Project number:

6-six sigma report

Project: Improvement short NG

Schedule ： 2024/05/29~ 2024/08/31

Leader:
Member:
 Project: Improvement T1 yield for A model

Project Statement: Project Goal :

WHAT ？ – improvement T1 test failure for A Short fail yield Goal ： From 0.17% to 0.02%
WHERE ？ – A Situation : solution short NG failure
WHEN ？ – 2024/05/29~ 2024/08/29 Target value:
WHOM ？ – MFG, FA, QA Short NG reduce 0.15%
WHAT EXTEND ？ – T
HOW DO I KNOW ？ – Data collection in IE Tools

Project Scope: CTQ:

Stencil aperture : Injection hole
Factory : VC SPI parameter : Pressing 36N
Customer: A
Model: A
Process: Sub and Main Line
Target ： 99.60%
Project Benefits :
SHORT FR 0.15% ； 4.85*20*20=1940 USD
Progressing:
Step D M A I C SHORT FR 0.02% ； 4.85*3*20=291 USD
COST =1940-291=1649USD/Month
Start 2024.05.29 2024.06.09 2024.07.06 2024.08.11 2024.08.26
~ ~ ~ ~ ~
Finish 2024.06.08 2024.07.05 2024.08.10 2024.08.25 2024.09.20

2
 T1 YIELD status (W14-W23)

WK 24 WK14 24 WK15 24 WK16 24 WK17 24 WK18 24 WK19 24 WK20 24 WK21 24 WK22 24 WK23
Input (pcs) 80,113 92,124 58,381 82,310 48,278 82,399 81,853 78,906 81,840 81,988
Def. qty (pcs) 377 387 257 395 227 363 352 339 295 385
Def. rate (ppm) 4,700 4,200 4,400 4,800 4,700 4,400 4,300 4,300 3,600 4,700
YR (%) 99.53% 99.58% 99.56% 99.52% 99.53% 99.56% 99.57% 99.57% 99.64% 99.53%

Yield rate was not achieved target 99.6% in the long time
 Defect type collection (W14-W23)

- About 80% defect caused by 3

issue Short, Remount and Dirty
- Worst issue in Short NG
- In period W14~W23, Short NG
Defect type
2500 100 has 1,334 pcs, defect rate is
2000 80
1,716 dppm  Estimate Short

Percent
NGQTY

1500 60

1000 40 NG reduction to 200 dppm 

500 20

0 0
Yield rate achieve 99.6%
NG ITEM

NG QTY 1334 332 209 146 107 73 49 38 31 25 22 34

Percent 55.6 13.8 8.7 6.1 4.5 3.0 2.0 1.6 1.3 1.0 0.9 1.4
Cum % 55.6 69.4 78.1 84.2 88.7 91.7 93.8 95.3 96.6 97.7 98.6 100.0
 Defect description

Solder material overflow cause Short NG in actual … … and displayed in X-ray record
Target setting

Imrprove Short NG to 200 ppm to enhance yield rate 99.6 %

6
 6-six sigma 专案成员

为了提升 T1 测试良率，减少 FAIL YEILD,

Leader 布置工作任务和规划改善方案 , 带领团队达成目标

MFG ME QA AOI FA SQE

按照参数设定表作业，修 按照参数设定表作业，修 维修分析 T1 不良品， 产线反馈材料异常及时处

管理员工按照 SOP 监督产线良率趋势，收集
改参数通知 PE 确认执 改参数通知 PE 确认执 收集不良品数据，分类 理，统计材料不良发生频
作业，及时反馈异常 T1 测试良率
行 行，收集直通率数据 总结 率

7
 Planning
JUN JUL AUG
Step WHAT WHERE WHO HOW
23W 24W 25W 26W 27W 28W 29W 30W 31W 32W 33W 34W

Weixuan_ Organization
-Plan & job distribution Team Zhang chart

D
Histogram
-Select the improved item All Team
Pareto chart

Xu_shuai,
Histogram
Chen_conghui
- Data collection & measurement TE , Dong_yanjun
Gage R&R
Bias Linearity
,Anh_Pham
M
Fish bone
- Summarize all data and find
All Team Caused and
improving method effect diagram

Weixuan_ - Flow chart

Zhang, - Process
- Experiment data analysis and
A EE/ME Yang_feng, mapping
verification Xu_xin, - Hypothesis
Cao_yu test
Implement improvements action
- ANOVA
I Control and recheck Improvement All Team
- Proportion
actions
Weixuan_
- Next action: continue improve and Zhang,
C All Yang_feng,
-Control chart
standardize good action
Xu_xin,
8
Plan Actual
 Defect type collection (W14-W23)
NG ITEM NG QTY Rate DPPM
SHORT 1334 55.6% 1,716
REMOUNT 332 13.8% 427
DIRTY 209 8.7% 269
EXCESS COMPONENT 146 6.1% 188
REFLASH 107 4.5% 138
SHIFT 73 3.0% 94
REVERT 49 2.0% 63
UP SIDE DOWN 38 1.6% 49
PCB OPEN 31 1.3% 40
MISSING 25 1.0% 32
REHEAT 22 0.9% 28
SHIFT TO SHORT 20 0.8% 26
WRONG PART 6 0.3% 8
Misaligned paste 3 0.1% 4 Must be
UNDERFILL NG 3 0.1% 4 improved
NO SOLDER 2 0.1% 3
Total 2400

In period W14~23, short NG has 1,334

pcs, defect rate is 1,716 dppm 
Estimate Short NG reduction to 200 dppm
 Yield rate achieve 99.6%
 Line inspection diagram

Post detection
 Confirm Gage R&R linearity bias

- Inspection including Visual test, AOI and Function test Check inspection matching by appearance method
- Appearance method applied to AOI machine and Visual inspector
- Use Gage R&R linearity and bias check matching  Inspection can be accepted
Day Sample qty AOI Visual Gage Linearity and Bias Study for Visual
Boxplot of C11

16
Reported by: % linearity =
1 5,000 12 9
Gage name: Tolerance:
14 Date of study: Misc: 3.8 < 5 
2 5,000 11 10 12
Predictor
Gage Linearity
Coef SE Coef P
Good
C11
Regression

matching
1.0 95% CI Constant 0.3261 0.7938 0.692

3 5,000 16 12
10
Data Slope -0.03804 0.06888 0.596
Avg Bias

8
S 0.590919 R-Sq 3.7%

4 5,000 9 10 6
0.5
Linearity 0.629543 % Linearity 3.8

Gage Bias
5 5,000 14 11 Reference Bias % Bias P
Histogram of C11
0.0 0
Average -0.1
7 0.0
0.6
0.0
*
*
% Bias = 0.6
6 5,000 7 5

Bias
< 5  Good
4
8 0.0 0.0 *
9 0.0 0.0 *
10 -1.0 6.0 *
7 5,000 8 9 3
-0.5 11 0.0 0.0 * matching
Frequency

12 1.0 6.0 *

8 5,000 11 10 2 14 0.0 0.0 *

-1.0
Percent of Process Variation

9 5,000 10 8 4
1

Percent
2
0 6 8 10 12 14 16
10 5,000 14 11 8 10
C11
12 14 16
Reference Value 0
Linearity Bias

Descriptive Statistics: C11

Variable N N* Mean SE Mean StDev Minimum Q1 Median Q3 Maximum

C11 20 0 11.150 0.617 2.758 7.000 9.000 11.000 14.000 16.000
 Location short NG (W14-W23)

Location Qty Cumulative

R14501 64 25.3%
U12007 36 39.5%
U11029 31 51.8%
CR25001 27 62.5%
C14500 17 69.2%
R14502 12 73.9%
U11000 11 78.3%
U14006 10 82.2%
U14003 9 85.8%
U24001 7 88.5%
D25027 6 90.9%
U12005 5 92.9%
U14004 5 94.9%
C13014 4 96.4%
U22003 3 97.6%
C24098 3 98.8%
Short NG by location, main defect at R14501, U12007, U11029 C24008 3 100.0%
 Comparison test location

- Compare line 1 vs line 2 to define Short NG location

Location
Line 1 Line 2
Total - Use data in Wk14~23 in both of line 1 vs 2, collect defect q’ty and location
defect defect
- Evaluation by Variance and Sample T test
R14501 36 28 64
U12007 19 17 36
Probability Plot of Line 1 defect Probability Plot of Line 2 defect
U11029 14 17 31 Normal Normal

CR25001 18 9 27
99 99
Mean 7.941 Mean 6.941
StDev 9.162 StDev 7.352

C14500 7 10 17 95
90
N
AD
17
1.665
95
90
N
AD
17
1.594
P-Value <0.005 P-Value <0.005

R14502 7 5 12 80 80
70 70

U11000 5 6 11

Percent

Percent
60 60
50 50

U14006 6 4 10
40 40
30 30
20 20
U14003 6 3 9 10 10

U24001 3 4 7 5 5

D25027 3 3 6 1
-20 -10 0 10 20 30 40
1
-10 0 10 20 30
Line 1 defect
U12005 1 4 5 Line 2 defect

U14004 3 2 5
C13014 3 1 4
U22003 1 2 3
C24098 1 2 3 P value in normality test < 0.005  Use Levene’s test index
C24008 2 1 3
 Comparison test location
Evaluation by Equal variance test and sample T test  No difference between Short NG location in line 1 vs 2
Variance short NG location
Boxplot of Line 1 defect, Line 2 defect
F-Test
40
Test Statistic 1.55
Line 1 defect P-Value 0.388
Levene's Test
Test Statistic 0.19
P-Value 0.666 30
Line 2 defect

5.0 7.5 10.0 12.5 15.0

95% Bonferroni Confidence I ntervals for StDevs

Data
20

Line 1 defect
10

Line 2 defect

0
0 10 20 30 40
Data Line 1 defect Line 2 defect

Test for Equal Variances: Line 1 defect, Line 2 defect Two-Sample T-Test and CI: Line 1 defect, Line 2 defect
95% Bonferroni confidence intervals for standard deviations
Two-sample T for Line 1 defect vs Line 2 defect
N Lower StDev Upper
Line 1 defect 17 6.55545 9.16154 14.8937
N Mean StDev SE Mean
Line 2 defect 17 5.26099 7.35247 11.9527 P value > 0.05  Line 1 defect 17 7.94 9.16 2.2
Not difference Line 2 defect 17 6.94 7.35 1.8
F-Test (Normal Distribution) between Line 1 vs
Test statistic = 1.55, p-value = 0.388 Difference = mu (Line 1 defect) - mu (Line 2 defect)
2 about Short NG
Estimate for difference: 1.00
Levene's Test (Any Continuous Distribution) location 95% CI for difference: (-4.82, 6.82)
Test statistic = 0.19, p-value = 0.666
T-Test of difference = 0 (vs not =): T-Value = 0.35
P-Value = 0.728 DF = 30
 Defect status
Location R14502 Location U11029

Location U12007 Location R14501

 Fishbone

设计 来料 印刷
Shield cover PCBA 变形 锡膏过厚
PCB
PAD 锡膏坍塌

钢网开孔 来料短路 印刷偏移

SHORT
风机速度

峰值高 受热移动

回流时长 维修拆解

炉温 维修
 Focus item

Analysis item Analysis method Tools Remark

- Comparison Stencil in line 1 and line 2 based in

Stencil type in 1. Stencil #43 in line 1
experiment 2 proportion
line vs line 2. Stencil #64 in line 2
- Use Hypothesis test to evaluate

- Comparison PCB part status in vendor COMPEQ

PCB status by
and AVARY 2 proportion
vendor
- Use Hypothesis test to evaluate

Cleaning - Comparison between Not wet and wet 1 time/5pcs 1. No Wet

2 proportion
condition - Use Hypothesis test to evaluate 2. Wet time/5 pcs

Shield status by - Comparison Shield status by vendor EW and SW

2 proportion
vendor - Use Hypothesis test to evaluate

1. #59,#43 : 36 N,
- Comparison Stencil using time between 4 type #59,
Stencil using Chi-square < 40,000 times
#43, #67, #66
times test 2. #67,#66 : > 38N,
- Use Hypothesis test to evaluate
> 40,000 times
 Confirmation stencil line 1 vs line 2

Line name Line1#43 stencil Line2#64 stencil - Use 2 proportion check

SPI data 85.23% 101.23%
variances between Stencil
T1 input 15000 15000
short QTY 5 13
effective in line 1(#43) vs
Fail rate 0.03% 0.09% line 2(#64)
- Test sample 15000 pcs in
Test and CI for Two Proportions both of line
Sample X N Sample p
1 5 15000 0.000333
2 13 15000 0.000867
P-Value = 0.059  Not
Difference = p (1) - p (2) difference between effective
Estimate for difference: -0.000533333
95% CI for difference: (-0.00108750, 0.0000208290) by Stencil 43/ line 1 and
Test for difference = 0 (vs not = 0): Z = -1.89 P-Value = 0.059
Fisher's exact test: P-Value = 0.096
Stencil 64/line 2  Eliminate
 Comparison test PCB status by vendor

PCB Vendor COMPEQ+#64 stencil AVARY#64 stencil

- Use 2 proportion check
SPI data 98.60% 99.30%
T1 input 15000 15000 variances between PCB
short QTY 35 10
effective in COMPEQ vs AVARY
Fail rate 0.23% 0.07%
vendor
- Test sample 15000 pcs in both

Test and CI for Two Proportions of PCB type

Sample X N Sample p
1 10 15000 0.000667
2 35 15000 0.002333 P-Value = 0.000 
Difference = p (1) - p (2) Difference between
Estimate for difference: -0.00166667
95% CI for difference: (-0.00254233, -0.000791005) effective by Vendor
Test for difference = 0 (vs not = 0): Z = -3.73 P-Value = 0.000
Fisher's exact test: P-Value = 0.000 COMPEQ and AVARY
 Comparison test Cleaning condition

Print parameter No Wet 5pcs/ Wet - Use 2 proportion check

SPI data 93.50% 88.50%
variances defect in changing of
T1 input 15000 15000
Cleaning condition between No
short QTY 12 5
Fail rate 0.08% 0.03% Wet vs 5 pcs/Wet
- Test sample 15000 pcs in both
Test and CI for Two Proportions of method
Sample X N Sample p P-Value = 0.089  Not
1 5 15000 0.000333
2 12 15000 0.000800
difference between
Difference = p (1) - p (2)
Estimate for difference: -0.000466667 Condition Wet and Not wet
95% CI for difference: (-0.00100523, 0.0000718973)
Test for difference = 0 (vs not = 0): Z = -1.70 P-Value = 0.089  But consider to improve
Fisher's exact test: P-Value = 0.143
because some difference
in experiment
 Confirmation shield vendor

Shield Vendor EW+ #64 stencil SW+#64 stencil - Use 2 proportion check
SPI data 100.12% 99.80% variances effective by Shield
T1 input 1200 5000
between vendor EW vs vendor
short QTY 0 3
SW
Fail rate 0.00% 0.06%
- Test sample 15000 pcs in both
Test and CI for Two Proportions of vendor

Sample X N Sample p
1 0 1200 0.000000
2 3 5000 0.000600
P-Value = 0.083  Not
Difference = p (1) - p (2)
Estimate for difference: -0.0006 difference between Stencil
95% CI for difference: (-0.00127875, 0.0000787477)
Test for difference = 0 (vs not = 0): Z = -1.73 P-Value = 0.083 64/ Vendor EW and vendor
Fisher's exact test: P-Value = 1.000 SW  Eliminate
 Confirmation stencil using time

Stencil No #59 #43 #67 #66

SPI data 82.95% 83.16% 100% 106.59% - Use Chi-square test check
T1 input 15000 15000 15000 15000
short QTY 2 1 12 15
variances effective by Stencil
Fail rate 0.01% 0.01% 0.08% 0.10% using time between #59, #43,
Chi-Square Test: 59, 43, 67, 66 #67 and #66
Expected counts are printed below observed counts - Test sample 15000 pcs in all
Chi-Square contributions are printed below expected counts
stencil type
59 43 67 66 Total
1 15000 15000 15000 15000 60000
14994.50 14993.50 15004.50 15007.50
0.002 0.003 0.001 0.004
P-Value = 0.000

2 2 1 12 15 30
Difference between
7.50 7.50 7.50 7.50
4.031 5.630 2.696 7.489 effective by at least one of
Total 15002 15001 15012 15015 60030 stencil type
Chi-Sq = 19.856, DF = 3, P-Value = 0.000  Focus on analysis
 Analysis conclusion
通过要因分析， SHORT issue 和两个因子相关，
PCB Vendor and Stencil using time
1 2
Stencil No #59 #43 #67 #66

SPI data 82.95% 83.16% 100% 106.59%

T1 input 15000 15000 15000 15000

short QTY 2 1 12 15
Fail rate 0.01% 0.01% 0.08% 0.10%

Stencil tension #59 and #43 : 35 ~38N,

Vendor 1 Vendor 2
< 40,000 times
PCB COMPEQ W=0.417MM ， Stencil tension #67 and #66 : >38N ,
PCB AVARY W=0.472MM ， > 40,000 times
Summarize : AVARY 厂商不良率低 Summarize : #59 and #43 不良率低
 Improvement action list

Improve item Improve method Result Remark

PCB 两家厂商存在外观和尺寸差异 , 测量尺寸 In SPEC, 推动厂商改善理由不 Dimension in spec. and

PCB Vendor
充足 , 已出货板还需要消耗使用 no change  Not apply

Input 20000pcs
Stencil aperture SH004 and R14501 位置进行缩孔 , 两个位置内缩 0.03MM, 减少锡量
NG:0

Input 10000pcs
增加湿擦功能， Before : 未启动湿擦， After: 5pcs 增加一次湿擦，更换钢
Wipe model NG:10 旧钢网
网清洁轨道
Fail rate:1000DPPM

重新定义 SPEC, 由 Min:35N, 改为 Min:36N

Stencil tension 隔离 4 张钢网
每班进行钢网检查，发现变形报废处理

Non re-use 报废钢网入 MES 系统，下次无法上线使用

报废 3 张钢网
stencil in MES 手动签字入报废库
 Changing Solder volume by reduce stencil aperture

Increase instant between … reduction Stencil - Short NG at SH004 and R14501 

SH004 and R14501 by … aperture Solder volume is over in appearance
- Consider to increase instant
between SH004 and R14501 
Reduce stencil aperture
- Stencil aperture parameter 0.23
mm at present
- Collect data by defect quantity in
experiment with Stencil parameter
Defect 0.25, 0.23, 0.20, 0.18 mm,
Day Samples
Apt 0.25mm Apt 0.23mm Apt 0.20mm Apt 0.18mm experiment in 1 week ( 6 days,
1 2,000 8 6 1 3
2 2,000 9 4 2 3
2,000 samples/time)
3 2,000 8 6 0 2 - Defect type include Short NG, poor
4 2,000 5 5 0 5
5 2,000 2 3 1 4 or no-solder and dirty
6 2,000 - 5 0 2 - Analyze by ANOVA
 Changing Solder volume by reduce stencil aperture
Normality test
Probability Plot of Apt 0.25mm Probability Plot of Apt 0.23mm Probability Plot of Apt 0.20mm Probability Plot of Apt 0.18mm
Normal Normal Normal Normal
99 99 99
99
Mean 4.833 Mean 0.6667 Mean 3.167
Mean 6.4
StDev 1.169 StDev 0.8165 StDev 1.169
StDev 2.881
95 N 6 95 N 6 95 N 6
95 N 5
AD 0.311 AD 0.544 AD 0.311
AD 0.400 90 90 90
90 P-Value 0.428 P-Value 0.092 P-Value 0.428
P-Value 0.213
80 80 80
80
70 70 70
70

Percent
Percent

Percent
60
Percent

60 60 60
50 50 50
50
40 40 40
40
30 30 30
30
20 20 20
20

10 10 10
10
5 5 5
5

1 1 1
1
2 3 4 5 6 7 8 -1 0 1 2 3 0 1 2 3 4 5 6
0 2 4 6 8 10 12 14
Apt 0.23mm Apt 0.20mm Apt 0.18mm
Apt 0.25mm

Defect by aperture 0.25 mm : Defect by aperture 0.23 mm : Defect by aperture 0.20 mm : Defect by aperture 0.18 mm :
P value = 0.213  Normal P value = 0.428  Normal P value = 0.092  Normal P value = 0.428  Normal

Variance Test
testfor Equal Variances for Defect qty - Data in Normality test is
Bartlett's Test Normal  Use Bartlett’s test
Apt 0.18mm Test Statistic 8.17
P-Value 0.043 index in evaluation for
Levene's Test
Test Statistic 1.16 Variances test
Stencil Aperture

P-Value 0.351
Apt 0.20mm - P value in variances test is
0.043 < 0.05  At least 1
Apt 0.23mm aperture parameter has
effective to defect quantity
difference
Apt 0.25mm
- Aperture 0.20mm data
0 2 4 6 8 10 12 has lowest Mean value
95% Bonferroni Confidence Intervals for StDevs
 Changing Solder volume by reduce stencil aperture
ANOVA and residence displayed
Residual Plots for Defect qty
Boxplot of Defect qty
Normal Probability Plot Versus Fits
- Data in Normality test is Normal
9 99

90
2  Use Bartlett’s test index in
8
evaluation for Variances test

Residual
Percent
0
7 50
-2
6 10
-4
- P value = 0.000 < 0.05  At
Defect qty

1
5 -4 -2 0
Residual
2 4 0.0 1.5 3.0
Fitted Value
4.5 6.0
least 1 aperture parameter has
4

3 8
Histogram Versus Order effective to defect quantity
2 6
2
difference

Frequency

Residual
0
1 4
-2
- Aperture 0.20mm data has
0 2

Apt 0.18mm Apt 0.20mm Apt 0.23mm Apt 0.25mm 0

-4 lowest Mean value and 95% CI
-4 -3 -2 -1 0 1 2 3 2 4 6 8 10 12 14 16 18 20 22
Stencil Aperture
Residual Observation Order
area not overlap with another
group  Select Aperture
One-way ANOVA: Defect qty versus Stencil Aperture
- 95% CI Overlap zone is - Hsu’s check with 0.20mm in improvement
Source DF SS MS F P small smaller is better 
Stencil Aperture 3 101.02 33.67 12.74 0.000 - No Overlap for all group Select Apt. 0.20 mm
Error 19 50.20 2.64
Total 22 151.22 Hsu's MCB (Multiple Comparisons with the Best)
S = 1.625 R-Sq = 66.80% R-Sq(adj) = 61.56% Family error rate = 0.05
Individual 95% CIs For Mean Based on Critical value = 2.20
Pooled StDev Intervals for level mean minus smallest of other level means
Level N Mean StDev ---+---------+---------+---------+------ Level Lower Center Upper ---+---------+---------+---------
Apt 0.18mm 6 3.167 1.169 (-----*----) +------
Apt 0.20mm 6 0.667 0.816 (-----*----) Apt 0.18mm 0.000 2.500 4.568 (------*-----)
Apt 0.23mm 6 4.833 1.169 (----*-----) Apt 0.20mm -4.568 -2.500 0.000 (-----*------)
Apt 0.25mm 5 6.400 2.881 (-----*-----) Apt 0.23mm 0.000 4.167 6.235 (-----------*-----)
---+---------+---------+---------+------ Apt 0.25mm 0.000 5.733 7.903 (---------------*------)
0.0 2.5 5.0 7.5 ---+---------+---------+---------
Pooled StDev = 1.625 +------
-3.5 0.0 3.5 7.0
 Changing Stencil using time
Variance for Stencil pressing
Probability Plot of C18
F-Test
Normal
Test Statistic 11.83
99 P-Value 0.001
Stc min 35N

Experiment 10,000 sample

Mean 1.5
StDev 1.701 Levene's Test
95 N 20 Test Statistic 6.08
AD 1.535 P-Value 0.024

using stencil #59, #43 with

90 Stc min 36N
P-Value <0.005
80
70 0 1 2 3 4

tension min 36N, and more

Percent
60 95% Bonferroni Confidence I ntervals for StDevs
50
40

one using stencil #67, #66

30
20
Stc min 35N
10

with using time min 38N 5

Stc min 36N

1
-3 -2 -1 0 1 2 3 4 5 6 0 1 2 3 4 5
C18 Data

Defect Defect
Time Q'ty
(Using 38N) (Using 36N)
- P value < 0.005 so data is - P value = 0.024 < 0.05 so
1 1,000 4 0
non-normal Stencil using time 36N has
2 1,000 2 1 - Use Levene’s P value for difference with 38N
3 1,000 5 0
Variance test - Variance of defect in case
4 1,000 1 0
- Use variance test to evaluate
5 1,000 5 1 Stencil using 36N is lower than
range between two type of
6 1,000 1 1 Stencil using 38N
using time
7 1,000 0 0  Use using 36N, less than
8 1,000 2 0 40,000 times for stencil
9 1,000 4 0
setting up
10 1,000 2 1
 Changing Cleaning condition

Experiment 10,000 sample using No wet condition and

Wet 5 pcs/time and record defect quantity
Defect Defect
Q'ty
(No Wet) (Wet 5pcs/time)
10,000 27 10

Test and CI for Two Proportions

Sample X N Sample p
1 24 10000 0.002400
2 10 10000 0.001000
- P value = 0.016 < 0.05 so apply Wet
5pcs/time has effective to defect status
Difference = p (1) - p (2) - 95% CI range greater than 0 so Mean defect
Estimate for difference: 0.0014
95% CI for difference: (0.000258291, 0.00254171) with Wet 5pcs/time less than No Wet
Test for difference = 0 (vs not = 0): Z = 2.40
P-Value = 0.016  Use Wet 5pcs/time in Cleaning condition

Fisher's exact test: P-Value = 0.024

 Short NG defect status

Before After

- Total defect rate reduced a little

- Short NG q’ty and rate strongly down
 Sigma tools in control phase
Apply P chart in control Short NG defect with Upper limit 200 ppm

From W23 to W35 After improvement, from W30 to W35

Control Short NG total Control Short NG after improvement

0.0020 0.0011
1 1
1 0.0010 1 UCL=0.0009968

1 1 0.0009
0.0015 UCL=0.001521
0.0008

Proportion
Proportion

_ 0.0007 _
P=0.001095 P=0.0006655
0.0010 0.0006

0.0005
LCL=0.000670
0.0004
0.0005 LCL=0.0003341
1
0.0003
1
1 1
1 0.0002
1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6
Sample Sample
Tests performed with unequal sample sizes Tests performed with unequal sample sizes
Yield status

改善前后对比，良率提升 0.10% ，八月份开始良率维持 99.60% 以上

Before After
Thanks…