300-mm: How will production software keep up?
03/01/1998
300-mm: How will production software keep up?
Julie Fraser, Industry Directions Inc., Newburyport, Massachusetts
As a semiconductor fabrication company manager or employee, what would you do to ensure everything was processed correctly if a single batch of product through a furnace could have a selling price of over $21 million? This is a question IC makers will face in the next few years as 300-mm wafers become available for processing.
While 300-mm fabs are a logical next step, there are some important differences between them and even automated 200-mm fabs. Fab executives and information systems staffs must consider how these differences will affect what is economical, logical, and necessary in these new fabs. Since the wafers are not yet available, and I300I standards are just coming together, we`ve made some assumptions based on calculations by Dave Ferneyhough of FASTech Integration (see table).
 |
At these prices, it`s clear no one in a fab can afford to make mistakes. Automation will rule for economic as well as ergonomic reasons (40-lb SMIF pods are not for people to carry). The economies for this new generation of wafers demand some differences in fab-level software from our current generation - in MES, scheduling, cell control, document management, and materials handling.
For 300 mm, fab software must better support overall equipment effectiveness (OEE). From a functional perspective, that means better validation of each move before processing begins, recipe download assurance, intra-bay materials handling and tracking, strong production scheduling, and rapid analysis of an exploding amount of process data. Yield management will be far more pressing than it has ever been. While software providers are moving in the right direction, their customers need to pay attention and push them to ensure all the pieces are ready when the 300-mm fabs are, as soon as two years to full production for some companies.
Functionality is only one hurdle. More pressing is the ability to integrate these software functions in a tight and coherent way. The SEMATECH CIM Framework is pointing in the right direction for this, but software development takes time. The challenge for software providers is to implement their own proprietary message-passing bus protocols throughout their software, and that from partners, in just 24 months. This change in architecture has been underway for years, but software vendors and users have much work ahead to stay afloat during the change to 300 mm and integrated architectures.
Software: Moving to integrated not monolithic
Semiconductor fab software must integrate more easily into a seamless-looking information environment. This means no paper is exchanged in the fab, no information is lost from one function or area of the fab to the next, and equipment and materials handling are sent the correct recipes, specifications, and instructions reliably - with minimal intervention under expected operating circumstances. This is the "paperless fab" in theory.
Once a 300-mm wafer gets made into chips with 25- and 18-nm geometries, the issue of particulate contamination becomes important. Even more to the point are the issues surrounding speed, error reduction, utilization, costs, ramp-up, and ongoing improvement speed. It`s clear that beyond what has traditionally been integrated into a single MES system - WIP tracking, recipe management, equipment communications, etc. - entire other spheres of function must pass information seamlessly to achieve the levels of productivity, accuracy, and OEE required with 300-mm wafer production.
So what is the software suite for paperless processing in a 300-mm fab? The CIM Framework components spell out the functional areas quite well: factory services; management of factory, materials, specifications, plans, schedules, and labor; and process and machine control. While these functions are available somewhere on the commercial market, they are usually not well integrated. No major supplier offers a packaged system with full MES, cell control, document or specification management, materials handling, scheduling, process engineering, maintenance, and test automation capabilities. Some key software providers are adding on to their suites each year, yet the systems are not yet smoothly integrated and full-featured.
Fully automated fabs have already driven software vendors to create message bus architectures that are like CIM Framework. MES vendors, materials handling vendors, and others have generated integration architectures that are useful today - as long as all of the products happen to be compatible.
MES vendors, as the traditional hub of information systems in semiconductor fabs, are a logical source of integration schemes. They are all moving in that direction. However, because of their significant installed base (and the complexity of the software involved) market leaders Consilium and Promis are replacing their old code over time with newer modules that can communicate easily in the Framework style of message passing. Camstar and FASTech have their own versions of component architecture that use some similar concepts to the CIM Framework, but some different technologies than the final specification may use. IBM has been working to develop a CIM Framework-compliant product, but its delivery date has been moved back. Texas Instruments, the model for CIM Framework components, has made few attempts at commercializing its software or marketing it publicly.
As standards are finally published for the CIM Framework and the underlying CORBA standards on which it is based, we expect the vendors to take on these projects with some vigor. Those who purchase fab software must pay attention to the standards development, and funding efforts to migrate these products, as rapidly as new 300-mm fabs need. As not all of the MES vendors are profitable and well capitalized, the effort to get onto a new architecture - even if it appears close to the one they`ve been working on - may be tougher than they will admit.
Most of the MES vendors have had to add functionality to their products to address fully automated fabs. The requirements become even more urgent as IC makers move to 300 mm. Toward the top of the list is effective single-wafer tracking, coupled with an ability to account for multiple products on a single wafer. So more die per wafer, tracking each wafer (not just each lot or pod) and mixing products on a wafer results in the MES handling orders-of-magnitude more incoming data than in less-automated fabs.
In addition to tracking each piece of equipment, the MES and cell controller must now model each bay of the equipment, to track results of each individual process for engineering analysis. Mechanisms for handling all of this data are distributed processing, distributed databases, and distributed applications. Moreover, data warehousing, on-line analytical processing, and similar technologies are expected to blossom.
Beyond these MES capabilities, the architecture for a 300-mm fab must include seamless looking integration with all of these functions:
document management,
cell and station control,
scheduling,
maintenance, and
materials handling and auto ID.
Document management. To get to a truly paperless environment, cleanroom technicians need instant access to the latest specifications, recipes, certification documents, and policies. In back-end operations, the test plans and bond diagrams play a role as well. Unless the document management system keeps up with the latest engineering changes, the risk of mis-processing a lot or furnace batch can be very expensive. By dividing up storage of specifications in a document management application from other process data in the MES, it also lessens the burden of data on the central MES. Any document management system should be interfaced right through the MES, so it becomes a natural part of the operator`s workday. Consilium and Documentum have developed this type of interface, and Documentum bought integrator WMI, with its semiconductor-focused ECN Fastrack application.
Cell and station control. This technology is one of the keys to linking MES to the automation and materials handling systems that will be ubiquitous for 300-mm processing. As equipment vendors include intelligence in each bay, the cell controller assists in translating that data for higher-level systems.
Scheduling. A host of scheduling technologies are being offered for semiconductor fabs, but few have been well integrated into the MES and materials handling systems to ensure full leverage. The entire focus of a scheduling system is on optimum processing for equipment utilization and effectiveness, taking into consideration all of the constraints of processing. 300-mm fabs will not be profitable without this
optimization of process resource utilization, and throughput of products. As systems mature, there should be tighter integration of scheduling to MES. The best path for the long term may be through acquisition, as Promis did several years ago.
Maintenance. While some of the MES products address equipment maintenance through a resource module, FASTech has gone the next step and purchased Midas. Sophisticated maintenance management software - using preventative and predictive logic - is commercially available, but not commonly integrated with other software products. The time is coming for these functions to become part of the semiconductor fab suite. They should integrate with scheduling and MES functions, and take cues from data collection and engineering analysis to develop appropriate strategies for 100% uptime.
Materials handling and auto ID. Materials handling may be the most critical piece of this seamless fab information handling environment. While the data explosion is not as evident at this level, the demands for tracking specific wafers through the entire
process, bay-by-bay, puts pressure on these systems to perform. The size and weight of 300-mm wafers means that highly automated systems will become the norm. Due to the value of each wafer and the chips on it, the materials handling system will need to perform many of the data collection functions as well as some materials management. In-situ testing results may pass through the materials handling system, and poor results could determine when to stop feeding materials to a particular process step. While the intelligence for much of this will rest with scheduling and MES, the materials handling system will hold a key to the success of 300-mm processing.
How to prepare
Certainly, the 13 companies involved in I300I - and those in Japan involved in J300 - are preparing for 300 mm. Some of these companies plan to bring 300-mm fabs into production in 2000. Even IC makers who don`t expect to move into 300 mm for several years beyond that should begin now to create the right environment to ensure success.
Internally, fab managers need to begin educating their workforces about the changes that are coming down the road. Those in the bunny suits are generally the most knowledgeable about the challenges they face today. If they are inspired to start thinking about how the economics and mechanics of a fab will work in the future, they will no doubt contribute to success, possibly sooner than the company`s first 300-mm fab begins operating.
The disciplines and mindset of ensuring top yield and OEE are permeating the industry, but there is always improvement available. Further, the new ideas fostered when fab managers bring accuracy, timeliness, and integration to new levels of urgency may begin to self-select the best teams for new automated operations.
There is only so much a semiconductor fab can do on its own. Fab managers and engineers must ask their software providers to demonstrate their capabilities in all of the areas listed above. Further, software buyers should ask about:
their suppliers` timeline for compliance to the SEMATECH CIM Framework, which appears to be the emerging standard for ensuring open interoperability;
their partnership intentions with providers of other functions; and
openness to the integration services companies who know each fab best.
Many alliances are being forged in the industry, and it is part of the semiconductor customer`s job to test how genuinely those alliances will meet the fab`s needs.
Because of the pressure on OEE in automated fabs, both 200- and 300-mm, the vendors must also work to eliminate downtime due to software. Software users need to configure, upgrade, and migrate software on-the-fly. The era of multiple-day shutdowns to move to the next version of an MES product are gone.
Chipmakers need to give something back to their vendors. These companies are usually a fraction the size of most semiconductor companies. The director of IT should provide assistance to them, by assigning someone to sit on advisory councils and co-funding product development efforts. Each company`s voice is heard most clearly when knowledgeable representatives are present. Every executive and employee will have a stake in the success of integrated fab software when that $21 million batch of wafers goes through the oven. Better to invest now, while the design process is early, than pray it works when the 300-mm fabs need to be in full production.
Julie Fraser is principal and director of market strategies for Industry Directions Inc., a market analyst firm that applies its manufacturing, supply chain, and IT industry intelligence to strategic custom projects. Industry Directions Inc., 6 Broad St., Newburyport, MA 01950; ph 800/635-2175, e-mail [email protected].