Analyzing the benefits of 300mm conveyor-based AMHS
07/01/2002
By Frederika Tausch, Asyst Technologies, Austin, Texas
Gabriel Gaxiola, Lawrence Hennessy, IDC, Tempe, Arizona
Overview
While the need for automation in 300mm fabs is not debated, the form and performance of such automation is still in question. Software simulation that compares conveyor-based continuous flow transport technology to conventional car-based wafer-lot delivery has detailed delivery time and throughput advantages to the former.
Differences exist between conveyor style and conventional monorail automation systems being installed in 300mm factories. We have studied the performance of both — continuous flow transport (CFT) and car-based — as interbay and intrabay automated material handling systems (AMHS), specifically a configuration to support the IDC 300mm virtual planning fab (VPF).
VPF is a "farm" fab configuration that supports 30,000 wafer starts/month [1]. The process flow is based on International Sematech's 130nm technology with the addition of processing for copper interconnect. We obtained process tools to simulate and associated data for our study from IDC's internal 300mm tool database. Our evaluations used AutoMod 10.0 software simulation.
The car-based system in our study has three discrete segments: interbay transport, intrabay transport, and stockers. A single car delivers lots between stockers for interbay transfer. Stockers are located at the head of bays and used for local buffer storage and pass-through between interbay and intrabay systems. The intrabay system is a traditional hoist system. In our simulation, interconnection between like bays via the intrabay hoist was allowed as defined by the process flow. We also included reasonable routing for both the interbay and intrabay systems.
 Figure 1. Asyst's FasTrack continuous flow transport (CFT) system. |
The CFT AMHS in our study consisted of asynchronous conveyors with turntables that move lots independently (Fig. 1). Lots use the conveyor as a means of delivery directly to a tool-loading module. The tool-loading module combines storage capability with intrabay automation to provide buffer storage for tools and direct tool loading. (In the discussion below, the stocker for the CFT system refers to this distributed bay storage.)
AMHS analysis
In our study, we first needed to develop a method and metrics to compare the two types of AMHS. Then, we needed to provide initial meaningful results that would suggest further analysis for AMHS comparison.
 Figure 2. a) Car-based and b) CFT-based lot movement. |
Metrics were a challenge to this study because the two systems operate differently (Fig. 2). With the car-based system, stockers are located at the head of a bay to provide intermediate storage for work in progress (WIP) prior to movement to the tool load ports. With the CFT system, intermediate WIP storage is located in buffers on the tool face. This leads to a significant difference when comparing intrabay delivery times.
Overall travel distance for the CFT, which travels into the bay, is longer than the car-based system. We chose not to use distance traveled as a metric in our analysis because of its static nature. These differences will become apparent below in the comparison of delivery and transport time results from the simulation.
In the end, we selected delivery time, transport time, throughput volume, throughput variability, and maximum throughput capability as our metrics:
- Interbay delivery time is measured from the time when a vehicle call is initiated at the source stocker to the time when the lot is placed on a shelf at the destination stocker.
- Interbay transport time is measured from a lot's removal from a shelf in the source stocker to its placement on the shelf at the destination stocker. (Transport time included two stocker robot cycles for the car-based system or the travel time on the rail for the CFT conveyor system to placement on the destination buffer system.)
- Intrabay delivery time is measured from the time when a vehicle is called at the source stocker or tool to the time when the lot is placed on the port or shelf of the destination tool or stocker.
- Intrabay transport time is measured from the time when a lot is removed from a shelf or port in the source stocker or tool to the time when the lot was placed on the shelf or port at the destination stocker or tool.
- Throughput variability is defined as the coefficient of variability (CV) — the standard deviation of the values observed divided by the average of those same values (a commonly used measure).
- Throughput is simply moves completed/hour.
From-to requirements for both the CFT and car-based system were generated per the layout and process flow. We conducted our simulation runs over eleven 24-hour days with the first day's data — during which the simulation reached a steady state — discarded.
Because CFT is an interconnected system and the car-based system uses stocker pass-through, overall system utilization has to be calculated differently. For the CFT system, we included conventional interbay moves plus intrabay moves in the simulation so that traffic effects of the intrabay moves weigh on system capability to meet throughput. The car-based system contains only conventional moves between bays, and data is reported only for moves that are traveling between bays for the interbay analysis.
In our tests, we simulated four high-volume bays — one diffusion bay and three photolithography bays. Both interbay and intrabay systems were evaluated at average throughput as defined by throughput and process flow requirements. A secondary set of runs was conducted to measure the maximum throughput capability of the systems.
 Figure 3. Interbay delivery time distribution. |
Interbay simulation results
For interbay-only moves, the CFT and car-based systems were both able to meet the required moves/hour (Table 1). Faster overall delivery times were realized with the CFT because the conveyor resource was not subject to waiting for vehicles as was the car-based system. This characteristic also leads to a tighter distribution of lot delivery times (Fig. 3). Transport-time data comparison shows the impact that overall distance traveled has on transport time of a lot during interbay travel. It is important here not to consider this a negative for the CFT system. Comparison of the average transport and average delivery time data shows no difference for the CFT system. This illustrates the always-available resource of the conveyor system.
 |
The always-available feature of the CFT system also demonstrated reduced variability in meeting output moves/hour. CV comparison shows a 300% reduction in variability of the output data for the CFT system. Maximum interbay throughput was tested on both the CFT and car-based transport systems. The results demonstrated that the CFT system was able to meet 2.5x the original throughput requirements with an average delivery time of 230 sec, while the car-based system was able to meet only 1.5x the original throughput requirements with an average delivery time of 352 sec. The car-based system became limited by stocker throughput capability and vehicle utilization requirements.
Intrabay simulation results
When we compared the intrabay output results from the two highest throughput bays (Table 2), on average the CFT system delivered lots 271 sec faster than the car-based system. In addition to an average faster delivery time, the average travel time is 156 sec faster for the CFT. This is because lots are being stored at the tool-face, whereas the car-based system locates lots in the stocker at the head of the bay. This results in a high concentration of moves being completed from the buffer to the tool load port in <1 min.
 |
Variability comparison for the intrabay system also showed more balanced delivery to tools. While the system is still dependent on the bay robot for delivery, the close proximity from storage to tool significantly helped dampen variability. The CV results for each bay are given in Table 3.
 |
Along with variability, we tested the maximum throughput of both systems, with similuations increasing throughput 1.5, 2.0, and 2.5 times. We did this to test the robustness of each system. In three of the four bays, the CFT was able to meet up to 2.5x the throughput requirement, while keeping delivery time <1 min. In contrast, the car-based system was able to achieve 2.5x the throughput for only two bays with an average delivery time of 6-8 min.
Conclusion
Our simulation work has shown that compared to a car-based wafer-lot delivery system, a conveyor system gives faster delivery times and imparts less variability in a manufacturing operation. One reason for this is that vehicles on a car-based system split their time in three modes — transporting lots, moving idle on the track, and moving to pick up a lot — as shown by the difference between average interbay transport and delivery time. A conveyor system is also able to provide a greater throughput bandwidth to support periods of elevated operation and reduced time to load port. In continued work, we are carrying out simulations that integrate a conveyor system into a manufacturing model to determine the effect these efficiencies have on factory cycle time, WIP management, and hot-lot support.
Acknowledgments
FasTrack is a trademark of Asyst Technologies Inc.
Reference
1. This generic factory configuration was developed by IDC as a 300mm factory planning tool. It is a combination of 300mm process and tool data.
Frederika Tausch received her BS from Southwest Texas State University. She is a senior simulation engineer in the Automated Material Handling Systems Department at Asyst Technologies, 11000 N. MoPac Expwy., Suite 100, Austin, TX 78759; ph 512/340-1030, fax 512/340-7530 , e-mail [email protected].
Gabriel Gaxiola received his MS in industrial engineering from Arizona State University. He works in the industrial engineering department at IDC.
Lawrence Hennessy received his BS in mechanical engineering from Temple University in Philadelphia, PA. He is the technical manager of material handling in IDC's industrial engineering group.