|
|
 |
|
|
Yield Analysis in the Semiconductor Industry
Analysis Challenges in the Semiconductor Industry and How the PV-WAVE Family of Products Delivers Solutions
A Business White Paper by Visual Numerics, Inc. - May 2002
Introduction
 The Semiconductor industry includes manufacturers and engineering data analysis solution providers interested in yield management solutions. Semiconductor companies are faced with tremendous time to market pressures due to product cycles getting shorter and shorter with each new generation of wafer technology and selling prices declining rapidly after new products are introduced.
To remain competitive, companies need to send their
products quickly to market with the highest yield production
possible. To achieve this, yield management has emerged as
an area of increased importance and plays a significant role in technology
development and continuous manufacturing process improvements. One
extremely important component of yield management is effective yield analysis in
order to achieve accurate yield prediction and ensure the right level of work in
progress (WIP).
The PV-WAVE family - PV-WAVE, JWAVE, and TS-WAVE - as well as our
customizable yield analysis technical frameworks and strong consulting expertise
in the semiconductor market offer tailored made solutions particularly suited to
meeting the yield analysis challenges of the semiconductor industry.
Yield Analysis Requirements in the Semiconductor Industry
Effective yield analysis is an integral component in driving quality products to
market as quickly as possible with optimal yield output - which maximizes
revenue. Once a product is in production, continuous improvement to the
process can further increase quality yield output. Rapid data collection and
reliable analysis reduces the response time to yield and process problems.
Whether you need to improve time to volume, quickly identify yield issues,
improve yield visibility, analyze split lot or lot quality, perform outlier or
tend analysis or accelerate your yield learning curve, the ability to quickly,
consistently, and flexibly analyze and visualize specialized data from your wafer
manufacturing processes significantly keeps operational costs down and
increases your return on investment (ROI).
Yield analysis is a vital component to yield management. Effective yield analysis
includes the following key components and allows accurate yield prediction and
ensures the right level of work in progress (WIP).
 Accurate and Reliable
High quality statistical techniques, statistical charting methods
and wafer maps are needed to meet the demanding
requirements of a yield analysis solution. Robust, reliable
numerical algorithms are essential to achieving accurate results
from yield analysis efforts.
Flexible, Easy to Use, and Customizable
A packaged solution seldom meets all of a user's requirements. A custom
solution for yield analysis allows each organization to have an integrated analysis
environment tailored to their unique requirements. It is important to have
technology that not only addresses immediate needs and, is also is easy to use,
flexible, customizable, and scalable to meet unanticipated future requirements.
Web-enabled for Collaboration and Sharing
With outsourcing and global companies access to information quickly from
anywhere in the world is increasingly important. A web-based solution allows
engineers to access engineering data in a collaborative environment. The
Internet speeds the flow of vital analysis to all interested parties.
Integration with various architectures and systems
Modern semiconductor companies, like other large high-tech organization often
employ enterprise-wide information architectures for managing their data
requirements. To provide the best productivity gains the yield analysis function
must work within the corporate information architecture. Data needs to be
gathered from diverse sources for analysis by the engineers. The analysis
environment must be capable of working in a wide variety of information
architectures and bring the data into the analysis system quickly and easily.
Challenges in Yield Analysis in the Semiconductor Industry
Yield analysts in the Semiconductor industry face manufacturing processes that
are highly specialized and unique in several significant ways:
- The process to build a semiconductor wafer is very long and complicated
and the amount of data collected in the modern fab is growing rapidly.
The manufacturing process for a wafer involves hundreds of steps and
can take from days to weeks. With fabs running 24 hours a day 7 days a
week, rapid detection of yield problems is crucial.
- The data generated by a modern fab is rapidly increasing as the metrology
instrumentation improves in speed and capability. Further, factory
throughput is continually increasing and this is another factor for the
generation of massive amounts of data.
- The process of building a semiconductor is often a global activity with
outsourcing occurring at many stages. Rapid analysis and information
flow to all participants in the design and manufacturing process is critical.
- Due to increasingly intense competition, many semiconductor companies
are making significant changes in their supply chain with increased
outsourcing and fabless integrated device manufacturing - to achieve
optimal production efficiencies.
- Traditional supply chain management products and techniques ignore the
vital and important area of yield analysis as part of the semiconductor
manufacturing process.
- Time to market pressures result in manufacturers using new chip designs
prematurely - before bugs are completely addressed - making it difficult
to achieve predictable yield output of quality wafers or chips.
Accurate yield analysis increases yield predictability, ensures quick response
time to yield problems, and reduces die loss. In addition, the right analysis tools
will allow manufacturers to monitor process tool performance, improve
maintenance schedules, and enhance control of process modules.
The Solution - the PV-WAVE Family of Products
The PV-WAVE family of products by Visual Numerics - PV-WAVE, JWAVE, and
TS-WAVE - delivers yield analysts specialized analysis and visualization tools
to optimize their custom manufacturing environments. Additionally, our team of
consultants has strong expertise across various industries including the
semiconductor market, and with your yield analysts, can build the most optimal,
custom solution tailored to meet your specific needs.
The PV-WAVE family of development solutions allows analysts to rapidly import,
manipulate, analyze and visualize your data. And, PV-WAVE Advantage
includes a sophisticated set of analysis routines based on the IMSL libraries. For
over 30 years, the IMSL libraries with its industry standard mathematical and
statistical functionality have provided robust and reliable numerical analysis
across several industries including the semiconductor market.
JWAVE provides a web enabled solution that allows teams to collaborate across
the enterprise and quickly access and understand what the yield data means,
from anywhere, at anytime. Using JWAVE and/or our Visual Numerics' highly
specialized consulting team with our customizable yield analysis technical
frameworks, we can quickly address your specific requirements in the wafer
manufacturing process.
TS-WAVE delivers an extensible, off-the-shelf solution that allows for custom
interactive time-series data analysis. It allows you to read, transform and plot
data in time history plots, tabular data views, X-Y plots, and contour plots.
Through a point and click interface you can customize the plots extensively by
modifying many plot attributes and adding annotation. It has built-in report
generation features and allows batch processing.
How the PV-WAVE Family Meets the Needs of the Semiconductor Industry
The process becomes streamlined, operationally efficient, and fast
Wafer fabrication is a complex, time consuming and often costly process. Using the statistical methods provided by the IMSL libraries to analyze yield data and make adjustments to the process as quickly as possible can save fabs thousands of dollars per batch.
Massive amounts of data can be managed and analyzed more effectively
PV-WAVE provides the tools necessary for gathering and analyzing large
amounts of data quickly and easily. PV-WAVE provides all of the tools
necessary to extract data from a database or read in textual or binary data files.
The array-oriented language makes working with the data simple. The charting
techniques allow the engineers to use visual analysis to help direct their
statistical investigation of the yield data.
Management of global activities through web-based solutions is possible
Using JWAVE to build a web-based application allows the engineers to have a
common interface to the data no matter what the platform or location. The
results of the analysis can be quickly and easily made available to other
departments in the organization without time-consuming data transmission and
report generation.
Effective yield analysis results in higher predictability and profitability
As chip designs are pushed to production faster and faster and the designs
become even more complex, effective analysis tools are vital. Yield analysis
must be sophisticated enough to meet the needs of the engineers in order to fully
understand all of the components of yield and defect detection. Using the IMSL
libraries for the analysis allows the application designers to expand the yield
analysis to new techniques without costly algorithm development. Using the
IMSL libraries through the PV-WAVE scripting language makes the development
of new application features a simple process. This rapid application development
environment allows the developers and the engineers to collaborate on the
development of a new technique and iterate to the optimal solution for their
unique environment and conditions.
How PV-WAVE Meets Yield Analysis Requirements in the Semiconductor Industry
The PV-WAVE Family provides solutions that are robust, flexible, scalable, and
easy to use in order to meet immediate needs quickly as well as address
unanticipated data analysis and visualization needs in the future. Several
flagship customers around the world use PV-WAVE as the one stop shop
solution for all their data analysis and visualization requirements.
 With JWAVE and our Visual Numerics' highly specialized consulting team, we can quickly
address your specific requirements for a customized yield analysis solution. Further, our
consulting experts have developed technical frameworks based upon JWAVE that can quickly
address your needs in the wafer manufacturing process.
The PV-WAVE scripting language is easy-to-use and its application specific
analysis algorithms and charting techniques can be leverage to build programs
quickly. Using automated metrology tools and testing equipment, data can be
generated at an astonishing rate. Collected data has no value if it cannot be
easily retrieved and analyzed. PV-WAVE provides a wide variety of data import
techniques, allowing data to come from databases or raw data files.
Data Analysis and Visualization
Trend analysis requires that data be filtered on a variety of criteria selected by
the user. Trend analysis can quickly identify what processes are driving high
yield results and what issues and problem areas might exist with low yield
results. The PV-WAVE language provides an easy methodology for working with
the parameters and quickly filtering the data sets for analysis.
Statistical and visual correlation is an important aspect of yield and failure
analysis. Analysis tools need to provide visual representations of the correlation
results. With sufficient data, graphical analysis can identify major failure modes
and identification of specific failure mechanisms.
For statistical analysis, PV-WAVE utilizes the reliable and trusted IMSL libraries.
PV-WAVE Advantage delivers powerful data analysis capabilities that include the
industry standard IMSL libraries of over 370 proven mathematical and statistical
routines. Yield analysts interested in effective yield management in the
semiconductor industry will routinely use IMSL for regression, correlation,
ANOVA, categorical analysis, nonparametric statistics and multivariate analysis.
For visual data analysis, the PV-WAVE Family delivers customizable analysis
and visualization functionality that can optimize yield analysis efforts leading to
operational efficiencies for higher productivity and profitability. PV-WAVE's
easy-to-use, interactive 2D, 3D, and OpenGL display capabilities along with
JWAVE's web-enabling functionality allow yield analysts to access, share, and
obtain meaningful insights to their data quickly and accurately.
Sample Charts and Mapping for Effective Yield Analysis
Effective Yield Analysis achieves accurate yield prediction and ensures the right
level of work in progress (WIP). Following is a brief overview of some of the
charting techniques available with the PV-WAVE Family that easily provide the
right analysis framework to increase yield predictability, ensure a quick response
time to yield problems, monitor process tool performance, improve maintenance
schedules, and reduce die loss.
Lot Quality Control
An important aspect of yield analysis is the statistical analysis of yield
parameters. A simple, intuitive interface tailored to your unique
requirements allows your analysts to focus on critical yield management
improvements and not spend valuable time on how to run applications or
get the data.
- Trend charts
- This type of chart gives you a graphical and statistical
overview of the selected lot data. The blue line shows the mean values for each lot. Each bar shows the min, max, and range for a single lot. The Upper Specification Limit (USL), the Lower Specification Limit
(LSL) and the target values identify specification values. The bottom of
the chart shows the statistical values: Cp, Cpk, Max, Min, and Avg.
- Histogram
-- The upper chart shows the histogram for all process IDs and the lower charts depict individual process IDs. Showing the total and component histograms in the same window helps highlight variations and which equipment (Process ID) could be affecting yield.
- Xbar-R and Xbar-S Charts
-- The X bar R chart allows you to compare variation in the data by displaying the mean and the range in a single display. As with the trend chart the Upper Specification Limit,
the Lower Specification Limit, and the target are displayed.
- Box Plot
-- A box plot provides an excellent visual summary of many important aspects of lot value distribution. With this chart, we can see the distribution is similar for all Process IDs (x axis labels) and you can delve deeper into outlier data for the problem process ID.
Wafer Mapping
Looking at wafer maps is another technique for finding yield problems and understanding defect patterns. With wafer maps, your analysts can drill down to individual wafer to look at specific failures or look at composite maps to identify meaningful trends and insights that can significantly improve the operational efficiencies of the wafer manufacturing process.
- Wafer Map
-- The Wafer Map display shows the defects for all of the selected lots superimposed. At the bottom of the wafer map are the yield percentages. You can subset the data and look at a single lot or a group of wafers. Or, you can subset the data by failure conditions. In this example, we can see that the DC failure rate is the highest, so this keeps the analysis focused on DC failures.
- Failure Map
- More detail is available on any one of the failure causes. Looking at the DC, most of the failures are around the edges, with a concentration in the lower left area. This interactive chart allows you to quickly and easily find the category causing the most defects.
- WAT Map
- This display allows you to take a look at WAT parameters. A simple right click on the display will change the parameter easily.
- Yield Map
- The yield detail map allows you to show failure causes for each die on the wafer and all the wafers in the lot at the same time. With this interface, you can change failure and category conditions.
- Failure Detail Map
-- You can display the wafer detail for the selected failure and category for any wafer by simply double clicking on that particular section of the yield map.
Customer Success Stories
| Customer: |
ICSI - Integrated Circuits Design Company |
| Situation: |
: ICSI required a web-based data analysis system that would allow
them to meet specific statistical and visual data analysis needs and
significantly increase performance and productivity. Engineers
were downloading the wafer acceptance test (WAT) or yield data to
their computers and then analyzing individual data using programs
such as Excel, which are not economical especially in terms of
time, efficiency and performance. |
| Solution: |
Using JWAVE, ICSI quickly built a sophisticated web-based data
analysis system meeting analysis and visual requirements and
allowing engineers to carry out real-time data and visualization
analysis. |
| Benefits: |
The JWAVE solution allows engineers to carry out real time
analysis and quickly find the accurate yield parameters to send to
the fab. ICSI states that JWAVE has provided tremendous savings
in time and increased productivity -- saving more than 50% in time
and efficiency.
|
| Customer: |
Winbond |
| Situation: |
Winbond had requirements to enhance their existing engineering
data analysis system. Upgrades were necessary to accomplish
faster processing of large amounts of data and graphics, improve
operational efficiencies, meet corporate e-strategy needs, and build
a web-based user interface. |
| Solution: |
Using JWAVE and customizable yield analysis technical
frameworks, Visual Numerics' expert consultants developed a new
custom-built, interactive yield management system. Using
JWAVE's web-enabled architecture, Winbond is now able to
integrate several modules under a single web portal including Lot
QC (LQC), yield-yield map, wafer acceptance test (WAT),
correlation, defect, work in progress (WIP), machine QC (MQC),
and others. Engineers can now use differently modules
concurrently to carry out interactive data analysis. |
| Benefits: |
Using JWAVE, Winbond can now take advantage of a web-enabled
solution with advanced technology for custom yield analysis.
Winbond can now use new streamlined methods for yield data
collection as well as quickly and easily determine factors affecting
yield. With the custom solution build upon JWAVE, Winbond has
improved the overall efficiency of yield analysis and significantly
increased the yield output of quality products.
|
Visual Numerics - A Partner not Just a Vendor
Visual Numerics has for over 30 years provided Trusted Numerics and Scientific
Graphic tools to thousands of users. The IMSL Family has the functionality you need now and in the future.
Visual Numerics has provided Trusted Numerics and Scientific Graphic tools to
many users in the semiconductor industry for years. The PV-WAVE Family has
the functionality you need now and in the future, including an open software
environment to allow your investment to be integrated with new technologies.
Visual Numerics is your partner, not just a software vendor. We have
unparalleled technical support that can answer the hard questions fast, and
consulting that can provide in-depth expertise and fast delivery of time-critical
solutions. Long time users can attest to the quality of support provided by Visual
Numerics, which can be very important when you have difficult problems or tight
schedules to meet. If you need specific expertise that is beyond what you have
available in-house, or need to rapidly develop a solution and do not have the
resources to develop it in-house, then the Technical Resources Group is
available for contract development and are there to make sure you are
successful.
 PV-WAVE has been an
important tool for yield analysis for years. We work closely with many customers and are
responsive to their particular needs and to the needs of the semiconductor industry. The PV-WAVE
Family provides you with standardized software that can provide rapid, high
quality, cost-effective solutions for many kinds of problems, and deliver a
positive return on your investment through maximizing your productivity.
|
|
|
