Integration Key for 300 mm Fab Automation
Clint Haris Brooks Automation Inc. Chelmsford, Mass. -- Semiconductor International, 6/1/2001
 | At a Glance | | |
| | ||
| |||
Pervasive fab automation is one of the themes that IC manufacturers have embraced as an enabler of cost-effective production. However, adding automation indiscriminately does not guarantee cost reductions for 300 mm fabs. In fact, just as 300 mm has been driven by the vision of cost-effective manufacturing, the use of 300 mm automation must be driven by the same philosophy.
One of the interesting trends emerging as a result of this focus on cost-effective automation is the tendency for companies to approach 300 mm automation holistically. Companies are realizing that it is not the individual automation components, but how they are integrated into one system that defines the success of a 300 mm project. This article will discuss some of the key integration challenges that highly automated fabs and tool OEMs are facing, as well as a few areas that are unique to the 300 mm era of fabs.
What's different
Many articles have been devoted to the challenges and the differences of 300 mm factories, but it is still worthwhile to restate the three main issues that chipmakers face in building a 300 mm fab:
Capacity— A 300 mm factory, in general, will produce 2.4 times the number of die as a 200 mm factory. Therefore, the first question that an IC manufacturer needs to answer before building a 300 mm fab: "Do I need the additional capacity?" The straightforward reason that there are not more 300 mm fabs in the world today is because the answer to this question has more often than not been "No."
The second, more challenging question related to the additional capacity brought by a 300 mm fab: "How will I fill the factory?" IC manufacturers with a high-volume, low-mix product portfolio such as DRAM producers have an easy answer to this question. Unfortunately, the high-volume, high-mix product companies such as ASIC manufacturers have more difficulty solving this issue. The companies planning for a high mix of products in their 300 mm fabs are being driven toward more sophisticated automation requirements such as the capability of supporting multiple lots in one carrier, as well as a closer coupling between fab operation and capacity planning systems to address this issue.
Cost— 300 mm factories are expected to cost upwards of $3B. Although advanced 200 mm factories are also becoming more expensive, the additional 20-40% premium for a 300 mm fab makes the endeavor daunting to all but the largest IC makers. Government subsidies, joint ventures, and guaranteed capacity by customers are examples of methods that are being used worldwide to finance 300 mm fab plans.
Automation— 300 mm factories require a much higher level of automation for two reasons. First, the weight of the 300 mm wafer carriers (front-opening unified pods, or FOUPs) exceeds the recommended limits set forth by numerous international organizations for workers to repetitively move. In fact, several 300 mm pilot production facilities have already reported incidents of operator injury caused by manually handling FOUPs. The second reason that fabs are migrating toward full automation is because of the vision that increased automation will lead toward increased factory efficiency.
New in automation
 |
|
|
1. Of the next 11 300 mm fabs being built, eight of them (73%) are opting for a new next-generation MES and fabwide automation architecture to support requirements. |
This leads to the question: What's so new and different about 300 mm fab automation that is driving IC manufacturers to completely replace their automation systems for 300 mm? The answer to this question lies not in the individual components of the automation system but in their integration.
Integration importance
Historically, companies have focused on each component individually and made selections based on best-of-breed evaluations. Following this selection process, it has been the responsibility of the chip manufacturer to integrate the pieces. The success or failure of an IC manufacturer's automation system has always been directly tied to the integration of the various automation components.
Fabs that are judged to have best-in-class automation systems are not identified as best in class because they have installed one software package or one particular type of automation hardware. Fabs are judged to be best in class because they have seamlessly integrated their automation systems in such a way that the whole is greater than the sum of its parts (Fig. 2).
|
|
|
2. It is important to integrate automation systems in such a way that the whole is greater than the sum of its parts. |
Increasingly, fabs as well as tool OEMs are turning toward selecting pre-integrated automation solutions from one or at most two automation vendors to realize their aggressive timelines. The reasoning behind this philosophy is twofold: 1) Companies can reduce the time required to integrate and test all of their automation systems by selecting a unified automation architecture; and 2) By outsourcing the automation implementation, companies are left with more time and resources to focus on their core businesses.
Tight integration must not come at the expense of flexibility, however. While a close-knit system of automation components is required, it is equally critical that individual elements can be removed, added or changed as a fab evolves or business models change.
To understand how 300 mm fabs (as well as highly automated 200 mm fabs) are integrating what were often disparate automation systems, it is useful to review some real-world examples of factory automation integration issues. The following five case studies are examples of new or challenging aspects of automation integration currently being worked on in 300 mm fabs/pilot lines today.
Case study #1: Integrating disparate databases
The first critical automation integration issue faced by 300 mm factories is the management or integration of data that were previously stored in disparate databases. For example, the successful automation of a 300 mm fab will require real-time association of data related to the individual process tools, their load ports, their chamber availability, their scheduled maintenance plans, the material control routing, carrier management data and single-wafer tracking information (to name a few).
An example of the advantages of an integrated database methodology is illustrated with a 300 mm fab that uses an SPC violation following a post-lithography inspection to automatically trigger the maintenance management system to take the litho cell out of service. This in turn automatically dispatches the reticles and material to another cell, and finally causes a report to be generated detailing which wafers have been through the cell in the past 24 hours.
To achieve this level of functionality, a majority of fabs are selecting a next-generation MES system with a unified database that supports the advanced requirements of 300 mm such as single-wafer tracking (e.g. Brooks Automation's FACTORYworks MES).
Those fabs that are not taking this approach must add additional databases and barnacle applications to increase the functionality of their legacy MES systems.
Extending legacy MES systems is feasible given enough time, money and resources. However, some fabs with legacy systems are increasingly running into difficulties not only with ensuring transparent connectivity between databases, but also with the more critical concern of maintaining data integrity/consistency.
These challenges will most likely result in the transition of all fabs using legacy MES architectures to MES systems that have been designed from the ground up to support 300 mm and its increasingly complex data requirements.
Case study #2: Integrating material control, MES, scheduling
The 300 mm requirement of full intrabay delivery is another area where a fab's automation systems must be tightly integrated to function properly. A 300 mm material control system (MCS) is much more dependent on the other CIM systems because decisions that were formerly made by operators are now made by a combination of the MES, the scheduling and dispatch systems, the reticle management system, etc. In other words, if one considers the two primary interface layers of the MCS — the interface to the AMHS equipment, and the interface to the CIM system — the CIM system interface is much more critical in 300 mm factories than it has been in 200 mm ones.
Some 300 mm AMHS vendors are currently bundling homegrown proprietary MCS as part of their package deal to the fabs. The potential advantage to this approach is that the AMHS software and hardware comes from one vendor. However, one of the emerging difficulties with this philosophy is that most AMHS vendors only have experience with the AMHS equipment interface layer, not the increasingly important CIM system layer of the MCS interface.
A second issue with the approach of using the AMHS vendor's MCS is that the fab is more often than not locked in to using the hardware from that particular AMHS vendor for any and all fab expansions. This restriction is contrary to the general industry guideline of MCS interoperability between the various vendors' interbay, intrabay and stocker systems (a requirement that is spelled out in detail in the SEMI standards E82 (IBSEM) & E88 (Stocker SEM)).
| |
| 3. The idea of integrating sorters and stockers is gaining momentum with some of the early 300 mm adopters. |
The MCS is an excellent example of an automation component that has evolved from its 200 mm usage as basically a stand-alone system to a 300 mm model in which it is part of the overall integrated automation solution. However, fabs are increasingly realizing that this tight integration of the MCS component of the automation system must not come at the expense of the ability to change or add different AMHS hardware.
Case study #3: Integrating sorters and stockers
Another example of disparate automation systems being merged together into a unified system for 300 mm is illustrated in the integration of sorters and stockers. The idea, which first appeared in 200 mm SMIF fabs some years ago, is gaining momentum with some of the early 300 mm adopters. In fact, one company that ordered Brooks' 300 mm MapTrax sorters chose for ~70% to be fully integrated with stockers (Fig. 3).
There are four driving factors behind the desire to integrate sorters with stockers:
- Cycle-time reduction— Carriers spend a significant amount of their fab time waiting in stockers for tools to become available. Ultimately, minimizing this wait time is one of the best ways for a fab to decrease cycle time. However, an alternative cycle-time reduction method is to use the time that the carriers spend in the stocker for another purpose — sorting.
- Contamination control— Because almost all 300 mm fabs are planning to support copper processing (certainly all non-DRAM fabs will), fabs are planning their contamination control strategies to be integrated into the automation systems. One strategy for isolation is to designate individual bays or tools as only allowing particular types of carriers. An integrated sorter/stocker can be used to exchange wafers from one carrier type to another at various stages of the process, therefore ensuring that contaminated carriers do not enter inappropriate sections of the fab.
- Test wafer storage— Another use for sorter/stockers is as a test wafer or, more generically, non-productive wafer (NPW) dispenser. 300 mm fabs are requiring the complete tracking and control of all NPWs and, by storing all wafers in stockers and only dispensing individual wafers or groups of wafers as needed, sorter/stockers serve as a valuable element of a fabwide NPW methodology.
- Footprint reduction— Finally, an integrated sorter/stocker uses less cleanroom space than separate systems do. There will always be a need for individual sorters in a fab, but the deployment of a few sorter/stocker systems in a fab can ultimately result in cleanroom space savings.
Case study #4: Integrating APC and station controllers
Fabwide APC deployment is another thrust that, while not specific to 300 mm, is certainly viewed as critical in 300 mm fabs because of the increased value of each processed wafer. Increasingly, APC is moving from a stand-alone initiative within certain process areas to a tightly integrated component of the automation methodology of 300 mm fabs.
One logical integration point for APC is at the station controller level. APC is predicated on the ability to collect equipment-generated data, perform the appropriate analysis, and feed back any resultant process condition changes to the process tool. To ensure that this is done consistently, it is essential that the data collected and the process changes generated are derived from, and delivered to, the factory floor equipment in an automated fashion. The station controller employed to connect the MES to the factory floor equipment is the natural conduit for this connection to the process tool.
The station controller is the repository for the definition of both the data variables available for collection and the run state model for the process tool. The timely and appropriate collection of these along with the current run state context (e.g. recipe, step, MES lot or batch, etc.) is essential to allow the appropriate APC analysis of the raw data. Tight integration of the APC application with the station controller therefore allows for both the raw data collection and essential run context tagging.
The station controller is also the remote control conduit to the rest of the CIM system. In a fully automated 300 mm fab, it is essential that there is only one owner of this conduit to avoid conflicting control requests. Having collected and delivered the run context tagged data to the APC application, it is then the station controller's job to update the process condition with the result of the SPC analysis. This could be an amendment to recipe conditions (e.g. etch time), prohibition of further use of a particular process for production until engineering intervention, prohibition of further production use of the equipment entirely, or some other tool-based activity required by the APC analysis. Close integration of the APC application with the station controller therefore is mandatory to allow the 300 mm fab to operate in a "lights out" or at least "semi-lights out" mode.
Case study #5: Integrating hardware, controls, standards
Close integration of what were once disparate automation systems is an issue that 300 mm tool OEMs must face as well. For example, the new SEMI standards that define the method in which 300 mm tools accept, verify and process material in a fully automated environment (e.g. E87 — Carrier Management Standard, E40 — Substrate Tracking, E40 — Process Jobs, E94 — Control Job Management, etc.) impose an increased level of automation systems integration. This is because the new 300 mm standards reach much deeper into the tool functionality than the 200 mm standards ever did.
|
|
Moreover, the 300 mm standards not only cover the operating functionality of individual components of a tool (e.g. the equipment front-end module (EFEM), the process chambers, the control architecture, etc.), but also how these components interact together (Fig. 4).
Because of the increased complexity of the new 300 mm information and control standards as well as the fact that they have been written with a "black-box" philosophy (in other words, they describe what is needed, but they don't describe how it should be done), more and more tool suppliers are starting to outsource the automation integration challenge to third-party companies.
This drive toward automation outsourcing at the tool OEM level is beneficial for primarily three reasons: 1) The tool OEM can focus efforts on its core competence/process technology; 2) The IC manufacturers will ultimately receive more "standardized" automation interfaces in the tools delivered to their factories; and 3) The tool OEM can deliver an integrated automation solution much faster than if it attempts it alone.
The companies best suited to support this drive toward automation outsourcing are those that have core competencies in 300 mm tool hardware (i.e. robotics, loadports, mini-environments, etc.), 300 mm factory software (i.e. MES, station controllers, MCS, etc.) and 300 mm tool control (i.e. embedded controllers, SECS/GEM interfaces, APC, etc.).
Automation integration key
Automation is a key enabler of the 20-40% cost reduction that IC manufacturers expect as they migrate toward 300 mm. The success of 300 mm automation is directly tied to how tightly one is able to achieve integration between the various components of the factory system. Several components that were once disparate (e.g. sorters and stockers, MCS and MES, SEMI standards and tool control, etc.) must share information, components and functionality in order to work well in an integrated 300 mm automation system.
Any fab or tool supplier can achieve automation integration given enough time, money and resources. However, the successful 300 mm implementers will obtain cost-effective automation by taking a ground-up, fully integrated approach toward the development and deployment of their automation systems.
Clint Haris is responsible for 300 mm strategic marketing at Brooks Automation. Prior to joining Brooks in the fall of 2000, he worked for Motorola for six years, responsible for 300 mm automation on the technical staff. While there, he was part of the team responsible for producing the world's first patterned 300 mm wafers, the first working transistors on 300 mm wafers and the first 300 mm production pilot line, Semiconductor300. Haris has a B.S. and M.S. from Cornell University (Ithaca, N.Y.).
