Apte08c Lect8c RM PHD COEP Taguchi Method Matrix Expts Factor Effects Audio 28slides Ver04

DoE / Taguchi Method
OA based Experiments
Factor Effect Plots
and
Analysis for Best Settings and Predicted Results
By
Prakash R. Apte
COEP, Pune
Lecture 8c to FY PhD scholars at COEP

May 2020
E-mail: apte1947@gmail.com
Web-page: www.ee.iitb.ac.in/~apte
OUTLINE
 OA based Experiments
 Simplified Polysilicon Deposition Process
(using L9)
 Analysis for Factor Effect Plots,

Best Settings and Best Results
May 2020 Apte 8c: Research Methodology using TRIZ & Taguchi Methods Factor Effects & Analysis --2
8 STEPS IN TAGUCHI METHOD
for a simplified
POLYSILICON DEPOSITION PROCESS
with 4 variables
8 STEPS IN TAGUCHI METHODOLOGY
Simplified Polysi Deposition Process:
Reduce Defects and Increase Thickness Uniformity
1. IDENTIFY THE MAIN FUNCTION,

SIDE EFFECTS, AND FAILURE MODE
2. IDENTIFY THE NOISE FACTORS,
TESTING CONDITIONS, AND QUALITY CHARACTERISTICS
3. IDENTIFY THE OBJECTIVE FUNCTION TO BE OPTIMIZED
4. IDENTIFY THE CONTROL FACTORS AND THEIR LEVELS
5. SELECT THE ORTHOGONAL ARRAY MATRIX EXPERIMENT
6. CONDUCT THE MATRIX EXPERIMENT
7. ANALYZE THE DATA,
PREDICT THE OPTIMUM LEVELS AND PERFORMANCE
8. PERFORM THE VERIFICATION EXPERIMENT
AND PLAN THE FUTURE ACTION

MAIN FUNCTION : Simplified Polysilicon Deposition Process

Desired Thickness = 3600 A
SIDE EFFECTS : No Side-Effects are considered
FAILURE MODE : Defects in Polysilicon Layer

Defect Density < 10 / sq cm
Pressure
Gauge Wafers
3-Zone Furnace
Gas To
Inlet Pump
Schematic Diagram of a low pressure reactor

Temperature, Pressure,
Variation in :
Settling Time, Cleaning Method
Testing Conditions : Wafers at 2 Ends and Centre

3 Wafers, 3 points each
Quality Characteristics : Defect Density
Step 2: Noise Factors and Their Levels
LEVELS
NoIsE Number of
NoIsE Level 1 NoIsE Level 2 NoIsE Level 3
FACTORS LEVELS
X-axis
End1 Centre End2 3
Locations
Y-axis
low middle top 3
Locations
X X X
X X X
X X X
END 1 CENTRE END 2
Testing conditions : 3-points on 3-wafers

OBJECTIVE FUNCTION :
n
1
(1) Defect Density ---> DEFCETS = – 10 Log10 ---
n y2
SMALLER - THE - BETTER i
i=1
Factor Effects & Analysis --

May 2020 Apte 8c: Research Methodology using TRIZ & Taguchi Methods 10

Control Factors = 4
Temperature, Pressure, Settling Time and Cleaning Method
Levels of each Control factor = 3

Step 4: Control Factors and Their Levels
LEVELS
CONTROL
FACTORS
1 2 3
Temperature (◦C) To-25 To To+25
Pressure (mTorr) Po-200 Po Po+200
Settling Time (min) to to+8 to+16
Cleaning Method none CM2 CM3
Starting Levels are indicated by underlined levels



1 FOR MEAN AND
DEGREES OF FREEDOM = 2 EACH FOR 4 FACTORS = 8 Total = 9
ORTHOGONAL ARRAYS WITH 3 - LEVEL FACTORS :

NO. OF FACTORS 2-4 5-7 8-13
ORTHOGONAL ARRAY
L9
L9 L18 L27
L9 Orthogonal Array
Control Factors
Expt. No. A B C D
1 1 1 1 1
2 1 2 2 2
3 1 3 3 3
4 2 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3 1 3 2
8 3 2 1 3
9 3 3 2 1

L9 Array with CF assignments
Control Factors Assigned to columns
Pressure Settling Time Cleaning
Expt. No. Temperature ◦C
(mTorr) (min) Method
1 To-25 Po-200 to None
2 To-25 Po to+8 CM2
3 To-25 Po+200 to+16 CM3
4 To Po-200 to+8 CM3
5 To Po to+16 none
6 To Po+200 to CM2
7 To+25 Po-200 to+16 CM2
8 To+25 Po to CM3
9 To+25 Po+200 to+8 none



---> CONDUCT THE 9 EXPTS. OF L9 ARRAY
---> IN EACH EXPT. MEASURE THE DEFECTS
FOR THE 9 NOISE CONDITIONS
OF WAFER LOCATION (IN X-AXIS) AND TEST POINTS (IN Y-AXIS)
X X X
X X X
X X X
END 1 CENTRE END 2
Testing conditions : 3-points on 3-wafers

L9 Orthogonal Array Measured Data Objective
Function

Pressure
(mTorr)
Settling
Time (min)
Cleaning
Method
y1 y2 y3 y4 y5 y6 y7 y8 y9 ηdefect
1 To-25 Po-200 To None 14 16 12 15 13 12 16 14 13 ??
2 To-25 Po to+8 CM2 4 4 4 3 3 4 4 3 3 ??
3 To-25 Po+200 to+16 CM3 20 17 17 20 21 19 18 21 22 ??
4 To Po-200 to+8 CM3 18 20 15 18 18 19 20 19 16 ??
5 To Po to+16 None 141 156 129 161 120 134 154 148 148 ??
6 To Po+200 To CM2 1122 1107 1119 1143 992 978 1071 1063 1159 ??
7 To+25 Po-200 to+16 CM2 126 119 121 138 111 124 136 120 125 ??
8 To+25 Po To CM3 1413 1258 1318 1317 1237 1611 1597 1468 1223 ??
9 To+25 Po+200 to+8 None 1995 2092 2026 1708 1947 1827 2210 1906 1700 ??
Ov-Mean = ??


3. IDENTIFY THE OBJECTIVE FUNCTION TO BE
OPTIMIZED
5. SELECT THE ORTHOGONAL ARRAY MATRIX
EXPERIMENT
ASSUMING ADDITIVITY
OPTIMUM FACTOR LEVELS
PREDICT
FACTOR EFFECTS
PLOTS PREDICTED IMPROVEMENT
May 2020 Apte 8c: Research Methodology using TRIZ & Taguchi Methods 8 Steps in Taguchi Method --19
19

Evaluate the OBJECTIVE FUNCTION for each of the 9 rows :

9
1
(1) Defect Density ---> DEFCETS = – 10 Log10 ---
n y2
SMALLER - THE - BETTER i
i=1

Calculate the Objective Function for each
row
L9 Orthogonal Array Measured Data Objective

Function

Pressure
(mTorr)
Settling
Time (min)
Cleaning
Method
y1 y2 y3 y4 y5 y6 y7 y8 y9 ηdefect
1 To-25 Po-200 To None 14 16 12 15 13 12 16 14 13 -23.00
2 To-25 Po to+8 CM2 4 4 4 3 3 4 4 3 3 -11.00
3 To-25 Po+200 to+16 CM3 20 17 17 20 21 19 18 21 22 -26.00
4 To Po-200 to+8 CM3 18 20 15 18 18 19 20 19 16 -25.00
5 To Po to+16 None 141 156 129 161 120 134 154 148 148 -43.00
6 To Po+200 To CM2 1122 1107 1119 1143 992 978 1071 1063 1159 -61.00
7 To+25 Po-200 to+16 CM2 126 119 121 138 111 124 136 120 125 -42.00
8 To+25 Po To CM3 1413 1258 1318 1317 1237 1611 1597 1468 1223 -63.00
9 To+25 Po+200 to+8 None 1995 2092 2026 1708 1947 1827 2210 1906 1700 -66.00
Ov-Mean = -40.00

FACTOR EFFECTS
 EFFECT OF A FACTOR LEVEL IS DEFINED AS
" THE DEVIATION IT CAUSES FROM OVERALL MEAN , µ “
FACTOR EFFECT OF TEMPERATURE LEVEL 1 :

 A1 OCCURS IN EXPTS. 1, 2 AND 3, mA1 = 1/3 * (η 1 + η 2 + η 3 )
 FACTOR EFFECT OF A1 = a1 = mA1 - m




 REPEAT FOR ALL OTHER FACTORS AND ALL LEVELS

– mB1, mB2 , mB3, mC1, mC2, mC3, mD1, mD2, Md3
Find OvMean= mean of obj func = 1/9 * (η1 + η2 + . . . + η9) = -40

 Ready to make Factor Effect Plot
Factor Effect Plot with Error-Bars
S /N R a t io ( s m a lle r- t h e - b e t t e r) in d B
a1 = 20 b1 = 10 c2 = 6 d2 = 2
Factor Effect Plot for Surface Defects

0.00
O v -M ean
P o+200
Po-200
T o+25
T o-25
to + 1 6
none
CM3
CM2
to + 8
To
Po
to
-10.00
a1
b1
-20.00 c2 d2
-30.00
-40.00
-50.00
-60.00
Control Factor Levels
BEST levels are

To-25, Po-200, to+8, CM2
BEST RESULT (at best levels)
OPTIMUM CONDITIONS : [ A1 B1 C2 {D2} ]
SURFACE DEFECTS ηopt = OvMean + a1 + b1 + c2 ± 3dB

= (-40) + 20 + 10 + 6 ± 3dB
= -4 dB ± 3dB = -7 dB to -1 dB
SURFACE DEFECTS Yopt = 10^(-ηopt/20)= 10 ^ (7/20) = 1 to 3

defects
(VERIFICATION) UNDER OPTIMUM CONDITIONS :

SURFACE DEFECTS = 1 TO 10 / Sq. cm.
PREDICTED AND MEASURED RESULTS MATCH WELL

NOMINAL RESULT (at starting
levels)
NOMINAL CONDITIONS : [ A2 B2 C1 D1 ]
SURFACE DEFECTS ηnom= OvMean + a2 + b2 + c2 + d1 ± 3dB

= (-40) + (-3) + (1) + (-9 ) + (-4 ) ± 3dB
= -55 dB ± 3dB = -52 dB to -58dB
SURFACE DEFECTS Ynom=10^(-ηopt/20)=10^(-52/20)= 400 to 800

defects
(Predicted Results) UNDER NOMINAL CONDITIONS :
(Predicted Results) UNDER OPTIMUM CONDITIONS :

VERIFICATION Experiments needed to check whether

PREDICTED AND MEASURED RESULTS MATCH WELL or NOT
PREDICTION AND
VERIFICATION
STARTING CONDITIONS : ( A2 B2 C1 D1 )
SURFACE DEFECTS = 600 / Sq. cm. (400 TO 800 /sq cm)
OPTIMUM CONDITIONS : (PREDICTION) [ A1 B1 C2 {D2} ]

SURFACE DEFECTS = 1 to 3 / Sq. cm.
(VERIFICATION) UNDER OPTIMUM CONDITIONS :

SURFACE DEFECTS = 1 to 10 / Sq. cm.
PREDICTED AND MEASURED RESULTS MATCH

WELL


8. PERFORM THE VERIFICATION EXPERIMENT

AND PLAN THE FUTURE ACTION
ADOPT
RESULTS MATCH WELL WITH PREDICTION NEW
SETTINGS
End of RM Lecture 8c

P.R. Apte :Integration of TRIZ Innovation and Taguchi
10 Sep 2014 Optimization EDM & Polysi Case Studies- 29
DoE / Taguchi Method
Analysis of Variance (ANOVA)
Analysis for % Factor-Effects and Error-Bars
Best Settings and Predicted Results
By
Prakash R. Apte
COEP, Pune
Lecture 8c to FY PhD scholars at COEP

May 2020
E-mail: apte1947@gmail.com
Web-page: www.ee.iitb.ac.in/~apte
OUTLINE
 OA based Experiments
 Simplified Polysilicon Deposition Process
(using L9)
 Analysis for Factor Effect Plots,

Best Settings and Best Results

Good Plot with Error-Bars
a1 = 20 b1 = 10 c1 = 6 d1 = 2

0.00
O v -M ean
P o+200
Po-200
T o+25
T o-25
to + 1 6
none
CM3
CM2
to + 8
To
Po
to
-10.00
a1
b1
-20.00 c1 d1
-30.00
-40.00
-50.00
-60.00
Apte : WS L : Taguchi Method to QUICKLY get World’s BEST

9 June 2016 Results Matrix Expt 32
Good Plot without Error-Bars
r- tr-ht eh -eb- eb tet et tr)e r)inind Bd B
a lle
a1 = 20 b1 = 10 c1 = 6 d1 = 2
sm
S /NR aR taiot io( s(m a lle

0.00
O v -M ean
P o+200
P o-200
T o+25
T o-25
to + 1 6
none
CM2
CM3
to + 8
To
Po
to
S /N
-10.00
-10.00
a1
b1 c1
-20.00
-20.00
d1
-30.00
-30.00
-40.00
-40.00
-50.00
-50.00
-60.00
-60.00
-70.00


Step 2: Noise Factors and Their Levels
LEVELS
NoIsE FACTORS NoIsE Level 1 NoIsE Level 2 NoIsE Level 3 Number of LEVELS
X-axis Locations End1 Centre End2 3
Y-axis Locations low middle top 3

Step 4: Control Factors and Their Levels

LEVELS
CONTROL FACTORS
1 2 3
Temperature ◦C To-25 To To+25
Pressure (mTorr) Po-200 Po Po+200
Settling Time (min) to to+8 to+16
Cleaning Method none CM2 CM3

L9 Orthogonal Array
Control Factors
Expt. No. A B C D
1 1 1 1 1
2 1 2 2 2
3 1 3 3 3
4 2 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3 1 3 2
8 3 2 1 3
9 3 3 2 1

L9 Array with CF assignments
Temperature Pressure Settling Time Cleaning
Expt. No. ◦C (mTorr) (min) Method
1 To-25 Po-200 to None
2 To-25 Po to+8 CM2
3 To-25 Po+200 to+16 CM3
4 To Po-200 to+8 CM3
5 To Po to+16 none
6 To Po+200 To CM2
7 To+25 Po-200 to+16 CM2
8 To+25 Po To CM3
9 To+25 Po+200 to+8 none

Good Plot without Error-Bars
r- tr-ht eh -eb- eb tet et tr)e r)inind Bd B
a lle
a1 = 20 b1 = 10 c1 = 6 d1 = 2
sm
S /NR aR taiot io( s(m a lle

0.00
O v -M ean
P o+200
P o-200
T o+25
T o-25
to + 1 6
none
CM2
CM3
to + 8
To
Po
to
S /N
-10.00
a1
b1 c1
-20.00
d1
-30.00
-40.00
-50.00
-60.00
-70.00


Factor Effects
LEVELS
CONTROL FACTORS 1 2 3
Temperature ◦C a1=20 a2= -3 a3= -17
Pressure (mTorr) a1=10 b2=1 b3= -11
Settling Time (min) c1= -9 c2=6 c3=3
Cleaning Method d1= -4 d2=2 d3=2

MATRIX EXPERIMENTS
USING ORTHOGONAL ARRAY
 DESCRIPTION OF 'CVD' PROCESS UNDER STUDY

 MATRIX EXPERIMENT AND CONCEPT OF
ORTHOGONALITY
 ANALYSIS OF DATA OBTAINED FROM MATRIX EXPT.
 ADDITIVE MODEL
 ANALYSIS OF VARIANCE ( ANOVA ) AND ANALOGY
BETWEEN ANOVA AND FOURIER ANALYSIS
 USE OF ADDITIVE MODEL FOR PREDICTION AND
DIAGNOSIS
17-19 Oct 2016 Apte : DoE / Taguchi Method : ROBUST Technology Development Matrix Expt - 40
CVD PROCESS FOR POLYSILICON DEPOSITION
IN VLSI TECHNOLOGY
 CONTROL PARAMETERS ( FACTORS )

 TEMPERATURE ( A ), PRESSURE ( B ),
SETTLING TIME ( C ) AND CLEANING METHOD ( D )
 EACH PARAMETER HAS 3 SETTINGS ( LEVELS ) ONE

OF WHICH IS THE STARTING LEVEL
 AIM IS TO MINIMISE THE SURFACE DEFECTS
 MATRIX EXPERIMENTS ARE PERFORMED SO AS TO

OBTAIN OPTIMUM LEVELS OF FACTORS
MATRIX EXPERIMENTS USING L9 ARRAY
 ONLY 9 EXPERIMENTS

OUT OF A TOTAL OF 34 = 81
 FOUR CONTROL FACTORS ASSIGNED TO
THE FOUR COLUMNS OF L9
 EACH ROW REPRESENTS AN EXPERIMENT
HAVING A COMBINATION OF CONTROL
FACTOR LEVELS
e.g. ROW #3 ( EXPT. #3 ) --> A1 B2 C3 D3
and ROW #8 ( EXPT. #8 ) --> A3 B2 C1 D3
L9 ORTHOGONAL ARRAY
EXPT. NO. 1 2 3 4
. A . . B . . C . . D .
1 A1 B1 C1 D1
2 A1 B2 C2 D2
3 A1 B3 C3 D3
4 A2 B1 C2 D3
5 A2 B2 C3 D1
6 A2 B3 C1 D2
7 A3 B1 C3 D2
8 A3 B2 C1 D3
9 A3 B3 C2 D1
L9 ORTHOGONAL ARRAY
 COLUMNS ARE ORTHOGONAL – IN THE SENSE THAT

" FOR ANY PAIR OF COLUMNS, ALL
COMBINATIONS OF FACTOR LEVELS
OCCUR AN EQUAL NUMBER OF TIMES "
 COLUMNS ARE BALANCED,

WHICH ALSO IMPLIES ORTHOGONALITY
 SIX PAIRS OF COLUMNS ( 4C2 )
? CAN YOU VERIFY ORTHOGONALITY FOR A FEW PAIRS ¿
L9 ORTHOGONAL ARRAY
EACH COMBINATION APPEARS only ONCE
EXPT. NO. 1 2 3 4
. A . . B . . C . . D .
1 B1 C1
2 B2 C2
3 B3 C3
4 B1 C2
5 B2 C3
6 B3 C1
7 B1 C3
8 B2 C1
9 B3 C2
L9 ORTHOGONAL ARRAY
with MEASURED SN-RATIO
EXPT. NO.
1 2 3 4 SN-RATIO η
. A . . B . . C . . D . ( in dB )
1 A1 B1 C1 D1 η1
2 A1 B2 C2 D2 η2
3 A1 B3 C3 D3 η3
4 A2 B1 C2 D3 η4
5 A2 B2 C3 D1 η5
6 A2 B3 C1 D2 η6
7 A3 B1 C3 D2 η7
8 A3 B2 C1 D3 η8
9 A3 B3 C2 D1 η9
FACTOR EFFECTS
 EFFECT OF A FACTOR LEVEL IS DEFINED AS
" THE DEVIATION IT CAUSES FROM OVERALL MEAN , m "
 FACTOR EFFECT OF TEMPERATURE LEVEL 3 :
– A3 OCCURS IN EXPTS. 7, 8 AND 9
– mA3 = 1/3 * (η 7 + η 8 + η 9 )
– FACTOR EFFECT OF A3 = a3 = mA3 - m

 REPEAT FOR ALL FACTORS AND ALL LEVELS
– mA1 , mA2 , mB1 , . . . . , mD2 , mD3
TABULAR AND GRAPHICAL REPRESENTATION
OF FACTOR EFFECTS
 NUMERICAL VALUES ARE GIVEN IN A TABULAR FORM
or
 GRAPHICAL REPRESENTATION IS CONVENIENT FOR

DRAWING QUALITATIVE INFERENCES AND CHOOSING
THE OPTIMUM LEVELS OF FACTORS
L9 ORTHOGONAL ARRAY AND EXPERIMENTER'S LOG
TEMPERATURE COLUMN VARIABLE SHOWS HOW TO ENTER THE EXPT. LOG
TEMPERATURE PRESSURE SILANE CLEANING METHOD

EXPT. NO.
A B C D
1 T0-25--> 1 1 1 1
2 T0-25--> 1 2 2 2
3 T0-25--> 1 3 3 3
4 T0-----> 2 1 2 3
5 T0-----> 2 2 3 1
6 T0-----> 2 3 1 2
7 T0+25--> 3 P0-200--> 1 S0------> 3 CM2-----> 2

8 T0+25--> 3 P0------> 2 S0-100--> 1 CM3-----> 2
9 T0+25--> 3 3 2 1
EXPT #7 AND #8 SHOW THE EXPERIMENTER'S LOG
SELECTING OPTIMUM FACTOR LEVELS
 PRIMARY GOAL – TO MINIMISE DEFECTS

– TO FIND OPTIMUM LEVELS
 BEST LEVELS
FOR TEMPERATURE  A1  TO - 25
FOR PRESSURE  B1  PO - 200
FOR SETTLING TIME  C2  to + 8
FOR CLEANING METHOD  D2  CM2
or  D3  CM3
 OPTIMUM SETTINGS  MINIMUM DEFECTS
 A1 B1 C2 D2 or A1 B1 C2 D3  how few ?
ADDITIVE MODEL
AN ADDITIVE MODEL ( ALSO CALLED SUPERPOSITION MODEL )
 RELATIONSHIP BETWEEN RESPONSE VARIABLE AND
CONTROL FACTORS
 EFFECTS OF CONTROL FACTOR LEVELS IS ESTIMATED
SEPARATELY
 PREDICTION BY SIMPLY ADDING INDIVIDUAL FACTOR
EFFECTS
THE ADDITIVITY IS INFLUENCED BY THE CHOICE OF

 QUALITY CHARACTERISTICS, S/N RATIO, CONTROL
FACTORS / LEVELS
BENEFITS OF ACHIEVING ADDITIVITY :

 PREDICTIONS VALID FOR LAB AND MANUFACTURING
CONDITIONS
ADDITIVE MODEL (CONTD.)
VARIABLES SEPARABLE MODEL

PROVIDES JUSTIFICATION FOR CALCULATING
FACTOR EFFECTS AS AVERAGES OF APPROPRIATE
EXPTS. ( AS WAS DONE EARLIER )
ADDITIVE MODEL AS AN APPROXIMATION
= + ai + b j + ck + dl +...+
DEVIATION CAUSED BY LEVEL Ai FROM µ IS
GIVEN BY ai , BY Bj FROM µ AS bj ETC.
 IS THE ERROR DUE TO ADDITIVE

APPROXIMATION AS WELL AS ANY ERROR IN
MEASUREMENT.
ADDITIVE MODEL (contd.)
 INDIVIDUAL FACTOR EFFECTS CAN BE LINEAR,

QUADRATIC OR OF HIGHER ORDER
 CROSS - PRODUCT TERMS INVOLVING TWO OR

MORE FACTORS ARE NOT ALLOWED
 SUM OF DEVIATIONS CAUSED BY 3 LEVELS OF

EACH FACTOR IS ZERO
e.g. FOR FACTOR A  a1 + a2 + a3 = 0
 THE MEAN OF FACTOR EFFECTS OF ‘ALL’ FACTORS IS

EQUAL TO 'µ'
L8 ORTHOGONAL ARRAY
1 2 3 4 5 6 7
EXPT. NO.
A B C D E F G
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 1 2 1 2
6 2 1 2 2 1 2 1
7 2 2 1 1 2 2 1
8 2 2 1 2 1 1 2
L9 ORTHOGONAL ARRAY
EXPT. NO. 1 2 3 4
. A . . B . . C . . D .
1 A1 B1 C1 D1
2 A1 B2 C2 D2
3 A1 B3 C3 D3
4 A2 B1 C2 D3
5 A2 B2 C3 D1
6 A2 B3 C1 D2
7 A3 B1 C3 D2
8 A3 B2 C1 D3
9 A3 B3 C2 D1
L18 ORTHOGONAL ARRAY
EXPT. 1 2 3 4 5 6 7 8
NO. . A .. B . C . D . E .. F .. G . H .
1 1 1 1 1 1 1 1 1
2 1 1 2 2 2 2 2 2
3 1 1 3 3 3 3 3 3
4 1 2 1 1 2 2 3 3
5 1 2 2 2 3 3 1 1
6 1 2 3 3 1 1 2 2
7 1 3 1 2 1 3 2 3
8 1 3 2 3 2 1 3 1
9 1 3 3 1 3 2 1 2
10 2 1 1 3 3 2 2 1
11 2 1 2 1 1 3 3 2
12 2 1 3 2 2 1 1 3
13 2 2 1 2 2 1 3 2
14 2 2 2 3 3 2 1 3
15 2 2 3 1 1 3 2 1
16 2 3 1 3 3 3 1 2
17 2 3 2 1 1 1 2 3
18 2 3 3 2 2 2 3 1
ACHIEVING ADDITIVITY
 RELATIONSHIP BETWEEN CONTROL FACTORS AND
PERFORMANCE CHARACTERISTIC MUST BE
ADDITIVE
 ADDITIVITY ENSURES STABILITY OF DESIGN
UNDER LABORATORY , MANUFACTURING AND
CUSTOMER USAGE
> DESIGN IS ROBUST
 ADDITIVITY DEPENDS ON
– CONTROL FACTOR SELECTION
– QUALITY CHARACTERISTIC SELECTION
 LOG FORM OF OBJECTIVE FUNCTION

 IMPROVES ADDITIVITY
ESTIMATION OF FACTOR EFFECTS
USING ADDITIVE MODEL
 EFFECT OF SETTING TEMPERATURE AT LEVEL 3

mA3 = 1/3 * (η7 + η8 + η9 ) = m + a3
 EFFECT OF SETTING CLEANING METHOD AT LEVEL 3

mD3 = 1/3 * (η3 + η5 + η9 ) = m + d3
etc. etc.
ANALYSIS OF VARIANCE (ANOVA)
 RELATIVE MAGNITUDE OF FACTOR EFFECTS 

‘QUALITATIVELY’ FROM TABLES or GRAPHS
 ANOVA GIVES A ‘QUANTITATIVE’ MEASURE FOR

THE RELATIVE MAGNITUDES OF DIFFERENT
FACTOR EFFECTS
 ANOVA IS NEEDED FOR ESTIMATING THE
ERROR VARIANCE FOR THE FACTOR EFFECTS and
VARIANCE OF THE PREDICTION ERROR
ANOVA - SUM OF SQUARES
9
2
 GRAND TOTAL G = i
T
SUM OF SQUARES i=1
 SUM OF SQUARES SM = 9 * m2
DUE TO MEAN
 TOTAL SUM OF SQUARES =

9
G
T
- S M
2
also =
i
- 9 * m2
i=1
ANOVA - SUM OF SQUARES
DUE TO FACTORS
 SUM OF SQUARES DUE TO FACTOR ' A '

SSQ_A = 3 ( a 12 + a22 + a32 )
 SUM OF SQUARES DUE TO FACTOR ' B '
2 2 2
SSQ_B = 3 ( b1 + b2 + b3 )
 SUM OF SQUARES DUE TO FACTOR ' C '
SSQ_C = 3 ( c 12 + c 22 + c32 )
 SUM OF SQUARES DUE TO FACTOR ' D '
2 2 2
SSQ_D = 3 ( d1 + d2 + d3 )
DEGREES OF FREEDOM and VARIANCE RATIO 'F'
 NUMBER OF INDEPENDENT PARAMETERS OR LEVELS

ASSOCIATED WITH A
– MATRIX EXPERIMENT
– FACTOR
or
– SUM OF SQUARES
 TOTAL DEGREES OF FREEDOM DECIDE WHICH
ORTHOGONAL ARRAY IS SUITABLE
 DETERMINE  SUM OF SQUARES DUE TO ERROR
 DETERMINE THE VARIANCE RATIO ' F ' FOR A FACTOR
MEAN OF SUM OF SQUARES DUE TO A FACTOR

F = ---------------------------------------------------------------
MEAN SQUARE ERROR
ERROR VARIANCE
AND POOLING OF ERRORS
 ERROR VARIANCE –
2 SUM OF SQUARES DUE TO ERROR
σe = --------------------------------------------------------
DEGREES OF FREEDOM FOR ERROR
 POOLING OF ERRORS
– WHEN DEGREES OF FREEDOM FOR ERROR ARE ZERO
or
– WHEN ONE OR MORE FACTORS HAVE A SMALL
EFFECT ON THE OBJECTIVE FUNCTION
CONFIDENCE INTERVALS FOR FACTOR EFFECTS
 ERROR VARIANCE –
2 SUM OF SQUARES DUE TO ERROR
σe = --------------------------------------------------------
DEGREES OF FREEDOM FOR ERROR
 FOR 95% CONFIDENCE INTERVAL –

 CHOOSE TWO STANDARD DEVIATIONS FOR ERROR AS THE
CONFIDENCE INTERVAL
 CONFIDENCE INTERVAL = ± 2 [ SQRT (ERROR VARIANCE) ] dB
VARIANCE RATIO ' F ' and
INTERPRETATION OF ANOVA TABLES
 VARIANCE RATIO ' F ' –

MEAN SUM OF SQUARES DUE TO A FACTOR
F = --------------------------------------------------------------
ERROR MEAN SQUARE
 F < 1 – FACTOR EFFECT IS SMALLER THAN THE

ERROR OF THE ADDITIVITY MODEL
 F > 2 – FACTOR EFFECT NOT QUITE SMALL
 F > 4 – FACTOR EFFECT IS QUITE LARGE
SUM OF SQUARES DUE TO A FACTOR

% CONTRIBUTION = ------------------------------------------------
TOTAL SUM OF SQUARES
PREDICTION OF IMPROVEMENT
 OPTIMUM QUALITY FOR , say , A1 B1 (C2 D2)
 ηOPT = m + a1 + b1
 ηOPT = m + (mA1 - m) + (mB1 - m) NO CHANGE
 IMPROVEMENT OVER NOMINAL (A2 B2 C2 D2)


ηOPT - ηNOM = (a1 - a2) + (b1 - b2)
CONFIRMATION EXPERIMENTS
 TO VERIFY THAT THE PREDICTION IS CORRECT
 TO VERIFY WHETHER ADDITIVE MODEL IS SATISFIED

– IF YES > ADOPT THE ' NEW ' SETTINGS
– IF NO > FAILURE OF ADDITIVE MODEL
and
INDICATION OF STRONG INTERACTION
AMONGST TWO OR MORE FACTORS
therefore
PLAN FUTURE ACTION
INTERACTION AMONGST FACTORS
 DESIRABLE – NO INTERACTIONS
– SATISFY ADDITIVITY
 SYNERGISTIC INTERACTION – LEADS TO SOME ERROR

– BUT IT IS ACCEPTABLE
 ANTI-SYNERGISTIC INTERACTION –
> FAILURE OF ADDITIVE MODEL
> SELECT APPROPRIATE S / N RATIO
> PERFORM FRESH MATRIX EXPERIMENTS
TO STUDY SPECIFIC INTERACTIONS
EXAMPLES OF ORTHOGONAL ARRAYS
L8 ORTHOGONAL ARRAY
1 2 3 4 5 6 7
EXPT. NO.
A B C D E F G
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 1 2 1 2
6 2 1 2 2 1 2 1
7 2 2 1 1 2 2 1
8 2 2 1 2 1 1 2
L9 ORTHOGONAL ARRAY
EXPT. NO. 1 2 3 4
. A . . B . . C . . D .
#1 1 1 1 1
#2 1 2 2 2
#3 1 3 3 3
#4 2 1 2 3
#5 2 2 3 1
#6 2 3 1 2
#7 3 1 3 2
#8 3 2 1 3
#9 3 3 2 1
L18 ORTHOGONAL ARRAY
EXPT. 1 2 3 4 5 6 7 8
NO. . A .. B . C . D . E .. F .. G . H .
1 1 1 1 1 1 1 1 1
2 1 1 2 2 2 2 2 2
3 1 1 3 3 3 3 3 3
4 1 2 1 1 2 2 3 3
5 1 2 2 2 3 3 1 1
6 1 2 3 3 1 1 2 2
7 1 3 1 2 1 3 2 3
8 1 3 2 3 2 1 3 1
9 1 3 3 1 3 2 1 2
10 2 1 1 3 3 2 2 1
11 2 1 2 1 1 3 3 2
12 2 1 3 2 2 1 1 3
13 2 2 1 2 2 1 3 2
14 2 2 2 3 3 2 1 3
15 2 2 3 1 1 3 2 1
16 2 3 1 3 3 3 1 2
17 2 3 2 1 1 1 2 3
18 2 3 3 2 2 2 3 1
FOR DRILLING HOLE IN A CAST ALLOY WHEEL
SPEED FEED RATE DRILL LENGTH COOLANT FIXTURE RESPONSE

EXPT. NO.
A B C D E S/N
1 1500 15 4 BLUE OLD 42.18

2 1500 15 7 WHITE NEW 53.96
3 1500 18 4 WHITE NEW 48.48
4 1500 18 7 BLUE OLD 48.98
5 2000 15 4 BLUE NEW 44.20

6 2000 15 7 WHITE OLD 42.19
7 2000 18 4 WHITE OLD 41.58
8 2000 18 7 BLUE NEW 56.40
COMPUTER TUNING for HIGH PERFORMANCE
1 2 3 4 5 6 7 8
EXPT. NO.
F B C D E A G H
1 2 a 4 1/5 0 4&1 400 a
2 2 a 3 1/4 3 4&2 500 b
3 2 a 3.5 1/3 8 4&2 600 c
16 0 c 4 1/3 3 4&2 400 b
17 0 c 3 1/5 8 4&2 500 c
18 0 c 3.5 1/4 0 4&1 600 a
ROLE OF ORTHOGONAL ARRAYS
 OPTIMIZATION BY STUDYING ONE FACTOR AT A TIME

DOES NOT GIVE ANY IDEA ABOUT INTERACTIONS
 ABSENCE OF ADDITIVITY IS POINTED OUT
BY THE FAILURE OF VERIFICATION EXPERIMENT
WHICH IN TURN POINTS TO LARGE INTERACTIONS
 DEVIATION FROM ADDITIVITY QUANTIFIES LARGENESS OF
INTERACTIONS RELATED TO MAIN EFFECTS
 RECHECK THE CHOICE OF CONTROL FACTORS AND
QUALITY CHARACTERISTICS
§ IF THE VERIFICATION EXPERIMENT SUCCEEDS,
THE PRODUCT / PROCESS IS PROVED TO BE
ROBUST ( AGAINST NOISE FACTORS )
Thank You
S /N R a tio (s m a lle r-t h e - b e t te r) in d B

0.00
O v -M ean
P o +2 0 0
Po-200
T o +2 5
T o-25
to +1 6
none
CM2
CM3
to +8
Po
To
to
-10.00
-20.00
a1
b1
-30.00 c1 d1
-40.00
-50.00
-60.00
-70.00

a1 = 20 b1 = 10 c1 = 6 d1 = 2

0.00
O v -M ean
P o+200
Po-200
T o+25
T o-25
to + 1 6
none
CM2
CM3
to + 8
To
Po
to
-10.00
a1
-20.00 b1 c1 d1
-30.00
-40.00
-50.00
-60.00
-70.00

Good plot
a1 = 20 b1 = 10 c1 = 6 d1 = 2

0.00
O v -M ean
Po+200
P o-200
T o+25
T o-25
to + 1 6
none
CM2
CM3
to + 8
To
Po
to
-10.00
a1
b1 c1
-20.00
d1
-30.00
-40.00
-50.00
-60.00
-70.00


Apte08c Lect8c RM PHD COEP Taguchi Method Matrix Expts Factor Effects Audio 28slides Ver04

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Apte08c Lect8c RM PHD COEP Taguchi Method Matrix Expts Factor Effects Audio 28slides Ver04

Uploaded by

Copyright:

Available Formats

DoE / Taguchi Method

Lecture 8c to FY PhD scholars at COEP

 Analysis for Factor Effect Plots,

1. IDENTIFY THE MAIN FUNCTION,

1. IDENTIFY THE MAIN FUNCTION,

MAIN FUNCTION : Simplified Polysilicon Deposition Process

SIDE EFFECTS : No Side-Effects are considered

FAILURE MODE : Defects in Polysilicon Layer

Schematic Diagram of a low pressure reactor

1. IDENTIFY THE MAIN FUNCTION,

Testing Conditions : Wafers at 2 Ends and Centre

Quality Characteristics : Defect Density

END 1 CENTRE END 2

Testing conditions : 3-points on 3-wafers

1. IDENTIFY THE MAIN FUNCTION,

Factor Effects & Analysis --

1. IDENTIFY THE MAIN FUNCTION,

4. IDENTIFY THE CONTROL FACTORS AND THEIR LEVELS

Levels of each Control factor = 3

Factor Effects & Analysis --

Temperature (◦C) To-25 To To+25

Pressure (mTorr) Po-200 Po Po+200

Settling Time (min) to to+8 to+16

Cleaning Method none CM2 CM3

Starting Levels are indicated by underlined levels

1. IDENTIFY THE MAIN FUNCTION,

5. SELECT THE ORTHOGONAL ARRAY MATRIX EXPERIMENT

ORTHOGONAL ARRAYS WITH 3 - LEVEL FACTORS :

Factor Effects & Analysis --

Factor Effects & Analysis --

1. IDENTIFY THE MAIN FUNCTION,

6. CONDUCT THE MATRIX EXPERIMENT

END 1 CENTRE END 2

Testing conditions : 3-points on 3-wafers

Factor Effects & Analysis --

Expt. No. Temperature ◦C

2 To-25 Po to+8 CM2 4 4 4 3 3 4 4 3 3 ??

3 To-25 Po+200 to+16 CM3 20 17 17 20 21 19 18 21 22 ??

4 To Po-200 to+8 CM3 18 20 15 18 18 19 20 19 16 ??

Factor Effects & Analysis --

1. IDENTIFY THE MAIN FUNCTION,

1. IDENTIFY THE MAIN FUNCTION,

Evaluate the OBJECTIVE FUNCTION for each of the 9 rows :

Factor Effects & Analysis --

L9 Orthogonal Array Measured Data Objective

Expt. No. Temperature ◦C

2 To-25 Po to+8 CM2 4 4 4 3 3 4 4 3 3 -11.00

3 To-25 Po+200 to+16 CM3 20 17 17 20 21 19 18 21 22 -26.00

4 To Po-200 to+8 CM3 18 20 15 18 18 19 20 19 16 -25.00

Factor Effects & Analysis --

 A1 OCCURS IN EXPTS. 1, 2 AND 3, mA1 = 1/3 * (η 1 + η 2 + η 3 )

 FACTOR EFFECT OF A1 = a1 = mA1 - m

FACTOR EFFECT OF TEMPERATURE LEVEL 2 :

 A2 OCCURS IN EXPTS. 4, 5 AND 6, mA2 = 1/3 * (η 4 + η 5 + η 6 )

 FACTOR EFFECT OF A2 = a2 = mA2 - m

FACTOR EFFECT OF TEMPERATURE LEVEL 1 :

 A3 OCCURS IN EXPTS. 1, 2 AND 3, mA1 = 1/3 * (η 1 + η 2 + η 3 )

 FACTOR EFFECT OF A1 = a1 = mA1 - m

 REPEAT FOR ALL OTHER FACTORS AND ALL LEVELS

Find OvMean= mean of obj func = 1/9 * (η1 + η2 + . . . + η9) = -40

Factor Effect Plot for Surface Defects

Control Factor Levels