Error propagation

Differentiation
Terminology and notation
The process of finding the rate of change of given function is called differentiation. The function is said to be
differentiated. If y is a function of the independent variable x, we say that y is differentiated with respect to
(w.r.t.) x.
d
rdinar differentiation (denoted by )
dx
 There is a notation for writing down the derivative of a function. If the function is y ( x) , we denote the
dy
derivative of y by , pronounced dee y by dee x .
dx
 nother notation for the derivative is simply y , pronounced y dash.
dy
 Similarly if the function is y (t ) we write the derivative as or y , pronounced y dot .
dt
df  d n 
 If we have a function f(x) derivative of f with respect to x is written as .  Rule : x  nx n 1 
dx  dx 
Derivative rules
d d
k  0, where k is a constant, x 1
dx dx
d du d n du
au  a u  nu n 1
dx dx dx dx
d 1 du d u du
ln u  e  eu
dx u dx dx dx
d du d du
sin u  cos u , cos u   sin u
dx dx dx dx
d du d du
tan u  sec 2 u cot u   csc 2 u
dx dx dx dx
df dg
 The derivative of f x  g x is 
dx dx
df
 The derivative of k f x is k
dx
xamples:
df
(1). f x  x3 ,  3x 2
dx
df
(2). f x  5x2  8x  9 ,  10 x  8
dx
(3). f x, y , z  3 x 4  3 y  z
df dy dz df dx dz df dx dy
 12 x3  3  ,  12 x 3 3  12 x3  3 1
dx dx dx dy dy dy dz dz dz

1
 df df
(4). If f x, y  sin x  y , Find  , and 
dx dy
df d df  dx dy  df  dy 
 cos x  y x y  cos x  y     cos x  y 1  
dx dx dx  dx dx  dx  dx 
df d df  dx dy  df  dx 
 cos x  y x y  cos x  y     cos x  y   1
dy dy dy  dy dy  dy  dy 

Partial differentiation
ntroduction
Partial differentiation (denoted by curly dee , ,)
x
Suppose we have function of several variables f(x, y, z). Then the partial derivative of f(x, y, z) with respect to
f
x is defined to be the ordinary derivative of f(x, y, z) with respect to x with all other variables treated as
x
constant.
 In engineering, it sometimes happens that the variation of one uantity depends on changes taking place
in two, or more, other uantities. For example, the volume V of a cylinder is given by V   r 2 h . The volume
will change if either radius r or height h is changed. The formula for volume may be stated mathematically as
V  f r , h which means V is some function of r and h . Some other practical examples include:

l
(i) time of oscillation, t  2 i.e. t  f l , g
g
(ii) tor ue I i.e.  f I,
1
(iii) resonant fre uency f r  i.e. f r  f L, C , and so on.
2 LC
V
 If V   r 2 h then means the partial derivative of V with respect to r, with h remaining constant .
r
V d 2
Thus,  h r  2 rh
r dr
V
Similarly, means the partial derivative of V with respect to h , with r remaining constant . Thus,
h
V d V V
  r2 h   r2 , and are examples of rst order partial derivatives. ,
h dh r h
 First order partial derivatives are used when nding the total differential, rates of change and errors for
functions of two or more variables.
Examples:
( ). f x, y , z  3 x 4  3 y  z
f f f
Solution:  12 x3 , 3 1
x y z
( ) f x, y, z  3 x 2  4 xz  yz
f f f
Solution: = 6x 4z =z = 4x y
x y z
2
( ) f x, y , z  z x e 2 y

f f f
Solution: = z e2y = z x 2e2y = x e2y
x y z
f f
(4) If f x, y  sin x  y , Find  , and 
x y
f f f f
Solution:  cos x  y x y  cos x  y ,  cos x  y x y   cos x  y
x x x y y y
f f
(5). If f x, y  x 2  2 xy  3 y 2 , Find  , and 
x y
f f
Solution:  2 x  2 y, and  2 x  6 y
x y
f f f
(6) f x, y, z  3 x 2  4 xz  yz Solution: = 6x 4z =z = 4x y
x y z

( ) f x, y , z  z x e 2 y ,

f f f
Solution: = z e2y = z x 2e2y = x e2y
x y z
f f
(8) f x, y  xe xy , Find  , and 
x y
f f f
Solution:  xe xy xy  e xy x  xe xy y  e xy  e xy xy  1
x x x x x
f f f
 xe xy xy  xe xy x  x 2 e xy
y y y y

z z
Problem 1. If z  5 x 4  2 x3 y 2  3 y find (a ) and (b)
x y
z
Solution (a) To find , y is kept constant
x
z  5 x 4  2 x3 y 2  3 y

z
 5x4  2 y 2 x3  3 y 1  20 x3  6 x 2 y 2
x x x x
z
(b) To find , x is kept constant
y

z  5 x 4  2 x3 y 2  3 y

z
 5x4 1  2 x3 y2  3 y  4 x3 y  3
y y y y

3
 z z 1
Problem 2. Determine and when z 
x y x2  y 2

1
1 
Solution z   x2  y2 2

x2  y2

z 1 
3
1 2 
3
x
  x2  y2 2 x2  y 2   x  y2 2 2x 
x 2 x 2 x2  y 2
3

z 1 
3
1 2 
3
y
  x2  y 2 2 x2  y 2   x  y2 2 2y 
y 2 y 2 x2  y2
3

l
Problem 3. The time of oscillation, t ,of a pendulum is given by t  2 where l is the length of the
g
t t
pendulum and g the free fall acceleration due to gravity. Determine and
l g
t  2  1
Solution For , t 
 g 
l2
l  

t 2 1
2 1  12 
 l2   l 
l g l g 2 lg
1
t 
For , t  2 l g 2
g

t 
1
 1 3  l
 2 l g 2
 2 l    g 2   
g g  2  g3

1
Problem 4. The resonant fre uency f r in a series electrical circuit is given by f r  . Show that
2 LC
fr 1

L 4 LC 3
1
 1
1  2
Solution . f r   L
2 LC  2 C 
 1 2  1
1 3
fr 1  1   1  1 fr
   L 2
  
 2 L  
L 2 LC  2 C  L 2 LC  2 C   L 4 CL3

x z z z 1 z x
Problem 5. If z  , find and ( nswer  ,  2)
y x y x y y y
y 1 z z
Problem 6. If z  x3 y 2  2  , find and
x y x y
z 2y z 1 1
( nswer  3x 2 y 2  3 ,  2 x3 y  2  2 )
x x y x y
z z
Problem . If z  cos 3 x sin 4 y, find and
x y
z z
( nswer  3sin 3 x sin 4 y,  4 cos 3x cos 4 y )
x y
4
 Error propagation
 The calculation of an error from the errors on other uantities is called error propagation.

Let us start by considering a simple experiment, to measure the time it takes to a metal ball to fall through a given
distance. The switch that releases the ball also starts the clock. The micro switch stops the clock when the ball strikes it.

magnet

metal ball

digital clock
00.000

micro switch

Figure 1.

 Suppose we make few measurements. Some typical numbers might be: 3.46 s. 3.459 s, 3.463 s, 3.465 s
etc. ll the numbers are different. This is known as phenomenon of errors. Whenever, we repeat a
measurement, using sufficiently precise apparatus, the results will not always be the same. lways a small
spread in the values. This spread can never be reduced to zero.
 We can never know the true value. If we use a clock that only counts seconds (instead of 0.001 s) all
values are e ual to 3 s. No spread but not very useful. We can use better e uipment. Lasers and photocells
could be used to start and stop the clock, and everything could be mounted in a vacuum. We might get results
like 3.462 5 s. 3.46266 s, 3.462 0 s, etc. The spread of number is much smaller, but it still exists.
re uenc diagram

t
(t  , t  

Figure 2.

5
The plot of measured values (table 1) is shown in figure 2. axis coves all possible measurements. axis
records the number of times each measurement occurred. For example, 3.461 occurred 16 times, so x y
coordinates (3.461, 16).
 The fre uency diagram gives a visual picture of how the measurements are distributed. We can easily
see that the center of the distribution is around 3.462 s. Figure 2 is called fre uency distribution, distribution
curve, or histogram etc. The point at 3.4 6 s can be ignored (bad data).
ean value
The mean value (or average) of any set of measurements is just the sum of the measurements divided by the
number of measurement.
n
i 1 i
x
x = x = (1)
n
x1  x2  x3    xn
x = x =
n
From the measurements (table 1) excluding 3.9 6 s.
3.461  3.463    3.462
t  =
99
It is easy to group the values together using the fre uency diagram.
1 3.456  2  3.45  4  3.458  
t  =
99
3.456 occurs once, 3.45 twice, 3.458 four times etc.
If fl be the fre uency that any particular value t occurs, we can write
fl  t
t  = l
(2)
n
Where n  fl , Using e uation (2),  t  = 3.4619 s
l

tandard deviation

xperiment 1
xperiment 2

Figure 3.
 Suppose that two separate experiments are done to measure the same uantity. The fre uency diagrams
from the two experiments might look like figure 3.
 Both have the same mean but second experiment has larger spread than the first experiment.
 xperiment 1 is better than the experiment 2. We would like a method by which we could calculate
a number, which express this fact. The method is to calculate the standard deviation ().
ere we will just state how to calculate it and use it for our purposes. It is represented by sigma ().
6
 n 2
x  xi
x  i 1
(3)
n 1
x means standard deviation for the distribution of x values. It is the width for the distribution since it
depends on the distances of the individual measurements fro the mean.
 From the table 1, using  t  = 3.462 s.
1/ 2
1(3.462  3.456) 2  2(3.462  3.45 ) 2   
l     l = 0.00239 s
 98 
It has the same units as the units on the measurements.
 The meaning of  is that if we move a distance  l on either side of the mean value, we will form an
2
interval that contains about of the measurements. (  t   l ,  t   l ) = (3.4596, 3.4644).
3
So the standard deviation is a measure of the precision of the experiment.
eneral formula for error propagation
We want to calculate the error on some uantity , which is a function of several variables, = (x, y, .. )
then
2 2
 Z   Z 
 z    x     y    (1)
 x   y 
If is a function of only one variable, x
dZ
z  x (2)
dx
bsolute value is used to ensure that z is always positive. uation (1) is valid when
x y
(1)  1 ,  1
x y
(2) x, y are independent.
The majority of calculations will fall into two categories (a) addition and subtraction (b) multiplication and division

ddition and su traction

Let Z  ax  by  cu 

Z Z Z
Where x, y, u, . are variables and a, b, c, . are constants. ,  a,  b,  c . Using e uation (1)
x y z
2 2 2
z  a x  b y  c u   (3)

This e uation is the error formula to use adding and subtracting uantities. We can say the errors are added in
uadrature, which means add the s uares, then, take the s uare root.
Examples
Example . z  x  y
Where x = 3.6  0.03 and y = 2.14  0.02
z = 3.6 – 2.14 = 1.53
2 2
z =  x2   y2 z = 0.03  0.02 z = 0.036
 z = 1.53  0.036
Example . R = 3L 4M
Where L = 63.6  0.2 and M = 19.  0.3
R = 3(63.6) 4(19. ) = 269.6
2 2 2 2
R = 3 L  4 M R = 3  0.2  4  0.3 R = 1.3

 R = 269.6  1.3
ultiplication division or raised to power
Consider Z  Ax n y  m
Z Z
 A n x n 1 y  m  A x n (m) y  m 1
x y
s e uation (1)
1/ 2
 Z  2  Z
2

z =    x      y  
2 2

 x   y 
1/ 2
z =  Anx n1 y  m   x 2  Ax n m y  m1   y 2 
2 2

 
It will be complicated when there are more than two variables, to simplify, divide it s both sides by .
1/ 2
z  Anx n1 y  m 2  Ax n m y  m1 
2


=  n m 
 x  
2
n m   y 
2

Z  Ax y Ax y 
    
1/ 2
 n  2 m
2

z =     x      y 2 
2

 x   y 
If we have more variables like Z  Ax n y mu  p
1/ 2
 n  2 m
2
 p
2 
z =     x      y      u 2  
2 2
(3)
 x   y u 
For single variable
1/ 2
 n  2  n
z =     x 2  z =    x (4)
 x   x
Example . A  LW
1/ 2 1/ 2
 1  2  1
2
   L  2   W  2 
 =     L
2
   W 
2
 =      
 L  W    L   W  

Example . Z  x2
1/ 2
 2  
2
2
z =    x   z =  x z = 2x x
 x   x
Example . Z y
1/ 2
 1 1  
2
1 1
 z  Z     y   z =  y z =  y
 2 y   2y 2 y

8
 1
Example . V   r3
3
1/ 2
 3  
2
3 V
 V  V    r   V   r
 r   r

g
Example . iven that f 
2s

g = 980.6 cm / s2 s = (0. 6  0.02) cm / g1/2 = (0.040  0.001) g / cm

(1) Calculate f and its error
(2) Which variable contributes most of the error, s or .

980.6
f  f = 103.01 / s
2  0. 6  0.040

g 1
For f, as f = s 1/2
2
1/ 2
 2
1 2   
2
2 s 
 f  f (1)     ( )    
  s  2   
1/ 2 1
 0.02 2 1  0.001 2   0.000625  2
 f  103.01        103.01 0.000692  
 0. 6  4  0.04    4
1/ 2
 f  103.01 6.9252 104  1.5625 104  ( B)
1/ 2
 f  103.01102 8.48  f  103.01 2.9133 102  f  300.105 102  f  3.00105 s 1

 f = (103.01  3.00) s 1
(2) s contribute more than .(clear from e uation (B))
Example . Z  A cos , where is a constant and has error.
1/ 2
 Z  2  dZ
 Z      2  Z  
   d

z = sin  (Note  must be in radians ( radians = 180o))

Let = 16, = (36.3  0.2) degree
= 16 cos(36.3) =12.895

 = 0.2 x = 0.0035 radian
180
 z = 16 sin(36.3) (0.0035) z = 0.0332  =12.895  0.0332
Example . Z  xyw
Let x = 3.6  0.1 y = 2.  0.2 w = 10.2  0.8
= (3.6)x(2. ) 10.2 = 19.92

9
 1/ 2 1/ 2
 z   y   x  w 
2
 z   2.  0.1  3.6  0.2  0.8 
2 2 2 2 2
 x  y
   
1/ 2
 z  0.0 29  0.5184  0.64 z = 1.1  =19.92  1.1

Table 1. Measurements of fall time taken for the falling ball experiment. histogram of the data is shown in figure 2.
Time (s) Time (s) Time (s) Time (s) Time (s) Time (s)
3.461 3.460 3.461 3.456 3.461 3.461
3.463 3.464 3.465 3.464 3.465 3.463
3.460 3.461 3.466 3.462 3.463 3.462
3.465 3.462 3.462 3.458 3.460 3.45
3.463 3.464 3.463 3.4 6 3.464 3.465
3.461 3.461 3.460 3.461 3.461 3.460
3.463 3.459 3.466 3.465 3.465 3.461
3.464 3.460 3.461 3.462 3.462 3.463
3.460 3.45 3.462 3.459 3.459 3.458
3.458 3.462 3.460 3.458 3.463 3.462
3.462 3.46 3.463 3.462 3.464 3.459
3.468 3.460 3.462 3.463 3.461 3.461
3.463 3.463 3.459 3.461 3.459 3.460
3.459 3.461 3.462 3.464 3.462 3.46
3.464 3.464 3.461 3.460 3.463 3.462
3.462 3.462 3.463 3.466 3.460
3.466 3.463 3.466 3.462 3.464
Table 2.
time fre uency time fre uency
3.460 12 3.46 2
3.461 16 3.468 1
3.462 18 3.4 6 1
3.463 15 3.459
3.464 10 3.458 4
3.465 6 3.45 2
3.466 5 3.456 1

10
 Determine the volume of a c linder using Vernier aliper
In 1631, a French instrument maker, Pierre Vernier, devised a way to improve the precision of length
measurements. The Vernier aliper, commonly called a Vernier, consists of a main engraved scale and a
movable jaw with an engraved Vernier scale (figure 1). The span of the lower jaw is used to measure length
and is particularly convenient for measuring the diameter of a cylindrical object. The span of the upper jaw is
used to measure distances between two surfaces, such as the inside diameter of a hollow cylindrical object.
 Vernier caliper is used to measure internal dimensions, external dimensions, and depth of an object

Figure 1. Vernier Caliper

Zeroing
Before making a measurement, one should check the zero of the Vernier Caliper with the jaws completely
closed. It is possible that through misuse the caliper is no longer zeroed and thus gives erroneous readings
(systematic error). If this is the case, a zero correction should be made for each reading.
 In zeroing, if the Vernier zero lies to the right of the main scale zero, measurements will be too large
and the error is taken to be positive. In this case, the zero correction is made by subtracting the zero reading
from the measurement reading.
 Similarly, if the error is negative, or the Vernier zero lies to the left of the main scale zero,
measurements will be too small, and the zero correction must be added to the measurement readings.

igure Zeroing and error. The zero of the Vernier Caliper is checked with the jaws closed. (a) ero error.
(b) Positive error, 0.05 cm.

11
 ow to read a Vernier aliper
 The main scale is calibrated in centimeters with a millimeter least count, and the movable Vernier scale
has 10 divisions that cover 9 divisions on the main scale. The function of the Vernier scale is to assist in the
accurate reading of the fractional part of the scale division, thus increasing the precision.
The leftmost mark on the Vernier scale is the zero mark (lower scale for metric reading and upper scale for
inches). The zero mark is often unlabeled.
 measurement is made by closing the jaws on the object to be measured and reading where the zero
mark on the Vernier scale falls on the main scale (figure 3). In figure 3, the first two significant figures are read
directly from the main scale. The Vernier zero mark is past the 2 mm line after the 1 cm major division mark,
so there is a reading of 1.2 cm (12 mm).
 The next significant figure is the fractional part of the smallest subdivision on the main scale. This is
obtained by referring to the Vernier scale markings below the main scale.
If a Vernier mark coincides with a mark on the main scale, then the Vernier mark number is the fractional part
of the main scale division. In the figure, this is the third mark to the right of the Vernier zero, so the third
significant figure is 3 (0.03 cm (0.3 mm)). Finally, since the 0.03 cm reading is known exactly, a zero is added
as the doubtful figure for a reading of 1.230 cm or 12.30 mm.
 owever, a mark on the Vernier scale may not always line up exactly with one on the main scale, and
we say there is a change of phase between two successive Vernier markings. In that situation take the middle
of the range. Thus a 5 would be put in the thousandth of a centimeter digit (hundredth of a millimeter digit).
Example (calculating the least count, and reading the Vernier scale on a Vernier caliper.)

Figure 3. n example of calculating the least count, and reading the Vernier scale on a Vernier caliper.
1 mm
Least count ( L.C )   0.1 mm
10
Main scale reading = a = 12 mm
Vernier coincidence (aligned mark) = n = 3
Vernier scale reading = b = n x least count = 3x 0.1 = 0.3 mm
Total reading = a b = 12.30 mm = 1.230 cm

12
Volume of a c linder
V   r 2l , where l is the length and r is the radius of the cylinder
smallest main scale division
Least count ( L.C )   1/20
total number of vernier scale division
Length of the cylinder:

Main scale reading Vernier coincidence Vernier scale reading Total reading
a (mm) n b = n x least count (mm) l = (a b) (mm)

26 12 0.6 26.6

Length of the cylinder = l = (mm)

Diameter of the cylinder:

Main scale reading Vernier coincidence Vernier scale reading Total reading
a (mm) n b = n x least count (mm) d = (a b) (mm)

Diameter of the cylinder = d = (mm)

Radius of the cylinder = r = (mm)

V   r 2l
Volume in V  mm3
Volume in V  m3

Use:  r   l   0.05 mm (reading error)

Standard deviation (error propagation)

1/ 2 1/ 2
 n  2 m
2
  2  2 1
2

 V  V     x 2      y 2   V     r 2      l 2 
x  y  r  l 
   

Volume  V   V 

13
 Determine the volume of a sphere using micrometer (screw gauge)
The most widely used device for the accurate measurement of small dimensions is the micrometer
(screw gauge), as shown below. It is particularly convenient in measuring the diameters of thin wires
and the thicknesses of thin sheets.

Figure 1. Micrometer
 The axial line on the sleeve main scale serves as a reading line.
 The movement mechanism of the micrometer is a carefully machined screw with a pitch of
0.5 mm. The pitch of a screw, or the distance between screw threads, is the lateral linear distance the
screw moves when turned through one rotation.
 One complete turn of T (rotating scale) is 0.5 mm on main scale and the numbers on the S
(main scale) show the number of millimeters. Since there are 50 divisions on the T (rotating scale).
Lengths can be measured to hundredth's of a millimeter.
pitch of the screw 0.5
Least count ( L.C )    0.01 mm
total number of divisions on the rotating scale 50

Example of a micrometer reading:

Figure 2. Micrometer reading

The S (main scale) reading  6.5 mm.
Number of rotating scale divisions coinciding with the reading line  48.
Least count  0.01 mm
Rotating scale reading 48  0.01  0.48 mm
The complete reading  6.5  0.48  6.98 mm

14
Volume of a sphere
4
V   r 3 , where r is the radius of the sphere
3
Diameter of the sphere:

Main scale reading Number of rotating scale Rotating scale reading Total reading
divisions coinciding
a (mm) with the reading line b  n  least count (mm) d  a  b (mm)
n

Diameter of the sphere = d = (mm)

Radius of the sphere = r = (mm)

4
V   r3
3

Volume in V  mm3

Volume in V  cm3

Use:  r   0.005 mm (reading error)

Standard deviation (error propagation)

1/ 2
 3  2  3 
 V  V     r 2   V   r 
 r   r 

Volume  V   V 

15
 Precautions and care
 Clean the ends of the anvil and spindle before use.
 Check the instrument for zero error (systematic error) reading before use.
 While taking measurements, the thimble is turned until the object is gripped very gently
between anvil and spindle.
 Use the ratchet to apply a consistent pressure on the object. The ratchet permits the jaw to be
tightened on an object with the same amount of force each time.

Zero error.
 If on touching the tip of the spindle with the anvil, the zero mark of the rotating scale
coincides with the zero mark of the reading line, the instrument is free from zero error, figure 3 (a).
 But if the zero mark of the rotating scale remains below the reading line of the main scale, the
zero error is said to be positive. To find it, note the division of the rotating scale which coincides
with the reading line of the main scale. Now multiply this number with the least count which gives
the zero error. In figure 3(b) the 2 division of rotating scale coincides with the reading line. If the
least count of the screw gauge is 0.01 mm, then zero error is  2  0.01  0.02 mm.
 On the other hand, if the zero of the rotating scale goes above the reading line of the main
scale, the zero error is said to be negative. To find it, note the division of the rotating scale which
coincides with the reading line of the main scale. Subtract it from the total number of divisions on
the rotating scale and then multiply it with the least count. In figure 3(c) 46 division of rotating scale
coincides with the reading line and the total number of divisions on the rotating scale are 50 and the
least count is 0.01 mm, then the zero error   (50  46)  0.01  0.04 mm.
 To find the correct reading, the zero error with its proper signs is subtracted from the
observed reading. Thus,
Correct reading = observed reading – zero error (with sign)

Figure 3. (a) No zero error (b) Positive zero error (c) Negative zero error

16