Factors Influencing Semiconductor Package Migration

Transcription

1 Factors Influencing Semiconductor Package Migration by Tom Strothmann and Kevin Kan Tempe, AZ, USA STATS ChipPAC, Inc Originally published in the International Wafer Level Packaging Conference Proceedings, Santa Clara, California, USA, October 27-30, Copyright The material is posted here by permission of the SMTA - The Surface Mount Technology Association. By choosing to view this document, you agree to all provisions of the copyright laws protecting it.

2 FACTORS INFLUENCING SEMICONDUCTOR PACKAGE MIGRATION Tom Strothmann and Kevin Kan STATS ChipPAC, Inc Tempe, AZ, USA ABSTRACT Semiconductor package selection and subsequent migration between package technologies is a key decision that is faced by both customers and suppliers in the industry today. Chip suppliers must understand the benefits of new package technologies and the available supply chain to ensure design wins into new products. Contract packaging suppliers must accurately predict market demand for new technologies to ensure the required infrastructure is ready to support it. Incorrect assessment of market demand results in poor allocation of capital or the inability to seize market opportunities on the part of the suppliers. The adoption of new package technology is a balance of competing forces that include cost, manufacturing complexity, package size, package performance, infrastructure momentum, and multiple sources of supply. Because of this complexity, it is often difficult to predict the tipping point for transition from one package type to another.. Keywords: Semiconductor Package Migration, QFN package migration, WLCSP package migration, infrastructure momentum.. INTRODUCTION The pace of package migration in the semiconductor industry is determined by many competing factors. As a general rule:. Package migration is driven by cost and the least expensive package that provides adequate performance will win the design.. Exceptions to this rule are when form factor or performance considerations dominate the product requirement. Technically elegant semiconductor packages offered to the industry that are not in line with this rule have endured limited market acceptance. In this paper, two specific examples of WLCSP adoption will be discussed. The initial rate of adoption of the WLCSP technology and the case of package migration from QFN to WLCSP are discussed as case studies to identify and understand factors that influence semiconductor package migration.. THE CASE OF WLCSP ADOPTION A good example of the timing of package transition can be found in the initial adoption rate of the Wafer Level Chip Scale Package. Wafer Level Chip Scale Packages (WLCSP) were introduced to the industry as a commercial package in March of 1998 with the announcement of Flipchip Technologies UltraCSP package. The package started volume production in 1999 and was projected to experience a CAGR of more than 80% from 2001 through 2006, quickly becoming a significant package in the industry. It was arguably the least expensive package possible since all packaging is done while the wafer is intact and the economies of a mature semiconductor wafer technology could be leveraged to enable the process. The package had the benefits of the smallest possible package size, the best parametric performance, and potentially lowest cost. The actual growth of the WLCSP did not meet expectations and given the high capital cost of implementing wafer level technologies and many contract packaging suppliers struggled with the decision to implement this technology. As an example of this differential between expectation and actual growth, in 2003 WLP volume had only reached 2,876M units, which was well below expectation and the product penetration was limited to small die such as IPD and analog products. Since 2003 the unit growth has been significant with 2009 volume projected at 12,500M units resulting in a CAGR of 28%. 1 It is suggested the factors surrounding package implementation be evaluated to understand the reasons why WLCSP had slower than expected initial growth and did not capture significant Table 1 Factor Description 1 Package Cost The actual cost of creating the package 2 Assembly Cost The cost of placing the device in the final application 3 Form Factor The device package outline and the board space consumed 4 Performance Speed or thermal performance advantage to other packages 5 Reliability Standard JEDEC reliability test for TC, HTS, and drop test (mobile package) 6 Infrastructure Existing production facility with depreciated infrastructure Momentum 7 Market Ease of handling, confidence in package design, confidence with assembly Acceptance 13

3 market share from competing package technologies. General factors that influence adoption of new packaging technologies are proposed in Table 1 and discussed in more detail below.. Factor 1, Package Cost: The actual cost of the packaging process must be competitive with alternative package technologies. This is particularly challenging for a new technology where the infrastructure cost of the technology must be absorbed by the relatively low volume of the new package technology. Rapid adoption of the new technology demands that the total package cost be lower or very competitive with existing options for packaging technology.. Factor 2, Assembly Cost: The total cost to the end user, including board and underfill costs if needed. Factor 3, Form Factor: The total size of the package (X,Y,Z) and the amount of board space it consumes. This becomes particularly critical in advanced hand held products but is not as important in many other applications.. Factor 4, Performance: The package must be evaluated against current package benchmarks for thermal dissipation and electrical performance. Superior performance can justify a higher cost if it provides capability that is in demand in the target market.. Factor 5, Reliability: Package must meet al requirements for TC, HTS, THT, and drop test (mobile products). Factor 6, Infrastructure Momentum: Existing product has established infrastructure which may be fully depreciated. The cost of a new technology must compete directly with the established product. This becomes particularly challenging when the new technology has high capital cost of entry.. Factor 7, Market Acceptance: In the case of WLCSP it was critical to achieve market acceptance of the bare die package. Semiconductor packages had previously provided layers of protection for the silicon die to ensure mechanical and environmental durability. The bare die package challenged these existing packaging paradigms.. implementation of WLCSP technology is summarized in Table 2. The most significant factors suppressing the adoption of WLCSP historically were assembly cost and infrastructure momentum (Factors 2 & 6). The total cost of implementing the package was not competitive with other package types due to the high cost of boards and the limited experience and infrastructure supporting bare die assembly. Although the bare die package cost was attractive, it was the total cost of package implementation that suppressed the adoption of the package. The reliability performance (Factor 5) is considered to be neutral in this evaluation since the performance was in line with market requirements, although it did impose some limitations on the board selection that affected cost.. Early in the life of the package, form factor and performance requirements (Factors 3 & 4) could not override the total package cost considerations, and the adoption was further suppressed by the reluctance of the market to adopt the bare die format as a new technology (Factor 7). This same balance of factors would have continued to suppress adoption if there was not an overridding motivation for reduced form factor in new consumer products.. In spite of this slow start, WLCSP now has strong demand and is on track to support volume of 12,500M units in 2009, supporting the CAGR of 28%. The current growth of WLCSP has increased dramatically due to the reduction in board cost (Factor 2), the placement of infrastructure (Factor 6), and the demands of form factor and performance in advanced mobile devices requiring smaller size with ever increasing levels of functionality (Factors 3 and 4). The rapid growth in advanced smart phones and other mobile products such as MP3 players has aggressively driven both the form factor and performance factors to achieve the recent unprecedented growth in demand for WLCSP product. Based on the understanding of these factors, the key question becomes will the growth of WLCSP erode the volume in other package technologies?. The evaluation of these factors as they relate to the initial Table 2 Factor Evaluation Comment 1 Package Cost Very Good Die processing was done in wafer form, the economies of standard wafer processing can be applied, and the cost of gold wire is eliminated. 2 Assembly Cost Poor Bare die must be handled and flipped for assembly requiring specialized equipment. The required pitch and board material cannot be optimized for lowest cost. 3 Form Factor Very Good Smallest possible package 4 Performance Very Good Lowest package parasitic loss 5 Reliability Medium Initially, reliability declined as the number of I/O s increased, supporting a practical limit of <70. (Current technology allows >120 I/O s.) 6 Infrastructure Momentum 7 Market Acceptance Poor Medium High cost of entry for suppliers limits growth of supply chain bumping and bare die assembly. Assembly capability for bare die handling and placement in conjunction with customer acceptance of the bare die package. 14

4 Table 3 Factor Evaluation Comment 1 Package Cost Very Good High density leadframe design, high efficiency manufacturing throughput and economies-of-scale drive lower packaging cost. 2 Assembly Cost Good Mature EMS infrastructure for surface mount. 3 Form Factor Very Good CSP with >80% die-to-package ratio 4 Performance Good Low inductance capability 5 Reliability Good MSL1 capable with enhanced packaging methods. 6 Infrastructure Very Good Medium cost of entry (vs WLCSP) for suppliers spur growth of supply chain. Momentum 7 Market Acceptance Very Good Cost effective and compact form factor, mobile applications market segment PACKAGE MIGRATION FROM QFN TO WLCSP Can the example of initial WLCSP adoption be used to project future package migration? Does the growing acceptance of WLCSP impact of this growth of other package technologies? Since the QFN package is the closest standard package to WLCSP, will product migration continue through QFN to WLCSP? These are significant questions for the assembly and test supplier seeking to create the proper infrastructure to support market share growth. The characteristics of the two package platforms will be compared and evaluated to answer these questions.. THE CASE FOR QFN PACKAGES QFN packages are experiencing some of the highest growth rates in the industry today as other package types migrate to the smaller form factor and lower cost available with this package technology. Growth in QFN is further spurred by the explosion of wireless applications where board space is premium, more so in mobile handheld devices. QFN is forecasted to grow from 15 billion units (2009) to 35 billion units in 2013, supporting a CAGR of 24%. In addition to lower inductance capability, QFN offers a wide range of I/O possibilities (4 - >100 I/O) and design flexibility, including a quick time to market. The shorter product life in mobile market, and low cost packaging technology has also supported a disposable product which has further stimulated volume growth.. QFN advantages Cost effective protective overmolded CSP Thin & compact form-factor Well proven market & technical solutions Compatible with mature EMS infrastructure Prevalent markets Analog devices (Standard & Application specific): 70% of TAM Special purpose logic Standard logic The growth of QFN packages is driven by the beneficial characteristic listed above as well as the migration of older package technologies to QFN. This is evident in the lower growth rate of leaded package types where the I/O count allows for package conversion.. If the same factors are evaluated for the QFN package, the results in Table 3 are obtained. COMPARATIVE PACKAGE ATTRIBUTES WLCSP Attributes True Chip scale package Smallest, Thinnest, High I/O density Bare die handling capability required at EMS Mobile applications/market segment WLCSP Key Market Drivers Form Factor, Performance, and Cost are the key drivers for the market segment Form Factor has driven the Mobile Phone market with increasing functionality in new products Form factor has also driven WLCSP implementation in advanced disk drive products Increasing number of I/O s in small form factor allow for reduction of chip counts in products and promotes adoption of the technology Use of advanced dielectrics allows higher speed and use of WLCSP designs in Bluetooth and RF applications Projected growth in the WLCSP market, broken out by I/O count is shown in the table below. The high CAGR in the I/O range of is consistent with the growth in the mobile applications and the cost cross-over with QFN discussed later in this paper. 2.. I/O Count CAGR ,868 13, % % , % QFN Attributes Near CSP packaging Over molded protected silicon Compatible with mature EMS infrastructure Broad application space; 70% Analog 15

5 X WB QFN Package Size (mm) Max I/O Y (0.5mm pitch) PCB area X Y WLCSP equivalent (mm) Max I/O (0.5mm pitch) PCB area % Board space reduction % % % % % % QFN Key Market Drivers Growing in popularity driven by wireless application growth Less inductance vs substrate packages Miniaturization Portability Broad I/O Range Projected growth in the QFN market, broken out by I/O count is shown in the table below. 2 Table 4 It is apparent from the comparative evaluation that both QFN and WLCSP have sweet spots in the packaging space and although there is some overlap it is not extensive. Both packages offer excellent value in their respective market space, and have wide acceptance in the industry. Package migration will occur but it will be at the expense of older leaded package types where the I/O count allows for package conversion to QFN. The migration of QFN packages to WLCSP is expected to be offset by the migration of the older packages into QFN.. I/O Count CAGR ,186 17, % ,886 9, % ,258 4, % COMPARATIVE ANALYSIS OF PACKAGES Form factor comparison is shown in Table 4. Board space reduction is achieved in all WLCSP types Beyond I/O >36, WLCSP able to achieve >70% board space area reduction Analysis of package cost by I/O, shown in Figure 1, indicates there is a cost cross-over from QFN to WLCSP in the range of 36 I/O (8 I/O QFN used as cost index 1.0 baseline). I/O sweet spot range for WLCSP is >36 I/O sweet spot for QFN ranges 4-32 CONCLUSION Evaluation of the implementation of a new packaging technology requires a macro view that includes the entire supply chain. Seven factors have been proposed that must be evaluated to understand the growth of a given package platform. In the evaluation of competing technologies, it should be remembered that electronics assembly is aggressively cost driven and that the package that offers adequate technical performance at the lowest cost will win the design.. In the case of evaluating the migration from QFN to WLCSP, both package types have significant positive attributes and segments of the market that are well served. Consequently, although there is some overlap in the application space, both packages will continue to experience significant growth. The migration that occurs from QFN WB QFN & WLCSP, Cost Index vs I/O Cost Index I/O WB QFN WLCSP Figure 1 16

6 packages to WLCSP packages will be offset by the migration of other package types to QFN. In product applications where form factor is not critical, QFN packages will continue to dominate for the foreseeable future. Investment in infrastructure to support both QFN and WLCSP should be continued to support the projected growth, however the investment in older package types should be carefully scrutinized.. In summary: Many applications will continue with standard mainstream QFN (based on market research data) QFN remains cost competitive at low/mid I/O Infrastructure momentum has been overcome with broad acceptance of the package types and the demand of new applications WLCSP has more cost advantage >36 I/O QFN/WLCSP are complimentary packages, and will be used for same silicon Final package selection (WLCSP or QFN) driven by end customer application and design ACKNOWLEDGEMENTS The authors would like to thank Blake Gillett and Scott Sikorski for their valuable insights and support in the evaluation of QFN and WLCSP package migration. REFERENCES [1] TechSearch International, Inc. [2] Electronic Trends Publications, The Worldwide IC Packaging Market, 2009 Edition 17