EE290H F05 Spanos

Robust Design

A New Definition of Quality.

The Signal-to-Noise Ratio.
Orthogonal Arrays.

Lecture 9: Robust Design 1

EE290H F05 Spanos

The Taguchi Philosophy

Taguchi starts with a new definition of Quality:

Quality
Qualityisisrelated
relatedto
tothe
thetotal
totalloss
lossto
tosociety
societydue
dueto
to
functional
functional and
andenvironmental
environmental variance
varianceof
ofaagiven
givenproduct
product

Taguchi's method focuses on Robust Design through use of:

• S/N Ratio to quantify quality
• Orthogonal Arrays to investigate quality

Lecture 9: Robust Design 2

EE290H F05 Spanos

Meeting the specs vs. hitting the target

better quality worse quality

m-5 m m+5
Lecture 9: Robust Design 3
EE290H F05 Spanos

Quadratic Loss Function:

Fig 2.3 pp 18 from

Quality Engineering Using Robust Design
by Madhav S. Phadke
Prentice Hall 1989

L(y) = k (y - m)2
Lecture 9: Robust Design 4
EE290H F05 Spanos

Quadratic Loss Function on Normal Distribution

Average quality loss due to µ and σ:

Fig 2.5 pp 26

2 2
E(Q) = k [(µ-m) + σ ]

Lecture 9: Robust Design 5

EE290H F05 Spanos

Exploiting non-linearity:

Fig 2.6 pp 28

Lecture 9: Robust Design 6

EE290H F05 Spanos

Parameters are classified according to function:

Fig 2.7 pp 30

Lecture 9: Robust Design 7

EE290H F05 Spanos

Orthogonal Arrays
During Regression Analysis, an orthogonal arrangement of
the experiment gave us independent model parameter
estimates:

b = (XTX)-1XTy V(b) = (XTX) -1σ2

Orthogonal arrays have the same objective:

For every two columns all possible factor combinations
occur equal times.

L4(23) L9(34) L12(211) L18(21 x 37)

Lecture 9: Robust Design 8

EE290H F05 Spanos

Simple CVD experiment for defect reduction

max n = -10 log 10(MSQ def)

Lecture 9: Robust Design 9

EE290H F05 Spanos

Simple CVD experiment for defect reduction (cont)

Using the L9 orthogonal array:

Lecture 9: Robust Design 10

EE290H F05 Spanos

Estimation of Factor Effects (ANOM)

m = 1 η1+η2+η3+...+η9
9
mA 1 = 1 η1+η2+η3
3
mA 2 = 1 η4+η5+η6
3
mA 3 = 1 η7+η8+η9
3
...
mB2 = 1 η2+η5+η8
3
...
mD3 = 1 η3+η4+η8
3

η A i,Bj,Ck,Dl = μ+αi+βj+γk+δl+e
Σ αi = 0 Σ βi = 0 Σ γi = 0 Σ δi = 0
Lecture 9: Robust Design 11
EE290H F05 Spanos

Analysis of CVD defect reduction experiment

Fig 3.1 pp 46
Tab 3.4 pp 55

Lecture 9: Robust Design 12

EE290H F05 Spanos

ANOVA for CVD defect reduction experiment

9
Grand total sum of squares: Σ ηi2 = 19,425 (dB)2
i=1
9
Sum of squares due to mean: Σ m2 = 15,625 (dB)2
i=1
9
Total sum of squares: Σ ηi-m 2 = 3,800 (dB)2
i=1
3
Sum of squares due to A: 3 Σ mA i-m 2 = 2,450 (dB)2
i=1

9
Sum of squares due to error: Σ ei2 = ??? (dB)2
i=1
Lecture 9: Robust Design 13
EE290H F05 Spanos

ANOVA for CVD defect reduction experiment (cont)

Lecture 9: Robust Design 14

EE290H F05 Spanos

Estimation of Error Variance

The experimental error is estimated from the ANOVA

residuals.

It is then used to estimate the error of the effects and to

determine their significance at the 5% level.

Lecture 9: Robust Design 15

EE290H F05 Spanos

Confirmation Experiment
Once the optimum choice has been made, it is tested by
performing a confirmation run.
This run is used to "validate" the model as well as confirm
the improvements in the process.

Variance of prediction (for the model)

1 σe 2 = 1 + 1 - 1 + 1 - 1 σe 2
n0 n nA 1 n nB1 n

σpred 2 σe
2
= n + nσe
2
0 r

This gives us +/-2σ limits on the confirmation experiment.

Lecture 9: Robust Design 16

EE290H F05 Spanos

The additive model

Since we assumed additive model, we must make sure
that there are no interactions:

Fig 3.3 pp 63, enlarged 120%

Lecture 9: Robust Design 17

EE290H F05 Spanos

Example: Large CVD experiment.

Objectives:

a) reduce defects n = -10 log10(MSQ Def)

2 2
b) maximize S/N of rate n'= 10 log10(µ / σ )

c) adjust poly thickness to a 3600 Å target.

Lecture 9: Robust Design 18

EE290H F05 Spanos

Choosing the Control Factors

Tab 4.6-7 pp 88-90

Lecture 9: Robust Design 19

EE290H F05 Spanos

Using the L18 orthogonal array...

Tab 4.3 pp 78, enlarged 120%

Lecture 9: Robust Design 20

EE290H F05 Spanos

Data summary for large CVD experiment:

Tab 4.5 pp 85, enlarged 120%

Lecture 9: Robust Design 21

EE290H F05 Spanos

Data analysis for large CVD experiment (cont)

Fig 4.5 pp 86, enlarged 120%

Lecture 9: Robust Design 22

EE290H F05 Spanos

ANOVA table for large CVD experiment: η

Lecture 9: Robust Design 23

EE290H F05 Spanos

ANOVA table for large CVD experiment : η’

Lecture 9: Robust Design 24

EE290H F05 Spanos

ANOVA tables for large CVD experiment: η’’

Lecture 9: Robust Design 25

EE290H F05 Spanos

Combined Prediction Using the Additive Model

Tab 4.6-7 pp 88-90

Lecture 9: Robust Design 26

EE290H F05 Spanos

Verification for large CVD experiment

Tab 4.10-11 pp 92

for further reading: Quality Engineering Using Robust Design by Madhav S. Phadke
Prentice Hall 1989
Lecture 9: Robust Design 27
EE290H F05 Spanos

“Inner” and “Outer” Arrays

• Often one want to improve performance based on some “control”

factors, in the presence of some “noise” factors.
• Two arrays are involved: the inner array explores the “control”
factors, and the entire experiment is repeated across an array of the
noise factors.
• Inner arrays are typically orthogonal designs
• Outer arrays are typically small, 2-level fractional factorial designs.

Lecture 9: Robust Design 28

EE290H F05 Spanos

Why use S/N Ratios?

• They lead to an optimum through a monotonic function.

• They help improve additivity of the effects.

S ⎛Y 2 ⎞
Nominal is best: = −10 log⎜⎜ 2 ⎟⎟
N ⎝s ⎠
S ⎛1 1 ⎞
Larger is better: = −10 log⎜⎜ ∑ 2 ⎟⎟
N LTB ⎝ n i Yi ⎠
S ⎛1 2⎞
Smaller is better: = −10 log⎜ ∑ Yi ⎟
N STB ⎝n i ⎠

Lecture 9: Robust Design 29

EE290H F05 Spanos

Taguchi vs. RSM

Taguchi RSM
Small number of runs Explicit control of Interactions
Engineering Intuition Statistical Intuition
“Complete” package Training Issues
Additive Models More General Models
Orthogonal Arrays Fractional Factorials
A “Philosophy” A Tool

Lecture 9: Robust Design 30

EE290H F05 Spanos

Design of Experiments
Comparison of Treatments
Blocking and Randomization Analysis
Reference Distributions
ANOVA
MANOVA
Factorial Designs
Two Level Factorials
Blocking
Fractional Factorials
Modeling
Regression Analysis
Robust Design
Lecture 9: Robust Design 31