Chapter 4.

Roundoff and
Truncation Errors

Prof. Kiho Park

Numerical Methods 1
Errors: Accuracy and Precision
• Accuracy refers to how closely
a computed or measured
value agrees with the true
value.
– Inaccuracy (bias): systematic
deviation from the truth
• Precision refers to how closely
individual computed or
measured values agree with
each other.
– Imprecision (uncertainty): the
magnitude of the scatter

Numerical Methods 2
Errors: Definitions
• True error (Et) = true value - approximation

• True fractional relative error =

(true value - approximation)/(true value)
• Normalizes the error using the best available estimate of
the true value.
a = 100 x (approx. error)/(approximation) %

Numerical Methods 3
Errors: Definitions
• For iterative processes:
a = 100% x (present approx. - previous approx.)
/(present approx.)
• Stopping criterion: whether the absolute value of the
percent relative error is lower than a pre-specified
tolerance (s)

Numerical Methods 4
Errors: Example 4.1
• A Maclaurin series expansion for ex:
2 3 4 n
x x x x x
e 1  x      
2! 3! 4! n!
• Estimate e0.5 and compute the true and approximation
percent relative errors when each new term is added.
Terms Result et % ea %
1 1 39.3
2 1.5 9.02 33.3
3 1.625 1.44 7.69
4 1.645833333 0.175 1.27
: : : :
True value = 1.648721270700128….
Numerical Methods 5
Roundoff Errors
• Roundoff errors arise because digital computers
cannot represent some quantities exactly.
• Two major facets of roundoff errors involved in
numerical calculations:
– Digital computers have size and precision limits on their
ability to represent numbers.
– Certain numerical manipulations are highly sensitive to
roundoff errors.

Numerical Methods 6
Roundoff Errors: Computer Number
Representation
• The binary representation of the decimal integer
-173 on a 16-bit computer:
(10101101) 2 27  25  23  2 2  20 128  32  8  4  1
(173)10

• A range of integers for a 16-bit computer

(0111 111) 2 214  213   2 1 0 15
…  2  2  2 32,767 2  1
(0000 000) 2 0
(1111 111) 2 214  213   2 1 0
…  2  2  2  32,767  (2  1)
15

(1000 000) 2  32, 768

-32768 ≤ integer ≤ 32767 (-2 to 2 -1)
15 15

Numerical Methods 7
Roundoff Errors: Floating-Point
Representation
• Fractional quantities are typically represented in
computers using floating-point format:
 s  be

s = the significand
b = the base of the number system being used
e = the exponent
– This is done so computer memory is not wasted on storing
useless non-significant zeros.

0.005678 5.678  10 3

(not 0.005678 10 0 )
Numerical Methods 8
 Roundoff Errors: Floating-Point
Representation
• Example 4.2: Suppose a hypothetical base-10 computer with a
5-digit word size.
– one digit: the sign (s1)
– two digits: the exponent, one of them for its sign (so, do)
– two digits: mantissa (d1, d2)

s1 d1.d 2  10s d
0 0

• Roundoff errors: 2  5 0.03125  3.110  2 0.031

0.03125  0.031
0.008
0.03125
Numerical Methods 9
Roundoff Errors: Floating-Point
Representation
• By default, MATLAB has adopted the IEEE double-
precision format in which eight bytes (64 bits) are
used to represent floating-point numbers:

n = ±(1+f) x 2e

– The sign is determined by a sign bit

– The mantissa f is determined by a 52-bit binary number
– The exponent and its sign e is determined by an 11-bit
binary number

Numerical Methods 10
Roundoff Errors: Arithmetic
Manipulations of Computer Numbers
• Assume a hypothetical decimal computer with a 4-
digit mantissa and a 1-digit exponent
1.557  0.04341 ?
 0.1557 101  0.004341 101 0.160041 101 (?)
 0.1600 101
– Subtractive cancellation
0.7642 103
 0.764110 3 normalize  0.1000 10 0
0.0001103

Numerical Methods 11
Roundoff Errors: Floating-Point
Representation
• Large computations : Even though an individual
roundoff error could be small, the cumulative effect
over the course of a large computation can be
significant.
• Adding a large and a small number 0.4000 10 4
0.000000110 4
0.400000110 4

• Smearing - Smearing occurs whenever the individual

terms in a summation are larger than the summation
itself.
– e.g. (x+10-20) - x = 10-20 mathematically,
but, in MATLAB, when x=1, (x+10-20)-x = 0 !
Numerical Methods 12
Truncation Errors
• Truncation errors are those that result from using an
approximation in place of an exact mathematical
procedure.
– e.g.
dv v vti1 vti 
 
dt t ti1  ti



Numerical Methods 13
Truncation Errors: The Taylor Series
• The Taylor theorem states that any smooth function can be
approximated as a polynomial. The Taylor series provides a
means to express this idea mathematically.
f ' ' xi  2 f (3) xi  3 f ( n ) xi  n
f xi 1   f xi  f ' xi h  h  h  ...  h  Rn
2! 3! n!

f ( n 1) () n 1
Rn  h
(n  1)!

Numerical Methods 14
Truncation Errors: The Taylor Series
• Example 4.3: Use Taylor series expansions with n = 0 to 6 to
approximate f(x) = cos x at xi+1 = p/3 on the basis of the value of
f(x) and its derivatives xi=p/4
f ( xi ) 2
– 2nd order approximation: f ( xi 1 )  f ( xi )  f ( xi )h  h
2!

2
          cos( / 4)   
f   cos   sin       
3 4  4   12  2  12 
       
  0.707106781
         
     0.521986659
            
0.497754491

Numerical Methods 15
Truncation Errors: Numerical
Differentiation
• Forward difference approx.
' f (x i1 )  f (x i )
f (x i )   O(h)
h

• Backward difference approx.

' f (x i )  f (x i 1 )
f (x i )   O(h)
h
• Centered difference approx.

' f (x i1 )  f (x i 1 )
f (x i )   O(h 2 )
2h
Numerical Methods 16
Total Numerical Errors
• The total numerical error is the summation of the truncation
and roundoff errors.

Numerical Methods 17
Summary
• Accuracy vs. precision
• Learning how to quantify error
• Roundoff errors occur because digital computers have a
limited ability to represent numbers.
– floating-point numbers have limits on their range and precision
• Truncation errors occur when exact mathematical
formulations are represented by approximations.
– Taylor series to estimate truncation errors

Numerical Methods 18