Taguchi Method and SPC system in Semiconductor Manufacturing Chin-Yuan Chen 陳志遠 Quality Manufactory Advance Semiconductor Engineering Manufacturing Company, Ltd Fisher_Chen@aseglobal.com April 30, 2010 Page 1 Quality Manufactory Abstract: This presentation will cover the following topics, Taguchi Method Application in Assembly and Testing Site, SPC appliation system in Assembly process. • First, to present basic process introduction of semiconductor of assembly and testing. Taguchi Methodology and examples of assembly/test will be shown. To assist engineers to find root casue and enhance process capability, reduce variance, improve process capacity. Our company offer complete statistic training to new engineers. • However, SPC methodology is most important tool for process control. To understand SPC concept, QC tools and OCAP(Out of Control Action Plan) become basic request of engineers. • JMP software Page 2 Quality Manufactory The statistical software in ASE Page 3 Quality Manufactory The statistical software in ASE Page 4 Quality Manufactory Content: 1. Assembly/Test brief introduction 2. Taguchi Method in test site - Enhance wrinkle bump yield 3. Taguchi Method in Assembly site - Gold Wire Diameter Reduction Evaluation 4. SPC in Assembly site Page 5 Quality Manufactory 1. Assembly/Test brief introduction Page 6 Quality Manufactory What common point they had? Page 7 Quality Manufactory Page 8 Quality Manufactory Page 9 Quality Manufactory Page 10 Quality Manufactory ASE Vision:The Comprehensive Solution Final Test Module, Board Assembly & Test (DMS) Assembly Wafer Bumping/Probing Foundry Materials Circuit Design Engineering Test Page 11 Quality Manufactory Brief Assembly Process 研磨 黏晶片 晶片切割 缺點檢查 . .. . . . ... .. . 晶圓 (研磨後) 已切割 缺點檢查 Page 12 Quality Manufactory Brief Assembly Process 上銀膠 黏晶粒 銲線 . .. . . . ... .. . Page 13 Quality Manufactory Brief Assembly Process 銲線缺點檢驗 金線直徑(20um)約頭髮直徑(40um)的一半 Page 14 Quality Manufactory Brief Assembly Process 封膠 正印 植球 Page 15 Quality Manufactory Brief Test Process 測試 Page 16 Quality Manufactory Brief Test Process Contact pushor Contact blade Dut socket Loadboard Handler TESTER TEST HEAD Interface 晶圓 Wafer Page 17 Quality Manufactory 2. Taguchi Method in test site Page 18 Quality Manufactory Bump Status Check Wrinkle bump issue Low Yield Normal bump Wrinkle bump Page 19 Quality Manufactory Test Pin Test Pin Normal Bump Wrinkle Bump Page 20 Quality Manufactory Normal Bump Wrinkle Bump If you are engineer, then what will you improve? Page 21 Quality Manufactory Hint Good Yield Good Contact Page 22 Quality Manufactory Normal Bump Wrinkle Bump Contact deeper (Enhance Over Drive) Page 23 Quality Manufactory Taguchi Method Page 24 Quality Manufactory Taguchi Method - Analysis L9 (34 )直交表 To choose L9 Taguchi experiment matrix and consider two response(y) as above table. Page 25 Quality Manufactory Taguchi Method- Analysis Page 26 Quality Manufactory 3. Taguchi Method in Assembly site Page 27 Quality Manufactory History 1. Customer want cost saving with device, and to reduce gold wire diameter. 2. Current diameter is 25um, customer request to reduce as 20 um. 3. Current composition is Au, customer need to replace by Au-Pb. Page 28 Quality Manufactory Structure 652 golden wire Page 29 Quality Manufactory Taguchi Matrix Page 30 Quality Manufactory Taguchi Analysis Result On Desirability Functions The desirability functions are smooth piecewise functions that are crafted to fit the control points. The minimize and maximize functions are three-part piecewise smooth functions that have exponential tails and a cubic middle. The target function is a piecewise function that is a scale multiple of a normal density on either side of the target (with different curves on each side), which is also piecewise smooth and fit to the control points. These choices give the functions good behavior as the desirability values switch between the maximize, target, and minimize values. For completeness, we implemented the upside-down target also. JMP doesn't use the Derringer and Suich functional forms. Since they are not smooth, they do not always work well with JMP's optimization algorithm. Page 31 Quality Manufactory 4. SPC in Assembly site Page 32 Quality Manufactory Page 33 Quality Manufactory Page 34 Quality Manufactory height Quantil es 50 55 60 65 Normal(62.55, 4. 24234) 70 100.0% maximum 99. 5% 97. 5% 90. 0% 75. 0% quart ile 50. 0% median 25. 0% quart ile 10. 0% 2.5% 0.5% 0.0% minimum Page 35 Moments 70. 000 70. 000 69. 975 68. 000 65. 000 63. 000 60. 250 56. 200 51. 025 51. 000 51. 000 Mean St d Dev St d Err Mean upper 95% Mean lower 95% Mean N Sum W gt Sum Varianc e Sk ewnes s Kurt os is CV N Miss ing Fi tted Nor mal 62. 55 4.2423385 0.6707726 63. 906766 61. 193234 40 40 2502 17. 997436 -0. 781526 0.8768985 6.7823157 0 Parameter Estimates Type Parameter Loc at ion £g D ispersion£m Esti mate Lower 95% U pper 95% 62. 55 61. 193234 63. 906766 4.2423385 3.4751583 5.4473132 Goodness-of-Fi t Test Shapiro-W ilk W Tes t W Prob<W 0.953459 0.0998 Quality Manufactory N ot e: H o = The dat a is f rom the N ormal dist ribut ion. Small p-v alues reject H o. 授權書 範例內容(Enhance wrinkle bump yield, Gold Wire Diameter Reduction Evaluation) 的原作者之授權 Page 36 Quality Manufactory 攜帶物品 八吋晶圓(破片)乙片 PBGA(正印已處理過)四顆 PBGA 空基版四條 Page 37 Quality Manufactory