PHY1002 Homework on Error Analysis

1 Significant figures (2 pts) should expect (δδ) = 0. At the end of the day, using the
definition of the standard error σ, we can obtain
v
t ∂ƒ 2 ∂ƒ 2
u   
1.1 Determine the significant figures σ = σ2 + σ2
∂ ∂
Write down the number of significant figures for the fol- Exercise problems: Now find the standard error σ in
lowing numbers  = ƒ (, ) as a function of the errors in σ and σ for the
(a) 1.00101 ( ) following functions:
(a)  =  +  (0.5 pts)
(b) 1.0110 × 10−3 ( )
You can find the absolute uncertainty of the sum
(c) 0.0010011 ( ) (or difference) is the root square sum of the

(d) 3.140 ( ) individual absolute uncertainties when adding (or

subtracting).
(e) 16700. ( )
(b)  =  ×  (0.5 pts)
(f) 1.68 × 104 ( )
(c)  = /  (1 pt)
1.2 Significant figures in calculations You can find that the relative uncertainty of the prod-
uct (quotient) is the root square sum of the individual
Express the results of the following calculations with the relative uncertainties.
correct significant figures.
(d)  =  2 (1 pt)
(a) 3.1416 × 0.28/ 2.34
(e)  =  exp(c) with c constant. (0.5 pts)
(b) 123.62 + 7.1 − 5.33
2 Propagation of Uncertainty (Error) (f)  = 1/  (0.5 pts)

3 Snell’s Law (2 pts)

Let us first define the standard deviation s. Suppose we

perform N measurements 1 , 2 , · · · , N with the average Through the equation n1 sin θ1 = n2 sin θ2 , the Snell’s
̄. Then the deviation of each measurement is given by law relates the incident angle θ1 of a ray traveling in a
δ =  − ̄ with  = 1, 2, · · · , N. The standard deviation s is medium of index n1 to the refraction angle θ2 of the same
light ray in the medium of refraction index n2 . Find n2 and
vP
u N
t
=1
(δ )2
s= its uncertainty from the following measurements
N−1
When we report the average value of N measurements, θ1 = 22.0◦ ± 0.2◦
the uncertainty we should associate with this average θ2 = 16.3◦ ± 0.2◦
value is the standard error. n1 = 1.000 (assumed to be exact)
v
u PN
t =1 (δ )
2
s u Note that δθ must be converted into radians when you
σ= p =
N N(N − 1) compute the uncertainty. In calculus, we only use radians
The standard error is smaller than the standard deviation in trigonometrical functions.
4 Simple Pendulum (2 pts)
p
by a factor of 1/ N, since the statistical uncertainty can
be reduced by large number of measurements. Also it is
1 XN
useful to write σ2 = δ2 ≡ δ2 . Suppose you determine the acceleration of gravity g =
N(N − 1) =1  4π 2 L
Suppose we want to determine a quantity  = ƒ (, ), by measuring the oscillation period T of a pendu-
T2
which depends on  and . We want to know the error in lum with length L. Determine the value and uncertainty
 = ƒ (, ) if we measure  and  with errors σ and σ . of g from the following measurement
Using the Taylor expansion, we can obtain the law of the
T = 2.01 ± 0.02 s, L = 1.000 ± 0.002m.
error propagation as follows
If you wish to improve the above measurement
∂ƒ 2 ∂ƒ 2 ∂ƒ ∂ƒ
     
(δ)2 = (δ)2 + (δ)2 + 2 (δδ) significantly, which part of the measurement (T or L) do
∂ ∂ ∂ ∂
you want to improve? Why?
If the measurements of  and  are uncorrelated, then,
on the average, we should expect to find equal distribu-
tions of positive and negative values for this term, and we

PHY1002, Lecturer 2022, Page 1