FabTime Cycle Time Management Newsletter Abstracts

FabTime Cycle Time Management Newsletter Abstracts - Volume 5 (10 issues)


Management Behavior and Fab Cycle Time (Issue 5.10)

Because we have extensive subscriber discuss this month, our main article is relatively short. We turned for inspiration to the responses to our cycle time issues survey (in which we have been asking people “What is the biggest cycle time problem in your fab?”). We noticed a number of responses pointing towards management behaviors that influence variability in the fab, at least from the perspective of people working in the fab. We’ve chosen to highlight these responses, and discuss their impact on our “Traffic Cop” cycle time management style recommendations.

In this issue we have an announcement about the latest version of the software, as well as a short recap of a recent industry conference. Our software user tip of the month concerns methods for updating home page chart data. In the subscriber discussion forum we have several responses to last month’s questions about paperless cleanrooms and the effect of linked tools on 300mm cycle times, as well as a new question about benchmarking for “single strand” toolsets.



Analyzing Capacity Using MES Data (Issue 5.09)

In our main article this month, we discuss using data from the fab manufacturing execution system (MES) to perform static capacity analysis. FabTime is in the business of taking data from the MES, and using it to provide information to the people who manage wafer fabs. Our software takes updates from the MES in near real-time, and stores the data in a separate database, making a digital dashboard of charts available via web browser. Recently, we have been working with our customers to use this data to help them plan capacity. The primary advantage of this approach is that most of the data is already available and automatically updated to reflect current fab conditions. This lets planners spend their time generating and running scenarios, rather than performing data entry to keep standalone capacity models up to date.

This month we have an announcement related to past issue abstracts (to make it easier for you to find references to topics previously discussed in the newsletter). Our software user tip of the month is about generating a list of hot lots. We also have subscriber discussion related to 300mm cycle times (in response to last month’s issue) and paperless cleanrooms.



Real-Time Alerting based on Fab Conditions (Issue 5.08)

In our main article this month, we discuss real-time alerts sent to semiconductor wafer fab users to notify them of particular conditions in the fab. We review pros and cons of using these types of alerts at all, and then describe several examples in detail (including hot lot queue delay, early warning of lots due to reach time limit, and critical tool idle with wip available). Finally, we solicit subscriber feedback on the general usefulness of alerts, and on other types of these warning messages that might be useful in semiconductor fabs.

This month we have several announcements, as well as a considerable amount of subscriber discussion. Our FabTime user tip of the month describes how to add a chart from a shared home page tab to a user’s own home page. Subscriber discussion topics include: capacity planning for time links between process steps, understanding 300 mm cycle time performance, assessing the impact of downtime on fab performance, setting targets for WIP and turns, and defining fab loading in the presence of multiple near-bottlenecks. We also have several responses to our question about the oldest continuously operating wafer fab.



Quantifying the Effect of Tool Downtime (Issue 5.07)

In our main article this month, we discuss metrics for measuring the effect of tool downtime. If we could eliminate downtime from our semiconductor wafer fabs, we could increase throughput (where the constraint tools have any downtime at all), and improve cycle time at the same time. In this article, we make a first pass at quantifying this impact more formally, by measuring the increased operation-level cycle time for lots that are in queue when a tool goes down. We believe that better understanding the cycle time cost from specific downtime events could be helpful for fabs in deciding where to focus tool improvement efforts.

This issue includes subscriber discussion related to capacity planning, Dynamic X-Factor, WIP Utilization, and metrics for measuring the effect of tool downtime. Our FabTime user tip of the month concerns setting a default time of day for newly generated interval-based charts.



Cycle Time Constrained Capacity (Issue 5.06)

In our main article this month, we discuss increasing semiconductor wafer fab throughput through improvements in cycle time constrained capacity. The idea is that fabs always have a buffer of planned idle time on tools, designed to keep cycle times from getting out of control. Through variability reduction, fabs can sometimes squeeze this buffer, without increasing cycle time. In an up market, this can lead to increased sales, from the same equipment set. The financial benefit from this can be substantial, and provides a clear justification for variability reduction / cycle time improvement efforts.

In this issue’s subscriber discussion forum, we have discussion concerning several previously introduced topics: WIP Utilization Percentage, Dynamic X-Factor, and the Closest-to-Completion-Time Dispatch Rule. Our FabTime user tip of the month concerns setting filter defaults.



WIP Utilization Percentage (Issue 5.05)

In our main article this month, we propose a new wafer fab metric for tracking shift-level use of individual tools by operators, called WIP Utilization%. This metric was developed jointly by Frank Chance of FabTime and Jimmy Martin of Analog Devices. We define WIP Utilization% as Productive Time / (Productive Time + Standby WIP Waiting Time). This is similar to our definition of Utilization, which is Productive Time / (Productive Time + Standby Time). However, in the denominator, we only include the standby time in which WIP is waiting for the tool. WIP Utilization% will approach 100% if, whenever WIP is waiting, and a qualified tool is available, the WIP is processed as soon as possible. Driving WIP utilization to 100% generally minimizes per-visit cycle times through the tool, and helps to maximize shift-level throughput. This metric overcomes several shortcomings of the standard utilization definition as a shift-level metric for operators in semiconductor wafer fabs.

In this issue’s subscriber discussion forum, we have several responses to last month’s subscriber discussion question, about breaking up standby time according to whether or not WIP is available. We also have a new subscriber discussion question about the closest-to-completion-time dispatch rule.



Presenting Fab Performance Data (Issue 5.04)

In our main article this month, we discuss ideas for presenting semiconductor wafer fab performance data. These ideas are based in part on concepts proposed by Edward Tufte, author of “The Visual Display of Quantitative Information”. Tufte’s suggestions include using quantitative metrics for data graphics and integrating text with graphical data in charts. We have included a detailed FabTime-generated example of improving an Excel-generated chart, and summarized a few recommendations from both FabTime and Edward Tufte. We hope that you find it interesting.

We have no new subscriber contributed discussion topics this month. However, we have included a sample of responses to our newsletter sign-up question: What is the biggest cycle time problem in your fab?” and we have posed a topic ourselves for future discussion.



Dynamic X-Factor Revisited (Issue 5.03)

In our main article this month, we revisit the topic of dynamic x-factor, a metric for semiconductor wafer fabs that we first described back in issue 4.08. Dynamic x-factor is a point estimate that looks at the total wafers that you have in your fab, divided by the wafers that are currently being processed on tools. In this article, we look further into what dynamic x-factor can tell us about how a fab is operating, with emphasis on evaluation of shift change coverage policies and comparison of relative performance across modules.

Community announcements in this issue include two calls for papers for conference sessions related to semiconductor manufacturing applications. Subscriber discussion topics include wafer holds, cycle time and yield, operator utilization, and dynamic x-factor.



Cycle Time and Yield Revisited (Issue 5.02)

Subscriber discussion topics for this month include nine responses to last month’s topic of Cycle Time and Yield. These responses point out some significant omissions in our article. Therefore, instead of introducing a new main article, we’ve chosen to revisit the topic of cycle time and yield, and very briefly summarize the additional points made by contributing subscribers. This article is a companion article to Issue 5.01, and we recommend reading the two of them together.



Cycle Time and Yield (Issue 5.01)

This month’s main article is about the interaction between cycle time and yield. We’ve always cited yield improvement as a potential benefit from cycle time improvement, and people we talk with about this generally agree. However, because the actual data tends to be proprietary in nature, references on this topic are scarce. Therefore, we’ve decided to open the topic for discussion here, and summarize a few references that are available. We hope that you’ll find the discussion interesting.

Subscriber discussion topics for this month include responses to Issue 4.09 (WIPHours Metric) and Issue 4.11 (Cycle Time and Factory Size).