Other Abstracts

FabTime Cycle Time Management Newsletter Abstracts - Volume 9 (10 Issues)

Improving Cycle Time during a Downturn, Redux (Issue 9.10)

In this issue we have one announcement about the next Fab Owners Association meeting. Our software user tip of the month is about the distinction between different types of filters for excluding data in FabTime. We have one response to last month’s subscriber discussion question about the use of lean techniques in wafer fabs. In our main article this month, we revisit an article that we wrote back in 2001, about methods for making fundamental cycle time improvements during a downturn (so that the fab will be in good shape to achieve great cycle times when things improve). We really do believe that, if you're in a niche that’s being hit by this economy, as most people are, this is a time to hunker down and work on low-cost improvements. If you don’t have great cycle times now, you certainly won’t have great cycle times when demand comes back up.

WIP Bubbles in Wafer Fabs (Issue 9.09)

In this issue we have an announcement about the upcoming Winter Simulation Conference, and a subscriber discussion question about success stories in implementing lean and six sigma techniques in wafer fabs. Our software user tip of the month is about looking at variation from a WIP goal. Continuing this theme of looking at WIP variation, our main article discusses WIP bubbles in wafer fabs. A WIP bubble is a larger-than-normal buildup of WIP at a particular point in the line. WIP bubbles result in large queues in front of a few tools, while other tools, sometimes even bottleneck tools, remain idle. A common goal in fabs is to smooth out the WIP bubbles, so that all production areas remain relatively busy. Smoothing of WIP bubbles improves cycle time by reducing arrival variability throughout the fab. In this article, we discuss techniques for avoiding WIP bubbles in the first place (where possible) and for coping with them when they do arise.

Tool State Calculations for Cluster Tools in Fabs (Issue 9.08)

In this issue, we have a community announcement about a deadline extension for abstracts for the Advanced Semiconductor Manufacturing Conference. Our FabTime software user tip of the month is about separating out the components of non-process time (queue time, hold time, etc.). We have no subscriber discussion this month. However, we do suspect that this month’s main article will inspire some discussion for the next issue. We have asked occasional contributor Professor Scott Mason to write about tool state calculations for cluster tools. Professor Mason discusses two primary methods for estimating overall cluster tool performance, one based on logical rules and the other based on averages (possibly weighted) across chambers. He gives several examples, and shows through these examples how different the overall results can be depending on the calculation method used. He concludes that the wide disparity of results begs the question: how are cluster tool E10 states computed in your fab?

How To Extend the Life of Your Fabs? Measure, Monitor, and Control (Issue 9.07)

In this issue we have a brief announcement about the Fab Owners Association. Our software user tip of the month is about setting up home page tabs that update on a monthly (instead of a weekly or daily) basis. In our subscriber discussion forum we have a response to last month’s article about measuring line yield, and an update on a previous discussion question about short-term fab simulation.

For our main article this month we have a special treat - a guest article by Dick Deininger and Rebecca Taylor of Taylor-Deininger Partners. The article is about measuring, monitoring, alarming and control actions that cut costs and increase productivity, to extend the lifespan of existing fabs. The article is specifically geared towards older fabs that are make products that "do not demand bleeding edge technology to be profitable... Taylor-Deininger Partners has developed a modeling capability to demonstrate the value of implementing remote monitoring in a number of application areas. The model analyzes current Fab wafer losses, die yield losses, gas usage, electrical usage, water usage, and air handling. It then estimates the value of implementing measurement, monitoring and alarming in the highest value areas. It can show how losses and usage of key materials can be reduced, thus improving productivity. The model shows that remote dynamic monitoring helps identify problems before they can adversely affect product. This paper describes a number of these situations with demonstrated savings in a number of ultra clean facilities involved in semiconductor manufacturing as well as laboratories."

Definitions for Short-Term Line Yield Metrics (Issue 9.06)

In this relatively brief issue we have no community announcements. Our FabTime user tip of the month is about using the SQL filter to remove outliers from a down tools list. We have one subscriber contribution, concerning the use of Dynamic X-Factor at On Semiconductor in Gresham. In our main article this month, we review potential definitions for short-term line yield metrics, definitions that take into account the particular complexities of wafer fabs. We are seeking input from the newsletter community because we would like to include one or more detailed definitions for line yield in FabTime's metrics lexicon. We believe that these definitions will be useful to the community as a whole.

Paper vs. Electronic Lot Travelers (Issue 9.05)

In this issue we have a community announcement about the third issue of Fab Engineering & Operations Magazine (a publication that’s not affiliated with ours, but that we think our readers will enjoy). Our FabTime user tip of the month is about using FabTime’s software to generate a list of lots ahead of or behind schedule, according to planned cycle times at future operations.

We have one new subscriber discussion question in this issue, about the use of short-term simulation. We also received several detailed responses to last month’s question about the transition between paper and electronic lot travelers. In fact, these responses are so thorough, and varied, that we’ve decided to convert them into this month’s main issue. This is a slightly unusual approach to our main article, but one that we think will prove useful to anyone facing this paper to electronic traveler transition. The strength of these responses also shows what a valuable resource this subscriber community can be. We are very grateful to the subscribers who took time to respond to this topic and to all of you who have helped us with prior discussions. We welcome further feedback on paper vs. electronic lot travelers.

Dynamic X-Factor and Shipped Lot X-Factor (Issue 9.04)

In this issue we have a brief summary of upcoming industry conferences in our community announcements section. Our FabTime software user tip of the month is about using the home page chart alert functionality. We have one subscriber discussion question, about the transition from paper to electronic travelers, for which we could use your input. Our main article this month is about the comparison between dynamic x-factor (a point estimate measured as total WIP divided by WIP running on tools) and shipped lot cycle time x-factor. We show that although in the long run, DXF can be used to predict x-factor, various issues sometimes make it difficult to draw exact comparisons between this week’s DXF and some future week’s shipped lot x-factor value.

Batch Loading Policies for Wafer Fabs (Issue 9.03)

In this issue, we have community announcements about the second issue of Fab Engineering and Operations Magazine and a milestone reached by the Fab Owners Association. Our software user tip of the month describes how to use FabTime's new Queue Limit Lot List chart, which shows the non-held lots in queue that have exceeded, or are in danger of exceeding, a user-specified threshold.

We have one subscriber discussion question this month, about batch loading rules. In responding to this question, we realized that it has been more than five years since we last discussed batching in detail in the newsletter. Therefore, we decided to discuss batching in our main article this month. Specifically, we review the cycle time benefits of a greedy vs. a full batch policy, with examples, and also provide a simple rule of thumb for using look-ahead information in the batch formation decision. We welcome subscriber feedback, especially about experiences with greedy vs. full batch policies and incorporating look-ahead information into the batch loading decision.

Manual Lot Transfer in Wafer Fabs (Issue 9.02)

In our main article, we discuss lot transfer between operations for non-automated fabs. Although material handling in automated fabs has gained considerable attention in the literature, we believe that lot transfer is also having a significant impact on cycle time in less automated fabs, and that this topic is relevant for many of our newsletter subscribers. Behaviors such as the use of carts for lot transfer and the use of performance incentives for operators that do not reward the movement of material between areas can lead to higher than anticipated cycle times. For those fabs that are experiencing delays due to lot transfer, we recommend working towards reducing transfer batch sizes between steps, either by physically purchasing smaller carts, or by changing the way that operators are assigned or measured. If a full-scale change in carts or operating practices is not possible, we recommend identifying the specific areas in which material movement issues are causing cycle time, and implementing changes in those areas first. We discuss these potential solutions in detail, and welcome feedback. We also have one community announcement - a call for papers for the 2008 MASM conference. Our software user tip of the month is about showing and hiding data table columns in FabTime. We have no subscriber discussion this month.

Our Top Recommendation for Cycle Time Improvement: Tackle Single Path Operations (Issue 9.01)

We have one brief community announcement this month, about the winter Fab Owners Association meeting. Our FabTime software user tip of the month is about eliminating time spent with particular hold codes or owner codes from Operation Cycle Time Trend and Pareto charts. We also have a subscriber discussion response from Dov Kotlar of Tower Semiconductor (one of our software customers) to some previous questions that we raised about cycle time benchmarking.

In our main article this month we return to a topic that we have discussed before, but that continues to pose challenges for people who manage wafer fabs: single path operations. We review the different types of single path operations, and focus on those that stem from tool dedication. We present a rule of thumb for estimating the potential impact of going from single path to dual path for a given operation, and discuss two particularly insidious forms of tool dedication: soft dedication due to operator preferences; and process restrictions for new operations. In both cases, we recommend strategies for identifying and eliminating the single path operations. We believe that this is one of the highest benefit low-cost changes that an existing fab can make to improve cycle time.