ECE 510 Lecture 3

Functions
Reliability Functions, T&T 2.1-6, 9
Distributions, T&T 3.1-4, 4.1-4, 5.1-3

Scott Johnson
Glenn Shirley
 Reliability Functions

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 2

 Reliability Functions
H t    ht  dt 
t
• Functions of time 0
– CDF(x)  F(t) 1

F t    f t  dt 
t
• Survival function S(t) = 1-F(t)
0
• PDF(x)  f(t)
f t  
fraction of ORIGINAL population that fails in dt S t   1  F t 
dt
t
dF t  dS t  0
 
dt dt
• Hazard function h(t)
f t 
ht  
fraction of CURRENT population that fails in dt
ht  
dt S t 
f t  dS t  1 d ln S t 
  
S t  dt S t  dt

• Cum hazard function H(t)

f t   F t 
d
H t    ht  dt
t

0
dt
S t   exp H t 
F t   1  exp H t  0 t

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 3

 Exercise 3.1a
• Calculate H(t), S(t), and F(t) for the given human mortality data, and plot
h(t), S(t), and F(t). The data is given as h(t) for each age, that is, the
probability of a living person dying at the given age. Use a sum to
approximate the integral for H(t).

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 4

 Exercise 3.1a Solution, Part 1

H t    ht  dt
t

S t   exp H t 
F t   1  exp H t 

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 5

 Human Mortality Graphs

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 6

 Reliability Indicators
1

0.5
S(t)

0 t

t50 MTTF

• Mean time to failure (MTTF)

 N 
MTTF   t f t  dt    S t  dt
1
N
t
j 1
N
0 0

• Median time to failure (t50) is the solution of

S t50   0.5
– Time at which half of the initial population fails

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 7

 Exercise 3.1b
• Find the mean and median times to failure for the human mortality data
set from the last exercise

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 8

 Exercise 3.1b Solution

MTTF=76.4 t50=80

• Sum S(t) to get MTTF

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 9

 Reliability Measures: DPM
• Metric designed for low fail rates
• DPM = Defects Per Million

Typical target
% pass % fail DPM at end of life

99 1 10,000
Typical target
99.9 0.1 1000 at t=0

99.95 0.05 500 Typical range for

semiconductor
99.99 0.01 100 reliability

99.999 0.001 10

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 10

 Reliability Measures: FIT
• FIT = Failures In Time
• FIT is a fail rate, fails per billion device hours
– FIT = DPM per 1,000 hours
• DPM is a fail total, fails per million total devices
– DPM = FIT * hours / 1,000

200 FIT for 1,000 hours = 200 DPM

200 FIT for 10,000 hours = 2,000 DPM
200

FIT
time
0
0 1,000 10,000
(hours)
1 year

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 11

 Reliability Indicators: AFR
h

AFR

0 t
t1 t2

• AFR, Average Fail Rate

ht  dt H t   H t  ln S t   ln S t 
t2

AFRt1 , t 2 
t1
  2 1 1 2
t 2  t1 t 2  t1 t 2  t1

– If t in hours, units are fail fraction per hour

– Multiply by 109 for units of FIT

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 12

 Exercise 3.1c
1. Plot the hazard function in FIT
2. Find the AFR (in FIT) for:
– The 10-year range from ages 6 to 15
– The 10-year range from ages 71 to 80
– The 10-year range from ages 91 to 100
– The entire 100-year range from ages 1 to 100

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 13

 Exercise 3.1c Solution

91-100
71-80
1-100

6-15
/ (24*365) * 10^9

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 14

 The Bathtub Curve
Failure Rate

Total Constant
External
Use Time
< 1 year 5-15 years

•Infant Mortality (IM) from latent reliability defects

•Wearout from reliability mechs like oxide wearout
•Constant from external effects like radiation
•Many versions of this graph – it is a very important concept
14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 15
 The End

14 Jan 2013 ECE 510 S.C.Johnson, C.G.Shirley 16