FabTime Cycle Time Management Newsletter Abstracts

FabTime Cycle Time Management Newsletter Abstracts - Volume 8 (10 issues)

A Fab Cycle Time Improvement Checklist (Issue 8.10)

In this month’s newsletter we are pleased to announce a FabTime case study that was published by one of our suppliers. Our software tip of the month is about displaying chart data tables directly on FabTime home page tabs. We have a plethora of subscriber discussion in this issue, including a response to an ongoing topic about modeling cluster tool behavior, two responses to a question about managing combined production and development fabs, and two detailed responses to last month’s article about cycle time benchmarking.

In our main article this month, written by Frank Chance, we propose a fab cycle time improvement checklist. The idea is to help codify cycle time improvement practices, so that they become repeatable. The seven items in the checklist include identifying baseline cycle time metrics, finding metrics that indicate current and future cycle time problems, and looking for root causes. These, and other steps, are discussed.

Definitions for Cycle Time Benchmarking (Issue 8.09)

In this issue we have a brief followup to an earlier announcement, to remind you about the upcoming launch of the Fab Engineering & Operations Magazine, as well as a job change announcement from V.A. Ames. Our FabTime tip of the month is about the use of new formatting controls to enable smaller home page charts. We have subscriber discussion about managing production and development activities in the same fab, loading and managing batch tools, and varying lot sizes in the fab.

Because we have quite a bit of subscriber discussion this month, we bring you a relatively short main article. We discuss some of the challenges of calculating cycle time benchmark data. Specifically, we review the two primary metrics currently used for benchmarking across fabs and technologies, X-factor and days per mask layer (DPML), and discuss specific computational issues that apply to each one. We also discuss the conversion ratio between the two metrics. Our hope is that this article will spur further discussion, which will in turn help people who are looking to benchmark and improve their cycle times.

Wafer Fab Flow Control via Order Release (Issue 8.08)

This month we have community announcements about a FabTime demo offer and some upcoming industry meetings. Our FabTime software tip of the month is about using FabTime to predict when lots will complete processing in the fab. We also have some subscriber discussion regarding batch tools, a continuation of an earlier topic.

In our main article this month, written by Professor Scott Mason, we discuss the impact that the way lots are released into a wafer fab can have on performance. We provide an overview of workload (flow) control terminology, and then briefly discuss both push- and pull-based methods. Finally, after discussing recent results from case studies, we conclude by returning to last month’s newsletter topic, dispatching in wafer fabs, to discuss advanced dispatching strategies for linked process steps. We conclude with three recommendations for evaluating lot release policies used in fabs.

Scheduling and Dispatching in Wafer Fabs (Issue 8.07)

In this issue, we have a community announcement about a new electronic publication that we think will be of particular interest to subscribers of this newsletter. It’s targeted to established fabs, rather than focusing only on the bleeding edge of technology. Our FabTime user tip of the month concerns exploiting the archive of past FabTime tips from inside the software. We have two subscriber responses to last montht’s issue – one about holding batch tools idle, and the other about cluster tools.

Our main article this month comes from our esteemed guest contributor (introduced last month), Professor Scott Mason of the University of Arkansas. Professor Mason is a national expert on dispatching, scheduling and manufacturing performance improvement for wafer fabs. This month, Professor Mason discusses scheduling and dispatching. He provides an overview of scheduling and dispatching terminology, discusses the state of the practice with respect to fab dispatching, briefly outlines FabTimet’s dispatching functionality, and then presents some case study results from across the industry describing the positive impact that effective dispatching can have on a fab. We hope that you find this article useful, and we welcome your feedback.

Cluster Tools in Wafer Fabs (Issue 8.06)

In our community news section this month, we summarize information about four upcoming conferences that have relevance for fab manufacturing performance improvement. Our FabTime software tip of the month is about restricting FabTime data access for individual users. We also have a subscriber discussion topic, introduced by Walt Trybula, related to last month’s question about fab utilization. This month’s main article, about cluster tools, was written by Professor Scott Mason of the University of Arkansas, a national expert on dispatching, scheduling and manufacturing performance improvement for wafer fabs. He provides an overview of cluster tools, including a discussion of some of their pros and cons, and then discuss approaches (both academic and practical) for modeling and analyzing cluster tools in order to develop estimates of tool capacity and cycle time. He also shows, by example, the way that adding a chamber can sometimes increase capacity, while decreasing cycle time, for a cluster tool, by reducing blocking.

Conquering WIP Bubbles (Issue 8.05)

Our FabTime software tip of the month this issue is about using FabTime’s new tool qualification charts to identify single-path operations in the fab. In our subscriber discussion forum, we have an extended discussion with Dov Kotlar of Tower Semiconductor about metrics for measuring fab utilization. In our main article this month we tackle the subject of WIP bubbles. People ask us occasionally: “how do I manage WIP bubbles in the fab?”. A WIP bubble is a large pile of WIP, usually in queue at a particular tool-group or small set of tool-groups. WIP bubbles occur due to a variety of causes, the most notable of which is extended downtime on a one-of-a-kind tool. In this article, we discuss common causes of WIP bubbles, methods for avoiding them, early WIP bubble indicators, and potential methods for mitigating their effect. Several of the latter involve making dispatching decisions that encompass information about downstream operations. We hope that yo’ll find this article useful, and we would love to hear your feedback.

Sources of Variability in Wafer Fabs (Issue 8.04)

In this issue we have a software user tip of the month about copying user accounts, and two subscriber responses to last month’s issue about estimating operation-level cycle times. In our main article this month, we address sources of variability in wafer fabs. Variability is one of the main causes of fab cycle time. Variability affects the shape of the operating curve of cycle time vs. tool utilization. By reducing variability, we can move the knee of the operating curve for a fab, achieving a lower cycle time at the same throughput rate. Variability reduction is a relatively inexpensive way to improve cycle time, because it does not require the purchase of capital equipment, or any reduction in starts. However, in order to reduce variability in your fab, you need to be able to identify the specific sources of variability. In this article, we review some of the major sources of variability in fabs, and suggest several general methods for reducing it. We then discuss in detail metrics that you can use for quantifying and identifying specific variability problems in your fab.

Estimating Planned Operation Cycle Times (Issue 8.03)

This month we have two community announcements and one response to last month’s article about making morning meetings more effective. Our FabTime software user tip of the month is about tracking cumulative hold time across lots. In our main article this month we discuss the reasons for needing planned operation-level cycle time values, and review several potential methods for generating them. Methods discussed include using a straight multiple of theoretical, across all steps, using queueing or simulation models to estimate step-specific values, and using actual historical data. We then discuss some technical issues related to the use of actual data, specifically the selection of using mean or median value from a set of actual observations. We hope that you will find this discussion useful.

What Makes an Effective Morning Meeting? (Issue 8.02)

In our main article this month, written by Frank Chance, we discuss the ever-popular, but rarely examined in the literature, fab morning meeting. Nearly all fabs that we know hold a daily morning production meeting. Our hypothesis is that fabs that hold effective morning meetings are also likely to be effective at achieving their manufacturing goals. A morning meeting is effective if it routinely achieves it stated purpose, whether that is to distribute information, hold individuals accountable, make decisions, brainstorm solutions, or another purpose specified by fab management. In this article, we examine the purposes for morning meetings, and explore behaviors that may make meetings ineffective. Our goal is to motivate you to examine and improve the effectiveness of your morning production meetings.

In this issue we also have a call for papers for the 2007 MASM conference, to be held in Scottsdale, Arizona in September. Our software user tip of the month involves using FabTime to track on-time delivery performance. We have no subscriber discussion, though we have been having some informal discussions with people about wafer size transitions and benchmarking, which may be reflected in future issues.

Highlighting Cycle Time Problems for New Products (Issue 8.01)

In our main article this month we address a problem that we’ve heard mentioned at several companies. The issue is that new, low volume products often incur long cycle times, because the traditional performance measures in fabs allow them to slip through the cracks. We present a series of recommendations for mitigating this effect. We welcome your feedback.

Our software tip of the month is about how to cross-slice move and WIP data in a single chart data table (to look at, for example, WIP by priority within each area on the same page). We have several subscriber discussion topics. Anonymous subscribers wrote in with new questions concerning justifying additional capacity to management, and identifying and analyzing your own fab’s top three cycle time problems. We also have a response from David Jimenez of WWK to a question posed last month about labor modeling for fabs.