Modeling vehicle-free transport to enable elimination of bottlenecks
11/01/2007
Executive Overview
Automated material handling systems (AMHS) are critical to the infrastructure of a 300mm fab. Modern AMHS must support movement demands that dynamically range from 1000 FOUP moves/hr up to 10,000. Demand in a single bay can exceed 1000 moves/hour. Many fabs are now interested in reducing their lot size to reduce cycle time and inventory. If the current 300mm Prime Initiative effort aligns semiconductor manufacturers and equipment suppliers to enable production of single wafer lots, movement demand will increase 25-fold. AMHS should never be a bottleneck to fab output, therefore very high capacity AMHS is required to support 300mm Prime. Results of a comparison study of the capacity of various AMHS layouts and a strategy for eliminating bottlenecks are presented.
In order to determine the AMHS capacity required to meet a specific requirement, the entire fab should be modeled to understand the movement demands. With a clear picture of the movement demand, an AMHS layout can be created based on the rates of flow. The next step is to identify AMHS bottlenecks, eliminate them, and maximize transport capacity. Below are the results of detailed simulation models of factory applications that have been studied to predict the capacity, average, and peak delivery times for normal and hot-lots.
Models have been generated for two cases: 1) an operating facility requiring improved operational performance, and 2) a new facility. In many current AMHS configurations, the key bottlenecks are the AMHS interface points, and directional changes.
AMHS interface points include stocker inputs, stocker cranes, stocker outputs, and the load port interfaces on the process and metrology equipment. FOUPs wait at these interface points an indeterminate time for an overhead transport (OHT) vehicle to arrive for a pick up. Each interface point can cause a back-up, quickly building a movement queue that chokes the system. The queues take a long time to clear-much like an auto traffic accident can cause a backup that lasts long after the crashed vehicle is removed from the road.
Many fabs are now exploring vehicle-free transport (VFT), such as conveyor solutions, to achieve a higher rate of flow. The added capacity absorbs the dynamic demand to allow for more deterministic delivery times. To be an effective solution, the hot lot delivery times must also meet customer requirements. Figure 1 shows that for this example interbay application, a VFT solution must exhibit a speed >2m/sec to meet the customer required delivery times.
To determine the required layout and speeds for the example of Fig. 1, the following analyses were performed. A fab model was created and tested using AutoMod modeling software. In the initial model, OHT vehicles are used to move FOUPs between process equipment and stockers, and in some cases, directly from tool to tool. Next, a VFT system was added under the existing OHT so that FOUPs can be moved between areas of the fab and between fabs by the VFT system. The VFT moves FOUPs from OHT to stocker, stocker to OHT, stocker to stocker, and even OHT to OHT.
 Figure 2. The capacity for the interbay example of Fig. 1 is related to the transport speed. |
The model was progressively iterated with different VFT speeds to study the effect on the average, maximum, and standard deviation of delivery times. As shown in Fig. 1, the delivery time is normalized to the customer’s requested average delivery time and shown as a function of system speed in meters/second. Based on these results, we determined that for a given number of moves, a VFT speed of 2-3m/sec is necessary to achieve acceptable delivery times. The capacity for this interbay example is also related to the transport speed as shown in Fig. 2.
The small VFT system modeled in Figs. 1 and 2 achieves a capacity of 28,000 moves/day. Key factors to reaching this capacity included the system speed of 3m/sec, efficient I/O hardware and software, and the elimination of potential bottlenecks at directional changes such as curves (90º) and jogs (<90º).
 Figure 3. A new 300mm full fab model used to evaluate the impact of 3m/sec VFT speed in a full fab implementation. |
Small jogs in the AMHS path are often required to avoid obstacles or fit a system into tight spaces. Turn tables used for directional changes cause the FOUP to stop. A curve that does not make the FOUP stop eliminates these killer bottlenecks.
 Figure 4. Normalized delivery time for three different AMHS configurations, using 25-wafer and 13-wafer lot sizes. |
To evaluate the impact of 3m/sec VFT speed in a full fab implementation, a new 300mm full fab model was developed (Fig. 3). A 20K wafer start per month factory module and full process flow were modeled using three different AMHS configurations. The results are shown (Fig. 4) for 25 wafers per FOUP and for 13 wafers per FOUP. The first AMHS configuration was a conventional global OHT that required 180 vehicles to deliver 13-wafer FOUPs. By adding a 3m/sec VFT interbay system, we could remove 60 OHT vehicles, reducing the total to 120 vehicles, and still achieve a delivery time reduction of 9% for 25 wafers and 12% in the 13-wafer case. By utilizing the full fab 3m/sec VFT with integrated loading functions for the 13-wafer case, the delivery time was reduced by 26%, and the maximum delivery time was reduced by 30%.
A faster AMHS with fewer bottlenecks can reduce the overall fab cycle time, especially during peak load cycles. These benefits are well-known and accepted. Additionally, if the delivery time variations are reduced, then deliveries become more predictable and process equipment can be more accurately scheduled.
In a vehicle-based system, there is delivery variability due to transport vehicle availability (the unpredictable wait for a taxi cab) or the traffic jams that result from concentrated traffic or incidents. The system’s ability to recover is based on the number of vehicles available. In a VFT system, the capacity and move predictability does not depend on vehicle count. Such a system is significantly more predictable and better equipped to absorb perturbations in demand.
Conclusion
Modern 300mm fabs can benefit significantly from a fast-VFT based AMHS that eliminates bottlenecks to provide shorter, more deterministic delivery times. These improved wafer transport times can improve overall fab cycle time, increase equipment utilization, and require less work-in-process inventory for the same fab output.
Acknowledgments
AutoMod is a trademark of Applied Materials. This article has been revised from a previous version published in the September issue of SST Taiwan.
For more information, contact Michael Brain at Aquest Systems Corp., 683 W. Maude Ave., Sunnyvale, CA 94085-3535 United States; ph 408/530-2507, e-mail [email protected].