Understanding Equipment Cost-of-Ownership
Beyond equipment purchases, increased device complexity and manufacturing performance requirements are impacting COO.
Darren L. Dance, David W. Jimenez, Alan L. Levine, Wright Williams & Kelly, Pleasanton, Calif. -- Semiconductor International, 7/1/1998
 |
At a Glance | | |
|
| ||
| |||
Cost-of-Ownership Survey
The use of cost-of-ownership (COO) has expanded significantly since its introduction in early 1990. Initially used in making purchase decisions, increasing semiconductor complexity and manufacturing performance requirements are driving rising semiconductor equipment costs. Simultaneous increased demands for ICs and lower selling prices are creating the need for increased productivity. This has meant an expansion in applications of COO and overall equipment effectiveness (OEE) analysis, a subset of COO. Understanding the true relationships between COO and OEE and knowing how each should be used can lead to improvements in equipment availability and performance, reduced maintenance and service costs, and continuous improvement of equipment designs.
What is COO?
COO is the full cost of embedding, operating and decommissioning a processing system in a factory environment.1 Since SEMATECH began developing a COO model in 1990, COO standards have been published by Semiconductor Equipment and Materials International (SEMI).With a few details about costs and productivity, users can determine the life-cycle cost of owning a semiconductor tool. Useful terms are as follows:
- Fixed costs: Costs incurred once and usually associated with the acquisition and incorporation of equipment into the factory (includes purchase, installation and facility expenditures).
- Operating costs: Costs incurred during equipment use (such as materials, labor, repair, utility and overhead expenses).
- Throughput: The number of wafers per hour the process system delivers to the factory.
- Composite yield: The operational yield of the tool (may include breakage, misprocessing and defects).
- Utilization: The ratio of production time compared to total available time as defined by SEMI E10.2
- Yield loss cost: A measure of the value of wafers and die scrapped through operational losses and defects.
COO calculations use yield models to relate IC yield to circuit and process parameters such as device geometry and particle density. With these details, COO provides an objective analysis for evaluating decisions and a systematic focus on issues that might otherwise be overlooked.
What is OEE?
OEE is a measure of the productivity of manufacturing equipment. Originally developed in Japan as part of total productive maintenance (TPM) to improve overall factory operations, OEE is the efficiency metrics of availability, performance and quality. Equipment availability includes both scheduled and unscheduled equipment downtime, regardless of cause. Performance encompasses standby or idle time, minor interruptions (E10 assists) and reduced equipment production speed. Quality covers loss of productivity that is due to rework and yield losses.
 |
| Fig. 1. Equipment cost and OEE are just two bodies of information needed in determining COO. |
How do COO and OEE relate?
Performance, availability and quality factors in OEE are important input to COO (Fig. 1).3 However, OEE does not take into account any cost-related factors or cost impacts. Both OEE and cost factors are included in COO. Since equipment purchase decisions have long-term impacts on the profitability of semiconductor companies, the more complete estimate of COO forms a better model for analysis of these decisions.
For equipment that is already installed in manufacturing, OEE provides a useful metric for performance analysis. Even if a new tool provided significant improvement in operating costs, it is unlikely to replace an existing tool with satisfactory performance. Tracking OEE will help identify improvements in equipment availability, performance and rate of quality. However, lower OEE does not guarantee lower COO.
The operating costs of installed equipment can also benefit from COO analysis. One user of COO identified photoresist waste as having a significant impact on COO. Reducing the photoresist dispense rate lowered COO, but had no impact on OEE.
Applications of COO and OEE
For many users, buying a new process tool is their first COO application. For tool suppliers, responding to a request for price quotation (RFQ) is often the first. The IC manufacturer either collects specific input to his own COO model, or the supplier provides a complete model. Equipment suppliers and end-users often cooperate on COO analyses. Comparing equipment COO and OEE models is simpler if all models comply with SEMI guidelines; however, clarifying assumption differences is still important.
Implementing COO for new purchases is an important application; however, using COO and OEE to evaluate the long-term benefit of manufacturing modifications may be the biggest value to the organization. This can be done through equipment benchmarking, comparative and bottleneck analysis, project prioritization and process optimization.
Equipment benchmarking
Equipment benchmarking allows the semiconductor equipment supplier or user the ability to gauge their products to industry requirements and best-of-breed performance and processes. The areas of greatest impact on benchmark results are yield, productivity, maintenance and consumables.
A successful COO and OEE benchmark requires good data about fab operating parameters for different IC types (logic, memory, ASIC) and detailed best-of-breed performance specifications. Equipment and material data must represent actual performance and not just paper specifications. These data requirements extend to other types of analysis. Data for COO and OEE analysis often come from historical data analysis, the most frequently used source of data for the semiconductor industry. It is a quick, simple method of gathering productivity data under processing conditions. Common sources of historical data are manufacturing execution systems (MES), work-in-process (WIP) tracking systems, equipment and engineering logs, and spare parts ordering history.
 |
| Fig. 2. Comparisons for OEE (a) and COO (b) indicate greater productivity for on-site service, but at what cost? |
Other common methods for collecting COO and OEE data include work sampling and time studies. Work sampling uses random sampling techniques derived from statistical theory. With careful work sampling, the proportion of time an operation spends in each of the E10 equipment states can be estimated.
Time studies are often called 'stopwatch' studies because of their reliance on direct time measurements of operations parameters. These studies are most useful for estimating throughput rates and cycle times. Video methods can automate both direct time studies and work sampling analyses using a video camera focused on the operation of interest.
Comparative analysis
While benchmarking allows suppliers to gauge individual products and processes against the best-of-breed in each area of interest, comparative analysis provides information on current strengths and weaknesses.
Often the most frequent use of COO and OEE, comparative analysis is used by suppliers to compare themselves to other suppliers and by end-users in purchase decisions. IC manufacturers can also compare themselves with other fabs or manufacturing operations.
Comparisons of COO and OEE are common in equipment, materials and services evaluation. For example, a service contract providing an on-site service technician may cost more than a service contract providing a guaranteed four-hour response time. However, on-site service results in higher equipment availability and improves OEE. Improved OEE is illustrated in Figure 2a, and an increase in COO, including the impact from availability, is shown in Figure 2b. Does the improvement in OEE justify the higher COO? Comparing the two metrics and considering the needs of the user will help clarify the correct decision.
| Table 1. Relative Impact of Improvements | ||
| COO | OEE | |
| Baseline analysis | $5.24 | 74% |
| Reduction in scheduled maintenance | $5.23 | 75% |
| Reduction in equipment cost | $5.22 | 74% |
| Reduction in cost of consumables | $5.21 | 74% |
| Increase in throughput | $5.14 | 78% |
Project prioritization
Which project will provide the largest cost benefit on productivity and manufacturing? The answer can be found by prioritizing engineering projects. Project evaluation starts with a baseline COO analysis on a fully loaded single machine. Once a baseline COO is run, projects of interest can be evaluated. By modifying one parameter at a time, the impact of the specification changes can be determined using COO and OEE. This is illustrated in Table 1, which compares the impact of 5% improvements in scheduled maintenance requirements, cost of consumables, equipment costs and throughput rate on COO and OEE.
Armed with wafer performance information, cost performance and cost of implementation, the return on investment (ROI) can be estimated. In this case, the implementation costs would include equipment and material purchases and other costs incurred by the end-user. This ROI technique is repeated for each engineering project to provide insight into which improvements will provide the largest returns to both the IC manufacturer and the equipment or material supplier: 
Optimization
Optimization is an important tool for COO and OEE analysis. Optimization in this case is not a technical optimum; it is a cost optimum. Essentially, it asks: What is the most cost-effective method to do the job? An optimization analysis is a means for evaluating changes to existing equipment and materials.
 |
| Fig. 3. Optimal lamp replacement strategy is determined by using COO analysis. |
Figure 3 illustrates the use of COO to determine the optimal lamp replacement strategy for a stepper.
As lamp intensity degrades over time with use, the throughput of the stepper is reduced. However, lamps are expensive, and the maintenance time to replace a lamp reduces production availability. This complex optimization study requires information about the lamp intensity curve over time and the cost of replacing the lamp. The analysis estimated average production throughput and annual maintenance impacts for several lamp replacement frequencies. The throughput rate, maintenance downtime and lamp cost drive changes in COO. The optimum lamp replacement strategy will minimize financial and COO losses. If the lamp is not replaced often enough, manufacturing will lose money because of performance losses. If the lamp is replaced too frequently, manufacturing will lose money because of lost availability and higher lamp costs.
Sometimes optimization only requires improved operating methods. The enhancements yielding the lowest COO performance will most likely be given priority. The least expensive materials and equipment for today's process, however, may not have the lowest cost over the life of the equipment.
| COO Modeling for Equipment Engineering Maurice Cloutier, Like device manufacturers, equipment suppliers use cost-of-ownership (COO) modeling to evaluate system costs. Performing COO analysis and benchmarking, they use the models to make design improvements to lower COO and to demonstrate those improvements to customers. Using the data generated, suppliers are able to prioritize engineering projects and implement improvements that will have the greatest effects on minimizing COO and maximizing wafer output for their customers. COO data can also be used to better understand the cost implications of various system integration and process decisions for device manufacturing. Throughput improvements assumed an even higher engineering priority at Lam (Fremont, Calif.) after data acquired using the TWO-COOL COO model indicated its significant impact on COO. After conducting a sensitivity analysis using typical performance parameters, it became clear that raw throughput has a much greater impact on COO and wafer output per week than other variables. For example, a 6.5% increase in throughput, from 46 wph to 49 wph, on a metal etch system will yield an improvement of 6.1% in wafer output per week and a reduction in COO of 4.6%. However, on the same tool, a mean time between failure (MTBF) improvement from 250 hours to 500 hours (50%) will yield only a 1% improvement in wafer output and less than a 0.05% reduction in COO. COO modeling to quantify alternatives is another way of assisting customers with process and system integration decisions. A device manufacturer, for example, may etch a number of dissimilar films such as a BARC/TEOS/WSix/poly gate stack in a poly etch system in situ or choose a three-chamber ex situ approach. Comparisons indicate that the in situ process provides higher throughput, 42 wph compared to 33 wph for ex situ, has much lower capital requirements since fewer modules are needed and thus has $1 lower COO per wafer. When customers are given such cost and process information, they can make informed decisions for specific applications. COO modeling, more than a tool for purchasing cost analysis, allows equipment suppliers to focus engineering efforts on the critical factors that impact equipment COO and develop lower-cost systems. |
Bottleneck analysis
OEE is most critical for bottleneck tools. Some have suggested it is critical only for bottlenecks. While that view may be overly narrow, OEE is clearly a powerful tool for systems that can limit factory productivity. Given a choice, a bottleneck ought to be at the most expensive processes or tools in a factory. Looking at it from the opposite perspective makes this clear. It would be ridiculous to limit the capacity of a $1 billion factory because one did not purchase enough $1000 microscopes. In wafer fabs, the bottleneck is most frequently at the exposure tools. In assembly houses, frequently the testers are the bottleneck. In mask manufacturing, the mask writer or inspection systems are typically the bottleneck.
Figure 4 illustrates how the bottleneck tool (shown in green) limits the capacity of this process. Both OEE and COO are important in analyzing the productivity of this tool as well as for critical tools (shown in yellow), which could becomes bottlenecks because of reliability or schedule changes. COO is less important for a noncritical tool (shown in gray) but is very useful in understanding the cost of the total process sequence. OEE is unimportant for noncritical tools since they are not a capacity constraint, and any changes could divert resources from critical and bottleneck tools.
Clearly, optimizing expensive bottleneck processes for cost will have the greatest leverage on profit. COO is the most effective way to measure that cost optimum. It is not surprising that in each of the cases cited above (steppers for wafer fab, etc.), mix-and-match strategies have been applied to reduce COO for those process steps. What becomes clear from this perspective is that OEE is not separate from COO. COO measures the cost impact of OEE performance improvements. When an OEE improvement project is undertaken, COO should also be analyzed.
Benefits of standards
Standardization is critical. Many models have been developed, yet few comply with existing and proposed standards. This is especially critical as both suppliers and IC manufacturers jointly use this information. Ignoring the standards can result in significant inefficiencies. For example, an IC manufacturer having three different models to test the exact same process ended up with three different results. One man-week was spent trying to reconcile the models, without success. Another IC manufacturer indicated that the number of different versions of COO equaled the number of users -- each person had created his own custom version. Add the number of versions developed by suppliers, and you create a Tower of Babel. Clearly, this makes cooperative work nearly impossible.
Standards, including E10, E35 and the proposed OEE standard, have gone a long way to mitigate this problem. COO and OEE benefit from clear definitions of terms. This is especially important since many of these terms are shared among suppliers and IC manufacturers, providing communication among all parties. They can speak the same language, comparing similar data and costs using standard software and equations.
 |
| Fig. 4. Tools that create bottlenecks, such as the 'green tool,' are clearly identified during process sequence COO analysis. |
References
