College of Engineering and Computer Science

Computer Science Department

Computer Science 106
Computing in Engineering and Science
Last update: April 11, 2006 Instructor: Larry Caretto

Binary Numbers and Computer Representation of Numerical Data

Introduction

Most people have heard that a computer is a binary device, but many students in beginning
programming courses are not familiar with the details of binary numbers. Because all data is
stored in a computer as a binary representation, that is a series of ones and zeros in the
electronic circuits of the computer, it is necessary to understand the structure of binary numbers
to understand the limitations of the computer representation of data.

These notes first provide a background on binary numbers then discuss how these numbers are
used to represent various data types in C++.

Binary numbers

The interpretation of binary numbers is a specific application of the general approach to

representing numbers in any base. In out usual base ten, we recognize that the number 132
represents 1 times 102 plus 3 times 101 + 2 times 100. We can create a formula for representing
this number in base ten and then generalize if to other bases. To start with we write the three
digits in our number (the 1, the 3, and the 2 in the number 132) as the symbol, di, for the digit in
position i. We number the digits from right to left, starting with the symbol d0 for the rightmost
digit, which is 2 in this example of the number 132. We then say that the first digit to the left of d0
is d1, which is 3 in this example. The next digit to the left is d2, which is 1 in our example. We can
represent a general three-digit number by the symbols d2d1d0. Since this is a base-ten number,
we can calculate the value of the number as d2(102) + d1(101) + d0(100). Note that the range of
digits for a base ten number is zero to nine.

How can we make this more general? We want to represent a number, in any base, with an
arbitrary amount of digits in the number. For a base-ten number with N digits, we can write the
following equation for the value of the number.

Here, we have written the number on the left as a sequence of digits and on the right as the
calculation, which gives its value. Note that the subscripts for the digits start with zero for the
rightmost digit and range to N-1 for the leftmost digit. This range from zero to N-1 gives us a total
of N digits. We can rewrite the equation above using the usual sigma notation for sums.

In this equation, we use ten because that is the base of our usual number system. To consider
other number systems such a binary (base 2), octal (base 8), or hexadecimal (base 16), we use
the same formula, but we replace the factor to ten by the appropriate base, b.

Jacaranda (Engineering) 3333 Mail Code Phone: 818.677.6448

E-mail: lcaretto@csun.edu 8348 Fax: 818.677.7062
 When want to formally represent the base that we are using for the number, we include it as a
subscript following the number. Thus the base-ten number shown above as 132 could be
formally written as 13210 to emphasize that this is a decimal (base ten) number.

Note that the value of any “digit”, dk in a given numbering system must be less than the base b.
We know that decimal (base ten) digits range from zero to 9; similarly octal digits range from 0 to
7 and binary digits range from 0 to 1. Hexadecimal digits range from 0 to f, where we have the
following correspondence between hexadecimal digits and base ten numbers: a16 = 1010; b16 =
1110; c16 = 1210; d16 = 1310; e16 = 1410; f16 = 1510. We can write our example number 13210 in the
following bases: 13210 = 100001002 = 2048 = 8416.

The table below shows the numbers in four bases (decimal, binary, octal and hexadecimal) for
numbers from 010 to 2310.

Numbers from 010 to 2310 in Four Bases

Decimal 0 1 2 3 4 5 6 7 8 9 10 11
Binary 0 1 10 11 100 101 110 111 1000 1001 1010 1011
Octal 0 1 2 3 4 5 6 7 10 11 12 13
Hex 0 1 2 3 4 5 6 7 8 9 10 a
Decimal 12 13 14 15 16 17 18 19 20 21 22 23
Binary 1100 1101 1110 1111 10000 10001 10010 10011 10100 10101 10110 10111
Octal 14 15 16 17 20 21 22 23 24 25 26 27
Hex b c d e 10 11 12 13 14 15 16 17

You can see from studying this table that there is a convenient relationship between binary
numbers and octal or hexadecimal numbers that results in their use for computer notation. The
table shows that the binary values for 810 and 1610 are, respectively, 10002 and 100002. Since 108
= 10002, one octal digit is equivalent to three binary digits. You can see this pattern in the table.
The last three binary digits in 08 to 78 are repeated for as the last three binary digits for 108 to 178
and for 208 to 278. For numbers between 108 and 178, the repeating pattern of three binary digits
is preceded by a 1, the binary representation for the leading octal 1 in this sequence of numbers.
Similarly, in numbers between 208 and 278, the repeating pattern of three binary digits is
preceded by a 10, the binary representation for the leading octal 2 in this sequence of numbers.

For binary numbers the only digits we can have are zero and one. Thus our general-base
equation is written as follows for binary (b = 2) numbers.

How large a number can we represent on a computer if we use 16 binary digits? (A binary digit is
usually called a bit and eight binary digits or bits are called a byte.) The minimum value of our
16-bit binary number is zero and the maximum occurs when all the bits are one. With N = 16, our
sum ranges from k = 0 to k = 15 and we get the following result when all sixteen bits are one.

Jacaranda (Engineering) 3333 Mail Code Phone: 818.677.6448

E-mail: lcaretto@csun.edu 8348 Fax: 818.677.7062
In any case like this one, where all the binary bits are one, we can add one to the number to
obtain the following result.1

We can express this result as follows: the maximum value of a binary number with N digits is
2N – 1. If N = 16, we obtain the result shown previously: the maximum number that we can
represent is 216 – 1 = 65 53510. If we wanted to use these 16 bits to represent both positive and
negative numbers, we could interpret the electronic bits in the computer so that the binary 0 was
the lowest number and 10000000000000002 = 32 76810 would represent zero. In this case our
computer circuit would have to subtract 32 76810 from the binary bits stored in memory to give us
our actual number. Since our 16-bit number ranges from 0 to 65 53510 in the computer memory,
subtracting 32 76810 from this range gives us a range of -32 76810 to 32 76710 for our 16-bit
number when we want to have negative numbers.

Of course choosing more or less than 16 bits would affect the range of our number. In the next
section we discuss the various C++ data types and show how their allowed ranges are governed
by the way in which the numbers are represented as binary information.

C++ data types

C++ has two basic kinds of numerical data types: integer numbers without decimal points and
“floating point” numbers which have a decimal point. (Some accounting systems have “fixed
point” systems that have a fixed range for decimal numbers.) The C++ standard allows the size
of integer data types to be set by the compiler vendors, but there are a set of rules that govern
the relationship among the sizes of various data types. The names of the data types shown
below are common to all C++ implementations. The ranges shown for each type are those used
in Visual C++. However, these ranges are common for many other compilers. (Note that the int
data type, according to the C++ standard, can have a range anywhere between the short and the
long):
short a 16-bit integer data type with range -32 768 to 32 767
1
Note that adding binary numbers is done according to the following addition rules: 0 + 0 = 0; 1 +
0 = 0 + 1 = 1; 1 + 1 = 10. When we add 1 to 11 we get 100. You can check this by using decimal
equivalents: 12 = 110, 112 = 310 and 1002 = 410. We can also do this by adding the binary numbers
directly, using the conventional idea of carrying a digit when the addition of two single-digit
numbers produces a two-difit number Doing this in binary we have the addition shown below:
11 Carried digits when addition of 1 + 1 gives 10
11 One of the original numbers 112 = 310
1 The other original number 12 = 110
100 The sum 1002 = 410
In the first column the sum of 1 + 1 gives 10 so we enter a zero in the sum and carry a 1. In the
second column the 1 or the original first number and the 1 carried gives 10 again so we enter
another zero in the sum and carry a 1. In the final column, there are no entries from the original
numbers so we add 1 + 0 giving 1. We can generalize this to adding 1 to another binary number
that is a string of ones:
111111 Carried digits when addition of 1 + 1 gives 10
111111 One of the original numbers 1111112 = 6310
1 The other original number 12 = 110
1000000 The sum 10000002 = 6410
The general result is the following: when we add 1 to a binary number that consists of N
consecutive ones, the result is a binary number with N+1 digits, having a one in position N+1 and
all its remainig digits are zero.
Jacaranda (Engineering) 3333 Mail Code Phone: 818.677.6448
E-mail: lcaretto@csun.edu 8348 Fax: 818.677.7062
 int a 32-bit integer data type with range -2 147 483 648 to 2 147 483 647
long a 32-bit integer data type with range -2 147 483 648 to 2 147 483 647
unsigned short a 16-bit integer data type with range 0 to 65,535
unsigned int a 32-bit integer data type with range 0 to 4 294 967 295
unsigned long a 32-bit integer data type with range 0 to 4 294 967 295

Note that the range for the short and the unsigned short are just the numbers we that we found
above when analyzing the 16-bit binary number. For the 32-bit number, the maximum range is
from 0 to 232 -1 = 4 294 967 295. This is the range for the unsigned long and unsigned int. When
we want negative integers, we use half this range for negative numbers and half for positive
numbers and zero.

Floating point data types are represented in the binary equivalent of a power notation. That is we
can represent numbers like 6.02x1023 and 1.38x10-16 in terms of their power of ten (called the
characteristic) and their prefactor (called the mantissa). We also require a bit to tell us if the
number is positive or negative. The formats of floating point numbers used in computer are
governed by an international standard that started as IEEE standard 754 in 1985. In this
standard a double precision floating point number is stored in eight bytes or 64 bits. These 64
bits are used as follows: 1 sign bit, 11 bits for the (binary) mantissa, and 52 bits for the
characteristic.

With 11 bits for the mantissa, we can handle a range from 0 to 211 – 1 = 2047. The maximum
value in this range is 22047  10616. This range is usually divided into a range between 10-308 and
10308. The binary characteristic will always start with a 1. This 1 is not stored so the effective
number of bits for the mantissa is really 52 +1 = 53. Fifty-three binary bits can store a number up
to 253-1 = 9.007x1015. Thus the characteristic can store almost sixteen significant decimal digits.

C++ has three floating point types float, double, and long double. Each of these data types
represents a limited range of positive and negative numbers and zero. The range for float is
shown below.

-3.402823466x1038 to -1.175494351x10-38, 0, and 1.175494351x10-38 to -3.402823466x1038

The range for type double is shown next:

-1.7976931348623158x10308 to -2.2250738585072014x10-308, 0, and

2.2250738585072014x10-308 to 1.7976931348623158x10308

In Microsoft Visual C++, there is no difference between a double and a long double. Other
compiler vendors, who do offer a difference, usually have has a range of approximately
3.4x10-4932 to 3.4x10-4932.

The float data type occupies 4 bytes or 32 bits; the double type occupies 8 bytes or 64 bits; the
long double described in the previous paragraph occupies 10 bytes or 80 bits. The number of
bits for the mantissa defines the exponent range and the number of bits for the exponent defines
the number of significant figures. The float type represents about seven significant figures
exactly. Double and long double (in systems where long double uses more than 8 bytes for
storage) represent about 15 and 19 significant figures, respectively.

Because of the way that integer data types are stored on the computer you can do something like
the following.

short x, y = 32767; // sets x to maximum int

Jacaranda (Engineering) 3333 Mail Code Phone: 818.677.6448

E-mail: lcaretto@csun.edu 8348 Fax: 818.677.7062
 x = y + 1; // can we exceed the maximum?

cout << y; // surprise! Y = -32768!

Adding one to the maximum integer gives the minimum integer. For floating point variables
exceeding the maximum or going below the minimum gives you an overflow or underflow error,
respectively. These numbers are represented on the computer as 1.#INF and -1.#INF,
respectively. Another number you may see displayed on your computer screen is -1,#IND; the
IND is an abbreviation for INDefinite; this results from attempted mathematical operations such as
dividing zero by zero, raising zero to the zeroth power, etc.

Jacaranda (Engineering) 3333 Mail Code Phone: 818.677.6448

E-mail: lcaretto@csun.edu 8348 Fax: 818.677.7062