This month’s main article is about the cycle time effect of changing factory size for semiconductor wafer fabs. Everyone knows that for a given fab, as start rates increase (as they seem to be doing for many fabs) cycle time is likely to also increase. What’s less obvious is the behavior that one of our subscribers pointed out in this month’s subscriber discussion forum: sometimes when start rates decrease, cycle time increases. This wouldn’t normally happen if there were no other changes in the fab. Utilization would go down, for tools and operators, and cycle time would almost surely go down. However, that’s not a realistic case. What really happens in many fabs is that when start rates go down, tools are turned off and staffing is reduced. The net result from this is that the bottleneck utilization of the fab may stay the same, or even increase. So, no cycle time payoff from the decreased start rate. What also happens is that the number of tools per tool group decreases, sometimes to the point of having one-of-a-kind tools in operation. This lack of tool redundancy is a key driver of cycle time (currently ranked third on FabTime’s cycle time problems survey, after downtime and bottleneck utilization), and is the primary subject of this article.
Subscriber discussion topics for this month include two responses to last month’s article about tool standby and productive time reporting. New topics include incorporating setup in equipment utilization calculations, understanding the cycle time effects of automated material handling and robotic systems, and understanding cycle time and “under-utilization” in fabs. This month also kicks off a new newsletter section: Cycle Time in the News.
This month’s main article is about using manufacturing execution system (MES) data to calculate fab performance measures. More specifically, we discuss the cycle time management benefits of tracking standby and productive time, in addition to tracking tool downtime states. Tool utilization, defined as Productive Time / (Productive + Standby Time) is the largest driver of operation-level cycle times. For this reason, we recommend reporting tool utilizations on a short-term (e.g. shift-level) basis, and automatically flagging situations where utilization approaches 100%. Fabs may be able to do proactive things, like reassigning operators, or deferring engineering or maintenance time, to nip short-term cycle time problems in the bud. To do this, however, fabs will need to ensure that their manufacturing execution systems either track productive and standby state changes directly, or generates them in some other manner.
We have no subscriber discussion topics in this issue.
This month’s main article is about metrics for identifying short-term bottlenecks in a semiconductor wafer fab fab. Last month we proposed the metric Dynamic X-Factor as a short-term indicator of overall fab performance. In this article, we focus more on tool-level performance metrics. The idea is to identify metrics that can be used at the start of the shift to highlight current or anticipated cycle time problems in the fab. We first discuss a few simple metrics, and their relative applicability to this problem. We then propose a simple calculation (WIP hours) for identifying short-term bottlenecks without performing simulation, by estimating the hours of work in queue for a toolset. We don’t have all the answers here, but we would like to start a discussion with the FabTime newsletter community about this. Ultimately, we want to work towards developing useful short-term metrics for identifying temporary bottlenecks in wafer fabs.
Subscriber discussion topics for this month include two responses to last month’s article about the performance metric Dynamic X-Factor, and new questions about managing in high-mix and R&D environments. We also have announcements about a new one-day version of FabTime’s cycle time management course, a Cost of Ownership task force meeting, and the acquisition of WWK by its management team.
This month’s main article is about the wafer fab performance metric Dynamic X-Factor. Dynamic X-Factor measures, on a point-in-time basis, how much of the WIP in the line is currently being worked on, instead of sitting in queue. If Dynamic X-Factor drifts upward, cycle time will probably start to increase in the future (because either there is more WIP, or WIP in the line is sitting more than it should be). Dynamic X-Factor is calculated by taking the total number of wafers in the fab and dividing by the number of non-rework wafers actually being processed. While Dynamic X-Factor works out to be the same as the regular cycle time X-Factor (cycle time / theoretical cycle time) on a long-term basis, Dynamic X-Factor is easier to calculate, and is more forward-looking than an X-Factor based on shipped lot cycle times. While there are some limitations to this metric, we think that it provides a useful indicator of current fab cycle time performance. We recommend its use for semiconductor fabs.
We have no subscriber discussion topics in this issue.
This month’s main article is about identifying real-time cycle time problems in a wafer fab. We wrote this article in response to an informal survey that we have been conducting about cycle time problems in semiconductor wafer fabs. The fourth-most common response to date has been real-time identification of cycle time problems (e.g. problem tools or operations). This is a nuts-and-bolts kind of topic that we’ve addressed only indirectly in this newsletter so far. In this issue, we propose metrics and methods for identifying cycle time problems in the fab on a short-term basis, so that they can be addressed and improved. Metrics discussed include operation-level cycle time, summed operation cycle time, inventory age, arrival coefficient of variation, and availability variability. We also touch on some more detailed methods for using real-time data to understand problems and improve operational decisions. Specifically, we focus on tool dedication, staffing decisions, batch loading policies, and maintenance schedules.
Subscriber discussion topics for this month include a response to last month’s main article about operators and cycle time, several responses to last month’s question about how companies calculate On Time Delivery percentage, a new question about the productivity of engineering staff, and a new question about wet bench capacity.
This month’s main article is about planning and managing operators in semiconductor wafer fabs. In looking over the past issues of this newsletter, we observed that we have had a considerable amount of subscriber discussion related to staffing. This discussion has primarily fallen into two categories: 1) operator modeling/planning and 2) operator management (including dedication, cross-training, and performance evaluation). The first category concerns understanding how many operators will be required, and how they will impact cycle time and throughput. The second category concerns managing operators once staffing levels have been determined, to minimize cycle time and maximize throughput. In this article, we summarize the subscriber discussion to date on operators, bringing it into one place, instead of scattered across two years of newsletter issues. We will also summarize FabTime’s thoughts on the operator-related questions, and highlight industry resources that we know of related to operators (software, papers, etc.).
This month’s main article is about arrival variability and cycle time in semiconductor wafer fabs. While working with our FabTime cycle time entitlement calculator (described in Volume 4, Number 3), we observed some interesting behavior for cases with a high degree of arrival variability. We found that arrival variability due to batching tended to have less of an impact on cycle time than other types of arrival variability for the cases that we investigated. In this article, we show examples generated from simulation models, and discuss the impact of this behavior on the formulas in our operating curve generator and entitlement calculator. We also introduce a modification to our operating curve generator that accounts for arrival batching.
This month’s main article is about the cycle time effect of equipment downtime. When we ask people what factors contribute to cycle time in their fabs, the number one response that we get is "downtime". Certainly equipment downtime is a fact of life in wafer fabs. In this article we review the reasons why downtime has such a significant influence on cycle time (utilization and variability). We also propose three steps for mitigating the effect of downtime on cycle time.
Subscriber discussion topics for this month include material handling system metrics and cycle time reduction; the metric mean time to recover; and the cycle time effects of integrated metrology in the lithography area.
This month’s main article is about cycle time entitlement for semiconductor wafer fabs. This newsletter has frequently addressed topics related to managing and improving cycle times, and the various metrics for reviewing historical cycle times and benchmarking cycle time performance. But what people who work in fabs really need to know is: what is a good cycle time for our fab, under our current constraints? And where should we focus our cycle time improvement efforts? Cycle time entitlement is FabTime’s answer to these questions. More formally, cycle time entitlement is the best achievable cycle time for a fab given short-term realities related to tool utilization, staffing, and downtime characteristics. In this article we define cycle time entitlement, and discuss ways of estimating it, ways of using it, and associated data issues.
Subscriber discussion topics for this month include responses to our article about quantifying availability variability and to last month’s subscriber question about train schedule batch policies, as well as a new question about estimating company-wide savings from cycle time reduction.
This month’s main article is about quantifying the variability of availability in a fab. Last month we discussed calculating coefficient of variation for interarrival times and process times. We could calculate the coefficient of variation of availability. However CV is a dimensionless metric that may not carry intuitive meaning for people. Instead, we discuss the metrics A80 and A20, recently described by Peter Gaboury in a Future Fab International article. A80 is the best availability reached within 80% of the periods in a set of periods (shifts, days, weeks, etc.), while A20 is the best availability reached (or exceeded) in at least 20% of the periods in a set. By tracking the spread between A20 and A80, and trying to reduce it, we can reduce the variability of availability, and hence improve cycle time. And by dealing with percentiles, we can use metrics that carry more meaning for people on an ongoing basis than CV values.
In this month’s subscriber discussion forum we have a response to last month’s article about process time variability, a question about the cost of having the entire fab down for a period of time, a question about a "train scheduling" batch loading policy, and some comments on wafer moves per operator.
Our main article this month is about quantifying variability in wafer fabs. We have talked many times about how wafer fab cycle time can be reduced by reducing fab variability. In this article, we describe a metric for quantifying this variability (coefficient of variation), and discuss how to calculate it for times between arrivals and for process times. We believe that by measuring variability, particularly relative levels of variability at individual tool groups and operations, readers will be better able to identify potential improvement areas.
In this month’s subscriber discussion forum we have responses from three subscribers to our recent topics regarding operator productivity.
Copyright ©1999-2023 FabTime Inc.