these conditions to get a bit sloppy, and to take the lower cycle times for granted. But then when start rates begin to increase, when customers are paying attention again, your cycle times will degrade rapidly. If you don’t have great cycle times now, you certainly won’t have great cycle times when start rates go back up. Therefore, we suggest using this time to focus on low cost cycle time improvement efforts, including setup/dedication policy investigation, process analysis, layout analysis, bottleneck analysis, OEE/TPM analysis, simulation model validation, system upgrades, and education.
A downturn is a tough time - stressful, hard on your stock portfolio, and filled with the specter of layoffs. But it does offer at least one potential benefit: time to think. Time to think about manufacturing issues like lot size and batch size policies. Time to think about tool dedication schemes, and layout changes. Time to get your fab in order, and drive your cycle times to a minimum, before the next upturn comes along.
Discussion topics in this issue include: a success story on cycle time reduction through batch size decision rule changes; and a clarification of the units in the P-K formula.
This article concerns possible changes to production lot sizes for cycle time improvement. For fabs running 50 wafer lots, changing to 24 or 25 wafer lots offers a potential cycle time reduction opportunity. However, there can be tremendous resistance to this idea, and there are a number of potential pitfalls. In this article, we first review the reasons for the cycle time reduction opportunity, and then discuss some of the pitfalls.
The justification of lot size reduction for cycle time reduction comes into play primarily due to time savings at per-wafer tools, which can include critical tools such as steppers and implanters. In addition to providing these direct cycle time benefits, smaller lot sizes also make a fab more flexible, more adaptive in the event of problems, and can reduce variability. However, there are a number of issues to consider before changing the lot size, any one of which might keep a lot size reduction from being worthwhile, or even render it detrimental. These include capacity, material handling, MES, and dispatching/complexity issues, and are discussed in detail in the full article.
We have no black-and-white recommendation to make concerning lot sizes and cycle time. Smaller lot sizes may reduce cycle time, and make a fab more flexible. However, reducing the lot size can cause problems with material handling, capacity, MES performance, and fab complexity, particularly during the transition period. We suggest then, that you consider lot size reduction to reduce cycle times, but that you consider it very carefully.
Discussion topics in this issue include: observations about time constraints and batch size decisions, and sequence dependent setups and batch size decisions; and a question about defining utilization at batch tools.
Batch tools are tools in which more than one lot may be processed at one time. They are generally used for very long operations, such as furnace bake operations. Processing time is usually independent of the number of lots in a batch, and once a batch process begins, it cannot be interrupted to allow other lots to join. From a local perspective, when a furnace is available and full loads are waiting, the decision to process a batch is obvious, since no advantage can be gained at that work area by waiting (although a decision may still be needed concerning which product type to process). However, when there is a furnace available and only partial loads of products are waiting, a decision must be made to either start a (partial) batch or wait for more products to arrive.
There are two problems with running a partial batch. One is that the unused capacity of the furnace will be “wasted.” The other problem is that lots that arrive immediately after the batch starts cannot be added to the batch, and might have to wait many hours until another furnace is available. There are also problems that stem from waiting to form a full batch. The lots that are waiting to be processed incur extra queue time while waiting for other lots to arrive. The furnace is held idle, driving down its efficiency. And full batches contribute more to variability after the furnace operation.
This article discusses policies for deciding when to form a partial batch, using simple numerical examples and simulation results. We conclude that for batch tools that are not highly loaded, forcing full or near-full batches can significantly increase local cycle times, as well as overall fab cycle times.
Copyright ©1999-2020 FabTime Inc.