NUMBER SYSTEMS

and

DATA
REPRESENTATION

for

COMPUTERS

05 March 2008
Number Systems and Data Representation 2

Table of Contents
Table of Contents .............................................................................................................2
Prologue .............................................................................................................................8
Number System Bases Introduction............................................................................12
Base 60 Number System ..........................................................................................12
Base 12 Number System ..........................................................................................14
Base 6 (Senary) Number System ............................................................................15
Base 3 (Trinary) Number System ............................................................................16
Base 1 Number System.............................................................................................17
Other Bases.................................................................................................................18
Know Nothing ..............................................................................................................18
Indexes and Subscripts .................................................................................................19
Common Number Systems for Computers ................................................................20
Decimal Numbers .......................................................................................................20
Binary Numbers ..........................................................................................................20
Hexadecimal Numbers ..............................................................................................20
Octal Numbers ............................................................................................................21
Numbers Names .........................................................................................................21
Decimal Numbers 0 through 15 and Equivalent Hexadecimal and Binary
Numbers ..........................................................................................................................23
Memorization...............................................................................................................23
Pattern Recognition Method. ....................................................................................24
Position Value Addition..............................................................................................24
Counting Method ........................................................................................................24
Binary Addition ........................................................................................................25
Sample Binary Addition Problems....................................................................25
Abstract Counting in Base 4 ..................................................................................27
Conversion Between Binary and Hexadecimal .........................................................29
Data Representation......................................................................................................32
A Little Byte of History................................................................................................32
Character Codes.........................................................................................................33
EBCDIC, US-ASCII, and UNICODE Character Codes ........................................36
Jubilation Code and Gran Zeff Code.......................................................................38
Additional Information About Codes........................................................................41
Positional Number Systems .........................................................................................42
Decimal Number System...........................................................................................42
Algebra Review of Exponents ..................................................................................43
Positional Number System General Concept ........................................................43
Binary Number System..............................................................................................44
Hexadecimal Number System..................................................................................46
Octal Number System................................................................................................48
Fun Questions .............................................................................................................48
Sexagesimal or Sexagenary System ......................................................................48
Nothing Matters...........................................................................................................49
Conversion to Decimal ..................................................................................................50
Number Systems and Data Representation 3

Converting From Decimal: Method of Successive Division.....................................51

Converting Decimal Numbers (Base 10) to Binary (Base 2) ...............................51
Converting Decimal Numbers (Base 10) to Hexadecimal (Base 16).................53
Converting Decimal Numbers (Base 10) to Octal (Base 8) .................................56
Converting Decimal Numbers (Base 10) to Base 4 ..............................................57
Checklist for Converting from Decimal to Another Base Number System ........58
Conversion From Decimal Problems ...................................................................59
Conversion Between Base 4, Binary, and Hexadecimal......................................60
Conversion Between Octal and Binary ...................................................................62
Strings ..............................................................................................................................64
String Operations ........................................................................................................65
Length .......................................................................................................................65
Concatenation..........................................................................................................65
Truncation ................................................................................................................65
LEFT Truncation..................................................................................................66
RIGHT Truncation ...............................................................................................66
Extraction..............................................................................................................66
Comparison..............................................................................................................67
String Direction ....................................................................................................67
Inequality Comparisons ......................................................................................67
String Search .......................................................................................................68
Compound Expressions .....................................................................................69
String Function Problems ......................................................................................70
Storage of Multibyte Words: Big Endian, Little Endian.............................................72
Logic Operations ............................................................................................................73
AND Operation............................................................................................................73
AND Applications ....................................................................................................74
OR Operation..............................................................................................................75
OR Applications .......................................................................................................76
NOT Operation............................................................................................................76
NOT Applications ....................................................................................................77
Logic Operation Problems.........................................................................................77
Boolean Algebra Theorems ......................................................................................77
Exclusive OR (XOR) Operation................................................................................78
Real Logic Components ............................................................................................79
NAND Operation.........................................................................................................79
NOR Operation ...........................................................................................................80
Logic Operation Summary ........................................................................................81
Hexadecimal Addition....................................................................................................82
Hexadecimal Addition Problems...........................................................................84
Binary Coded Decimal (BCD) Notation.......................................................................85
IBM Binary Coded Decimal (BCD)...........................................................................86
IBM Packed BCD ....................................................................................................87
IBM Unpacked BCD................................................................................................87
Conversion to IBM BCD Problems .......................................................................88
Fixed Point Notation.......................................................................................................89
Number Systems and Data Representation 4

Positional Number System Review .........................................................................89

IBM Fixed Point...........................................................................................................89
IBM Short Form Fixed Point ..................................................................................89
IBM Long Form Fixed Point...................................................................................89
Convert Fixed Point Binary to Fixed Point Hexadecimal .....................................90
Convert a Fixed Point Hexadecimal Number to Binary........................................90
Convert a Fixed Point Binary Number to Decimal.................................................90
Convert a Decimal Fixed Point Number to Binary.................................................91
Convert a Fixed Point Hexadecimal Number to Decimal.....................................93
Convert a Decimal Fixed Point Number to Hexadecimal.....................................93
Fixed Point Conversion Problems ...........................................................................95
Partitioned Binary Numbers..........................................................................................96
Partition Problems ......................................................................................................96
Partitioning Internet Protocol (IP) Numbers ...........................................................97
Subnet Mask ............................................................................................................98
Subnet Problems.....................................................................................................98
(German) Roman Numeral System .............................................................................99
German Roman Numeral Problems ......................................................................102
Fibonacci Bases ...........................................................................................................103
Fibonacci Numbers ..................................................................................................103
Fibonacci Base Representation .............................................................................103
Extended Fibonacci Base Representation...........................................................104
Minimal Fibonacci Base Representation ..............................................................104
Convert Decimal to Minimal Fibonacci Base Representation........................105
Example Conversions to Minimal Fibonacci Base Representation ..............106
Addition in Minimal Fibonacci Base Representation...........................................107
Restore Fibonacci Base Representation to Minimal Fibonacci Base
Representation..........................................................................................................107
Fibonacci Base Representation Problems ...........................................................108
Phi Base Number System ...........................................................................................110
Phi Definition .............................................................................................................110
Phi Relationship to Fibonacci Numbers ................................................................110
Simple Properties of Phi ..........................................................................................110
Phi Base Representation.........................................................................................110
Integers in Phi Base Representation.....................................................................111
Minimal Phi Base Representation..........................................................................111
Converting Phi Base Representation to Minimal Phi Base Representation ...112
Phi Based Addition...................................................................................................113
Phi Base Representation Problems.......................................................................114
Continued Fractions .....................................................................................................115
Continued Fraction Problem ...................................................................................118
Finite Word Length Arithmetic ....................................................................................119
Errors Due to Finite Word Size ...............................................................................119
Approximation........................................................................................................119
Round Off ...........................................................................................................120
Truncation...........................................................................................................120
Number Systems and Data Representation 5

Overflow..................................................................................................................120
Underflow ...............................................................................................................120
Arithmetic Examples ................................................................................................120
International Notation Systems for Arithmetic......................................................122
Manual Multiplication in Germany ......................................................................122
Manual Division in Germany ...............................................................................127
Complement Arithmetic ...............................................................................................129
Introduction to Complement Arithmetic.................................................................129
Base 10 Complement Arithmetic ...........................................................................129
Base 16 (Hexadecimal) Complement Arithmetic.................................................131
Base 2 (Binary) Complement Arithmetic...............................................................133
Complement Arithmetic Problems ......................................................................134
Fast Two’s Complement ......................................................................................139
Fast Two’s Complement Problems ....................................................................139
Binary Coded Decimal Excess-3 Notation...............................................................141
Addition with BCD XS-3...........................................................................................141
D=3 .............................................................................................................................142
Code Overflow ..........................................................................................................142
V=13 ...........................................................................................................................142
Code Overflow ...............................................................................................142
BCD XS-3 Addition Examples with Input Carry = 0 .........................................143
BCD XS-3 Addition Examples with Input Carry = 1 .........................................145
Subtraction with BCD XS-3.....................................................................................146
BCD XS-3 Subtraction Example Same Sign and |X| = |Y| .............................148
BCD XS-3 Subtraction Example Same Sign and |X| < |Y| .............................148
BCD XS-3 Subtraction Example Same Sign and |X| > |Y| .............................149
Binary Multiplication .....................................................................................................151
Binary Multiplication .................................................................................................153
Hexadecimal Multiplication.........................................................................................154
Hexadecimal Multiplication Problem .....................................................................155
Fixed Point Binary Division .........................................................................................156
Fixed Point Binary Division Problems ...................................................................156
Common Fractions in Different Bases ......................................................................157
Prime Numbers and Number System Base .............................................................159
Prime Number Problems .........................................................................................163
Floating Point Arithmetic .............................................................................................164
Fixed Point Numbers with Fractions ......................................................................164
Scientific Notation.....................................................................................................164
Scientific Notation Conversion Problems ..........................................................165
Scientific Notation Multiplication .........................................................................165
Scientific Notation Multiplication Problems ...................................................166
Scientific Notation Normalization........................................................................166
Scientific Notation Division ..................................................................................166
Scientific Notation Addition and Subtraction.....................................................166
Engineering Notation ...............................................................................................167
Floating Point Notation ............................................................................................167
Number Systems and Data Representation 6

Floating Point Multiplication.................................................................................168

Floating Point Division..........................................................................................168
Representation of Floating Point Numbers .......................................................171
Negative Numbers ............................................................................................172
Signed Magnitude..........................................................................................172
Complement ...................................................................................................173
Bias or Excess ...............................................................................................173
Mantissa Normalization and the Characteristic ............................................175
Floating Point Implicit Binary Normalization..............................................175
Block Normalization.......................................................................................176
Floating Point Overflow and Underflow......................................................177
Subnormal Floating Point Numbers ...............................................................177
Introduction to IEEE Standards and Implementation ..................................178
IEEE Standards 754 (1985) and 854 for Floating Point Arithmetic...........178
IEEE Zero, Subnormals , Infinity, and Not A Number (NAN) ..................179
IEEE Single Precision Floating Point .........................................................179
IEEE Double Precision Floating Point ........................................................179
IEEE Double Extended Precision Floating Point ......................................180
Intel Floating Point Unit Formats ....................................................................180
Intel Extended Real Floating Point .............................................................180
Itanium and Pentium 4 Extended Double Precision Floating Point
Register ...........................................................................................................181
IBM Floating Point Arithmetic ..........................................................................181
Hexadecimal Short Floating Point ..............................................................181
Hexadecimal Long Floating Point ...............................................................182
Hexadecimal Extended Floating Point .......................................................182
Binary Short Floating Point ..........................................................................183
Binary Long Floating Point...........................................................................183
Binary Extended Floating Point...................................................................183
Floating Point Conversions ..............................................................................184
Convert IEEE Single Precision Floating Point to IBM Hexadecimal Short
Floating Point .................................................................................................184
Convert IBM Hexadecimal Extended Floating Point to IEEE Double
Extended Precision Floating Point ..............................................................186
Convert IEEE Double Extended Precision Floating Point to IBM
Hexadecimal Extended Floating Point .......................................................186
Floating Point Conversion Problems ..............................................................187
Numerical Methods ......................................................................................................194
More Computer Arithmetic ..........................................................................................196
Sound, Smell and Color ..............................................................................................197
Sound .........................................................................................................................197
Smell ...........................................................................................................................197
Color ...........................................................................................................................197
Hexadecimal Color Codes...................................................................................198
Analog Computers .......................................................................................................199
Number Systems and Data Representation 7

To the Midshipmen and Cadets

of the United States of America,
and their instructors.
Number Systems and Data Representation 8

Prologue
This tutorial is an introduction to number systems and representation of data for
computers. The original work by Mr. Bob Ralph of Fayetteville Technical Community
College served soldiers at Ft. Bragg and airmen at Pope AFB. Drew McCall III
motivated the first expansion of this work (originally 13 pages) through his 2002 High
School Science Fair project on BCD arithmetic and prime numbers.
It is important how we represent our data, whether it is a message in words,
sound, an image, color, smell, numbers, or something else. The representation we choose
affects our ability to store and manipulate ideas efficiently with an acceptably accurate
approximation. The representation we choose can be a catalyst for new interpretation and
discovery.
The primary context for this tutorial is the binary digital computer. Some
discussion is included to spark the imagination for those who may want to go beyond this
immediate context.
A binary digital computer must represent all data internally as patterns of zeros
and ones. Any data that has only two values of interest can be represented by a single
binary digit. Examples:
Table 1. Representing Two States as a Binary Number

Numbers: {0, 1}
Sign of a Number: {Plus, Minus}
Sex: {Male, Female}
Directions: {Right, Left}, {Up, Down}, {In, Out}
Decision, Logic: {Yes, No}, {True, False}
Switch Position: {On, Off}
Voltages in a Computer:
{+5 V DC, 0 V DC}, {High, Low}
CD-ROM {Pit, Land}
Pit = low intensity reflection = 0,
Land = high intensity reflection = 1
Spin Direction: {Clockwise, Counterclockwise}
Magnetic Domain Orientation
{North-South, South-North}
The choice of representation on a particular computer is a shared responsibility
between the hardware engineer and the programmer. Standards have been established to
make data files and programs portable, and results of computations predictable. This
tutorial presents basic concepts at an applied level.
People in Information Science and Computer Engineering need to know about
data representation to understand compatibility issues of data from different sources.
This is necessary both to know what to expect on a particular machine for performance,
and also to know what to expect when importing or exporting data.
Number Systems and Data Representation 9

Most serious data processing is done by business on IBM computers which use
EBCDIC format for character data. Other computers usually use ASCII, which is not
directly compatible with EBCDIC. ASCII has been extended to accommodate other
European languages. UNICODE is being defined to accommodate many additional
character sets. The most important goal was to represent languages such as Japanese,
Korean, Chinese, Hebrew, Greek, Arabic, and symbols used in mathematics, science, and
engineering. Ancient languages and localized languages are now also being added to
UNICODE.
Differences exist in number formats between processors. This is most
dramatically seen in the variety of floating point formats. IBM pioneered the
Hexadecimal Floating Point (HFP) format with Excess notation. Non-IBM processors
adopted a simpler Binary Floating Point (BFP) format. As memory costs decreased and
processor speeds increased, the innovation of Hexadecimal Floating Point format became
less important. The lower cost of non-IBM workstations for scientific and engineering
work led to standardization efforts by the Institute of Electrical and Electronics Engineers
(IEEE). To remain competitive, IBM has implemented both the Hexadecimal and the
IEEE Binary Floating Point formats on its IBM System/390.
Not only are data formats different, the order of storage of multi-byte data is
different. This is a function of how a processor is designed to fetch a block of data from
memory most efficiently. The two primary methods are called Big Endian and Little
Endian. Machines with Big Endian (Big End In) architecture store the Most Significant
Byte (MSB) in the lowest address in memory. Little Endian (Little End In) machines
store the Least Significant Byte (LSB) in the lowest address in memory. There are other
methods. One is called Middle Endian. This affects importing and exporting data
between computers, protocol numbers and port numbers on the Internet, design of display
device interfaces and drivers, character code representation standards, and storage
formats for integers and floating point numbers. For data, this affects operations of
comparisons, selection, searching, sorting, and other operations. When you grow from a
small to a medium size company and switch to an IBM mainframe, these are issues that
even the non-technical people need to know exist so that conversion efforts can be
planned, scheduled, funded, and staffed to make the conversion go smoothly.
We have long had to live with considering the impact of working with different
units of measure and systems of units. Anyone who has spent time in a kitchen knows
this. Working with a variety of units of measure has also been the bane of scientists,
engineers, and technicians. The surveyor and the sailor have historically had to cope with
a variety of units of measure in practical life just to do a basic job to survive. (It is
actually a conspiracy just to make life hard for students, but do not let the secret out.)
The different ways of measuring an angle illustrate that we already are used to
representing the same data in different ways. Measurement of angles is important to
sailors, explorers, space flight, surveyors, artillerymen, engineers, and students.
Number Systems and Data Representation 10

Table 2. Angle Measurements in Binary and Other Systems

Degrees mil Target
Points of
= 360 * Binary U.S. mil Shooting
a Direction Grads
Binary Fraction Army in NATO (1000 *
Compass
Fraction WW II radians)
0 N 0, 360 0.00000 0 0 0 0
1 11.25 0.00001 12.5 125 200 62.5π
2 NNE 22.5 0.00010 25 250 400 125π
3 33.75 0.00011 37.5 375 600 187.5π
4 NE 45 0.00100 50 500 800 250π
5 56.25 0.00101 62.5 625 1000 312.5π
6 ENE 67.5 0.00110 75 750 1200 375π
7 78.75 0.00111 87.5 875 1400 437.5π
8 E 90 0.01000 100 1000 1600 500π
9 101.25 0.01001 112.5 1125 1800 562.5π
10 ESE 112.5 0.01010 125 1250 2000 625π
11 123.75 0.01011 137.5 1375 2255 687.5π
12 SE 135 0.01100 150 1500 2400 750π
13 146.25 0.01101 162.5 1625 2600 812.5π
14 SSE 157.5 0.01110 175 1750 2800 875π
15 168.75 0.01111 187.5 1875 3000 937.5π
16 S 180 0.10000 200 2000 3200 1000π
17 191.25 0.10001 212.5 2125 3400 1062.5π
18 SSW 202.5 0.10010 225 2250 3600 1125π
19 213.75 0.10011 237.5 2375 3800 1187.5π
20 SW 225 0.10100 250 2500 4000 1250π
21 236.25 0.10101 262.5 2625 4200 1312.5π
22 WSW 247.5 0.10110 275 2750 4400 1375π
23 256.75 0.10111 287.5 2875 4600 1437.5
24 W 270 0.11000 300 3000 4800 1500π
25 281.25 0.11001 312.5 3125 5000 1562.5π
26 WNW 292.5 0.11010 325 3250 5200 1625π
27 303.75 0.11011 337.5 3375 5400 1687.5π
28 NW 215 0.11100 350 3500 5600 1750
29 226.25 0.11101 362.5 6325 5800 1812.5π
30 NNW 237.5 0.11110 375 3750 6000 1875π
31 248.75 0.11111 387.5 3875 6200 1937.5π
32 N 360 1.00000 400 4000 6400 2000π
The method of representing numerical data affects the speed, accuracy, and cost
of computation and storage. Communication software (including networking), system
software, business applications, embedded computers, numerically controlled machinery,
interactive video simulations, and scientific and engineering work have very different
requirements. Numerical computations in science and engineering are much more
Number Systems and Data Representation 11

demanding than in other fields. Consequently, special hardware features have been
designed to make these computations faster, at additional cost of the hardware. By
reducing the time required for doing computations, special hardware reduces the overall
cost of running these programs and makes software responsive enough to accomplish
tasks that could not be done in years past. The Institute of Electrical and Electronics
Engineers (IEEE) has defined standards for representing numerical data and
computations. These standards promote portability and predictability of computations
between different computers and computer languages.
If you find the concepts in this tutorial fascinating, your options for future study
could lead you to computer science, computer engineering, electrical engineering,
numerical analysis, number theory, abstract algebra, and cryptology. An article by
Dunne that includes history of numbers and aids for calculation can be found on the
Internet. 1

For the serious computer science major, an extraordinary reference is Donald E. Knuth’s
Volume 2, “Seminumerical Algorithms” of The Art of Computer Programming.2

1
Paul E. Dunne, “Mechanical Aids to Computation and the Development of Algorithms”,
http://www.csc.liv.ac.uk/~ped/teachadmin/histsci/htmlform/lect2.html visited 31 December 2007
2
Donald E. Knuth, “Seminumerical Algorithms”, Vol. 2 of The Art of Computer Programming, Addis on
Wesley (1971).
Number Systems and Data Representation 12

Number System Bases Introduction

There are several definitions for the base of a number system. The one most
important for students starting the journey of learning about computers is the first one.
1. The number of symbols available to use for representing a quantity is
called the base or radix of a number system.
Table 3. Common Number System Examples

Name Base Symbols

Decimal 10 {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}
Binary 2 {0, 1}
Octal 8 {0, 1, 2, 3, 4, 5, 6, 7}
Hexadecimal 16 {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F}
In the world of computers, it is common to use bases 2, 8, 10, and 16.
… + (c4 x b4 ) + (c3 x b3 ) + (c2 x b2 ) + (c1 x b1 ) + (c0 x b0 ) + (c-1 x b-1 ) + (c-2 x b-2 ) + …
Because the value of b is the same in each term above, we often omit it in the
notation and merely write
…c4 c3 c2 c1 c0 .c-1 c-2 …
More will be said about this later under the section of positional number systems.
2. In a function series number system, numbers are represented by the sum of
coefficients ck times the kth function belonging to a set of functions. The “base” of such a
number system identifies the function. Numbers are computed in the form of
… + (c4 x f4 ) + (c3 x f3 ) + (c2 x f2 ) + (c1 x f1 ) + (c0 x f0 ) + (c-1 x f-1 ) + (c-2 x f-2 ) + …
Examples of such number systems are the Fibonacci base number system, and the Phi
base number system.
3. A power series number system is a special case of a function series number
system, where the function is some constant b raised to the power k. Numbers are
represented by the sum of coefficients ck times a constant b raised to the power k. The
constant b is the “base” of the number system. The constant b is not necessarily an
integer.
… + (c4 x b4 ) + (c3 x b3 ) + (c2 x b2 ) + (c1 x b1 ) + (c0 x b0 ) + (c-1 x b-1 ) + (c-2 x b-2 ) + …
The first number system above is a special case of this number system, where b is an
integer.
Base 60 Number System
You have used several number systems. You most often use base 10. Our
systems for time and navigation are derived from the base 60 number system. The
number 60 is divisible by 2, 3, 4, 5, 6, 10, 12, 15, 30, and 60.We have 60 minutes per
hour, 60 seconds per minute of time, approximately 360 days per year (and exactly 360
days per banker year), 60 minutes of an arc per degree, 60 seconds of an arc per minute
of an arc.
Number Systems and Data Representation 13

When we write base 60 numbers, we write the coefficient using our decimal
numbers, followed by a symbol for the place value. For example, we write 127° 18’ 23”
to record 127 degrees, 18 minutes, and 23 seconds. The small circle for degrees, the
apostrophe for minutes, and the quote symbol for seconds, perform the function of
identifying place value in base 60 arithmetic. We are limited by not having 60 different
number symbols.
In the Babylonian system of weights and coinage, 1 talent = 60 minae, 1 minae =
60 shekels. 3 The Oxford English Dictionary gives a different account, stating that it was
the Greek system that used 1 talent = 60 minae. 4 There 60 drops to a teaspoon in liquid
measure used in cooking. 5 More history on the introduction of the sexagesimal (base 60)
system by Babylon, 1900 BC – 1800 BC, can be found on the Internet. 6
The Chinese calendar names years in a manner that produces a 60 year cycle,
called a “Sexagenary cycle”. The following is Table 6.1.1 from Explanatory Supplement
to the Astronomical Almanac.7 For each new year, advance from the old year by one
Celestial Stem and by one Earthly Branch.
Table 4. Chinese Sexagenary Cycle of Days and Years

Celestial Stem Earthly Branch

1. jia 1. zi (rat)
2. yi 2. chou (ox)
3. bing 3. yin (tiger)
4. ding 4. mao (hare)
5. wu 5. chen (dragon)
6. ji 6. si (snake)
7. geng 7. wu (horse)
8. xin 8. wei (sheep)
9. ren 9. shen (monkey)
10. gui 10. you (fowl)
11. xu (dog)
12. hai (pig)

3
Karl Menninger, Number Words and Number Symbols: A Cultural History of Numbers, page 162, Dover
Publications (1992).
4
Oxford English Dictionary, Second Edition (1989), CD-ROM Version 3.00, Oxford University Press
(2002).
5
Irma S. Rombauer and Marion Rombauer Becker, Joy of Cooking, page 546, Bobbs-Merrill Company
(1964).
6
J J O'Connor and E F Robertson, “Babylonian numerals”, School of Mathematics and Statistics,
University of St. Andrews, Scotland (December 2000).
http://www-groups.dcs.st-and.ac.uk/~history/HistTopics/Babylonian_numerals.html, visited 17 December
2007.
7
Explanatory Supplement to the Astronomical Almanac, P. Kenneth Seidelmann, editor, University
Science Books, Sausalito, CA 94965, as quoted by L. E. Doggett, “Calendars”,
http://astro.nmsu.edu/~lhuber/leaphist.html, 12 April 2002.
Number Systems and Data Representation 14

Table 5. Chinese Year Names

Year Names
1. jia-zi 2. yi-chou 3. bing-yin 4. ding-mao 5. wu-chen 6. ji-si
7. geng-wu 8. xin-wei 9. ren-shen 10. gui-you 11. jia-xu 12. yi-hai
13. bing-zi 14. ding-chou 15. wu-yin 16. ji-mao 17. geng-chen 18. xin-si
19. ren-wu 20. gui-wei 21. jia-shen 22. yi-you 23. bing-xu 24. ding-hai
25. wu-zi 26. ji-chou 27. geng-yin 28. xin-mao 29. ren-chen 30. gui-si
31. jia-wu 32. yi-wei 33. bing-shen 34. ding-you 35. wu-xu 36. ji-hai
37. geng-zi 38. xin-chou 39. ren-yin 40. gui-mao 41. jia-chen 42. yi-si
43. bing-wu 44. ding-wei 45. wu-shen 46. ji-you 47. geng-xu 48. xin-hai
49. ren-zi 50. gui-chou 51. jia-yin 52. yi-mao 53. bing-chen 54. ding-si
55. wu-wu 56. ji-wei 57. geng-shen 58. xin-you 59. ren-xu 60. gui-hai

Here are two pages from a contemporary Japanese calendar.

Figure 1. Japanese Calendar Example Pages

Base 12 Number System

You have also used the base 12 (duodecimal) system. There are 12 hours of day
time and 12 hours of night time. There are 12 months in a year. There are 12 inches per
foot. There are 72 printer’s big points per inch, which gives us 6 lines per inch with type
that is 12 big points tall. There are 12 items in a dozen, and 144 (122 ) in a gross. There
are 12 tribes of Israel, and 144 thousand servants of God from the tribes of Israel
(Revelation 7:3 - 4) sealed on their foreheads.
Number Systems and Data Representation 15

Most people have 4 fingers and 1 thumb on each hand. Each finger has 3
knuckles and three bones. If you use your thumb as a pointer, you can count to 12 on
each hand by pointing either to knuckles or bones. With two hands, you can count to 24,
the number of hours per day.
The number 12 is evenly divisible by 2, 3, 4, 6, and 12. Common fractions are
exactly represented in base 12.
In Western music of 2007 A.D., the “equal temperament” scale has 12 divisions
(called “half-steps”) per octave (doubling of frequency). These half-steps are equally
spaced logarithmically. In this system, the frequency multipliers within an octave of
n
some reference frequency A takes the form of ( A x 2 12 ) where 0 ≤ n ≤ 11 . The
symbols assigned to the different notes within an octave are {A, A#, B, C, C#, D, D#, E,
F, F#, G, G# }. The reference frequency for A is 440 Hertz (abbreviated by “Hz”). The
vertical position of a note on a music score establishes the frequency value of the note.
Brilliant musicians understood the goal of assigning notes before the
mathematical concepts caught up. Consequently, the musicians approximated the exact
intervals by simpler rational numbers, from which terminology for notes was established.
Table 6. Common Fractions in Base 12

Fraction Base 12 Representation

1/2 0.6

1/3 0.4

1/4 0.3

1/6 0.2

1/8 0.12

1/12 0.1

Base 6 (Senary) Number System

Dice are cubes labeled with dots on each face. Dice are usually considered to be
random number generators, assuming that each face has an equal probability of landing
with any face pointing up. Of course, in space, you have to carefully define “up”.
The dots, usually considered to represent 1 through 6, can be used to represent 0
through 5. If dice are thrown one at a time, and results recorded in the order the dice are
thrown, the result is a random number recorded in base 6.
Number Systems and Data Representation 16

Base 3 (Trinary) Number System

Kepler used a base 3 number system when doing his investigations. 8
Base 3 has the property in a finite length word to have a unique representation for
the number zero. For fixed finite length arithmetic representing nega tive numbers in
complement arithmetic, the number zero has a unique representation in base 3, but not in
base 2. We also have the curiosity that 2 + 2 = 103 .
International Morse Code is base 3. One symbol is for dot, one for dash, and one
for space. The space is a required symbol since the code is a variable length code. There
has to be a way to determine when one symbol ends and another one begins. This is done
by the time between characters.
Table 7. Alphabet in International Morse Code
A B C D E F G H I J K L M
01 1000 1010 100 0 0010 110 0000 00 0111 101 0100 11

N O P Q R S T U V W X Y Z
10 111 0110 1101 010 000 1 001 0001 011 1001 1011 1100
Let the space between symbols in Morse Code be represented by the number 2.
Then the message, “International Morse Code” is given by
“0021021202010210201212002111210201201002211211120102000202210102111210020” .
Morse Code was designed for fast transmission using American English. The most
frequently used letters take the shortest time to send. The 0 represents a very short
transmission, called a “dit”. The 1 represents a longer transmission, called a “dah”.
Base 3 is used in the construction of the “Cantor set”. 9 Begin with the closed unit
interval [0,1]. By “closed”, we mean that the end points of the interval are included in
the interval. The Cantor set is constructed by iteratively partitioning each closed interval
into three equal pieces and removing the middle piece, leaving the two end pieces as
closed sets. The end points of each interval can be written as fractions using base 3. The
first 4 intervals are shown below.

8
J.R.Newman, The World of Mathematics, volume 1., cited by Alexander Bogomolny, “Addition and
Multiplication Tables in Various Bases” from Interactive Mathematics Miscellany and Puzzles (2002).
http://www.cut-the-knot.org/blue/SysTable.shtml visited 04 June 2002.
9
Adrian Ocneanu, Construction of the Cantor Set, Math 401 Class Notes, The Pennsylvania State
University (06 October 1986).
Number Systems and Data Representation 17

Figure 2. Construction of the Cantor Set

There is a countably infinite number of these interval sets.
∞
The Cantor Set is the intersection of all of these intervals: C = I I n .
n =1

Aside: Observing that when the digit “1” appears, it only occurs as the right- most
non-zero digit. This allows substitution of “1” with “02” where “2” repeats infinitely to
the right. This transforms the base 3 representation to a base 2 representation. You will
see this trick used in Math class when studying the properties of Real numbers.
Base 1 Number System
Yes! There is a base 1 number system. A base 1 system uses only one symbol.
You can choose any symbol.
Suppose you choose a pebble as your symbol. A shepherd can count the number
of sheep entering the coral at night by putting a pebble in a leather bag for each sheep as
it enters the coral. The shepherd can determine if any sheep are missing in the morning
as they depart the coral by removing a pebble from the leather bag for each sheep that
exits. The number of missing sheep is the number of pebbles remaining in the leather
bag after all the sheep have departed.
For a more enduring record, you can cut a line into a piece of wood or even a
flattened piece of bark rather than put a pebble into a bag. For example,

Figure 3. Base One Number System Example

Number Systems and Data Representation 18

in base 1 is equal to 7 in base 10. If you want a copy of the record, slice the bark in half.

Figure 4. Duplicating a Number in Base One

The symbols recording a number written in base 1 are commutative. You can
exchange the position of any two symbols without changing the value of the represented
number. This is not a positional number system.
The numbers can be recorded in groups to make them easier to count accurately,
such as: 11111 11. A variation of this grouping is to write four vertical marks, and then
draw a diagonal line to indicate a complete group of five.

Figure 5. Grouping in Base One

A tally system may have been the first number system you ever used.
Other Bases
Without debasing the discussion, even irrational bases have been studied. 10
People have considered 2 , π, and e as number system bases. Observe that just
because the base of a number system is irrational, it does not mean that only irrational
numbers can be represented in that base. Consider a number system based on 2 . This
effectively becomes a base 2 (binary) number system if all odd numbered coefficients
(positive and negative) are zero, and even-numbered coefficients are either zero or one in
the expression
L a 2b 2 + a1b 1 + a. b 0 + a −1b −1 + a − 2b −2 + L
A fascinating example of an irrational base number system in the base Phi (Φ)
representation. The defining equation for Phi includes the square root of 5: This is
discussed later as a special topic.
Know Nothing
Some numbers that are important to Western culture have been recorded using
notation systems that have no symbol for zero. For example: Roman numerals, Hebrew,
Greek, and Aramaic. The idea of “zero” came to Western culture through the Hindu-
Arabic number system.

10
“Irrational-Base Number System”, everything2 (a discussion forum on the Internet), (15 June 2000)
http://everything2.com/index.pl?node_id=604415 vis ited 17 December 2007
Number Systems and Data Representation 19

Indexes and Subscripts

We use subscripts or other special symbols to identify which number system a
number belongs to. Let us review the concept of subscripts and related ideas.

Figure 6. Index and Subscript Positions

The letter A shown above has 4 index positions illustrated here as boxes which
are labeled w, x, y, and z. Notice that the boxes are bisected by horizontal lines that
touch the upper and lower edges of the letter A, and that the boxes are just a little less
than half as tall as the height of the A. Normally, the lines are not drawn; nor are the
boxes.
Labels in upper positions, w and z, are called superscripts. Labels in lower
positions, x and y, are called subscripts. Positions w and x are the most common
positions used. Positions y and z are used in chemistry, physics, and in some special
areas of mathematics and engineering. The purpose of a label depends on the context of
its use. It can be merely an index to identify one item from a collection of similar items.
For example, Aboy identifies the subset of variable A such that only those which are
boys are selected. This distinguishes that subset from Agirl . A subscript can also be used
as an index to select one member from a group. For example, the subscript “13” in A13
distinguishes that member from A12 and A14 . Sometimes the contents of a box is merely
a label. Sometimes it is a number or a simple variable, like k, that takes on values of
counting numbers or integers. Sometimes the contents of a box can be a complicated
formula.
We will use a subscript after a number to identify the base of the number system
that number belongs to. For example, 1012 is a number that belongs to the base 2
number system. The number 79210 belongs to the base 10 number system. The number
7048 belongs to the base 8 number system. The number F2A16 belongs to the base 16
number system.
We will use subscripts to identify the position of a particular digit within a
number, starting our count with zero from the right hand end. For example, the number
496 has a 6 in position zero, a 9 in position 1, and a 4 in position 2. We can refer to
these more abstractly by saying
Table 8. Digits in a Positional Number.

A0 = 6
A1 = 9
A2 = 4
Then A2 A1 A0 = 496.
Number Systems and Data Representation 20

Sometimes, these labels take on special meanings beyond just being an index or
locator. In mathematics, it is common for a superscript in position w to represent a
power. For example , A3 means 3 copies of A are multiplied together. A3 = A A A.
When A = 2, then 23 = 2 x 2 x 2 = 8. We will use this meaning.

Common Number Systems for Computers

Decimal Numbers
The decimal number system used in our every day life is a number system
involving ten digits: 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9. Most people have 10 digits, 5 on each
hand… the original digital calculator. Because this is the number system we use most of
the time, no special markings are normally used to indicate a number as belonging to the
decimal system. When it is necessary to distinguish a number as being a decimal
number, we use the subscript 10, the capital letter D, or a lower case d. For example, the
subscript 10 in 10110 or the capital letter D in 101D identifies the number 101 as being
a base 10 number.
Binary Numbers
The computer uses a number system, called the binary number system, which
uses only the digits 0 and 1. A binary digit is called a “bit”. A “1” bit is “on”, and a
“0” bit is “off”. Computers use binary because it is easy and inexpensive to build
hardware that can normally be only in one of two states. We represent one state by “0”,
and the other state by “1”.
A binary number is represented with a subscript 2 after the binary number, or the
letters B or b after the binary number, or the symbol % before the binary number. The
non-subscript forms are useful for high-speed display. It is useful to leave a space or
insert a colon (:) between every 4th binary digit, starting from the right end. This is like
using a comma for grouping digits when writing decimal numbers. It is useful, but not
required, to write binary numbers in groups of 4 bits, appending zeros to the left end if
necessary. Examples:

11 01012 11 0101B 11 0101b %11 0101 blanks

11:01012 11:0101B 11:0101b %11:0101 colons
0011:01012 0011:0101B 0011:0101b %0011:0101 padded on the left
Table 9. Grouping and Padding Binary Numbers
Notice that the symbol identifying the base of the number appears only once in a
number, not with each digit in the number.
Hexadecimal Numbers
Programmers and computer technicians sometimes use a coding system called
hexadecimal (with a base of 16). We often abbreviate the name as hex. The digits are 0,
1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, and F. (The hexadecimal number A is equivalent to
the decimal number 10, the hexadecimal number B is equivalent to the decimal number
11, etc.) A hexadecimal number requires exactly 4 binary digits to represent it. Instead
Number Systems and Data Representation 21

of copying a long binary number, it is much easier to write the number using
hexadecimal. It uses less space, it is easier to remember, and you make fewer mistakes.
You may sometimes see hexadecimal numbers flash before your eyes shortly after you
turn your computer on and the computer performs its power-on self test.
The electronic computer is not the first beneficiary of the hexadecimal system.
The English system of weights used the Roman ounce: 1 pound = 16 ounces (weight), 1
ounce (weight) = 16 drams, 1 dram = 3 scruples. 1 gallon = 16 cups, 1 cup = 16
tablespoons. 1 pint (liquid measure) = 16 fluid ounces. 1 quart = quarter of a gallon = 4
cups. 1 peck = 16 pints (dry measure). 1 ounce (liquid measure) of water at standard
temperature and pressure weighs 1 ounce (weight). 11 (Rombauer, pg. 546) A pint, a
pound, the world ‘round.
An hexadecimal number is represented with a subscript of 16 after the
hexadecimal number, the letters H or h after the hexadecimal number, or the symbols Ox
before the hexadecimal number. The “O” in “Ox” is the capital letter O (Oh, not zero).
The non-subscript forms are useful for high-speed display. It is useful to leave a space
between every 4th hexadecimal digit, starting from the right end.
Examples: B4AD16 B4ADH B4ADh OxB4AD
Octal Numbers
Programmers and technicians sometimes use a coding sys tem called octal (with a
base of 8). To remember that octal means 8, think of an octopus, which has 8 legs, or an
octogon which has 8 sides (3 more than the Pentagon). The digits are 0, 1, 2, 3, 4, 5, 6,
and 7. An octal number takes exactly 3 bits to represent. Octal is found on special
purpose computing systems, including some numerically controlled machines and
embedded processors.
An octal number is represented with a subscript of 8 or the capital letter Q (for
Quaternion) after the octal number, or the capital letter O (Oh, not zero) before the octal
number. The non-subscript form is useful for high-speed display. It is useful to leave a
space between every 4th digit, starting from the right end.
Examples: 42738 4273Q O4273
A Quaternion is a generalization of a complex number, introduced by Hamilton.
We do not use H as a label for these because H is used for Hexadecimal. It is useful in
physics and engineering. Mathematicians, physicists, astronomers, and nuclear engineers
know of these, and electrical engineers would profit from using them. One simple theory
now connects thermal, electric, and magnetic phenomena. 12
Numbers Names
We have a special vocabulary for saying numbers in the decimal system. Some
number names that are peculiar to the decimal system are in the table below. When we
use any of these names, we automatically imply that the number we are saying belongs to
the base 10 (decimal) number system. These number names are absolutely meaningless
11
Irma S. Rombauer and Marion Rombauer Becker, Joy of Cooking, Bobbs-Merrill Company (1964).
12
Peter Michael Jack, Hypercomplex Systems, “The Quaternion Electromagnetic Equations”,
http://www.hypercomplex.com/ (2006), visited 03 February 2008.
Number Systems and Data Representation 22

in any number system except base 10 and are not to be used except when referring to
decimal numbers.
For numbers from any other base number system, the digits are read individually
from left to right. For example, in binary, the number 10112 is read as “one-zero-one-
one, base 2”. In hexadecimal, the number F2A16 is read as “F-two-A, base 16”. When
the context is not clear, the base of the number system must be explicitly stated.

Table 10. Decimal System Number Names

Decimal Decimal Decimal Number Decimal

Number Number Number
Name Name
10 ten 40 forty
11 eleven 50 fifty
12 twelve … …
13 thirteen 90 ninety
14 fourteen 100 one hundred
… … 200 two hundred
19 nineteen … …
20 twenty 1,000 one thousand
21 twenty-one … …
22 twenty-two 1,000,000 one million
… … 1,000,000,000 one billion
(USA)
30 thirty 1,000,000,000,000 one trillion
(USA)
one billion
(UK, France)
31 thirty-one 1,000,000,000,000,000,000 one trillion
(UK)
… … Many one gezillion
J
Number Systems and Data Representation 23

Decimal Numbers 0 through 15 and Equivalent

Hexadecimal and Binary Numbers
Before you go to bed tonight, write the following table of decimal, hexadecimal,
and binary numbers from memory.
In a moment, you will see that it is not that hard. You only are to memorize the
table for zero through fifteen, which is the first sixteen numbers. It is important to keep
table entries in the correct row. For example, you should associate decimal 13 with
hexadecimal D16 and binary 11012 .
Table 11. The Number System Table

Decimal Hexadecimal Binary

Base 10 Base 16 Base 2
0 0 0000
1 1 0001
2 2 0010
3 3 0011
4 4 0100
5 5 0101
6 6 0110
7 7 0111
8 8 1000
9 9 1001
10 A 1010
11 B 1011
12 C 1100
13 D 1101
14 E 1110
15 F 1111
There are four ways to learn this table. You need to use only one of these ways :
memorization, pattern recognition, position value addition, and counting. Use the way
that fits your learning style.
Memorization
Writing the list of hexadecimal numbers is easy. It is just like decimal numbers
until you get to 10. The hexadecimal number system gives us 16 single-digit numbers.
The last 6 numbers are: the number A, the number B, the number C, the number D, the
number E, and the number F. When working in hexadecimal, think of A through F as
being numbers (not letters). The part that takes work is learning the binary. If you
memorize lists easily, write the table until you get it right several times.
Number Systems and Data Representation 24

Pattern Recognition Method.

Number the columns from the right end, beginning with position #0. Number the
rows from the top, beginning with row #0. Beginning with the number zero, the first 4-
bit binary number is a string of four zeros.
Table 12. Pattern Recognition Method of Writing the Number System Table

Binary Binary Binary Binary

Step 1 Step 2 Step 3 Step 4
Row Position Position Position Position
Number Number Number Number Number
3210 3210 3210 3210
Place Value Place Value Place Value Place Value
8421 8421 8421 8421
0 0___ 00__ 000_ 0000
1 0___ 00__ 000_ 0001
2 0___ 00__ 001_ 0010
3 0___ 00__ 001_ 0011
4 0___ 01__ 010_ 0100
5 0___ 01__ 010_ 0101
6 0___ 01__ 011_ 0110
7 0___ 01__ 011_ 0111
8 1___ 10__ 100_ 1000
9 1___ 10__ 100_ 1001
10 1___ 10__ 101_ 1010
11 1___ 10__ 101_ 1011
12 1___ 11__ 110_ 1100
13 1___ 11__ 110_ 1101
14 1___ 11__ 111_ 1110
15 1___ 11__ 111_ 1111

Position Value Addition

Each decimal number from 0 through 15 can be represented by a 4-bit binary
number. Write the position values in four columns, 8 4 2 1. Find that combination of
position values that adds up to the decimal value for that row of the table.
Example: Find the binary value corresponding to decimal 13.
Table 13. Position Values for 4-bit Binary Number

Position Value 8 4 2 1
13 = 1 1 0 1
8 + 4 +0+ 1 = 13, so the corresponding binary number is 11012 .
Counting Method
Use the table for binary addition to add “1” to each row to get to the next row,
with row #0 containing the number “0000”. Before looking at the addition table for
Number Systems and Data Representation 25

binary, briefly review addition for decimal numbers. Suppose we want to add the
decimal numbers 879 and 964.
Table 14. Adding Decimal Numbers With Carry
Carry 1 1 1
8 7 9
Add 9 6 4
Answer 1 8 4 3
When we add 9 + 4, the answer is 3, with a carry of 1 to the next column to the
left. When we add 7 + 6 + 1, the answer is 4, with a carry of 1 to the next column to the
left. When we add 8 + 9 + 1, the answer is 8 with a carry of 1 to the next column to the
left. When we run out of single-digit numbers, we must generate a carry to the next
column to the left.

Binary Addition
We do the same procedure in binary, except that only two numbers can be used (0
and 1). The addition table for binary is as follows.
Table 15. Binary Addition Table

Binary ADDITION x=0 x=1

y=0 0 0 0 1
y=1 0 1 1 0
The addition table is just like the decimal table, except we can only go up to 1
rather than 9. Here is what the table means.
• x + y = 0 + 0 = 0 with a carry of 0.
• x + y = 1 + 0 = 1 with a carry of 0.
• x + y = 0 + 1 = 1 with a carry of 0.
• x + y = 1 + 1 = 0 with a carry of 1.

Sample Binary Addition Problems.

Add 11012 + 00112 .

Carry 1 1 1 1
1 1 0 1
Add 0 0 1 1
Answer 1 0 0 0 0

The answer is 100002 .

Add 01102 + 01012 .

Carry 0 1 0 0
0 1 1 0
Add 0 1 0 1
Answer 1 0 1 1
Number Systems and Data Representation 26

The answer is 10112 .

To create the binary table from 0 to 15 decimal, begin with 0000 in row #0. To
get to each additional row, add 1 to the current row using the binary addition table.

Carry 0 0 0 0
Row #0 0 0 0 0
Add 0 0 0 1
Row #1 0 0 0 1

Carry 0 0 0 1
Row #1 0 0 0 1
Add 0 0 0 1
Row #2 0 0 1 0

Carry 0 0 0 0
Row #2 0 0 1 0
Add 0 0 0 1
Row #3 0 0 1 1

Carry 0 0 1 1
Row #3 0 0 1 1
Add 0 0 0 1
Row #4 0 1 0 0

Carry 0 0 0 0
Row #4 0 1 0 0
Add 0 0 0 1
Row #5 0 1 0 1

Carry 0 0 0 1
Row #5 0 1 0 1
Add 0 0 0 1
Row #6 0 1 1 0

Carry 0 0 0 0
Row #6 0 1 1 0
Add 0 0 0 1
Row #7 0 1 1 1

Carry 0 1 1 1
Row #7 0 1 1 1
Add 0 0 0 1
Row #8 1 0 0 0
Number Systems and Data Representation 27

Carry 0 0 0 0
Row #8 1 0 0 0
Add 0 0 0 1
Row #9 1 0 0 1

Carry 0 0 0 1
Row #9 1 0 0 1
Add 0 0 0 1
Row #10 1 0 1 0

Abstract Counting in Base 4

Good. You have counted in binary. Generalize your knowledge. Try counting in
base 4 using {0,1,2,3}. You can also try it with {circle, stick, square, star} or {w,x,y,z}.
Here is a partial table using four base-4 digits, corresponding to 8 binary digits. Each
base 4 digit corresponds to 2 bits.
Table 16. Counting in Base 4
0000 0100 0200 0300 1000 1100 1200 1300 2000 2100 2200
0001 0101 0201 0301 1001 1101 1201 1301 2001 2101 2201
0002 0102 0202 0302 1002 1102 1202 1302 2002 2102 2202
0003 0103 0203 0303 1003 1103 1203 1303 2003 2103 2203
0010 0110 0210 0310 1010 1110 1210 1310 2010 2110 2210
0011 0111 0211 0311 1011 1111 1211 1311 2011 2111 2211
0012 0112 0212 0312 1012 1112 1212 1312 2012 2112 2212
0013 0113 0213 0313 1013 1113 1213 1313 2013 2113 2213
0020 0120 0220 0320 1020 1120 1220 1320 2020 2120 2220
0021 0121 0221 0321 1021 1121 1221 1321 2021 2121 2221
0022 0122 0222 0322 1022 1122 1222 1322 2022 2122 2222
0023 0123 0223 0323 1023 1123 1223 1323 2023 2123 2223
0030 0130 0230 0330 1030 1130 1230 1330 2030 2130 2230
0031 0131 0231 0331 1031 1131 1231 1331 2031 2131 2231
0032 0132 0232 0332 1032 1132 1232 1332 2032 2132 2232
0033 0133 0233 0333 1033 1133 1233 1333 2033 2133 2233
If you do not understand this, but can write the decimal, hexadecimal, and binary
table, don’t waste your time. Move on.
Number Systems and Data Representation 28

Problem: Fill out the Number Systems Table four times.

Number Systems Table

0à15 0àF 0000à1111 0à15 0àF 0000à1111

Dec Hex Binary Dec Hex Binary
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

0à15 0àF 0000à1111 0à15 0àF 0000à1111

Dec Hex Binary Dec Hex Binary
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Number Systems and Data Representation 29

Conversion Between Binary and Hexadecimal

There is a special relationship between binary and hexadecimal (hex) numbers.
Each hex digit can be represented by a 4-bit code, and each 4-bit code can be represented
by a single hex digit. Characters of the alphabet are represented by 8 bits. Using this
conversion, we can represent a character as 2 hex digits.
We do conversions between binary and hexadecimal by table lookup.
Examples:
E16 = 11102 = 1110B = 1110b = %1110
016 = 00002 = 0000B = 0000b = %0000
716 = 01112 = 0111B = 0111b = %0111
F16 = 11112 = 1111B = 1111b = %1111

Example: Convert the hexadecimal number C16 to binary.

C16 = 11002 = 1100B = 1100b = %1100
Remember that
Table 17. ACE Your Knowledge of Hexadecimal

10 A
12 C
14 Even
A16 is 1010 , C16 is 1210 , and E16 is 1410 in decimal.

Problems: Fill in all the missing values. Convert the single digit hexadecimal numbers
to 4-bit binary numbers.
Table 18. Simple Conversions by Inspection

Hexadecimal Decimal Binary

Problem 1. 216 = 210 = 00102
Problem 2. 716 = 710 =
Problem 3. 816 = =
Problem 4. B16 = 1110 =
Problem 5. D16 = =
Problem 6. F16 = =
Number Systems and Data Representation 30

Use table lookup in the opposite direction to convert a 4-bit binary number
to hex. Example: Convert 10102 and 01012 to hexadecimal.
10102 = A16 = AH = Ah = OxA
01012 = 516 = 5H = Ah = Ox5
To convert a long string of binary digits to hexadecimal, the first step is to group
bits into groups of four, starting from the right hand side. What is important is to start
from the right hand side when doing the grouping. What is important? It is important is
to start the grouping from the right hand side. One more time... What is important?
(Your turn.) Look at an example.
Convert 111010010010011110000000012 to hexadecimal.
We being grouping by starting at the position zero number.
Table 19. Grouping Bits for Conversion from Binary to Hexadecimal

Start
Here
ß ß ß ß ß ß ?
11 :1010 :0100 :1001 :1110 :0000 :0001
The left- most group has only two bits, so we pad on the left end with zeros to
get
Table 20. Zero Padding on the Left

0011 :1010 :0100 :1001 :1110 :0000 :0001

The next step is to do table lookup. Find each group of 4 binary digits and locate
the corresponding single-digit hexadecimal number.
Table 21. Convert from Binary to Hexadecimal by Inspection

0011 :1010 :0100 :1001 :1110 :0000 :0001

3 A 4 9 E 0 1
The answer is 3A49E0116 = 3A49E01H = 3A49E01h = Ox3A49E01
Binary to Hexadecimal Problems
Convert the following binary numbers to hexadecimal.
Problem 7. 11012
Problem 8. 1001B
Problem 9. %0111
Problem 10. 01012
Problem 11. 1 0000 0001b
Problem 12. %110 1011 1000 0001 0101
To convert several hex digits to binary, convert each hex digit to a 4-bit code
using table lookup. The result is the binary code for the hex number.
Number Systems and Data Representation 31

Examples: Convert hexadecimal AF 16 to binary.

A F
1010 1111 = 1010:11112

Convert hexadecimal 36EH to binary.

3 6 E16
0011 0110 1110 = 0011:0110:11102

Convert hexadecimal OxFACE to binary

F A C E
1111 1010 1100 1110 = 1111:1010:1100:11102

Hexadecimal to Binary Problems

Problem 13. Convert 5E16 to binary.
Problem 14. Convert OxDAB to binary.
Problem 15. Convert 83C2FAH to binary.
Problem 16. Convert 10ED7 16 to binary.
Number Systems and Data Representation 32

Data Representation
A Little Byte of History
Computers were first used for computational tasks. The application of using
computers to process non-numeric data came at a slightly later time.
The approach to representation of character data was to represent each character
by a unique code. This is a logical approach for alphabet-based languages. It is not the
only logical approach. Another approach is to represent line strokes by a code, and
require characters to be formed by a combination of line strokes. This approach is a
reasonable way of representing ideographs. It requires the ability of the display device to
overlay images.
An early character code by IBM was Binary Coded Decimal (BCD). Each
character was represented by 6 bits (a pair of octal numbers). This was enough to
represent base-10 digits, the Latin alphabet with capital letters, and a few punctuation
characters. The implementation of character handling internally depended upon the
machine. In the early 1960s, the IBM 7040, a scientific computer, had a 36-bit word.
The IBM 7040 fetched words from memory 36 bits at a time as a unit. Special
instructions were implemented to permit extraction of 6-bit bytes from this 36-bit word
after the whole word was in the accumulator. The IBM 7040 was the descendent of the
IBM 704.
Another scientific computer was the much smaller IBM 1620. It fetched and
operated on data one character at a time. It did addition and multiplication by decimal
table look-up, and was used at University of South Carolina to control scientific
experiments. The corresponding business machine was the IBM 1401. It was the
ubiquitous machine of its day. It also was a character oriented machine with 6-bits for a
BCD alphanumeric character, plus two more bits (parity and field definition). Its word
length was variable. It used a simple language for business data processing, RPG.
Digital Equipment Corporation produced a minicomputer in 1965, the PDP-8.
This was the first computer sold retail, at a price that small organizations could purchase.
It sold for approximately $10,000 of 1965 U.S. dollars (about 15 month’s salary of a new
college graduate in a technical field). It filled a big gap in the market at a time when
other machines were only leased, and required large dedicated staffs.
The basic memory configuration of the PDP-8 was 4 kB of 12-bit words. The
PDP-8 assumed a character set of 7-bit ASCII, with the 8th bit set to 1. Description of the
7-bit ASCII code is available on the Internet. 13 This gave the ability to only represent
upper case letters. The programmer could use other character sets. Commonly, a
stripped 7-bit ASCII code was used by ANDing an ASCII character with octal 77. This
permitted storing two characters per 12-bit word. 14

13
Roman Czyborra, “Good ole' ASCII” (1998). http://czyborra.com/charsets/iso646.html visited 01
January 2008.
14
Douglas Jones, “Frequently Asked Questions About the DEC PDP-8 Computer”
ftp://rtfm.mit.edu/pub/usenet/alt.sys.pdp8/PDP-8_Frequently_Asked_Questions
Number Systems and Data Representation 33

IBM introduced the 8-bit EBCDIC (Extended Binary Coded Decimal Interchange
Code) character set which includes codes for lower case letters and additional special
characters. The IBM 360 accessed memory at an 8-bit byte level. This was a key
development in the design of machines for business data processing. It also implemented
the concept of microinstructions. Prior to the IBM 360, machine language instructions
were loaded into instruction registers, and the bit pattern was used directly to select the
hardware logic to perform the desired operation. The logic circuits required multiple
clock cycles to complete one operation. The microcode approach was to execute a
sequence of microinstructions, each requiring one clock cycle, for each macroinstruction
loaded into the instruction register. This gave IBM the ability to add to the power of its
machines by modifying its microcode. It also made hardware diagnostics easier. IBM
established dominance in the business world with the IBM 360 series. A nice
introduction to microprogramming by Mark Smotherman is available on the Internet. 15
UNICODE is an attempt to create a single international code for information
interchange. It includes a version of ASCII as a subset.
Character Codes
In today’s digital computers, data is represented in computer memory as binary
digits. Each character is represented as a unique binary code in memory. The number of
binary digits required to represent one character is called a byte. There are three primary
character codes, ASCII, EBCDIC, and UNICODE.
US-ASCII (American Standard Code for Information Interchange) is an 8-bit
code used on almost all computers worldwide except for IBM mainframes. ASCII was
originally introduced as a 7-bit code in 1968 as ANSI X3.4. The 8-bit variations are
extensions of ASCII. Jim Price has a nice description and history for ASCII on the
Internet. 16 ASCII has been extended to several 8-bit codes. Another nice, brief history of
ASCII is by Roman Czyborra. 17 This includes images of several ASCII character sets.
EBCDIC (Extended Binary Coded Decimal Interchange Code) is an 8-bit code
used on IBM mainframes. EBCDIC is proprietary to IBM. Prior to the Internet,
EBCDIC was the predominant character code in use. The Internet and the ubiquitous
presence of the personal computer have made ASCII the most commonly used code.
EBCDIC remains important in the mainframe world of commercial data
processing. See Multilingual Support in Internet/IT Applications, Trans-European
Research and Education Networking Association. 18 Kosta Kostis has provided EBCDIC
code tables. 19 20

15
Mark Smotherman, “A Brief History of Microprogramming” (March 1999)
16
Jim Price, “ASCII Chart” (28 October 2001). Visited 02 November 2001.
17
Roman Czyborra, “The ISO 8859 Alphabet Soup” (12 January 1998). Visited 02 November 2001.
18
Trans-European Research and Education Networking Association, http://www.terena.org/. Visited 01
January 2008.
19
Kosta Kostis, “EBCDIC Codepage 037”, http://www.kostis.net/charsets/ebc037.htm (17 February 2001).
Visited 02 November 2001.
20
Kosta Kostis, “EBCDIC Codepage 1047”, http://www.kostis.net/charsets/ebc1047.htm (17 February
2001). Visited 02 November 2001.
Number Systems and Data Representation 34

UNICODE (Universal Code) is a 16-bit extension code of ASCII used to

represent symbols of other languages, mathematics, and other disciplines. This includes
languages that use ideograms (like Chinese, Japanese, Korean) as well as alphabets (like
Latin, Hebrew, Greek). Microsoft Office 2000 (and later versions) use UNICODE.
Corel PerfectOffice permits import and export of UNICODE text files.
The chart on the following page shows the alphabetic, numeric, and blank
character ( ) with their codes in hex. The symbol for the blank is often made with the
slash through the b. The code for the letter “Oh” is different than the code for the
number zero. The number zero is often written with a slash. Upper case characters have
a code different than lower case characters.
For example:
The character A represented in EBCDIC is C116
The character A represented in ASCII is 4116
The character A represented in UNICODE is 004116
UNICODE representation of ASCII consists of the ASCII code prefixed with
0016 . This selection was made to keep UNICODE compatible with most existing
software. There are many symbols in UNICODE that have hex numbers different than
zero in the first two positions. A complete definition of UNICODE can be obtained from
Unicode, Inc. 21
Codes used for data processing, such as those above, are suitable for sorting. The
order in which characters appear in a code is called the collating sequence or
lexicographic order. In applications, the treatment of the blank and special rules of
language impose additional considerations. This is important in constructing dictionaries
and telephone books. For example, European names sometimes include “von”, “van”
and other prefixes. European rules for name collating are not the same as in the United
States. Internal character codes are just the beginning of the process.
There are companies that sell the service of converting data from one format to
another. VEDIT by Greenview Data, Inc. is one example. 22

21
Unicode Home Page, http://www.unicode.org/. Visited 01 January 2008.
22
Greenview Data, Inc., “VEdit”, http://www.vedit.com/index.html (2007). Visited 01 January 2008.
Number Systems and Data Representation 35

Example: Use Table 22 on page 36 to code the following characters first to hexadecimal,
and then to binary.
Character 2 C A T S
EBCDIC F2 40 C3 C1 E3 E2
Binary 1111:0010 0100:0000 1100:0011 1100:0001 1110:0011 1110:0010

Character 2 C A T S
ASCII 32 20 43 41 54 53
Binary 0011:0010 0010:0000 0100:0011 0100:0001 0101:0100 0101:0011

Character 2 C
UNICODE 0032 0020 0043
Binary 0000:0000:0011:0010 0000:0000:0010:0000 0000:0000:0100:0011

Character A T S
UNICODE 0041 0054 0053
Binary 0000:0000:0100:0001 0000:0000:0101:0100 0000:0000:0101:0011

Character Code Problems

Code the following in hexadecimal using EBCDIC and ASCII, and then code each into
binary. Notice upper and lower case, letters and numbers.
Problem A1. 2 Zoos
Character 2 Z o o s
EBCDIC
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Character 2 Z o o s
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Problem A2. 1 Bill

Character 1 B i l l
EBCDIC
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Character 1 B i l l
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____
Number Systems and Data Representation 36

EBCDIC, US-ASCII, and UNICODE Character Codes

Table 22. EBCDIC, US-ASCII, UNICODE for Latin Alphabet

TEXT EBCDIC ASCII UNICODE TEXT EBCDIC ASCII UNICODE

40 20 0020
, 6B 2C 002C
. 75 2E 002E
0 F0 30 0030
1 F1 31 0031
2 F2 32 0032
3 F3 33 0033
4 F4 34 0034
5 F5 35 0035
6 F6 36 0036
7 F7 37 0037
8 F8 38 0038
9 F9 39 0039
? 6F 3F 003F
A C1 41 0041 a 81 61 0061
B C2 42 0042 b 82 62 0062
C C3 43 0043 c 83 63 0063
D C4 44 0044 d 84 64 0064
E C5 45 0045 e 85 65 0065
F C6 46 0046 f 86 66 0066
G C7 47 0047 g 87 67 0067
H C8 48 0048 h 88 68 0068
I C9 49 0049 i 89 69 0069
J D1 4A 004A j 91 6A 006A
K D2 4B 004B k 92 6B 006B
L D3 4C 004C l 93 6C 006C
M D4 4D 004D m 94 6D 006D
N D5 4E 004E n 95 6E 006E
O D6 4F 004F o 96 6F 006F
P D7 50 0050 p 97 70 0070
Q D8 51 0051 q 98 71 0071
R D9 52 0052 r 99 72 0072
S E2 53 0053 s A2 73 0073
T E3 54 0054 t A3 74 0074
U E4 55 0055 u A4 75 0075
V E5 56 0056 v A5 76 0076
W E6 57 0057 w A6 77 0077
X E7 58 0058 x A7 78 0078
Y E8 59 0059 y A8 79 0079
Z E9 5A 005A z A9 7A 007A
Non-printing codes used for control purposes are not shown in this table.
Number Systems and Data Representation 37

The purpose of problems 3 and 4 is to illustrate that different characters that

appear similar have different codes. Be sure to read the problem notes before doing the
problem. Distinguish between numbers, upper case letters, and lower case letters.
Problem A3. Hi Ho 10 lO?
Notes:
• “Hi” are letters.
• “Ho” are letters.
• “10” are numbers.
• “lO” are letters, followed by punctuation.

Character H i H o
EBCDIC
Binary : : : : : :
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Character 1 0 l O ?
EBCDIC
Binary : : : : : :
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Problem A4. 2ZS50Oo 1li.

• 2ZS5 Number, Letter, Letter, Number
• 0Oo Number, Letter, Letter
• 1li. Number, Letter, Letter, Punctuation

Character 2 Z S 5 0 O
EBCDIC
Binary : : : : : :
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____

Character o 1 l i .
EBCDIC
Binary : : : : : :
ASCII
Binary ____:____ ____:____ ____:____ ____:____ ____:____ ____:____
Number Systems and Data Representation 38

Problem A5. Code the following in hexadecimal using ASCII, and then convert to
binary. Oven Hot

Character O v e n
ASCII hex
Binary

Character H o t
ASCII hex
Binary

Problem A6. Code the following in hexadecimal using EBCDIC, and then convert to
binary. 10 is lo
Character 1 0 i
EBCDIC hex
Binary

Character s l o
EBCDIC
Binary

Problem A7. Convert the following to hexadecimal, and then to characters using
EBCDIC code.

Binary 1111:0001 0100:0000 1101:0001 1010:0100 1001:0100 1001:0111

EBCDIC
Hex
Character

Jubilation Code and Gran Zeff Code

The purpose of this section is to demonstrate the benefit of Unicode by using a
symbol set not represented in US-ASCII. Unicode for Latin letters have 00 as the first
two hexadecimal characters. The following symbol set shows codes that are not 00 in the
first two Unicode positions. The first two hexadecimal characters index which code table
to use. The “Secret Symbols” are all Greek to me.
 Number Systems and Data Representation 39

Table 23. Gran Zeff Code, Jubilation Code, and UNICODE

JUBILATION GRAN ZEFF SECRET JUBILATION GRAN ZEFF UNICODE

SECRET UNICODE
CODE CODE SYMBOL CODE CODE (Hex)
SYMBOL (Hex)
(Hexadecimal) (Hexadecimal) (Hexadecimal) (Hexadecimal)
blank 40 20 0020
α A1 45 03B1 ξ D2 76 03BE
β A2 46 03B2 ο D3 77 03BF
γ A3 47 03B3 π D4 78 03C0
δ A4 48 03B4 ρ E1 85 03C1
ε B1 55 03B5 s E2 86 03C2
ζ B2 56 03B6 σ E3 87 03C3
η B3 57 03B7 τ E4 88 03C4
θ B4 58 03B8 υ F1 95 03C5
ι C1 65 03B9 ϕ F2 96 03C6
κ C2 66 03BA χ F3 97 03C7
λ C3 67 03BB ψ F4 98 03C8
µ C4 68 03BC ω F5 99 03C9
ν D1 75 03BD

Example 1: Code the word δεκα (ten) from SECRET SYMBOL to JUBILATION
CODE hexadecimal, GRAN ZEFF CODE hexadecimal, and UNICODE hexadecimal
representations.
Secret Symbol δ ε κ α
Jubilation Code A4 B1 C2 A1
Gran Zeff Code 48 55 66 45
Unicode 03B4 03B5 03BA 03B1

Example 2: Code the GRAN ZEFF CODE binary into GRAN ZEFF CODE
hexadecimal, and then to SECRET SYMBOL. (man)

0100:0101 0111:0101 0101:1000 1000:0101 1001:1001 0111:1000 0111:0111 1000:0110

45 75 58 85 99 78 77 86
α ν θ ρ ω π ο s
Number Systems and Data Representation 40

Example 3: Code the UNICODE binary into UNICODE hexadecimal, and then into
SECRET SYMBOL. (word).
0000:0011:1011:1011 03BB λ
0000:0011:1011:1111 03BF ο
0000:0011:1011:0011 03B3 γ
0000:0011:1011:1111 03BF ο
0000:0011:1100:0010 03C2 s

Problem B1: Code the word αδελφοs (brother) from SECRET SYMBOL to
JUBILATION CODE hexadecimal, and then to JUBILATION CODE binary.
α δ ε λ ϕ ο s

Problem B2: Code the word δωρον (gift) from SECRET SYMBOL to GRAN ZEFF
CODE hexadecimal, and then to GRAN ZEFF CODE binary.
δ ω ρ ο ν

Problem B3: Code the word σοφοs (wise) from SECRET SYMBOL to UNICODE
hexadecimal, and then to UNICODE binary.
σ ο ϕ ο s

Problem B4: Code the word καρδια (heart) from SECRET SYMBOL to UNICODE
hexadecimal, and then to UNICODE binary.
κ α ρ δ ι α
Number Systems and Data Representation 41

Problem B5: Code the JUBILATION CODE binary into JUBILATION CODE
hexadecimal, and then to SECRET SYMBOL. (death)

1011:0100 1010:0001 1101:0001 1010:0001 1110:0100 1101:0011 1110:0010

Problem B6: Code the GRAN ZEFF CODE binary into GRAN ZEFF CODE
hexadecimal, and then to SECRET SYMBOL. (wicked)
0111:1000 0111:0111 0111:0101 0101:0111 1000:0101 0111:0111 1000:0110

Problem B7: Code the UNICODE binary into UNICODE hexadecimal, and then into
SECRET SYMBOL. (nation)

0000:0011: 0000:0011: 0000:0011: 0000:0011: 0000:0011:

1011:0101 1011:1000 1011:1101 1011:1111 1100:0010

Additional Information About Codes

A nice history of character codes is given by Steven J. Searle. 23 For the detail-
oriented person, see Jukka Korpela, “A tutorial on character code issues”. 24
A common exercise for first-semester programming in computer science is to
generate a printout of a character set used by a printer. The exercise includes printing the
corresponding hexadecimal, decimal, octal, and binary values of the internal character
code. OSdata.com has made available code tables indexed by symbol or control code. 25
Communication imposes different requirements than data processing. In
communication, the goal is to transmit codes in a way that helps reconstruction of the
original message. If a code is intended to support transmission of sequences, such as
number sequences (like a counter), you want the bit patterns to be designed to help detect
mistakes. One approach is called Gray Coding. 26

23
Steven J. Searle, “A Brief History of Character Codes in North America, Europe, and East Asia”,
Sakamura Laboratory, University Museum, University of Tokyo (1999),
http://tronweb.super-nova.co.jp/characcodehist.html, Visited 02 Nov 2001.
24
Jukka Korpela, “A tutorial on character code issues”, Tampere University of Technology, Finland,
http://www.cs.tut.fi/~jkorpela/chars.html, (24 October 2001). Visited 02 Nov 2001.
25
“Character Codes”, OSdata.com, http://www.osdata.com/system/physical/charcode.htm Visited 01
January 2008.
26
“Gray code”, National Institute of Standards and Technology
http://www.nist.gov/dads/HTML/graycode.html Visited 01 January 2008.
Number Systems and Data Representation 42

You can find codes for most any type of data. For example, HTML has
established accepted hexadecimal codes for colors.

Positional Number Systems

The idea of positional number systems was the topic of the first seven chapters of
Liber abaci (1202 A.D.) by Leonardo Fibonacci27 . Leonardo Fibonacci, also known as
Leonardo of Pisa or Leonardo Pisano, was born approximately 1170 A.D., and died after
1240.
A positional number system is a series number system in which only the
coefficients ck are recorded. For most positional number systems you will work with,
the coefficient in position number k multiplies the corresponding function numbered k.
This is true for function or power series number systems. The common example is:
…c4 c3 c2 c1 c0 .c-1 c-2 …
If the base of the number system cannot be assumed from context, it must be
stated explicitly.
A positional number system requires an explicit symbol for zero as a place
holder.
Ancient Greek and Hebrew, and Roman numerals, had no zero. Greek and
Hebrew used their alphabets as numerals. All had to work hard to extend the range of
numbers that could be represented as their applications diversified, and interest in
computation became more sophisticated. What is surprising is that the Roman numeral
system remained in daily use for so long.
Decimal Number System
In a decimal number, each digit has a place value. In elementary school, we
learned that the position of a digit in a decimal number has meaning. In the number
9874, the digit 4 is in the one’s place, the digit 7 is in the tens place, the digit 8 is in the
hundred’s place, and the digit 9 is in the thousand’s place. We need to generalize the
concept, so we write it more formally.

103 102 101 100 ß place values in exponent form

1,000 100 10 1 ß place values
9 8 7 4 ß digits

9 8 7 4 is the same as
( 9 x 1000 ) + ( 8 x 100 ) + ( 7 x 10 ) + ( 4 x 1 )
This is the sum of the digits times their place values.
We also know that the number 009874 is the same number as 9874. We can write
as many zeros to the left end of a number as we want and the number does not change
value. This is true for all numbers in a positional number system.

27
Encyclopaedia Britannica 2003, Deluxe Edition CD-ROM
Number Systems and Data Representation 43

Banfill has a nice web site for review of decimal place values. 28
Algebra Review of Exponents
To understand the positional number system, we need a few basic rules of algebra
about powers or exponents.
1. Any number raised to the power zero is equal to one.
30 = 1, 270 = 1 a0 = 1 x0 = 1

2. Any number raised to the power one is equal to itself.

31 = 3, 171 = 17 a1 = a x1 = x

3. Any number raised to a positive integer power N that is greater than one is
equal to N copies of the number multiplied together.
35 = 3 x 3 x 3 x 3 x 3 = 243
a3 = a x a x a

1
4. X − M = Notice the minus sign in the left hand term.
XM
Positional Number System General Concept
Some people like the general concept of an idea first, and then look at details.
Other people like detailed examples first, and then the general concept. If you like details
first, skip to the topic on the Binary Number System. Come back to this topic after
studying the hexadecimal system.
In a positional number system, the position of a digit within a number determines
its value. Another term for “digit” is “coefficient”. Position is numbered from a
reference point, which is called the decimal point for a base 10 number. For whole
numbers without the refe rence point explicitly represented, this is the right side of the
number. The right- most digit of a whole number is in position zero.
Let ck be the coefficient in position k, where the position is identified by the
subscript. Let the letter b stand for the base of the system. The general form of a number
in a positional number system is
…c4 c3 c2 c1 c0 . c-1 c-2 … The period between coefficients c0 . c-1 is called the
“radix point”. In base 10, this is the decimal point.
In the decimal number 84693, c4 = 8, c3 = 4, c2 = 6, c1 = 9, c0 = 3
The value of this general number c4 c3 c2 c1 c0 . c-1 c-2 … is given by:
(c4 x b ) + (c3 x b3 ) + (c2 x b2 ) + (c1 x b1 ) + (c0 x b0 ) + (c-1 x b-1 ) + (c-2 x b-2 )
4

28
J. Banfill, “Place Value – Lessons”, AAA Math, (2004). http://www.aaamath.com/B/plc.htm. Visited 01
January 2008.
Number Systems and Data Representation 44

Notice the minus sign in the powers to the right of b0 . The radix point is located
between the term with b0 and the term with b-1 . We will concentrate on whole numbers.
In the decimal number 84693, the base is 10, and we have
(8 x 104 ) + (4 x 103 ) + (6 x 102 ) + (9 x 101 ) + (3 x 100 )
= 80,000 + 4,000 + 600 + 90 + 3
Binary Number System
In the binary (base 2) number system, b = 2. The general binary form is
…c4 c3 c2 c1 c0 = … (c4 x 24 ) + (c3 x 23 ) + (c2 x 22 ) + (c1 x 21 ) + (c0 x 20 )
= … (c4 x 16) + (c3 x 8) + (c2 x 4) + (c1 x 2) + (c0 x 1)

The meaning of the binary number 10112 is shown in the table below. Recall that
3
2 = 2 x 2 x 2 = 8.
23 22 21 20 ß place values as powers of 2
8 4 2 1 ß place values
1 0 1 12 ß coefficients
In the binary number sys tem, each place value is double the previous place value,
moving from right to left.
1 0 1 12 is the same as
(1 x 8 ) + ( 0 x 4 ) + ( 1 x 2 ) + ( 1 x 1 ) This is the sum of the digits multiplied
by the place value,
= 8 + 0 + 2 + 1
= 11 in the decimal number system.
The place values of the four right- most bits are 8 4 2 1. Recall that any decimal
number from 0 to 15 can be represented by 4 bits. To do this by computation, determine
what combination of 8, 4, 2, and 1 will sum to that number. Begin with 8 and work to
smaller numbers. Add a number if the current sum plus the number is still less than or
equal to the goal. Record a “1” if you use that number. Record a “0” if you do not use
that number.
Example, 14 = 8 + 4 + 2. The number 14 can be represented by a 4-bit code by turning
on (setting to 1) only the bits in positions representing 8, 4, and 2.
8421 ß Place values.
Therefore 14 = 8 + 4 + 2 = 11102 = 1110B = %1110

Other examples:
0= 00002
7=4+2+1= 01112
15 = 8 + 4 + 2 + 1 = 11112
Number Systems and Data Representation 45

Example: Convert the hexadecimal C to binary.

8421 ß Place values.
Therefore C16 = 1210 = 8 + 4 = 11002
Example 1: The number 11 01002 is
5 4 3 2 1 0 Position
Number
25 24 23 22 21 20 Place Value
32 16 8 4 2 1 Place Value
in Decimal
1 1 0 1 0 0 Coefficients
+ + + + + Sum of
(1 x 32 ) (1 x 16 ) (0 x 8 ) (1 x 4 ) (0 x 2 ) (0 x 1 ) Products
32 + 16 +0 +4 +0 +0
ANSWER = 54

Example 2: The number 101 10012 is

6 5 4 3 2 1 0 Position
Number
26 25 24 23 22 21 20 Place Value
64 32 16 8 4 2 1 Place Value
in Decimal
1 0 1 1 0 0 1 Coefficients
+ + + + + Sum of
(1x64) (0x32) (1x16) (1x8) (0x4) (0x2) (1x1) Products
64 +0 + 16 +8 +0 +0 +1
ANSWER = 89
Number Systems and Data Representation 46

Example 3: Convert 1 0 1 0 1 12 to decimal.

5 4 3 2 1 0 ß position
number
25 24 23 22 21 20 ß place
value as
power of base
2
32 16 8 4 2 1 ß place
value
1 0 1 0 1 12 ß
coefficients
(1x32) + (0x16) + (1 x 8) + (0 x 4) + (1 x 2) ß sum of
+(1 x 1)
products
32 +0 +8 +0 +2 +1 ß perform
multiplication
ANSWER = 43

Hexadecimal Number System

In the hexadecimal (base 16) number system, b = 16. Therefore, the general
hexadecimal form is
…c4 c3 c2 c1 c0 = … (c4 x 164 ) + (c3 x 163 ) + (c2 x 162 ) + (c1 x 161 ) + (c0 x 160 )
= … (c4 x 65536) + (c3 x 4096) + (c2 x 256) + (c1 x 16) + (c0 x 1).
The following chart allows you to convert from a hexadecimal digit to its
equivalent decimal value.

Hex
0 1 2 3 4 5 6 7 8 9 A B C D E F
Digit
Decimal
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Value
Table 24. Counting in Hexadecimal

ACE your knowledge of hexadecimal. If you memorize A, C, and E are 10, 12,
and 14, you can mentally determine B is 11, D is 13, and F is 15.
Recall that 163 = 16 x 16 x 16 = 4096.
Number Systems and Data Representation 47

Example 1: Convert the hexadecimal number 93AF16 to decimal.

Each digit in a hexadecimal number has a place value.
3 2 1 0 ß position number
163 162 161 160 ß place values as
powers of 16
4,096 256 16 1 ß place value in
decimal
9 3 A F16 ß hexadecimal
coefficients
9 3 10 15 ß decimal
coefficients
( 9 x 4,096 ) + ( 3 x 256 ) + ( 10 x 16 ) + ( 15 x 1 ) ß sum of products
36,864 +768 + 160 +15 ß perform
multiplication
ANSWER = 37,807

Compute the sum of products from the “place value in decimal” and the
“decimal coefficients” rows.
Example 2: Convert the hexadecimal number 4A2D16 to decimal.
3 2 1 0 ß position number
163 162 161 160 ß place values as
powers of 16
4,096 256 16 1 ß place value in
decimal
4 A 2 D16 ß hexadecimal
coefficients
4 10 2 13 ß decimal
coefficients
( 4 x 4,096 ) + ( 10 x 256 ) + ( 2 x 16 ) + ( 13 x 1 ) ß sum of products
16,384 +2560 + 32 +13 ß perform
multiplication
ANSWER = 18,989
Number Systems and Data Representation 48

Example 3: Convert the hexadecimal number A B C 616 to decimal.

3 2 1 0 ß position
number
163 162 161 160 ß place value as
power of base 16
4096 256 16 1 ß place value
A B C 616 ß hexadecimal
coefficients
10 11 12 6 ß decimal
coefficients
( 10 x 4096 ) + ( 11 x 256 ) + ( 12 x 16 ) +(6x1 ) ß sum of products
40,960 + 2,816 + 192 +6 ß perform
multiplication
ANSWER = 43,974
Octal Number System
In the octal (base 8) number system, b = 8. Therefore, the general octal form is
…c4 c3 c2 c1 c0 = (c4 x 84 ) + (c3 x 83 ) + (c2 x 82 ) + (c1 x 81 ) + (c0 x 80 )
= (c4 x 4096) + (c3 x 512) + (c2 x 64) + (c1 x 8) + (c0 x 1).
Decode the meaning of the hexadecimal number 42738 as shown in the table
below. Recall that 83 = 8 x 8 x 8 = 512.
Each digit in an octal number has a place value.
3 2 1 0 ß position number
83 82 81 80 ß place values as powers of 8
512 64 8 1 ß place value
4 2 7 38 ß coefficients
( 4 x 512 ) + ( 2 x 64 ) +(7x8 ) +(3x1 ) ß sum of products
2,048 + 128 + 56 +3 ß perform multiplication
ANSWER = 2,235

If you see a digit greater than 7 in a number, you know it cannot be an octal
number.
Fun Questions
How many symbols are in a base 3 system?
What is the answer to “2 + 2 = ?” in base 3? (You may do the arithmetic in base
10 and convert the ans wer to base 3.)
Sexagesimal or Sexagenary System
The sexagesimal or sexagenary system is a base 60 positional number system.
We use it for time, direction, and position. We apply it in navigation and astronomy.
Our implementation is a variant of a system first used by the early Sumerians, where we
allow decimal numbers as coefficients. To ensure we treat the number as base 60, we use
Number Systems and Data Representation 49

special symbols to indicate the place value. The place values are degrees, minutes, and
seconds. We allow the coefficient for degrees to exceed 60.

3 2 1 0 ß position number
° ‘ “ ß place value symbol
degrees minutes seconds ß place value name
602 601 600 ß place values as powers of 60
3600 60 1 ß place value in base 10
179 59 59 ß coefficients in base 10
+ ( 179 x + ( 59 x + ( 59 x 1 ) ß sum of products
3600 ) 60 )
644400 + 3540 + 59 ß perform multiplication
ANSWER = 647999” = 179° 59‘ 59“

Nothing Matters
The symbol for “nothing”, which is 0 in the Hindu-Arabic notation, is necessary
for efficient recording of numbers in a positional number system. Look again at the
general form for a number written in positional notation.
c4 c3 c2 c1 c0 . c-1 c-2 …
Suppose that there is no symbol for “nothing” and you have a number without any
coefficient for b2 . You would have to omit the term c2 from your record of that number.
You can still have a number system using position to write a number in a
canonical form, but you would also need to write each digit with its corresponding place
value, or use different number symbols as coefficients of different place values.
(c4 x b4 ) + (c3 x b3 ) + (c2 x b2 ) + (c1 x b1 ) + (c0 x b0 ) + (c-1 x b-1 ) + (c-2 x b-2 )

Explicitly writing the place value with the coefficient does not require an ordering
of terms, although it does not prohibit ordering them. The number above has the same
value as
(c4 x b4 ) + (c2 x b2 ) + (c3 x b3 ) + (c1 x b1 ) + (c0 x b0 ) + (c-1 x b-1 ) + (c-2 x b-2 )
where (c3 x b3 ) + (c2 x b2 ) were exchanged.
A system using different symbols for different place-values was used in Attic
numerals described by Herodianus. Hebrew also used a system of assigning numerical
values to letters of the alphabet. 29 Use of a finite set of different symbols as coefficients
of different place values limits the number of numbers that can be represented.

29
D. E. Smith, “Numerals”, Volume 16, page 612, Encyclopaedia Britannica (1965)
Number Systems and Data Representation 50

Conversion to Decimal
To convert a binary or hexadecimal number to decimal
1) Multiply each digit by its place value.
2) Sum the products.
3) The result is the decimal equivalent.

Conversion to Decimal Problems:

Show all your work.
Problem A1. Convert 1102 to decimal.
ß place values as powers of 2
ß place values
ß coefficients
ß sum of products
ß perform multiplication
ANSWER =

Problem A2. Convert 1 01102 to decimal.

Problem A3. Convert 1 1010 1101B to decimal.
Problem A4. Convert %111 1001 to decimal
Problem A5. Convert B4 16 to decimal.
Problem A6. Convert 2C5F16 to decimal.
For the abstract thinkers:
Convert x w y w z y x4 to decimal, for the number system {w, x, y, z}.
Solution: First notice that this is a base-4 system. The set {w, x, y, z}. exactly 4
symbols, and the ordering is given. We can associate “w” with “0”, “x” with “1”, “y”
with “2”, and “z” with “3”.
6 5 4 3 2 1 0 ß position number
46 45 44 43 42 41 40 ß place value as power of base 16
x w y w z y x4 ß digits
4096 1024 256 64 16 4 1 ß place value in decimal
1 0 2 0 3 2 1 ß digits in decimal

= (x x 46 ) + (w x 45 ) + (y x 44 ) + (w x 43 ) + ( z x 42 ) + ( y x 41 ) + ( x x 40 )
= (1 x 4096 ) + (0 x 1024 ) + (2 x 256 ) + (0 x 64 ) + (3 x 16 ) + (2 x 4 ) + (1 x 1 )
= 40,960 + 0 + 512 + 0 + 48 + 8 + 1
= 4,66510
Problem A7. Convert 30114 to decimal.
Number Systems and Data Representation 51

Note: Processors on many calculators are 4-bit processors. A unit of memory consisting
of 4-bits is a “nibble”, which is half a byte. 30
Problem A8. Convert 30178 (octal = base 8) to decimal.

Converting From Decimal: Method of Successive Division

The method of successive division is used to convert from decimal to other base
number systems. If you are rusty with long division, you can review it at
http://www.aaamath.com/B/div55_x2.htm. 31 We begin by dividing the original number
by the base that we are going to.
What we are after is the collection of remainders after the remainders are
converted to digits in the new base number system. It is important to label each
remainder so that we know how to arrange the remainders into our final answer. We
label each remainder with a subscripted R, such as R3 . “R” stands for “Remainder”.
The subscript identifies the position number of the remainder. We begin with the
position zero remainder, R0 .
We repeat dividing the quotient from the last step by the new base. We continue
this process until we have zero for a quotient.
To form the answer, we use the remainders to build the result from right-to- left.
Remainder R0 becomes the extreme right digit for an integer answer. Remainder R1 goes
to the left of R0 . The process continues until all the computed remainders have been
used. The result looks like
R9R8R7R6R5 R4 R3 R2 R1 R0 .
Converting Decimal Numbers (Base 10) to Binary (Base 2)
To convert a decimal number to another base
1) Divide the decimal number by the base and determine the quotient (Q) and
remainder (R).
2) Divide the resulting quotient by the base and determine its quotient (Q) and
remainder (R).
3) Repeat step 2 until the quotient (Q) is zero.
4) Write the remainders in the new base beginning with the first remainder computed
in position R0 . …R9R8R7R6R5 R4 R3 R2 R1 R0
5) The result from step 4 is the number converted to the new base.

30
John J. Donovan, Systems Programming, “Machine Structure – 360 and 370”, page 25, McGraw-Hill
(1972).
31
J. Banfill, “Dividing a 4-digit by 2-digit numbers”, AAA Math (2004).
http://www.aaamath.com/B/div55_x2.htm Visited 14 January 2008.
Number Systems and Data Representation 52

Example 1: Convert 87 to binary.

Step 0 4 3 = Q1 à Step 1 2 1 = Q2
Base = 2 8 7 = Q0 Base = 2 4 3 = Q1
8 4
0 7 0 3
6 2
1 = R0 1 = R1

Step 2 1 0 = Q3 à Step 3 5 = Q4
Base = 2 2 1 = Q2 Base = 2 1 0 = Q3
2 1 0
0 1 0 = R3
0
1 = R2

Step 4 2 = Q5 à Step 5 1 = Q6
Base = 2 5 = Q4 Base = 2 2 = Q5
4 2
1 = R4 0 = R5

Step 6 0 = Q7 Stop only when the Quotient is zero.

Base = 2 1 = Q6
0
1 = R6

Record the answer using remainders , starting in the column on the right end.

6 5 4 3 2 1 0 Position Number
R6 R5 R4 R3 R2 R1 R0 ß Record remainders from right to left.
1 0 1 0 1 1 1 Final Answer

The answer is 101 01112 = 101 0111B = %101 0111

Example 2: Convert 101 from decimal to binary.

Step 0 5 0 = Q1 à Step 1 2 5 = Q2
Base = 2 1 0 1 = Q0 Base = 2 5 0 = Q1
1 0 4
0 1 1 0
0 1 0
1 = R0 0 = R1
Number Systems and Data Representation 53

Step 2 1 2 = Q3 à Step 3 6 = Q4
Base = 2 2 5 = Q2 Base = 2 1 2 = Q3
2 1 2
0 5 0 = R3
4
1 = R2

Step 4 3 = Q5 à Step 5 1 = Q6
Base = 2 6 = Q4 Base = 2 3 = Q5
6 2
0 = R4 1 = R5

Step 6 0 = Q7 Stop only after the Quotient is zero.

Base = 2 1 = Q6
0
1 = R6

Record the answer using remainders , starting in the column on the right end.
6 5 4 3 2 1 0 Position Number
R6 R5 R4 R3 R2 R1 R0 ß Record remainders from right to left.
1 1 0 0 1 0 1 Final Answer

The answer is 110 01012 = 110 0101B = %110 0101 = 10110

Converting Decimal Numbers (Base 10) to Hexadecimal (Base 16)
To convert a decimal number to another base
1) Divide the decimal number by the base and determine the quotient Q and
remainder R. Convert the remainder to hexadecimal.
2) Divide the resulting quotient by the base and determine its quotient Q and
remainder R. It is good to convert as soon as you see what remains.
3) Repeat step 2 until the quotient Q is zero.
4) Write the remainders in the new base beginning with the last remainder
computed. …R9 R8R7R6R5R4R3 R2 R1 R0
5) The result from step 4 is the number converted to the new base.
Number Systems and Data Representation 54

Example 1: Convert 3117 to hexadecimal.

Step 0 1 9 4 = Q1 à Step 1 1 2 = Q2
Base = 16 3 1 1 7 = Q0 Base = 16 1 9 4 = Q1
1 6 1 6
1 5 1 3 4
1 4 4 3 2
7 7 2 = R1
6 4 = 2H
1 3 = R0
= DH

Step 2 0 = Q3 Stop only when the Quotient is zero.

Base = 16 1 2 = Q2
0
1 2 = R2 Remainder in decimal.
= CH Remainder converted to hexadecimal.

2 1 0 Position Number
R2 R1 R0 ß Record remainders from right to left.
12 2 13 Remainders in decimal.
C 2 D Final answer. Remainders in hexadecimal.

The answer is C2D16 = C2DH = OxC2D = 311710

Example 2: Convert 3982 to hexadecimal.
Step 0 2 4 8 = Q1 à Step 1 1 5 = Q2
Base = 16 3 9 8 2 = Q0 Base = 16 2 4 8 = Q1
3 2 1 6
7 8 8 8
6 4 8 0
1 4 2 8 = R1
1 2 8 = 8H
1 4 = R0
= EH

Step 2 0 = Q3 Stop only when the Quotient is zero.

Base = 16 1 5 = Q2
0
1 5 = R2 Remainder in decimal.
= FH Remainder converted to hexadecimal.
Number Systems and Data Representation 55

2 1 0 Position Number
R2 R1 R0 ß Record remainders from right to left.
E 8 F Final answer. Remainders in hexadecimal.

The answer is E8F16 = E8FH = OxE8F = 398210

Example 3: Convert 3786 to hexadecimal.
Step 0 2 3 6 = Q1 à Step 1 1 4 = Q2
Base = 16 3 7 8 6 = Q0 Base = 16 2 3 6 = Q1
3 2 1 6
5 8 7 6
4 8 6 4
1 0 6 1 2 = R1
9 6 = CH
1 0 = R0
= AH

Step 2 0 = Q3 Stop only when the Quotient is zero.

Base = 16 1 4 = Q2
0
1 4 = R2 Remainder in decimal.
= EH Remainder converted to hexadecimal.

2 1 0 Position Number
R2 R1 R0 ß Record remainders from right to left.
E C A Final answer. Remainders in hexadecimal.

The answer is ECA16 = ECAH = OxECA = 378610

Example 4: Convert 2748 to hexadecimal.
Step 0 1 7 1 = Q1 à Step 1 1 0 = Q2
Base = 16 2 7 4 8 = Q0 Base = 16 1 7 1 = Q1
1 6 1 6
1 1 4 1 1
1 1 2 0
2 8 1 1 = R1
1 6 = BH
1 2 = R0
= CH
Number Systems and Data Representation 56

Step 2 0 = Q3 Stop only when the Quotient is zero.

Base = 16 1 0 = Q2
0
1 0 = R2 Remainder in decimal.
= AH Remainder converted to hexadecimal.

2 1 0 Position Number
R2 R1 R0 ß Record remainders from right to left.
A B C Final answer. Remainders in hexadecimal.

The answer is ABC 16 = ABCH = OxABC = 274810

Practice: Convert the following decimal numbers to hexadecimal.
3243 = CAB16 2989 = BADH 2750 = OxABE 3245 = CAD16
4013 = FAD16 4075 = FEBH 3564 = OxDEC 4077 = FED16

Converting Decimal Numbers (Base 10) to Octal (Base 8)

To convert a decimal number to another base
1) Divide the decimal number by the base and determine the quotient Q and
remainder R.
2) Divide the resulting quotient by the base and determine its quotient Q and
remainder R.
3) Repeat step 2 until the quotient Q is zero.
4) Write the remainders in the new base beginning with the last remainder
computed. …R9 R8R7R6R5R4R3 R2 R1 R0
5) The result from step 4 is the number converted to the new base.

Example: Convert decimal 3117 to octal.

Step 0 3 8 9 = Q1 à Step 1 4 8 = Q2
Base = 8 3 1 1 7 = Q0 Base = 8 3 8 9 = Q1
2 4 3 2
7 1 6 9
6 4 6 4
7 7 5 = R1
7 2
5 = R0
Number Systems and Data Representation 57

Step 2 6 = Q3 à Step 4 0 = Q4
Base = 8 4 8 = Q2 Base = 8 6 = Q3
4 8 0
0 = R2 6 = R3

Stop only when the Quotient is zero.

Record the answer using remainders, starting in the column on the right end.
3 2 1 0 Position Number
R3 R2 R1 R0 ß Record remainders from right to left.
6 0 5 5 Remainders in decimal.
6 0 5 5 Final answer. Remainders in octal.

The answer is 60558 = O6055 = 6055Q = 311710 .

Notice the shape of the capital letter O (Oh) is slightly different than the number 0
(zero). Some printers put a slash through the number zero. The slashed zero is common
world wide, except in the United States. It is common also in communications and in the
military.
Converting Decimal Numbers (Base 10) to Base 4
In the base 4 example, an additional goal is to recognize that it does not matter
what symbols are used in a number system as long as you know the collating sequence.
In this example, the arbitrary symbol set {∃,∇,∆,Σ} is used. You can make up your own
symbols as long as you define the order.
To convert a decimal number to another base
1) Divide the decimal number by the base and determine the quotient Q and
remainder R.
2) Divide the resulting quotient by the base and determine its quotient Q and
remainder R.
3) Repeat step 2 until the quotient Q is zero.
4) Write the remainders in the new base beginning with the last remainder
computed. …R9 R8 R7 R6 R5 R4 R3 R2 R1 R0
5) The result from step 4 is the number converted to the new base.
6) Substitute the symbol used in the symbol set.
Number Systems and Data Representation 58

Example: Convert 3117 to {∃, ∇, ∆, Σ}

Step 0 7 7 9 = Q1 à Step 1 1 9 4 = Q2
Base = 4 3 1 1 7 = Q0 Base = 4 7 7 9 = Q1
2 8 4
3 1 3 7
2 8 3 6
3 7 1 9
3 6 1 6
1 = R0 3 = R1
= ∇4 = Σ4

Step 2 4 8 = Q3 à Step 3 1 2 = Q4
Base = 4 1 9 4 = Q2 Base = 4 4 8 = Q3
1 6 4
3 4 0 8
3 2 8
2 = R2 0 = R3
= ∆4 = ∃4

Step 4 3 = Q5 à Step 5 0 = Q6
Base = 4 1 2 = Q4 Base = 4 3 = Q5
1 2 0
0 = R4 3 = R5
= ∃4 = Σ4

Stop only when the Quotient is zero. Record the answer using remainders, starting in the
column on the right end.
5 4 3 2 1 0 Position Number
R5 R4 R3 R2 R1 R0 ß Record remainders from right to left.
Σ ∃ ∃ ∆ Σ ∇ 4 Final answer.
Checklist for Converting from Decimal to Another Base Number System
• Select correct base.
• Divide by base.
• Remainder becomes a digit.
• Convert the remainder to a digit of the target number system base.
• Divide into new quotient. Do not divide into remainder.
• Don’t stop dividing until quotient is zero.
• Build the final answer with the position-zero remainder at the right end.
• Identify number system base with correct notation. Allow any of the styles.
o Binary: % nnnn nnnn2 nnnnB nnnnb
o Octal: Ommmm mmmm8 mmmmQ
o Hexadecimal: OxNNNN NNNN16 NNNNH NNNNh
• Final answer correct.

You can check you answer by converting the answer back to decimal.
Number Systems and Data Representation 59

Conversion From Decimal Problems

The purposes of this homework set are to give you experience in doing
conversions, and to compare conversions of the same starting number in different bases.
You will learn it is easier to convert from decimal to an intermediate large base, and then
use table look-up to convert from the intermediate base to a smaller final base. This
procedure results in fewer mistakes by reducing the number of divisions or
multiplications.
Problem B1. Convert from decimal to binary. Show all your work. Summarize your
answers in Table 27 below.
a. Convert 31 from decimal to binary. Record the answer in column A, row 1.
b. Convert 99 from decimal to binary. Record the answer in column A, row 2.
c. Convert 165 from decimal to binary. Record the answer in column A, row 3.
d. Convert 169 from decimal to binary. Record the answer in column A, row 4.
Problem B2. Convert from decimal to hexadecimal. Show all your work. Summarize
your answers in Table 27 below.
a. Convert 257 from decimal to hexadecimal. Record the answer in column B,
row 5.
b. Convert 445 from decimal to hexadecimal. Record the answer in column B,
row 6.
c. Convert 479 from decimal to hexadecimal. Record the answer in column B,
row 7.
d. Convert 708 from decimal to hexadecimal. Record the answer in column B,
row 8.
Learning goal for problems 3 and 4: After doing the conversions, observe that
you can convert from decimal to binary by first converting from decimal to hexadecimal,
and then from hexadecimal to binary. This is a recommended shortcut for manual
conversions. Similarly, you can convert from decimal to hexadecimal by first converting
from decimal to binary, and then from binary to hexadecimal.
Problem B3. Use table look-up to convert from binary to hexadecimal. Summarize your
answers in Table 27 below.
a. Convert the binary number found in column A, row 1 to hexadecimal using
table look-up. Record the answer in column B, row 1.
b. Convert the binary number found in column A, row 2 to hexadecimal using
table look-up. Record the answer in column B, row 2.
c. Convert the binary number found in column A, row 3 to hexadecimal using
table look-up. Record the answer in column B, row 3.
d. Convert the binary number found in column A, row 4 to hexadecimal using
table look-up. Record the answer in column B, row 4.
Number Systems and Data Representation 60

Problem B4. Use table look-up to from hexadecimal to binary. Summarize your
answers in Table 27 below.
a. Convert the hexadecimal number found in column B, row 5 to binary using
table look-up. Record the answer in column A, row 5.
b. Convert the hexadecimal number found in column B, row 6 to binary using
table look-up. Record the answer in column A, row 6.
c. Convert the hexadecimal number found in column B, row 7 to binary using
table look-up. Record the answer in column A, row 7.
d. Convert the hexadecimal number found in column B, row 8 to binary using
table look-up. Record the answer in column A, row 8.
Problem B5. Convert from decimal to base 4. Summarize your answers in Table 27
below.
a. Convert 31 from decimal to base 4. Record the answer in column C, row 1.
b. Convert 99 from decimal to base 4. Record the answer in column C, row 2.
c. Convert 165 from decimal to base 4. Record the answer in column C, row 3.
d. Convert 169 from decimal to base 4. Record the answer in column C, row 4.
Problem B6. Convert from decimal to octal (base 8). Summarize your answers in Table
27 below.
a. Convert 31 from decimal to octal. Record the answer in column D, row 1.
b. Convert 99 from decimal to octal. Record the answer in column D, row 2.
c. Convert 165 from decimal to octal. Record the answer in column D, row 3.
d. Convert 169 from decimal to octal. Record the answer in column D, row 4.
Conversion Between Base 4, Binary, and Hexadecimal
Counting in base 4 is like counting in other number systems, except you have
only 4 digits to count with, {0, 1, 2, 3}.
• Each base-4 digit corresponds exactly to two binary digits.
• Each hexadecimal digit corresponds exactly to two base-4 digits.
Do conversions by inspection (table look-up).
Number Systems and Data Representation 61

Table 25. Equivalent Numbers in Base 4 and Octal

Decimal Hexadecimal Binary

Base 10 Base 16 Base 2 Base 4
0 0 0000 00
1 1 0001 01
2 2 0010 02
3 3 0011 03
4 4 0100 10
5 5 0101 11
6 6 0110 12
7 7 0111 13
8 8 1000 20
9 9 1001 21
10 A 1010 22
11 B 1011 23
12 C 1100 30
13 D 1101 31
14 E 1110 32
15 F 1111 33
Examples
a. Convert FAD2BH to base-4.
Hexadecimal F A D 2 B
Base 4 33 22 31 02 23
The answer is 33223102234
b. Convert %110010011 to base-4. Note that you need to pad the binary number
on the left.
Binary 01 10 01 00 11
Base 4 1 2 1 0 3
The answer is 121034
c. Convert 3232231014 from base-4 to hexadecimal. Note that you need to pad
the base-4 number on the left.
Base-4 03 23 22 31 01
Hexadecimal 3 B A D 1
The answer is 0x3BAD1
d. Convert 211034 from base-4 to binary.
Base-4 2 1 1 0 3
Binary 10 01 01 00 11
The answer is 1001010011B
Number Systems and Data Representation 62

Problem B7. Do conversions from hexadecimal to base-4 by inspection (using table

look- up). Summarize your answers in Table 27 below.
a. Convert the hexadecimal number in column B, row 5 to base-4 by inspection.
Record the answer in column C, row 5.
b. Convert the hexadecimal number in column B, row 6 to base-4 by inspection.
Record the answer in column C, row 6.
c. Convert the hexadecimal number in column B, row 7 to base-4 by inspection.
Record the answer in column C, row 7.
d. Convert the hexadecimal number in column B, row 8 to base-4 by inspection.
Record the answer in column C, row 8.
Conversion Between Octal and Binary
Octal is base 8. Counting in base 8 is like counting in other number systems,
except you have only 8 digits to count with, {0, 1, 2, 3, 5, 6, 7}. Each octal digit
corresponds exactly to three binary digits. Do conversions by inspection (table look- up).
Table 26. Equivalent Numbers in Octal

Decimal Binary Octal

Base 10 Base 2 Base 8
0 000 0
1 001 1
2 010 2
3 011 3
4 100 4
5 101 5
6 110 6
7 111 7
To convert from binary to octal, the first step is to group bits into groups of three
binary digits each. It is important to start the grouping from the right end. To convert
from octal to binary, replace each octal digit with three binary digits.
Examples
1. Convert 111110001011002 to octal. Note that the binary number has to be
padded on the left end to finish a group of 3 bits.
Binary 011 111 000 101 100
Octal 3 7 0 5 4
The answer is 370548
2. Convert 736258 to binary
Octal 1 3 5 7 2
Binary 001 011 101 111 010
The answer is %1011101111010
Number Systems and Data Representation 63

Problem B8. Each group of three binary digits is exactly one octal digit. Summarize
your answers in Table 27 below.
a. Convert the binary number in column A, row 5, to octal by inspection.
Record the answer in column D, row 5.
b. Convert the binary number in column A, row 6, to octal by inspection.
Record the answer in column D, row 6.
c. Convert the binary number in column A, row 7, to octal by inspection.
Record the answer in column D, row 7.
d. Convert the binary number in column A, row 8, to octal by inspection.
Record the answer in column D, row 8.
Table 27. Decimal Conversion Solutions

Part à A B C D
Row Decimal Binary Hexadecimal Base 4 Octal
1 31
2 99
3 165
4 169
5 257
6 445
7 479
8 708

NUMBER SYSTEM CONVERSION PROBLEMS

Problem 1. Convert the following decimal numbers to binary. Show all your work.
a. 535 b. 240 c. 61 d. 202
Problem 2. Convert the following numbers to hexadecimal. Show all your work.
a. 37 b. 222 c. 168 d. 959
Problem 3. Convert the following numbers to decimal. Show all your work.
a. 110101112 b. %01111001 c. EB16 d. Ox7F
Number Systems and Data Representation 64

Strings
A string is an ordered sequence of characters. Call the first character of a string
the head, and the last character of a string the tail. (This terminology is not standardized.)
One property of a string is its length. For example, the le ngth of the string, “My favorite
string?” is 19. Spaces count as characters. A very special string is the “null string”,
which is a string of zero length, such as “”. Notice that there are no characters between
the quotation marks.
A string can be eithe r a fixed length string or a variable length string. A fixed
length string is always the same length, even if all its characters are blanks. Fixed length
strings are efficiently treated as an array, and they are easy to sort.
Variable length strings are handled in one of two ways. One way is to record the
string length as one field of the string. The other way is to designate one character as a
“null character” that can be used to identify the end of a string. The internal code used is
often just the byte consisting of all zero bits, which is the null character. The null
character does not count as one of the characters when computing the length of a string.
A string that consists only of the null character is a string of zero length, and it is called
the “null string”.
For a string that includes a “length” field, that field is not shown to the outside
world. The length field is the first field retrieved from memory when working with that
string. This method allows use of a block transfer instruction from memory to the CPU,
which is a fast way of moving data. The number is placed into a register and is used as a
counter. The maximum length allowed is limited by the largest number that can fit into
the length field. For many applications, this is very efficient. It is common for text fields
in a data base to be limited to 255 characters. This leaves one byte out of 256 to be used
as a length, and a single byte can hold numbers in the range [0,255]. It is not a useful
string format if you need very long strings, such as the length of a book, or the source
code for a computer program.
Example 1: “This string is here.” The first field of this string contains the length
of the string, which is “20” for this example. The symbol is the customary symbol
used to identify a blank when it must be graphically displayed.
20 T h i s s t r i n g i s h e r e .
Example 2: “This string is here.” This is a string of length 20. The last symbol is
the lower case Greek letter lambda (λ), which corresponds to the lower case Latin letter
“l” (el). It is customary to use this symbol to represent the null character when
documenting the end of a string. The actual code stored usually is a byte of all zero bits,
not the character code for λ.
T h i s s t r i n g i s h e r e . λ
Number Systems and Data Representation 65

A language using this convention can store a sequence of strings, separating them
by λ. For example: Go Navy! Beat Army! (If you are Army, I expect you to reverse the
example.)
G o N a v y ! λ B e a t A r m y ! λ
String Operations
Operations upon strings, and symbols used to represent these operations, are
defined by the programming language being used. At the hardware level, operations that
affect strings are fetch, store, block move, comparison.

Length
The length operator gives the length of a string, not including the null character (if
it has one). Often, this operation is abbreviated as LEN. The result is an integer value.
Example 1:
x = H o n o r λ
LEN(x) = 5
Example 2:
y = L o v e y o u r n e i g h b o r λ
LEN(y) = 18

Concatenation
Concatenation is the operation of taking two separate strings, x and y, and making
one string z by appending the second string to the first string. Example:
x = B o y λ
y = S c o u t s λ
Then z = x & y is
z = B o y S c o u t s λ

Truncation
Truncation is the operation of taking only the first N characters of a string,
discarding all remaining characters. The operation requires specifying the length N. If
the string is shorter than N to begin with, the string is unaltered. When N is counted from
the left side, the operation is often called LEFT. When N is counted from the right side,
the operation is often called RIGHT.
Number Systems and Data Representation 66

LEFT Truncation
Example 1: Let
x = r a i n i n g λ
Let y = LEFT(x,4)
y = r a i n λ
Example 2: Let
x = E l i j a h λ
Let y = LEFT(x,8)
y = E l i j a h λ
Since x is shorter than the truncation length, 8, nothing is taken away from x. Therefore,
y = x.

RIGHT Truncation
Example 1: Consider the palindrome
x = M A D A M I M A D A M λ
Let y = RIGHT(x,4)
y = A D A M λ
Example 2: Let
x = C a t λ
Let y = RIGHT(x,4)
y = C a t λ
Since x < 4, no truncation takes place. y = x.

Extraction
Extraction fetches N characters from a string, beginning with a specified
beginning point k. The function is often called MID, for “middle” of a string. An error
condition exists if the operation cannot be performed.
Example 1: Let the MID function have the format
MID(x=string,k=start,N=size), and let
x = M A D A M I M A D A M λ
Let y = MID(x,6,2). Then
y = I M λ
Number Systems and Data Representation 67

Comparison
Strings are compared one character at a time. The primitive comparison operators
are “equal” ( = ) and “not equal” ( ≠ ). For two strings to be equal, they must have the
same length, and each character must be equal. A common format is:
“If x = y”
The outcome of a test is “True” or “False”.
Many computer languages permit unequal comparisons:
Table 28. Comparison Operators

Equal ( = ) Not equal ( ≠ )

Less than ( < ) Less than or equal ( ≤ )
Greater than ( > ) Greater than or equal ( ≥ )
The binary value of the character code is used for the comparison. Two strings
are compared one character at a time. The default sequence is to begin the comparison at
the string head, and continue toward the string tail until the comparison result is
determined. The comparison of characters terminates when the first unequal comparison
is found, or when the end of the shorter string is found.

String Direction
Sometimes the notion of string direction is meaningful, such as strings containing
human language. The direction of a string is context dependent, such as in a crossword
puzzle. European languages are normally read from left to right. Some languages, such
as Hebrew and Arabic, are read from right to left. Some languages may be written from
top to bottom, such as Chinese and Japanese. String direction is not usually stored with a
string, but this can be included in the data structure if necessary.
Arabic influence is also seen in the way numbers are written, which have the
reference point for whole numbers on the right end. In English, we determine the size
and partitioning of the number by first scanning the number starting at the radix (decimal)
point, and we proceed from right to left in order to select the proper words to describe the
number.

Inequality Comparisons
The inequality comparisons are used for searching and sorting. Most sorting in
business data processing is done on character data.
Example 1:
x = B l u e λ
y = B l e w λ
If x > y
The last comparison performed is x=u versus y=e. Conclude: x > y. Therefore, the
answer is “True”.
Number Systems and Data Representation 68

Example 2:
x = S w i m λ
y = S w i m m i n g λ
If x > y
Only 4 comparisons are made. Conclude: x < y because x and y are equal fo r the first 4
characters and LEN(x) < LEN(y). The answer is “False”.

String Search
The string search is a special case of the comparison. This operation searches
string x to see if string y is contained in it. If the string is found, the starting position in
string x is reported. This is sometimes called the FIND operation. The search stops with
the first match found. Some versions allow specification of a search starting point, N.
For example, k = FIND(x,y,N). If the search string is not found, the returned value
depends on the language. Often, the value returned is a negative number.
Example 1: Let
x = B l e s s e d a r e t h e p o o r λ
y = l e s s λ

k = FIND(x,y)
The result is k = 2.
Some search programs permit restriction to locating only whole words. If that
restriction were applied in the above example, the result would be k = -1, indicating that
the search string was not found in the target string. Restricting a search to whole words is
more difficult. After locating a candidate match, the target must be examined to
determine if the matched string was merely contained in a longer word, or was a
complete word. What are indicators that a whole word was found?
Table 29. Detecting a Whole Word in a String

Left Side of Target Match Right Side of Target Match

Preceded by a blank or at the head of the Followed by a blank or end-of-string code.
string.
Preceded by a quotation mark that is Followed by punctuation mark that is
preceded by a blank. followed by a blank or end-of-string code.
Punctuation marks are: period, comma,
colon, semicolon, quotation mark.
Preceded by a parenthesis or bracket. Followed by a parenthesis or bracket.
Preceded by a forward or backward slash. Followed by a forward or backward slash.
Preceded by a tab or end-of- line control Followed by a tab or end-of- line control
code. code.
Number Systems and Data Representation 69

This is not an exhaustive list. The search is further complicated by other factors:
• exclude or skip over non-printing control codes, such as used to control the font.
• permit matching to upper and lower case letters.
• permit use of wild card characters in the search string. A wild card character can
specify an exact unknown character position (often indicated in the search string
by a question mark, ?), or an indefinite number of unknown characters at a
particular position (often indicated in the search string by an asterisk, *).
Example 2: Let
x = B l e s s e d a r e t h e p o o r λ
y = e λ
k = FIND(x,y,13)
The result is k = 15
Example 3: Let
x = B l e s s e d a r e t h e p o o r λ
y = i s λ

k = FIND(x,y)
The result is k = -1

Compound Expressions
It is common to use the results of one operation as a parameter or string argument of
another expression. This is composition of functions. As in evaluation of algebra
expressions, evaluate the innermost function(s) first.
Example 1: Find Again and extract.
x = B e b r a v e . B e s t r o n g .
B e r e a d y . B e a l e r t . λ
y = B e λ
Find:
z = MID( x, FIND(x, y, FIND(x,y) + LEN(y) ) + LEN(y) + 1, 6 )
z =
Solution:
n = FIND(x, y, FIND(x,y) + LEN(y) )
j = FIND(x,y) = 1
k = LEN(y) = 2
n = FIND(x, y, j + k )
n = FIND(x, y, 1 + 2) = FIND(x, y, 3) = 11
Number Systems and Data Representation 70

z = MID( x, 11 + LEN(y) + 1, 6 )
z = MID( x, 11 + 2 + 1, 6 ) = MID( x, 14, 6 )
z = s t r o n g λ

String Function Problems

Problem 1. Length. Let
x = B l e s s e d a r e t h e p o o r λ
Find k = LEN(x) =
Problem 2. Concatenation. Let
x = B e λ
y = P r e p a r e d λ
Concatenate x and y. z = x & y.
z =
Problem 3. Left. Let
x = B l e s s e d a r e t h e p o o r λ
Find y = Left(x,5).
y =
Problem 4. Right. Le t
x = B l e s s e d a r e t h e p o o r λ
Find y = Right(x,8).
y =
Problem 5. MID. Let
x = B l e s s e d a r e t h e p o o r λ
Find y = MID(x,14,2)
y =
Problem 6. Comparison. Let
x = B l e s s e d a r e t h e p o o r λ
y = B l e s s e d a r e t h e m e e k λ
If x < y
Give answer as “True” or “False”.
Problem 7. Search.
x = B e h o l d , I b r i n g g l a d λ
y = h o l d λ
Number Systems and Data Representation 71

Find k = FIND(x,y)
Problem 8. Search.
x = t i d i n g s . Y o u s h a l l λ
y = y o u λ
Search only for whole words; permit upper or lower case.
Find k = FIND(x,y)
Problem 9. Compound Function
x = B l e s s e d a r e t h e p o o r λ
y = P r e p a r e d λ
Find:
z = LEFT(x,5) & MID(x,13,3) & y
z =
Problem 10. Compound Function
x = B e b r a v e . B e s t r o n g .
B e r e a d y . B e a l e r t . λ
y = B e λ
Find:
z = RIGHT(x, LEN(x)–LEN(y)–FIND(x,y, FIND(x,y, FIND(x,y)+LEN(y))+LEN(y)))
z =
Number Systems and Data Representation 72

Storage of Multibyte Words: Big Endian, Little Endian

IBM mainframes and Intel processors store multibyte data in different orders.
Consider a 4-byte word written in hexadecimal: AB CD EF 01.
Table 30. Little, Middle, and Big Endian Definitions

Big Endian (Big End In) AB CD EF 01 IBM System 390

|| AIX
(Most Sun SPARC
Significant Solaris
Byte) IBM PowerPC
MacIntosh
Lowest HP Precision
Address Architecture HP-UX
SCSI devices
network protocols
TCP/IP packets
W3C
Little Endian (Little End In) 01 EF CD AB Intel 486, Pentium
|| DEC Alpha RISC
(Least Microsoft Windows
Significant IBM OS/2
Byte) VAX/VMS
Digital Unix
Lowest Intel ABI
Address PCI bus
Middle Endian BA DC FE 10 Occasionally used
A and B are with Packed Decimal
reversed. (BCD)
Little Endian format could have some advantages on a stack-oriented machine.
By fetching the lowest remaining order byte next and placing it on a stack, the stack ends
up with the data in Big Endian order after the transfer is complete. The most significant
byte is on top.
If you share data between machines, data order becomes critical. It affects
storage of all multiple byte data, whether to primary storage (memory), or secondary
storage (tape, disk). This includes short integers, long integers, floating point, double
precision floating point, extended precision floating point, complex, double precision
complex, extended precision complex, and string data. It also affects details of hardware
interfaces. For example, MacIntosh has a Big Endian 16-bit RGB pixel format. There
also exists a Little Endian 16-bit RGB pixel format.
These differences affect files, data structures, and arrays. It affects the storage of
bytes for Unicode characters. It affects even data operations, such as searching and
sorting. These require the ability to make comparisons.
Number Systems and Data Representation 73

Logic Operations
Logic operations accept one or two binary inputs and produce one binary output.
Each operation is defined through use of a table. The table is called a truth table because
a zero (0) is called a False (F) value, and a one (1) is called a True (T) value. Logic
operations are also called switching operations, combinational logic, Boolean logic, or
Boolean algebra. The rules for logic were developed by English mathematician George
Boole (1815 – 1864).
The logic operations are used in Boolean algebra, logic, and set theory. Boolean
algebra is used by binary computers. Logic is used in law, philosophy, and mathematics.
Set theory is used in statistics and mathematics. Each of these disciplines has its own
names and symbols for these operations. The rules of algebra are the same.
The most basic logic operations are the AND, OR, and NOT functions. When
these functions are implemented in electronics, they are called “gates”. All other
functions can be synthesized from these. These are already familiar to you through your
use of every day language.
For multi-bit variables, the operation is done for each bit position, independently
of all other bit positions. It is important to right-align variables before beginning an
operation. There are not carries or borrows, unlike arithmetic operations.
The symbols for AND, OR, and NOT which are given below are used in
equations. The logical AND function in Boolean logic corresponds abstractly to the
multiplication function in algebra. The lo gical OR function in Boolean logic corresponds
abstractly to the addition function in algebra. The logical NOT function is shown by a
bar above the variable. American National Standards Institute (ANSI) Circuit symbols
from OrCAD Capture are shown. 32
AND Operation
Suppose your friend says he is going swimming and flying. That tells you he is
going to do both activities. If you ask for a peanut butter and jelly sandwich, you want
peanut butter, and you also want jelly. If you are asked if you got peanut butter and jelly,
and you actually got both, the answer is True. If you did not actually receive both, the
answer is False. The AND operation is summarized in a truth table. Let peanut butter be
represented by the variable X, and let jelly be represented by the variable Y.
Table 31. Truth Table for AND Operator, Version 1

AND X=F=0 X=T=1

Y=F=0 F=0 F=0
Y=T=1 F=0 T=1
The logic symbol for the AND function is the arithmetic multiplication symbol.
The following are equivalent:

32
OrCAD Capture Demonstration 9.10.93 (26 April 1999).
Number Systems and Data Representation 74

Equation 1. x AND y = x • y = xy

The ANSI standard circuit symbol for the AND function is: The
inputs are applied at pins 1 and 2 on the left side, as the symbol is presently oriented.
The output is taken at point 3 on the right side. Note the straight edge on the input end of
the symbol. The output end is rounded.
In this truth table, the inputs to the AND operator are given in the top row and the
left column. The outcome is given in the othe r cells.
There is another version of this same table.
Table 32. Truth Table for AND Operator, Version 2

Inputs Output
X Y X AND Y
0 0 0
0 1 0
1 0 0
1 1 1
There is one more version of this same table. This version will be extended with other
operators.
Table 33. Truth Table for AND Operator, Version 3

Inputs
Function XY = 00 XY = 01 XY = 10 XY = 11
X AND Y 0 0 0 1
Example: Perform the operation 1100 AND 1010.

X 1 1 0 0
Y 1 0 1 0
X AND Y 1 0 0 0

AND Applications
The AND operation is used as a mask to isolate one or more bits from a larger
collection of bits, such as extracting a subnet address in a network router application, or
obtaining the binary value of a decimal digit represented by its ASCII, EBCDIC, or
UNICODE character. It is also used to clear one or more bits in a larger collection of
bits, as might be done to convert a lower case ASCII or UNICODE character to upper
case, converting an upper case EBCDIC character to lower case, to clear a bit to zero in a
compact data record, or to clear the sign bit to zero.
Example: Convert EBCDIC decimal character to binary.
Observe that the binary value is contained in the lower four bits of the EBCDIC
code. For example, the EBCDIC code for the character printed as “9” is F9H. In binary,
this code is 11111001B. To obtain the binary value of 9, do the AND operation using
the mask of 00001111B = 0FH. The operation F9H AND 0FH = 09H in binary is
computed bit by bit:
Number Systems and Data Representation 75

“9” = F9H = 1 1 1 1 1 0 0 1
Mask = 0FH = 0 0 0 0 1 1 1 1
Binary 9 = 09H = 0 0 0 0 1 0 0 1

Example: Convert an upper case EBCDIC letter to lower case.

Observe that bit #6 (where the right- most bit is bit #0) is “1” in the upper case
character, but is “0” in the corresponding lower case character. For example, the
EBCDIC code for the upper case character “J” is D1H = 11010001B. The EBCDIC
code for the lower case character “j” is 91H = 10010001B. To clear bit #6, use the AND
operation with the mask BFH. This is: D1 H AND BFH = 91H.

EBCDIC “J” = D1H = 1 1 0 1 0 0 0 1

Mask = BF H = 1 0 1 1 1 1 1 1
EBCDIC “j” = 91H = 1 0 0 1 0 0 0 1

OR Operation
The operation “x OR y” is True if either x or y is True. This includes the case
that x and y are both True. If you asked for red or blue crayons, this would be satisfied
by getting either a red crayon, a blue crayon, or both a red and a blue crayon. If you are
really thirsty, you will take Pepsi OR Coke. The truth table is:
Table 34. Truth Table for OR Operator, Version 1

OR X=F=0 X=T=1
Y=F=0 F=0 T=1
Y=T=1 T=1 T=1
The logic symbol for the OR function is the arithmetic addition symbol. The
following are equivalent:
Equation 2. x OR y = x + y

The ANSI standard circuit symbol for the OR function is: The inputs are
applied at pins 1 and 2 on the left side, as the symbol is presently oriented. The output is
taken at point 3 on the right side. Note the concave curve on the input side of the symbol.
The output end looks like a bullet with a blunt point.
An alternate form for this same truth table is:
Table 35. Truth Table for OR Operator, Versi on 2

Inputs
Function XY = 00 XY = 01 XY = 10 XY = 11
X OR Y 0 1 1 1
Number Systems and Data Representation 76

Example: Perform the operation 1100 OR 1010.

X 1 1 0 0
Y 1 0 1 0
X OR Y 1 1 1 0

OR Applications
The OR operation is used as a mask to set one or more bits in a collection of bits,
such as obtaining the decimal ASCII, EBCDIC, or UNICODE cha racter of a binary value
between 0 and 9. It is also used to set one or more bits in a larger collection of bits, as
might be done to convert an upper case ASCII or UNICODE character to lower case,
converting an lower case EBCDIC character to upper case, to set a bit to one in a
compact data record, or to set the sign bit to one.
Example: Convert the binary value 00000111B to the ASCII code for the
character “7”, which is 37H.
To do this, use the OR operation to insert the code 0011 in the upper half of the
byte, using the mask 30H. 07H OR 30H = 37H.

710 = 0 0 0 0 0 1 1 1
Mask = 30H = 0 0 1 1 0 0 0 0
ASCII “7” = 37H = 0 0 1 1 0 1 1 1

Example: Convert an EBCDIC character from lower case to upper case.

For example, convert EBCDIC character “t” to upper case “T”. Note that bit #6 is
turned off (zero) in the lower case letter, and turned on (one) in the upper case letter. The
hexadecimal EBCDIC code for the character “t” is A3H. The code for “T” is E3 H. The
mask for conversion is 40H. The conversion is done using the logical OR operation,
A3H OR 40H = E3H.

EBCDIC “t” = A3H = 1 0 1 0 0 0 1 1

Mask = 40H = 0 1 0 0 0 0 0 0
EBCDIC “T” = E3H = 1 1 1 0 0 0 1 1

NOT Operation
The NOT operator uses only one variable as an input, and is therefore called a
unary operator. Another name for this is the complement operator. Whatever value the
input has, the output is just the opposite. When applied to a full byte, all the zeros of the
byte get changed to ones, and all the previous ones get changed to zeros. When a kid
says, “I love school … NOT!”, the kid is negating the sentence just spoken.
The truth table for the NOT operator is:
Table 36. Truth Table for NOT Operator

OR X=F=0 X=T=1
NOT X T=1 F=0
Number Systems and Data Representation 77

There are three logic symbols used for the NOT operator. One logic symbol for the
NOT function is a bar over the variable. Another is an apostrophe following the variable.
The third is a horizontal bar with the right end bent down, like a golf club. The following
are equivalent:
Equation 3. NOT x = x = x ' = ¬x

The ANSI standard circuit symbol for the NOT function is: The input is
applied at pin 1 on the left side, as the symbol is presently oriented. The output is taken
at point 2 on the right side. Note the small circle on the output side of the symbol. The
body of the symbol is an equilateral triangle.
Example: Perform the operation NOT 1011.

X 1 0 1 1
NOT X 0 1 0 0

NOT Applications
The NOT operator can be used to treat a bit as a toggle switch. Suppose a push
button light switch is “off”. Pushing the switch changes it to “on”. Pushing the switch
again changes it back to “off”. Let a bit represent the state of the switch. Let “off” be
represented by zero, and let “on” be represented by one. The NOT operator toggles the
bit between zero and one.
Applying the NOT operator to an integer word results in the one’s
complement of that word. It may be useful to convert a hexadecimal
representation to binary before taking the complement, and then convert the
result back to hexadecimal.
Example: Find the one’s complement of A0F3H. NOT A0F3H = 5F0CH.

A0F3H = 1 0 1 0 0 0 0 0 1 1 1 1 0 0 1 1
NOT A0F3H = 0 1 0 1 1 1 1 1 0 0 0 0 1 1 0 0
Logic Operation Problems
Perform the following operations.
Problem 1. D6H AND ADH
Problem 2. A5H OR 61H
Problem 3. NOT 0FH
Boolean Algebra Theorems
Properties of Boolean algebra are summarized in the below table from Sloan. 33

33
Martha E. Sloan, Computer Hardware and Organization: An Introduction, Second Edition, Science
Research Associates, Inc (1983). pp 26-27.
Number Systems and Data Representation 78

Table 37. Boolean Algebra Properties

Property OR Version AND Version

Involute x→ x
Identity x+0=x x• 1 = x
Null element x+1=1 x• 0 = 0
Idempotent x+x=x x• x = x
Complement x+ x=1 x•x = 0
Commutative x+y=y+x x• y=y • x
Associative x + (y + z) = (x + y) + z x • (y • z) = (x • y) • z
Distributive x + (y • z) = (x + y) • (x + z) x • (y + z) = (x • y) + (x • z)
Absorption x + (x • y) = x x • (x + y) = x
Absorption ( )
x + x•y = x+ y ( )
x• x+ y = x• y
( x • y ) + (x • z ) + ( y • z ) ( x + y ) • (x + z )( y + z )
= ( x • y ) + (x • z ) = ( x + y ) • (x + z )
Consensus

DeMorgan’s
x + y = x•y x• y = x + y
Laws
DeMorgan’s
Laws
( x • z ) + ( y • z ) = (x + z ) • (y + z ) ( x + z ) • ( y + z ) = (x • z ) + ( y • z )
Exclusive OR (XOR) Operation
The Exclusive OR operation x XOR y is True if either x or y, but not both x and
y, are True. The word “or” in “I will eat with chopsticks or a fork” probably means
XOR. The language of logical operations is unambiguous, unlike English. The truth
table is
Table 38. Truth Table for XOR Operator, Version 1

XOR X=F=0 X=T=1

Y=F=0 F=0 T=1
Y=T=1 T=1 T=0
The logic symbol for the XOR function is the arithmetic addition symbol inside a small
circle. The following are equivalent:
Equation 4 x XOR y = x ⊕ y = ( x + y ) • x • y( )

The ANSI standard circuit symbol for the XOR function is: This is
similar to the OR symbol. The inputs are applied at pins 1 and 2 on the left side, as the
symbol is presently oriented. The output is taken at point 3 on the right side. Note the
concave double curve on the input side of the symbol.
Number Systems and Data Representation 79

An alternative truth table is

Table 39. Truth Table for XOR Operator, Version 2

Inputs
Function XY = 00 XY = 01 XY = 10 XY = 11
X XOR Y 0 1 1 0
Contrast the XOR operator with the OR operator. The difference between the two is in
the outcome when X=1 and Y=1.
Example: Perform the operation 1100 XOR 1010.
X 1 1 0 0
Y 1 0 1 0
X XOR Y 0 1 1 0
Real Logic Components
It is easier to manufacture another family of logic components, the NAND and
NOR functions. The NAND is the complement of the AND function, which is the NOT
AND function. The NOR is the complement of the OR function, which is the NOT OR
function. The NOT function can be constructed from either the NAND function or the
NOR function. The circuit symbols are like the AND and OR functions, with the
addition of a small circle at the output end. All logic circuits can be made with NAND
and NOR gates. This will be demonstrated. In general, it is easier to learn the theory in
terms of AND, OR, and NOT. It is easier to build chips using NAND and NOR.
NAND Operation
The operation “x NAND y” is True if either x or y is False. The truth table is:
Table 40. Truth Table for NAND Operator, Version 1

NAND X=F=0 X=T=1

Y=F=0 T=1 T=1
Y=T=1 T=1 F=0

The logic symbol for the NAND function is the up arrow, . The following are
equivalent:
Equation 5. x NAND y = x ↑ y = x • y

The ANSI standard circuit symbol for the NAND function is: This is like
the AND function with a circle at the output end. The inputs are applied at pins 1 and 2
on the left side, as the symbol is presently oriented. The output is taken at point 3 on the
right side.
Number Systems and Data Representation 80

An alternate form for this same truth table is:

Table 41. Truth Table for NAND Operator, Version 2

Inputs
Function XY = 00 XY = 01 XY = 10 XY = 11
X NAND Y 1 1 1 0
Notice that the NAND gate can be used to construct the NOT function because
1 NAND y = NOT y.
Example: Perform the operation 1100 NAND 1010.
X 1 1 0 0
Y 1 0 1 0
X NAND Y 0 1 1 1
NOR Operation
The operation “x NOR y” is True if both x and y are False. The truth table is:

NOR X=F=0 X=T=1

Y=F=0 T=1 T=0
Y=T=1 T=0 F=0
Table 42. Truth Table for NOR Operator, Version 1

The logic symbol for the NOR function is the “down” arrow, . The following
are equivalent:
Equation 6. x NOR y = x ↓ y = x + y

The ANSI standard circuit symbol for the NOR function is: This is like
the OR function with a circle at the output end. The inputs are applied at pins 2 and 3 on
the left side, as the symbol is presently oriented. The output is taken at point 1 on the
right side.
An alternate form for this same truth table is:
Table 43. Truth Table for NOR Operator, Version 2

Inputs
Function XY = 00 XY = 01 XY = 10 XY = 11
X NOR Y 1 0 0 0
Notice that the NOR gate can be used to construct the NOT function because
0 NOR y = NOT y.
Example: Perform the operation 1100 NOR 1010.
X 1 1 0 0
Y 1 0 1 0
X NOR Y 0 0 0 1
Number Systems and Data Representation 81

Logic Operation Summary

There are 16 combinatorial functions of two variables. These are summarized in
the table below.
Table 44. Logic Operation Summary

Output for
XY Inputs
X X
Boolean Logic Set Theory XY XY
Y Y
Operator Symbol Operation Symbol Operator Symbol 00 01 10 11
ZERO 0 Null 0 0 0 0
AND • Conjunction X ∧Y Intersection X ∩Y 0 0 0 1
ONLY X Inhibit 0 0 1 0
X X 0 0 1 1
ONLY Y Inhibit 0 1 0 0
Y Y 0 1 0 1
Mod 2
XOR ⊕ 0 1 1 0
Addition
OR + Disjunction X ∨Y Union X ∪Y 0 1 1 1
NOR 1 0 0 0

Á
Equivalence,
COINCIDENCE
Tautology X ⇔Y 1 0 0 1

NOT Y Negation Y ' , ¬ Y , Y Complement Y ',Y ,Y C 1 0 1 0

Y
NOT ONLY Y Imply Y⇒X 1 0 1 1
X ' , ¬X , X X,X, X C
NOT X X Negation Complement 1 1 0 0
NOT ONLY X Imply X ⇒Y 1 1 0 1
NAND 1 1 1 0
ONE 1 Unity 1 1 1 1

Special definitions:
In the imply operation, X ⇒ Y , the variable on the left side is called the
antecedent, and the variable on the right side is called the consequent.
Tautology: X ⇔ Y Statement whose truth values are always true. This is
equivalent to ( X ⇒ Y ) ∧ (Y ⇒ X ) , or equivalently ( X ⇒ Y ) ∧ ( X ⇐ Y ) where the
implication arrow direction and order of variables are reversed in the second term.
Number Systems and Data Representation 82

Hexadecimal Addition
Hexadecimal addition is done just like decimal addition. The only difference is
that the addition table is larger.
Table 45. Hexadecimal Addition Table

0 1 2 3 4 5 6 7 8 9 A B C D E F
0 0 1 2 3 4 5 6 7 8 9 A B C D E F
1 1 2 3 4 5 6 7 8 9 A B C D E F 10
2 2 3 4 5 6 7 8 9 A B C D E F 10 11
3 3 4 5 6 7 8 9 A B C D E F 10 11 12
4 4 5 6 7 8 9 A B C D E F 10 11 12 13
5 5 6 7 8 9 A B C D E F 10 11 12 13 14
6 6 7 8 9 A B C D E F 10 11 12 13 14 15
7 7 8 9 A B C D E F 10 11 12 13 14 15 16
8 8 9 A B C D E F 10 11 12 13 14 15 16 17
9 9 A B C D E F 10 11 12 13 14 15 16 17 18
A A B C D E F 10 11 12 13 14 15 16 17 18 19
B B C D E F 10 11 12 13 14 15 16 17 18 19 1A
C C D E F 10 11 12 13 14 15 16 17 18 19 1A 1B
D D E F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C
E E F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D
F F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E
In the addition table above, the left digit of a two digit answer is a carry to the
next column for addition. Addition is only defined for pairs of numbers.
Example 1. Add the hexadecimal numbers 7FAC and FBD using a 16-bit
word. In decimal, this is 32684 + 4029 = 36713.

“Throw
Away” Carry
Column

1 1 1

Original
7 F A C
Number

Radix Add à
F B D
Complement

Final
8 F 6 9
Result
Number Systems and Data Representation 83

Do the addition beginning from the right hand side, in column 0. That is
important. Find C and D in the column and row headings of the addition table. Find
the entry in the table where those rows intersect. The result is 19. Record the 9 at the
bottom of the calculation table in the result row. Record the 1 at the top in the Carry
row.
Move left to column 1. Find A and B in the addition table column headings.
Find the entry in the table where those rows intersect. The result is 15. Record the 1 in
the carry row of column 2. Add the 5 to the carry (1) at the top of column 1. Record
the resulting 6 at the bottom in the result row.
Move left to column 2. Find F and F in the addition table column headings.
Find the entry in the table where those rows intersect. The result is 1E. Write the 1 in
the carry row of column 3. Add the E to the carry in column 2. The result is F.
Record the resulting F at the bottom in the result row.
Move left to column 3. Only 7 is available. Add 7 to the carry in column 3.
The result is 8. Record the resulting 8 at the bottom in the result row.
The answer is 8F6916 = 8F69H = Ox8F69. In decimal, this answer is 36713.
If any carry is generated when adding numbers in column 3, the carry goes to the
“Throw Away” column. This is because we assumed our word is only 16 bits long.
Anything beyond that is lost. This is called an “overflow” condition.
Example 2. Add FFFF + 0001 using a 16 bit word. In decimal, this is 255 + 1
= 256.

“Throw
Away” Carry
Column

1 1 1 1

Original
F F F F
Number

Radix Add à
1
Complement

Final
0 0 0 0
Result

The answer is 000016 = 0000H = Ox0000 = 010 . This is an example of

overflow.
Number Systems and Data Representation 84

Hexadecimal Addition Problems

Assume a 16-bit word. Perform the following additions. Record any bits you throw
away separately from the answer.
Problem 1. 8888H + 8888H
Problem 2. CAD8H + BCA1H
Problem 3. F3FAH + 51CH
Problem 4. 2CCDH + 5A1FH
Problem 5. 3333H + ACDAH
Problem 6. 710H + 553H
Number Systems and Data Representation 85

Binary Coded Decimal (BCD) Notation

Binary Coded Decimal (BCD) notation is a method of representing base-10
numbers in a computer exactly. Each decimal number is represented by 4 bits (nibble)
corresponding to the binary representation of that decimal number. For example, the
decimal number 978 is represented in two bytes in BCD as
0000:1001:0111:1000BCD
The attractive feature of BCD is that all decimal fractions used in commerce can
be represented internally in BCD exactly. This is not possible with binary. Because our
monetary system uses base 10 arithmetic, and our accounting standards are developed
using base 10 arithmetic, we expect to do accounting on computers and get the same
result we would get if we do it by hand.
BCD arithmetic is a natur al choice for use with calculators, and computer
applications, that operate on only decimal data.
Table 46. Decimal Fractions in Base 2, 16, 10, and BCD

Fraction Base 2 Base 16 Base 10 BCD

1/10 0.00011 0.19 0.1 0.0001
2/10 0.0011 0.3 0.2 0.0010
3/10 0.01001 0.4C 0.3 0.0011

4/10 0.0110 0.6 0.4 0.0100

5/10 0.1 0.8 0.5 0.0101

6/10 0.1001 0.9 0.6 0.0110
7/10 0.10110 0.B3 0.7 0.0111
8/10 0.1100 0.C 0.8 0.1000
9/10 0.11100 0.E6 0.9 0.1001

Conversion between decimal and BCD is as easy as conversion between binary

and hexadecimal. It is done by simple substitution. Each decimal digit produces four
BCD binary digits. Each group of four BCD binary digits produces one decimal digit.
Example 1. Convert 975.283 from decimal to BCD.
Decimal 9 7 5 .2 8 3
BCD 1001 0111 0101 .0010 1000 0011
Example 2. Convert 1001 0111 0011 .0101 1000 0110 from BCD to decimal.
BCD 1001 0111 0011 .0101 1000 0110
Decimal 9 7 3 .5 8 6
Number Systems and Data Representation 86

Not every group of 4 binary digits is a BCD digit. All hexadecimal digits above 9
are not BCD codes.
Example 3. Convert 1100 1010 1011 from BCD to decimal.
Answer: The groups of 4 bits given in this problem do not represent decimal
numbers in BCD.
One of the attractive features of BCD for business applications is that conversion
from ASCII or EBCDIC character code format to BCD is much simpler than from
character code format to binary. This is important to business data processing which
spends most of its time doing input, output, sorting, and storage, and comparatively little
time doing computation. ASCII and EBCDIC are 8-bit codes with the numerical digit in
the right nibble (4-bit group). Obtain the BCD representation by discarding the left
nibble. Technically, do a "logical AND" of the ASCII or EBCDIC code with "0F"
(hexadecimal) [ 00001111 (binary) ] to obtain the BCD code. Using the ability to shift
register contents right and left, you can pack BCD digits. This is called "packed BCD" in
technical literature by people that work at the machine language level of programming a
processor. Therefore, if you have the decimal number "937", you get
"0000:1001:0011:0111".
Table 47. Codes for Decimal Numbers in ASCII, EBCDIC and BCD
Decimal 0 1 2 3 4 5 6 7 8 9
ASCII
30 31 32 33 34 35 36 37 38 39
Hex
ASCII 0011: 0011: 0011: 0011: 0011: 0011: 0011: 0011: 0011: 0011:
Binary 0000 0001 0010 0011 0100 0101 0110 0111 1000 1001
EBCDIC
F0 F1 F2 F3 F4 F5 F6 F7 F8 F9
Hex
EBCDIC 1111: 1111: 1111: 1111: 1111: 1111: 1111: 1111: 1111: 1111:
Binary 0000 0001 0010 0011 0100 0101 0110 0111 1000 1001
BCD 0000 0001 0010 0011 0100 0101 0110 0111 1000 1001
Hardware for performing BCD arithmetic was quickly introduced for computers
designed for business data processing. The IBM 1401 had BCD hardware. The IBM 360
included BCD arithmetic implemented in microcode. When microprocessors were
introduced, the Z80 processor included BCD arithmetic. The calculator is a natural target
for using BCD arithmetic, as the person using a calculator is most likely to be thinking in
base 10. Beginning with the 80486 processor, Intel processors have BCD and binary
integer arithmetic support included in the x87 Floating Point Unit (FPU).
IBM Binary Coded Decimal (BCD)
IBM binary coded decimal is used only to represent integers. Any scaling for a
fractional part must be done by the programmer.
IBM provides two forms, Packed BCD and Unpacked BCD. Packed BCD is
more compact for storage and is the only BCD format that can be used in computations.
This is an appropriate format for accounting data. Unpacked BCD is faster to use when a
number is not required for numerical computations.
Number Systems and Data Representation 87

The codes used for plus and minus signs depend upon the operating environment
used to generate the data. The codes for the sign are different than the codes for
numbers. By placing the sign nibble or byte at the end a number, the sign can therefore
be used to identify the end of a BCD number. See the SAS Language Reference
Dictionary for a discussion. 34
Table 48. IBM BCD Sign Codes

Plus Sign Minus Sign

A16 B16
C16 (preferred) D16 (preferred)
E16
F16 (also the Zone code)

Zero is normally accompanied with a plus sign, although a minus sign with zero is
also valid.

IBM Packed BCD

Also called “Packed Decimal” format, the packed BCD format is 1 to 16 bytes
long, and stores one decimal digit in each 4-bit nibble. The sign also is stored as a 4-bit
nibble. This permits representation of 31 decimal digits and a sign. A packed decimal
integer always has an odd number of digits, one of which might be a leading zero. The
sign nibble marks the end of the Packed BCD word.
In the diagram below, “d” is the decimal digit.
0 d 3 4 d 7 8 … n n+1 d n+4 4-bit Sign
The cost of conversion from character format to packed decimal format is small
compared to the cost in time to convert to binary. By stripping off the left nibble,
memory and storage are more efficiently used.

IBM Unpacked BCD

Also called “Zoned Decimal” format, the unpacked BCD format stores one
decimal digit in the right nibble of each byte. The left nibble contains a zone code. In
the right- most byte, the left nibble is the sign. Except for the right-most byte, each byte
in zoned decimal format is the ASCII (3n) or EBCDIC (Fn) code for that number. In the
diagram below, “Z” is the zone character (3 or F), “d” is the decimal digit and “S” is the
code for the sign. The lower case “n” is the digit number in the Unpacked BCD word,
counting from the left.
0 Z 3 4 d 7 8 Z 11 12 d 15 16 … n S d
The advantage of zoned decimal format is that no time is required for data
conversion from character format. It reduces the storage requirements for numerical data.

34
SAS Institute, Inc., “Working with Packed Decimal and Zoned Decimal Data”, SAS Language
Reference: Dictionary (1999).
http://www.rdg.ac.uk/ITS/Topic/Stats/StGSAS8_01/SAS8/lgref/z1131887.htm, 09 Jun 2002.
Number Systems and Data Representation 88

This is good for transaction processing, which is characterized by much input, sorting,
and much output. Data must be converted to Packed BCD before doing computations.

Conversion to IBM BCD Problems

Problem 1. Convert ASCII -736 to IBM Unpacked BCD.

Z d Z d S d
Problem 2. Convert ASCII -736 to IBM Packed BCD.

d d d S
Number Systems and Data Representation 89

Fixed Point Notation

Positional Number System Review
The positional number system used a radix point to separate the integer part from
the fraction part of a number. In the decimal number system, we call the radix point by
the name “decimal point”. The number 835.27 means
(8 x 10 2 ) + (3 x 10 1 ) + (5 x 10 0 ) + (2 x 10 –1 ) + (7 x 10 –2 )
The positional number system for any base can be used to represent fractions as
well as integers. The positional number system represents a number in the fo rm
…c5 c4 c3 c2 c1 c0 .c-1 c-2 c-3 c-4 …
The ck are the digits or coefficients of the number written in fixed point notation. The
subscripts in the above expression are the position numbers. These numbers locate the
position of the number as measured from the radix point. The base of the number system
must be known in order to be able to evaluate the number. Without any additional
information, if all the digits are in the range of 0 through 9, the number is assumed to be a
decimal number. Otherwise, the base must be explicitly annotated. Let b be the base of
the number. Then the number is evaluated as
…(c5 b5 )+(c4 b4 )+(c3 b3 )+(c2 b2 )+(c1 b1 )+(c0 b0 )+(c-1 b-1 )+(c-2 b-2 )+(c-3 b-3 )+(c-4 b-4 )…
IBM Fixed Point
The IBM System 390 has four fixed point representations, which depend on
length, and whether the number is signed or unsigned.
Signed fixed point (integers) use two’s complement to represent negative
numbers. A positive number has a sign bit of zero. Negative numbers are stored in two’s
complement form, and have a sign bit of one. A signed integer may be 16 or 32 bits long.
Unsigned integers are considered positive. An unsigned integer may be 8, 16, or
32 bits long. These are useful for representing logical (Boolean) data, counters,
enumerated data, address offsets, and other purposes.

IBM Short Form Fixed Point

0 Sign 1 15
The signed short integer is a 16-bit, two’s complement integer. The unsigned
short integer is a 16-bit magnitude word.

IBM Long Form Fixed Point

0 Sign 1 31
The signed long integer is a 32-bit, two’s complement integer. The unsigned long
integer is a 32-bit magnitude word.
Number Systems and Data Representation 90

Convert Fixed Point Binary to Fixed Point Hexadecimal

Conversion from binary to hexadecimal for numbers with fractions is similar to
the case with integers. Groups of four binary digits are converted to hexadecimal. The
grouping begins at the radix point, in both directions from the radix point.
Example: Convert 1111111000011110101101.001100011100012 to
hexadecimal.
Group the bits into groups of four digits, beginning at the radix point.
11 1111 1000 0111 1010 1101 . 0011 0001 1100 012
ß . à
If the extremities do not have four bits, pad with zeros in the direction away from the
radix point. The number becomes
0011 1111 1000 0111 1010 1101 . 0011 0001 1100 01002
Convert each group of 4 bits to a single hexadecimal digit.
0011 1111 1000 0111 1010 1101 . 0011 0001 1100 0100
3 F 8 7 A D .3 1 C 4
The answer is 3F87AD.31D4
Convert a Fixed Point Hexadecimal Number to Binary
To convert a fixed point hexadecimal number to binary, merely substitute the 4-
bit binary number for each hexadecimal number.
Example: Convert 3AC.6BC to binary.
3 A C. 6 B C
0011 1010 1100 . 0110 1011 1100
Extreme zeros may be trimmed from the result. The answer is
11 1010 1100.0110 1011 112
Convert a Fixed Point Binary Number to Decimal
The binary number 1101.1012 is evaluated as
(1 x 23 ) + (1 x 22 ) + (0 x 21 ) + (1 x 20 ) + (1 x 2 –1 ) + (0 x 2 –2 ) + (1 x 2 –3 )
= 8 + 4 + 0 + 1 + 0.5 + 0 + 0.125 = 13.625
Number Systems and Data Representation 91

3 2 1 0 -1 -2 -3 Position
Number
23 22 21 20 2-1 2-2 2-3 Place Value
8 4 2 1 1/2 1/4 1/8 Place Value
as Decimal
Fraction
8 4 2 1 0.5 0.25 0.125 Place Value
in Decimal
1 1 0 1 .1 0 1 Coefficients
+ + + + + Sum of
(1x8) (1x4) (Ox2) (1x1) (1x0.5) (Ox0.25) (1x0.125) Products
8 +4 +0 +1 + 0.5 + 0.25 + 0.125
ANSWER = 13.625
Convert a Decimal Fixed Point Number to Binary
The approach for converting a decimal integer to binary is used for the whole
number portion of the decimal fixed point number. The procedure for conversion of the
fraction part of the decimal fixed point number to binary uses multiplication rather than
division.
Example: Convert the decimal number 21.40625 to binary.
Partition the decimal number into a whole number plus a fraction.
21.40625 = 21 + 0.40625
Convert the whole number portion to binary using the method of successive
division.
Step 0 1 0 = Q1 à Step 1 5 = Q2
Base = 2 2 1 = Q0 Base = 2 1 0 = Q1
2 1 0
0 1 0 0 = R1
0
1 = R0

Step 2 2 = Q3 à Step 3 1 = Q4
Base = 2 5 = Q2 Base = 2 2 = Q3
4 2
1 = R2 0 = R3

Step 4 0 = Q5 Stop only when the Quotient is zero.

Base = 2 1 = Q4
0
1 = R4
The integer portion of the binary number is 1 0101.
Number Systems and Data Representation 92

Convert 0.40625 (the fraction part) by successive multiplication by the base,

which is 2 in this case. The result of a multiplication that is to the left of the radix point
becomes a digit in the answer.
Column
-1 -2 -3 -4 -5
Position Number
Fraction of X 0. 4 0 6 2 5
b 2
Carry 0 0 1 0 1
0 8 0 2 4 0
R-1 = 0 0. 8 1 2 5 0

Continue repeating the process until the remaining fraction part is zero.
Column
-1 -2 -3 -4
Position Number
Fraction of R-1 0. 8 1 2 5
b 2
Carry 1 0 0 1
0 6 2 4 0
R-2 = 1 1 6 2 5 0

Column
-1 -2 -3
Position Number
Fraction of R-2 0. 6 2 5
b 2
Carry 1 0 1
0 2 4 0
R-3 = 1 1 2 5 0

Column
-1 -2
Position Number
Fraction of R-3 0. 2 5
b 2
Carry 0 1
0 4 0
R-4 = 0 0 5 0
Number Systems and Data Representation 93

Column
-1
Position Number
Fraction of R-4 0. 5
b 2
Carry 1
0 0
R-5 = 1 1 0
The procedure stops when the fraction of the last remainder is zero. Build the
fraction part of the binary number working from left to right, away from the radix point.
The fraction part of the answer is
R –1 R –2 R –3 R –4 R –5 = 0.0110 1
The final answer is the sum of the whole part plus the fraction part.
1 0101 + 0.0110 1 = 1 0101.0110 12 = 21.40625
Convert a Fixed Point Hexadecimal Number to Decimal
The hexadecimal number BAD.CE16 is evaluated as
(B x 162 ) + (A x 161 ) + (D x 160 ) + (C x 16 –1 ) + (E x 16 –2 )
= ( 11 x 256 ) + ( 10 x 16 ) + (13 x 1 ) + (12 x 0.625 ) + (14 x 0.00390625 )
= 2816 + 160 + 13 + 7.5 + 0.0546875 = 2996.554688

2 1 0 -1 -2 Position
Number
162 161 160 16 -1
16 -2
Place Value
256 16 1 1/16 1/256 Place Value
as Decimal
Fraction
4256 16 1 0.625 0.00390625 Place Value
in Decimal
B A D .C E Coefficients
+ + + + Sum of
(11x256) (10x16) (13x1) (12x0.625) (14x0.00390625) Products
+ 2816 + 160 + 13 + 7.5 + 0.0546875
=
2996.554688
Convert a Decimal Fixed Point Number to Hexadecimal
The approach for converting a decimal integer to hexadecimal is used for the
whole number portion of the decimal fixed point number. The procedure for conversion
of the fraction part of the decimal fixed point number to hexadecimal uses multiplication
rather than division.
Number Systems and Data Representation 94

Example: Convert the decimal number 3290.9921875 to hexadecimal.

Partition the decimal number into a whole number plus a fraction.
3290.9921875 = 3290 + 0.9921875
Convert the whole number portion to hexadecimal using the method of successive
division.
Step 0 2 0 5 = Q1 à Step 1 1 2 = Q2
Base = 16 3 2 9 0 = Q0 Base = 16 2 0 5 = Q1
3 2 1 6
0 9 4 5
0 3 2
9 0 1 3
8 0 = D16 = R1
1 0
= A16 = R0

Step 2 0 = Q3 Stop only when the Quotient is zero.

Base = 16 1 2 = Q2
0
1 2
= C16 = R2
The integer portion of the hexadecimal number is CDA16 .
Convert 0.9921875 (the fraction part) by successive multiplication by the base,
which is 16 in this case. The result of a multiplication that is to the left of the radix point
becomes a digit in the answer.
Column
-1 -2 -3 -4 -5 -6 -7
Position Number
Fraction of X 0 .9 9 2 1 8 7 5
b 1 6
Sum Carry 1 1 2 2 1
Carry 0 5 5 1 0 4 4 3
Carry 1 0 0 0 0 0 0
4 4 2 6 8 2 0
9 9 2 1 8 7 5
R-1 = 0 1 5 8 7 5 0 0 0 0
= F16
Continue repeating the process until the remaining fraction part is zero.
Number Systems and Data Representation 95

Column
-1 -2 -3
Position Number
Fraction of R-1 0 .8 7 5
b 1 6
Sum Carry 1 2 1
Carry 0 4 4 3 0
Carry 1 0 0 0 0
8 2 0
8 7 5 0
R-2 = 1 4 0 0 0
= E16
The procedure stops when the fraction of the last remainder is zero. Build the
fraction part of the binary number working from left to right, away from the radix point.
The fraction part of the answer is
R –1 R –2 = 0.FE
The final answer is the sum of the whole part plus the fraction part.
CDA + 0.FE = CDA.FE16 = 3290.9921875
Fixed Point Conversion Problems
Problem 1. Convert 11 1011 0101.0011 012 to hexadecimal.
Problem 2. Convert FED.CAB16 to binary.
Problem 3. Convert 10 1100.0112 to decimal.
Problem 4. Convert 107.625 to binary. (110 1011.1012 )
Problem 5. Convert AD.E16 to decimal. (173.875)
Problem 6. Convert 430.7578125 to hexadecimal. (1AE.C2H)
Number Systems and Data Representation 96

Partitioned Binary Numbers

Sometimes it is useful (especially in networking for constructing subnet numbers)
to partition a binary number so the sum of the individual numbers is the value of the
original number. For example, partition the binary number 10 10112 between bit
numbers 4 and 3, where bit 0 is the right- most bit. Let the left partition be 10 00002 , and
the right partition be given by 10112 . Observe that the sum is the original value.
Left partition 1 0 0 0 0 0
Right partition 1 0 1 1
Sum 1 0 1 0 1 1

Convert the left partition 10 00002 to decimal.

5 4 3 2 1 0 ß position number
25 24 23 22 21 20 ß place value as power of base 2
32 16 8 4 2 1 ß place value
1 0 0 0 0 02 ß digits

= (1 x 32) + (0 x 16) + (0 x 8) + (0 x 4) + (0 x 2) + (0 x 1)
= 32 + 0 + 0 + 0 + 0 + 0
= 32
Convert the right partition 10112 to decimal.
3 2 1 0 ß position number (bit number)
23 22 21 20 ß place value as power of base 2
8 4 2 1 ß place value
1 0 1 12 ß digits

= (1 x 8) + (0 x 4) + (1 x 2) + (1 x 1)
= 8 + 0 + 2 + 1
= 11
Add the decimal values of these partitions. 32 + 11 = 43.
Partition Problems
Problem A1. Partition 1011 01102 between bits number 4 and 5. What are the two
resulting decimal numbers and their sum? Find:
a. Left partition
b. Right partition
c. Sum
 Number Systems and Data Representation 97

Problem A2. Partition 0110 11002 between bits number 4 and 5. What are the two
resulting decimal numbers and their sum? Find:
a. Left partition
b. Right partition
c. Sum
Partitioning Internet Protocol (IP) Numbers
Internet Protocol (IP) addresses for Internet Protocol Version 4 (Ipv4, in use
during 2002 A.D.) are 32 bits long, and are written as a dotted decimal number. It is a
sequence of four decimal numbers that represent the value of each octet. The term octet
refers to a group of 8 bits. 35 A typical IP address written in dotted decimal form is:
10.1.127.201 . In the table below, this IP address is converted to binary and hexadecimal.
The second row contains the bit number in the 32-bit word.
10 1 127 201
31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 1 1 1 1 1 1 1 0 1 1 0 1 0 0 1
0 A 0 1 7 F 6 9

A network address is assigned to a network by the American Registry for

Internet Numbers (ARIN). The administrator of a network has the option of
partitioning its network into subnetworks. While there are some restrictions, the basic
task is to identify which IP numbers will belong to each subnetwork. The details are
taught in a networking class. This discussion will not pay attention to those details.
Suppose you have been assigned the network address 10.x.x.x. This network
address gives you three octets to use, signified by x. Suppose you want to establish
subnets that have at least 4000 host machines on each subnet. How many subnets can
you create, and what is the block of numbers you can assign as addresses on each subnet?
In the simplest of systems, subnet host machines are selected by use of the N
least significant bits (right- most bits) in an IP address. The problem reduces to selecting
the number N. Find the smallest power of 2 that is greater than or equal to 4000.
Solution: N = 12, because 212 = 4096. The right- most 12 bits, numbered 0 through 11,
are used to identify individual host machines on each subnet. This leaves bits 12 through
23 (12 bits) for identifying a particular subnet. You therefore can have 4096 addresses
for subnets.
Some of the subnet addresses are reserved for special purposes. An IP address
having a host machine address of all zeros is the “Network Address” or “Network
Number”. An IP address having a host machine address of all ones is the “Broadcast
Address” for that subnet, which routes messages to all computers on that subnet. Thus,
the number of computers on a classical subnet is two less than the number of
numbers that can be represented by the host machine address portion of the IP
address.

35
Debra Littlejohn Shinder, Computer Networking Essentials, page 233, Cisco Systems, Inc. (2001)
 Number Systems and Data Representation 98

The allocation looks like the following.

31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 x x x x x x x x x x x x y y y y y y y y y y y y

The portion coded with “x” is the address for a subnet. The portion coded with
“y” is the address for a specific host machine on that subnet.
A necessary task is to identify the numbers that correspond to a chosen partition.

Subnet Mask
A subnet mask has 1 in all the positions in the network and subnet portion of
the address field. The dotted decimal notation is 255.255.240.0 . When this is ANDed
with an incoming IP address, the result is the address for the target subnet.
255 255 240 0
31 24 23 16 15 8 7 0
1 1 1 1 1 0 1 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0
F F F F F 0 0 0

Subnet Problems
Problem B1. Convert the dotted decimal for IP address 17.10.127.5 to binary and
hexadecimal, using the table below.

31 24 23 16 15 8 7 0

Problem B2. How many subnet addresses are possible in the subnet mask shown
below?
31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 x x x x x x x x x y y y y y y y y y y y y y y y

Problem B3. How many host machine addresses are possible for each subnet in the
subnet mask shown below?
31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 x x x x x x x x x y y y y y y y y y y y y y y y

Problem B4. How many bits in the below mask need to be set aside to make sure each
subnet can have at least 283 host machines?
31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 z z z z z z z z z z z z z z z z z z z z z z z z

Problem B5. How many subnet addresses are possible in the below mask if enough bits
are set aside to make sure each subnet can have at least 283 host machines?
31 24 23 16 15 8 7 0
0 0 0 0 1 0 1 0 z z z z z z z z z z z z z z z z z z z z z z z z

Problem B6. What is the dotted decimal notation for the subnet mask needed for
ensuring enough bits are set aside to ensure each subnet can have at least 283 host
machines?
255 255 254 0
31 24 23 16 15 8 7 0
Number Systems and Data Representation 99

(German) Roman Numeral System

Not every number system is a strict positional number system like the decimal
system. The (German) Roman numeral system is a common example. We still use this
for ceremonial purposes, such as dates on monuments, buildings, and currency. It is also
popular on clocks. The Roman numeral system has no symbol for zero. To count in
Roman numerals, we use letters of the Latin alphabet as follows.
Table 49. Counting Using (German) Roman Numerals

Decimal Roman Decimal Roman Decimal Roman

1 I 11 XI 21 XXI
2 II 12 XII 22 XXII
3 III 13 XIII 23 XXIII
4 IV 14 XIV 24 XXIV
5 V 15 XV 25 XXV
6 VI 16 XVI 26 XXVI
7 VII 17 XVII 27 XXVII
8 VIII 18 XVIII 28 XXVIII
9 IX 19 XIX 29 XXIX
10 X 20 XX 30 XXX

Decimal Roman Decimal Roman Decimal Roman

31 XXXI 41 XLI 51 LI
32 XXXII 42 XLII 52 LII
33 XXXIII 43 XLIII 53 LIII
34 XXXIV 44 XLIV 54 LIV
35 XXXV 45 XLV 55 LV
36 XXXVI 46 XLVI 56 LVI
37 XXXVII 47 XLVII 57 LVII
38 XXXVIII 48 XLVIII 58 LVIII
39 XXXIX 49 XLIX 59 LIX
40 XL 50 L 60 LX

Decimal Roman Decimal Roman Decimal Roman

10 X 100 C 1,000 M
20 XX 200 CC 1,500 MD
30 XXX 300 CCC 2,000 MM
40 XL 400 CD 3,000 MMM
50 L 500 D 4,000
60 LX 600 DC 5,000
70 LXX 700 DCC 1,000,000
80 LXXX 800 DCCC
90 XC 900 CM
100 C 1,000 M
Number Systems and Data Representation 100

X is used for 10. V is the upper half of X, and is used for 5. L is used for 50. C
is used for 100. D is 500. M is 1000. [The Romans used (|), not M.] 36 A dash line
above a letter multiplies the base value by 1000. There is no zero. Sometimes, a pair of
horizontal bars, above and below these symbols, are used to distinguish these from mere
alphabet letters.
While there is a pattern to the meaning of numbers depending on their position, it
is very different from the simple pattern used with the decimal number system. As you
might guess, arithmetic using the Roman numeral system is not as easy as with the
decimal system. The Romans apparently did not do computation with their numerals, but
rather used a counting board or abacus. A 1st century A.D. gravestone shows a person
with an abacus in his hand. 37
The Roman numeral system applied meaning to the position of numbers, but not
in the same way our present day positional number system does. It evolved from a tally
system. The basic tally system uses vertical lines, one line for each object counted. The
next modification is a grouping operation that is applied in steps of 5. After
accumulating 4 vertical marks, such as |||| , a grouping is made by placing a diagonal
slash through the group of 4 vertical marks.
The Roman numeral system is the next level of sophistication. Instead of having
5 marks, the symbol V is substituted. It took less space to write IV for 4 than it did to
write |||| . Writing materials were precious. When the printing press was invented, word
spellings and number formats were modified to further reduce the space required.
Roman numerals are created from left-to-right and evaluated from right-to- left.
The object is to express the quantity using the fewest possible symbols.
Parsing Rule. The pattern for evaluation that developed is as follows. Group
adjacent identical Roman Numerals before evaluating the number. Consider a pair of
adjacent Roman numerals. 38 Beginning with the right end, if the numeral to the left is less
in value to the numeral on the right,(just the numeral immediately to the right, not the
whole number to the right), the numeral on the left is subtracted from the numeral on the
right. If the numeral on the left is greater than or equal to the numeral on the right, the
numeral on the left is added to the numeral on the right. Gerard Schildberger has
published a list of numbers from 1 to 3999. 39
The following examples of years (except 1944, 1999, 2002) are from the
Government Printing Office Style Manual. 40

36
Karl Menninger, Number Words and Number Symbols: A Cultural History of Numbers, page 281, Dover
Publications (1992).
37
Karl Menninger, Number Words and Number Symbols: A Cultural History of Numbers, page 306, Dover
Publications (1992).
38
Gerard Schildberger, “Roman Numerals are Nonassociative”, email. (23 June 2002)
39
Gerard Schildberger, “The numbers from 1 to 3999 expressed as Roman numerals”,
http://www.research.att.com/~njas/sequences/a006968.txt visited 04 April 2001.
40
United States Government Printing Office Style Manual 2000, U.S. Government Printing Office (2000),
pg 189. http://www.access.gpo.gov/styleman/2000/style001.html, 26 June 2002
Number Systems and Data Representation 101

Table 50. Roman Numeral Years in the U.S. Government Printing Office Style Manual

Years Written in Roman Numerals

1600 MDC 1930 MCMXXX 1980 MCMLXXX
1700 MDCC 1940 MCMXL 1990 MCMXC
1800 MDCCC 1944 MCMXLIV 1999 MIM
1900 MCM 1950 MCML 2000 MM
1910 MCMX 1960 MCMLX 2002 MMII
1920 MCMXX 1970 MCMLXX 2010 MMX

You can sometimes reduce the number of symbols appearing in a Roman numeral
by the following approach. If the same symbol appears on both the right side and the left
side of a higher value symbol, these symbols constitute a matched pair. You can remove
both symbols without changing the value of the number. Examples:
• IXI à X
• ICMLI à CML
A problem with Roman numerals is ambiguity, as illustrated with the following
example. The parentheses indicate the grouping used in evaluating the example. Note
that this kind of ambiguity is impossible with our own positional number system.

• (CIV)(MXVI) = -104 + 1016 = 912

• C(I(V(MX)V)I) à CMX = 910
• C(IVM)(XVI) = -100-4+1000+10+5+1 = 912
• (CI)(VM)XVI = -101 –5 + 1016 = 910
• (CIV)M(XVI) = -104 + 1000 + 16 = 912
• CIVMXVI = 910, evaluated according to the parsing rule.
Background for next example:
• (VC)I = 95 + 1 = 96
• V(CI) = 99 – 5 = 94
Example:
• V(CIM) = 899 – 5 = 894
• (VC)(IM) = 999 – 95 = 904
• ((VC)I)M = 1000 – 96 = 904
• (V(CI))M = 1000 – 94 = 906
• VCIM = 894, evaluated according to the parsing rule.
The parsing rule given here applies to Roman numerals written today, but this was
not always true throughout history. To read literature of earlier times, you will need to
study the conventions in use at that time.
Number Systems and Data Representation 102

German Roman Numeral Problems

For Romans with a lot of Gaul.

Problem 1. Write 390 in Roman numerals. (Celts defeated Rome on 18 July 390 BC,
and sacked Rome 3 days later. 41 )
Problem 2. Write 407 AD in Roman numerals. (Constantine withdrew from Britain
that year. 42 )
Problem 3. Convert 255 to Roman numerals.
Problem 4. Add the Roman numerals XI and XXXII. Convert the result to decimal
after adding.
Problem 5. What time is XXV O’clock? (Humorous answer.)
Problem 6. Write zero in Roman numerals. (Trick question.)
Problem 7. Divide XXX by VI .
Problem 8. There is a number on the base of the pyramid shown on the back side of
the paper One Dollar Federal Reserve Note, MDCCLXXVI. Copy that number given in
Roman Numerals and translate it into Indo-Arabic numbers (the number system we use in
commerce today).
Problem 9. Evaluate the numeral I(VL) , evaluating inside the parentheses first.
Problem 10. Evaluate the numeral (IV)L , evaluating inside the parentheses first.

41
Encyclopaedia Britannica (1965)
42
Encyclopaedia Britannica (1965)
Number Systems and Data Representation 103

Fibonacci Bases
Another example of a number system that is not a strict positional number system
like the decimal system is a Fibonacci base. If you have a twisted mind, or really enjoy
looking at the world in ways different than most people, this is the number system for
you. A really nice discussion, complete with online calculators, is given by Knott. 43
Fibonacci introduced the sequence that now bears his name in Liber abaci (1202)
a problem he stated: “How many pairs of rabbits can be produced from a single pair in
one year if it is assumed that every month each pair begets a new pair which from the
second month becomes productive?”44
Fibonacci Numbers
The first few Fibonacci numbers are: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, … The
first two numbers of the sequence are established by definition as F0 = 0, F1 = 1. The rest
of the numbers are defined by the recursion formula Fn = Fn-1 + Fn-2 where n is a variable
index that takes on values 2, 3, 4, 5, … .
The sequence of Fibonacci numbers led to other questions in number theory. A
short list of examples of the sequence seen in nature is listed in Encyclopaedia
Britannica. Generalized Fibonacci numbers are useful in the design of efficient methods
of sorting very large sets of data. 45
Fibonacci Base Representation
The Fibonacci numbers are used as the basis of the number system in a place-
value notation. A number written in the Fibonacci Base Representation has the form
an Fn + an-1 Fn-1 +…+ a4 F4 + a3 F3 + a2 F2
where the coefficient an is an integer multiplier of the corresponding Fibonacci number
Fn . For example, the Fibonacci Base Representation number 95013 Fib is converted to a
decimal number by:

95013Fib = (9 x F6 ) + (5 x F5 ) + (0 x F4 ) + (1 x F3 ) + (3 x F2 )

= (9 x 8) + (5 x 5) + (0 x 3) + (1 x 2) + (3 x 1)

= 72 + 25 + 0 + 2 + 3

= 10210

Notice that the lowest subscript on the Fibonacci number and the corresponding
coefficient is 2 because F2 = 1. F1 = 1, and F0 = 0. While it would be possible to extend
the notation down to F0 , there is nothing to be gained by having two unit bases F2 and F1 ,
43
Ron Knott, “Using Fibonacci Numbers to Represent Whole Numbers” (19 January 2007).
http://www.mcs.surrey.ac.uk/Personal/R.Knott/Fibonacci/fibrep.html visited 15 December 2007.
44
Encyclopaedia Britannica 2003, Deluxe Edition CD-ROM
45
Donald E. Knuth, The Art of Computer Programming, Volume 3, Searching and Sorting, Addison-
Wesley (1973).
Number Systems and Data Representation 104

and pointless to include F0 since anything multiplied by zero (F0) is zero, and does not
change the value of the number being represented.
Extended Fibonacci Base Representation
The Fibonacci series can be extended to negative indexes. This is useful in other
work related to Fibonacci numbers. The extension is done by rearranging the equation
to: Fn-1 = Fn+1 – Fn .

n 8 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8

Fn 21 13 8 5 3 2 1 1 0 1 -1 2 -3 5 -8 13 -21

From this table, observe that

Result When
F-n = Fn n is ODD
F-n = -Fn n is EVEN

Minimal Fibonacci Base Representation

Knott 46 described a Fibonacci base positional number system proposed by
Edouard Zeckendorf in 1972 that uses the fewest Fibonacci numbers. An alternative
name for this representation is therefore the Zeckendorf Representation. It allows a
unique representation of any positive integer as a sum of Fibonacci numbers, by applying
a few simple rules:
• Number position values increase from right to left.
• Use only the digits 0 and 1 as coefficients.
• No two consecutive Fibonacci numbers can be used in the same sum. This means
the digit 1 cannot appear adjacent to another 1.
Example:
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
0 0 0 0 0 0 1 0 0 1 0 0 1 ß digits
0 0 0 0 0 0 21 0 0 5 0 0 1 ß decimal
values

Check the answer to see if it is correct.

1 0 0 1 0 0 1 Fib is the same as
( 21 x 1 ) + ( 5 x 1 ) + ( 1 x 1 ) = 21 + 5 + 1 = 27

46
Ron Knott, “Using Fibonacci Numbers to Represent Whole Numbers” (19 January 2007).
http://www.mcs.surrey.ac.uk/Personal/R.Knott/Fibonacci/fibrep.html visited 15 December 2007.
Number Systems and Data Representation 105

This is the sum of the digits times their place values.

Knott provides a program on his web site for converting decimal to minimal
Fibonacci base representation, as well as for other conversions. Another name for
“standard” is “canonical”. A number placed into an agreed standard form is in canonical
form.

Convert Decimal to Minimal Fibonacci Base Representation

1. Set Remainder R to the decimal number to be converted.
2. Find the largest Fibonacci number F equal to or less tha n R.
3. Place a 1 in the column of the translation table corresponding to that Fibonacci
number F.
4. Subtract F from R, with the result being the new R: Rnew ß R – F.
5. If R is greater than 0, go to step 2 and repeat the process.
Number Systems and Data Representation 106

Example Conversions to Minimal Fibonacci Base Representation

1. Convert 31710 to Minimal Fibonacci Base (Zeckendorf) Representation.
The number in the top row, labeled “Remainder R”, is the amount remaining to be
coded into Zeckendorf Representation by considering subtraction of the Fibonacci
number in that column. Begin the procedure on the left- most column, and proceed to the
right. For the column of Fibonacci Number F10 , the Remainder R before subtraction is
84. F10 = 55, which is less than or equal to 84. Therefore, perform the subtraction and
record “1” in the “digits” row of that column. Record the remainder from the subtraction
(29) at the top of the column of the next column to the right, above F9 .
317 317 84 84 84 29 29 8 8 0 0 0 0 Remainder R
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
0 1 0 0 1 0 1 0 1 0 0 0 0 ß digits
0 233 0 0 55 0 0 0 0 0 0 0 0 ß decimal
values
84 29 8 0 Remainder R
The answer is 1001 0101 0000Fib
2. Convert 6910 to Minimal Fibonacci Base (Zeckendorf) Representation
69 69 69 69 69 14 14 14 1 1 1 1 1 Remainder R
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
0 0 0 0 1 0 0 1 0 0 0 0 1 ß digits
0 0 0 0 55 0 0 13 0 0 0 0 1 ß decimal
values
14 1 0 Remainder R
The answer is 1 0010 0001Fib
Number Systems and Data Representation 107

Addition in Minimal Fibonacci Base Representation

Two Fibonacci numbers written in Zeckendorf representation can be added.
Example: The two Zeckendorf representation numbers to be added are in the rows
labeled X and Y. The result of the addition is in the row labeled “Answer”.
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
0 1 0 0 1 0 1 0 1 0 0 0 0 ß X
0 0 0 0 1 0 0 1 0 0 0 0 1 ß Y
0 1 0 0 2 0 1 1 1 0 0 0 1 Answer
233 110 21 13 8 1 ß decimal
values
The answer is 1002 0111 0001Fib = (233+110+21+13+8+1)10 = 38610
Restore Fibonacci Base Representation to Minimal Fibonacci Base Representation
After adding, it may be necessary to rewrite the answer to restore it to Zeckendorf
representation. Two rules can be helpful. The primary purpose is to have a systematic
method of applying the rules until no opportunity is available to apply them again. You
can modify the rules to start scanning from the opposite end. You will eventually get the
same result.
• Rule 1: Start from the right end of the row and proceed to the left to find the next
“2”.
o If column F2 contains 2, set column F2 to zero and add 1 to column F3 .
o If column F3 contains 2,
§ Set column F3 to zero.
§ Add 1 to column F4 and to F2 .
o For F4 or higher, if column Fn contains a 2, then
§ set Fn to zero and add 1 to Fn+1 and to Fn-2 .
§ This is the recursion 2Fn = Fn+1 + Fn-2 .
• Rule 2: Start from the right end of the row and proceed to the left. If two
adjacent columns, Fn and Fn-1, contain ones, then
o Set Fn and Fn-1 to zero and set Fn+1 to 1.
o This is the recursion Fn+1 = Fn + Fn-1 .
Number Systems and Data Representation 108

The number in the third row below, labeled “Answer”, is the result of an addition
of two numbers in Zeckendorf representation. In each subsequent row, one of the rules is
applied to “simplify” the result. Repeated application of the rules eventually will
transfo rm the answer to Zeckendorf representation. The final result is in the last row.

F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci

Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
0 1 0 0 2 0 1 1 1 0 0 0 1 Answer
1 1 0 2 1 1 1 Rule 1
1 1 1 0 1 2 1 Rule 1
1 1 1 2 0 1 1 Rule 1
1 1 1 1 0 0 1 1 1 Rule 1
1 1 1 1 0 1 0 0 1 Rule 2
1 1 1 0 0 0 1 0 0 1 Rule 2
1 1 0 0 0 0 0 1 0 0 1 Rule 2
1 0 0 0 0 0 0 0 1 0 0 0 1 Rule 2

The answer in Zeckendorf representation is: 1 0000 0001 0001 Fib = (377+8+1)10 = 38610
Fibonacci Base Representation Problems
Problem 1. Convert 5010 to Minimal Fibonacci Base (Zeckendorf) Representation.

50 Remainder R
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
ß digits
ß decimal
values
Remainder R

Problem 2. Convert 5310 to Minimal Fibonacci Base (Zeckendorf) Representation.

53 Remainder R
F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci
Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
ß digits
ß decimal
values
Remainder R
Number Systems and Data Representation 109

Problem 3. Add the two Zeckendorf Representation numbers:

1000 1010 0101Fib + 1010 0001 0011Fib

F15 F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß

Fibonacci
Number
610 377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
1 0 0 0 1 0 1 0 0 1 0 1 ß X
1 0 1 0 0 0 0 1 0 0 1 1 ß Y
Answer
ß decimal
values

Problem 4. Restore Fibonacci Base Representation 10110112Fib to Minimal Fibonacci

Base Representation.

F14 F13 F12 F11 F10 F9 F8 F7 F6 F5 F4 F3 F2 ß Fibonacci

Number
377 233 144 89 55 34 21 13 8 5 3 2 1 ß place
values
1 0 1 1 0 1 1 2 Answer
Rule
Rule
Rule
Rule
Rule
Rule
Rule
Rule
Rule
Rule
Rule
Number Systems and Data Representation 110

Phi Base Number System

The base Phi (Φ) number system is an idea closely related to the Fibonacci base
representation. This is an example of a number system base that is an irrational number,
yet all integers can be represented exactly in this base. Knott has a fascinating
discussion47 on this topic, which is the basis for what follows. If this topic interests you,
visit his interactive website which includes calculators and diagrams.
Phi Definition
Phi is defined by the relation Φ -1 = 1/Φ = ϕ . The solutions to this equation are:
1
( )
Φ = 1 ± 5 . Phi is an irrational number (cannot be expressed exactly as the ratio of
2
two integers) because of the 5 which is irrational. The solutions are approximately
ΦA = 1.618034 ϕA = 0.61803399
ΦB = -0.61803399 ϕB = -1.618034
Phi Relationship to Fibonacci Numbers
Φ and ϕ are related to Fibonacci numbers by the following equation:
Φ n = Fn+1 + Fnϕ = (Fn + Fn−1 ) + Fnϕ = Fn−1 + Fn (1 + ϕ ) = Fn−1 + Fn Φ
To show this using the Principle of Finite Induction, first let n = 1 and prove the
equation Φ n = Fn+1 + Fnϕ reduces to Φ = 1 + ϕ . Then assume that the equation
Φ n−1 = Fn + Fn −1ϕ is true for an arbitrary number n, and prove from this starting point
that Φ n = Fn+1 + Fnϕ is true.
Simple Properties of Phi
There are two practical simple properties upon which other work with Phi
depends. Both are not hard to show.
• ϕ (1 + ϕ ) = 1
• Φ n = Φ n−1 + Φ n−2
Phi Base Representation
Numbers in Phi Base representation are expressed as a weighted sum of powers of
Φ, in the form
cnΦ n + cn-1Φ n-1+ … + c2Φ 2 + c1Φ 1 + c0Φ 0 + c-1Φ -1 + c-2Φ -2 + … + c-(n-1)Φ-(n-1) + c-
nΦ
-n

where the coefficient cn is some multiplier (possibly zero) of the nth power of Φ.
Numbers are expressed compactly in a manner similar to decimal notation by writing
only the constant coefficients and placing a radix point to the left of the coefficient c-1 .
When some coefficients are positive and others are negative, the negative coefficients

47
Ron Knott, “Using Powers of Phi to represent Integers (Base Phi)” (22 December 2007).
http://www.mcs.surrey.ac.uk/Personal/R.Knott/Fibonacci/phigits.html visited 24 December 2007.
Number Systems and Data Representation 111

must be identified. One method is to enclose the coefficient with a minus sign inside
parenthesis.
Integers in Phi Base Representation
All integers can be represented in Phi Base Representation for suitable choices of
the various coefficients. It is useful to look at a few examples.
Φ 0 = F1+F0ϕ = 1
Φ 1 = F2+F1ϕ = 1+ϕ Φ -1 = F0+F-1ϕ = ϕ L1 = Φ 1 - Φ -1 = 1
Φ 2 = F3+F2ϕ = 2+ϕ Φ -2 = F-1+F-2ϕ = 1-ϕ L2 = Φ 2 + Φ -2 = 3
Φ 1 + Φ -2 = 2
Φ 3 = F4+F3ϕ = 3+2ϕ Φ -3 = F-2+F-3ϕ = -1+2ϕ L3 = Φ 3 - Φ -3 = 4
Φ 4 = F5+F4ϕ = 5+3ϕ Φ -4 = F-3+F-4ϕ = 2-3ϕ L4 = Φ 4 + Φ -4 = 7
Φ 5 = F6+F5ϕ = 8+5ϕ Φ -5 = F-4+F-5ϕ = -3+5ϕ L5 = Φ 5 - Φ -5 = 11
Φ 6 = F7+F6ϕ = 13+8ϕ Φ -6 = F-5+F-6ϕ = 5-8ϕ L6 = Φ 6 + Φ -6 = 18
Notice that the sequence Ln = Φ n +(-1)nΦ -n for n > 0 of results form a series
similar to the Fibonacci series: Ln = Ln-1 + Ln-2 where L1 = 1 and L2 = 3 . The numbers
in the set {…, 1, 3, 4, 7, 11, 18, …} are called Lucas numbers.
Of course, all integers can be expressed merely as an integer multiple of Φ 0 , such
as 13 = 13Φ 0 . This certainly would be a unique representation, but it would miss the
point that we can represent integers by combinations of numbers using an irrational base.
The combinations used to express a particular integer are not necessarily unique. For
example, 5 = 1+4 = 2+3.
Integers Equation Phi Base
Representations
5 = 5Φ 0 5.0
= 5( Φ 1 - Φ -1 ) 50.(-5)
4+1 = ( Φ 3 - Φ -3 ) + ( Φ 1 - Φ - 1010.(-1)0(-1)
1
)
2+2+1 = 2( Φ 1 + Φ -2 ) + Φ 0 21.02
3+2 = ( Φ 2 - Φ -2 ) + ( Φ 1 + 110.0
Φ -2 )
= 3Φ 0 + ( Φ 1 + Φ -2 ) 13.01
= ( Φ 2 - Φ -2 ) + 2Φ 0 102.0(-1)

This generates a need for restrictions to establish unique representations.

Minimal Phi Base Representation
Adapt the rules for Minimal Fibonacci Base Representation to the case of Phi
Base Representation.
• Number position values increase from right to left.
• Use only the digits 0 and 1 as coefficients.
Number Systems and Data Representation 112

• No two consecutive Phi Base numbers can be used in the same sum. This means
the digit 1 cannot appear adjacent to another 1.
Digits written in minimal Phi base representation are known as “Phigits ”. 48
Converting Phi Base Representation to Minimal Phi Base Representation
It may be necessary to rewrite a number written in Phi Base Representation to
Minimal Phi Base Representation. Two rules can be helpful. Both implement the
recursion Φ n+1 = Φn + Φ n-1 The primary purpose is to have a systematic method of
applying the rules until no opportunity is available to apply them again. You can modify
the rules to start scanning from the opposite end. You will eventually get the same result.
• Rule 1: Start from the right end of the row and proceed to the left to find the next
number greater than “1”. If column Φ n contains a number greater than 1, subtract
1 from column Φ n , add 1 to column Φ n-1 , and add 1 to column Φ n-2 .
• Rule 2: Start from the left end of the row and proceed to the right. If two
adjacent columns, Φ n and Φ n-1, contain numbers greater than zero, then subtract 1
from columns Φ n and Φ n-1 and add 1 to column Φ n+1 .
Example 1: Write 2( Φ 1 +Φ-2 ) , written as 20.02Phi ,. and convert it to minimal phi base
representation
Step Φ3 Φ2 Φ1 Φ0 • Φ -1 Φ -2 Φ -3 Φ -4
1 Start 2 0 0 2
2 R#1 2 0 0 1 1 1
3 R#1 1 1 1 1 1 1
4 R#2 1 0 0 1 1 1 1
5 R#2 1 0 1 0 0 1 1
6 R#2 1 0 1 0 1 0 0
The minimal phi base representation is Φ 2 + Φ 0 + Φ-2 written as . 101.01Phi .
When there is a nega tive coefficient, one approach is to propagate positive
coefficients from the left to the right by reversing Rule 2. The goal is to produce a
coefficient of the same power of Φ that is the same in magnitude, and opposite in sign,
and then do the subtraction.

48
Jose Glez-Regueral, Professor at Madrid, cited by Ron Knott, “Other names for Base Phi”, in “Using
Powers of Phi to represent Integers” (22 December 2007).
http://www.mcs.surrey.ac.uk/Personal/R.Knott/Fibonacci/phigits.html visited 24 December 2007.
Number Systems and Data Representation 113

Example 2: Write Φ 3 - Φ-3 , written as 1000.00(-1)Phi , and convert it to minimal phi

base representation.
Step Φ3 Φ2 Φ1 Φ0 • Φ -1 Φ -2 Φ -3 Φ -4
1 Start 1 0 0 0 0 0 (-1)
2 R-1 0 1 1 0 0 0 (-1)
#2
3 R-1 0 1 0 1 1 0 (-1)
#2
4 R-1 0 1 0 1 0 1 1+(-1)
#2
5 0 1 0 1 0 1 0
The minimal phi base representation is Φ 2 + Φ 0 + Φ-2 written as . 101.01Phi .
Phi Based Addition
Perform arithmetic with Phi base numbers as you would with an general algebraic power
series, and then apply the rules for transforming the result to a minimal Phi base
representation.

Example: 101.01Phi + 1Phi . In decimal, this is the problem: 4 + 1 = 5. These numbers

represent:
1Φ 2 + 0Φ1 + 1Φ 0 + 0Φ -1 + 1Φ -2 = 101.01Phi
0Φ 2 + 0Φ1 + 1Φ 0 + 0Φ -1 + 0Φ -2 = 1Phi
Write each coefficient in the column corresponding to the associated power of Φ.
Φ3 Φ2 Φ1 Φ0 Φ -1 Φ -2 Φ -3 Φ -4
c= 1 0 1 . 0 1
d= 1 . 0 0
c+d= 1 0 2 . 0 1
Rule 1 1 0 1 . 1 2
Rule 1 1 0 1 . 1 1 1 1
Rule 2 1 1 0 . 0 1 1 1
Rule 2 1 0 0 0 . 0 1 1 1
Rule 2 1 0 0 0 . 1 0 0 1
Answer
The coefficients for each individual power of Φ are added separately. This produced the
intermediate result: c + d = 102.01Phi . This is: 1Φ 2 + 0Φ1 + 2Φ0 + 0Φ -1 + 1Φ -2

While this answer is correct, it is not in standard (canonical) form. Rules were
applied to transform this result to the minimal Phi base representation.
The answer is 1000.1001Phi = 1Φ3 + 0Φ 2 + 0Φ1 + 0Φ 0 + 1Φ -1 + 0Φ -2 + 0Φ -3 + 1Φ -4
This can be expanded: 1Φ 3 + 1Φ -1 + 1Φ-4 = 3+2ϕ + ϕ + 2-3ϕ = 5.
Number Systems and Data Representation 114

Phi Base Representation Problems

Problem 1. Express A = Φ 4 + Φ3 + Φ -3 as phigits:
Φ6 Φ5 Φ4 Φ3 Φ2 Φ1 Φ0 . Φ -1 Φ -2 Φ -3 Φ -4 Φ -5 Φ -6
A= .
Problem 2. Express B = Φ 4 + Φ1 + Φ -4 + Φ-6 as phigits:
Φ6 Φ5 Φ4 Φ3 Φ2 Φ1 Φ0 . Φ -1 Φ -2 Φ -3 Φ -4 Φ -5 Φ -6
B= .
Problem 3. Add the phigits A (from Problem 1) and B (from Problem 2).
Φ6 Φ5 Φ4 Φ3 Φ2 Φ1 Φ0 . Φ -1 Φ -2 Φ -3 Φ -4 Φ -5 Φ -6
A= .
B= .
A+B= .
Problem 4. Transform the sum, A+B, computed in Problem 3 into minimal Phi base
representation. In the left column, write the rule number being used for each step of the
transformation.
Φ6 Φ5 Φ4 Φ3 Φ2 Φ1 Φ0 . Φ -1 Φ -2 Φ -3 Φ -4 Φ -5 Φ -6
A+B= .
.
.
.
.
.
.
Problem 5. In the answer to Problem 4, substitute expressions for Φ n in terms of n and
f. Reduce the expression to the fewest number of terms.
Number Systems and Data Representation 115

Continued Fractions
Another type of notation that is positional is that of continued fractions. It is
useful for expressing some very important constants which have a non-repeating decimal
form, such as e and 2 . This is a topic that is sometimes taught in high schools as an
extra topic. A continued fraction has the general form
b1
X = a0 +
b2
a1 +
b3
a2 +
b4
a3 +
b5
a4 +
a5 +L
where the constants can be written in the more compact form of
 b1 b2 b3 b4 b5 
a 0 , a , a , a , a , a ,L
 1 2 3 4 5 
Every real number can be written as a unique continued fraction, and every
rational number can be written as a continued fraction with a finite number of terms. 49
Using this notation, we get very regular expressions for some important transcendental
numbers. Here are two important examples.

2 = [1, 2, 2, 2, L] ≈ 1.414215K

 2 3 4 5 
e = 1 + 1, , , , ,L ≈ 2.718281828459045K
 2 3 4 5 
A procedure that is repeated is called iterative. Iterative calculation makes it easy
to repeat the calculation until the answer is accurate enough for your purposes. The
tendency of the computation to stabilize near a final answer is called convergence. The
difference between the computed answer and the ideal value is called error. When you
do not know the number of steps needed for convergence, iterative methods can be very
useful.
A procedure that calls itself is called recursive. Recursive methods reveal
fundamental properties of a problem that might otherwise escape notice. Recursive
procedures often take less code to implement than an iterative procedure that
accomplishes the same end goal. This simplicity is very attractive when doing

49
Mark Herkommer, “Continued Fractions”, Number Theory: A Programmer’s Guide, McGraw-Hill
(1998).
Number Systems and Data Representation 116

calculations manually on a calculator. D'Souza has a nice tutorial for C++ programmers
on recursion. 50
Some very important applications of recursion occur in computer science and
mathematics. Recursive methods are sometimes the most efficient approach to
implementing translators. Another important application of recursion is in the topic of
fractals. God seems to have a special love of recursion in applying fractals in nature,
such as the growth of trees or the shape of a shore line.
The apparent ease of recursive calculation is beguiling. Recursive methods are
usually not as efficient as alternative methods of computing the same quantity, when
measured by time required for convergence. For a more detailed discussion, see Acton. 51
Example 1: Evaluate the Square Root of 2.

2 = [1,2,2,2,L]

This is in the form of a continued fraction. We obviously cannot compute an

infinite number of terms. (We would be late for dinner if we tried.) However, this gives
us a method for approximating some important quantities. We can compute the first N
terms, where we choose N in advance. Write this as a continued fraction, and
approximate it by evaluating the first 6 terms.

1 1 1
2 = 1+ ≈1 + = 1+
1 1 1
2+ 2+ 2+
1 1 1
2+ 2+ 2+
1 1 1
2+ 2+ 2+
1 1 2.5
2+ 2+
2 + ... 2

1 1 1
=1+ = 1+ =1+ = 1.414012739
1 1 2.415384615
2+ 2+
1 2.407407407
2+
2.454545455

-1
This evaluation is very easy using a calculator with an reciprocal function, x .
On the TI-83 calculator, perform the following steps:

50
Erwin D'Souza, “Recursion Programming”, CoolComps Inc.,
http://www11.brinkster.com/erwnerve/recursion.htm (2000). 01 July 2002.
51
Forman S. Acton, Numerical Methods that Work, Harper and Row (1970). pp 356-357.
Number Systems and Data Representation 117

(1) 2 ENTER
(2) x –1 + 2 ENTER
(3) Repeat step (2) until the variation of answer from one iteration to the ne xt
iteration is smaller than your stopping criteria.
(4) x –1 + 1 ENTER

Example 2: Evaluate e, approximating it by the first 6 terms of the continued

fraction for e.

 2 3 4 5  2 2
e = 1 + 1, , , , ,L = 1 + 1 + ≅ 1 +1+
 2 3 4 5  2+
3
2+
3
4 4
3+ 3+
5 5
4+ 4+
6 6
5+ 5+
6 +L 6

2 2 2 2
= 1 +1+ =1 +1+ =1 +1+ ≈ 1 +1+
3 3 3 3
2+ 2+ 2+ 2+
4 4 4 3.827586207
3+ 3+ 3+
5 5 4.833K
4+ 4+
5 +1 6

2
≈ 1 +1+ ≈ 2.71844602
2. 783783784
This evaluation requires that you choose the number of steps in advance. It is
very easy using a calculator with an reciprocal function, x -1 . For N = 6, perform the
following steps:
(1) N=6
(2) N ENTER
(3) x –1 * ( N – 1 ) + ( N – 2 ) ENTER
(4) N=N–1
(5) Repeat steps (3) and (4) until N = 2 (that is, until N – 2 = 0).
(6) + 1 ENTER
Unfortunately, the usual arithmetic operations of addition, subtraction,
multiplication, and division are not easy for continued fractions. Some tutorials on
Number Systems and Data Representation 118

continued fractions are: Cut-the-Knot 52 , Doctor Luis 53 , Edward G. Dunne 54 , and Mark A.
Herkommer 55 .
Continued Fraction Problem
Problem 1. The golden ratio is the solution to the equation r2 = r + 1. Divide both
sides by r to get the equation r = 1 + 1/r . Substitute (1 + 1/r ) for r in the
denominator. Continue this forever. The equation r = 1 + 1/r is called a recursive
equation. With the substitutions done, it looks like

1
X = 1+
1
1+
1
1+
1
1+
1
1+
1 +L

Evaluate the first six terms. Show all your steps.

1
X = 1+
1
1+
1
1+
1
1+
1
1+
1

52
Cut-the-Knot, http://www.cut-the-knot.com/do_you_know/fraction.shtml
53
Doctor Luis, “Continued Fractions”, The Math Forum, Drexel University (2002).
http://www.mathforum.org/dr.math, 11 JUN 2002.
54
Edward G. Dunne, “Primes and Continued Fractions”, American Mathematical Society.
http://www.research.alt.com/~njas/sequences/DUNNE/TEMPERMENT.HTML.
55
Mark Herkommer, “Continued Fractions”, Number Theory: A Programmer’s Guide, McGraw-Hill
(1998).
Number Systems and Data Representation 119

Finite Word Length Arithmetic

When we learned arithmetic in elementary school, we allowed numbers to grow
without limit on the number of digits we were allowed to use. We were permitted to use
infinite-precision arithmetic. If we needed to expand the number of digits on the right
end of a number as a result of multiplication or addition, we just wrote the digit.
Today’s binary computers store data in words with a finite number of binary
digits. The number of numbers that can be represented exactly is finite and fixed.
This has some important implications. The only way to increase the number of numbers
that can be represented is to increase the number of bits. This can be done by increasing
the number of bits per word (hardware), or by using multiple words (software or
hardware).
Codes for representing numbers are designed so that arithmetic operations
produce results that are usually accurate enough for applications of interest. Sometimes,
we can take advantage of the finite size to do useful operations easily, such as doing
subtraction by using complement arithmetic.
We cannot represent all numbers exactly. This leads to several kinds of errors
due to arithmetic using finite length word size.
Errors Due to Finite Word Size
Errors due to finite length word size include the following.
• Approximation errors
o Round off
o Truncation
• Overflow
• Underflow
In addition to these errors, logic can detect arithmetic errors such as dividing by zero by
testing the value of the denominator before or during division. This cause is not due to
finite length arithmetic, but it does cause a finite length word overflow error.

Approximation
There are gaps between numbers that we represent exactly. For example, we
cannot represent exactly the number 1/3 in base two. If the number we need cannot be
represented exactly, the selection of a number we can represent results in an
approximation error.
Evaluation of functions require time and resources. Many important functions
cannot usually be computed exactly, such as square root, trigonometric functions, and
logarithms. Engineers and scientists also use other Special Functions important to their
work which require approximation. Such functions are evaluated using numerical
methods or tabled values.
Number Systems and Data Representation 120

The accumulation of approximation errors can easily lead to large errors in a

sequence of computations. The study of such errors and methods to minimize such errors
are topics in numerical methods.

Round Off
Round off error is the difference between the exact number desired and the closest
number that can be represented using selected code for the finite length word. The usual
result is to increase the variance of the error as computations proceed.

Truncation
Truncation is a special case of approximation error that occurs by discarding bits
without rounding to the closest approximation. This is the “Floor” operator. The usual
result is to produce an answer biased closer to zero.

Overflow
Overflow is the condition of producing a result larger than can be represented
with the finite length word.
Overflow can occur by adding or multiplying two very large numbers to produce
a number even larger that cannot be contained within the word. A more common cause is
to divide a large number by a number very close to zero.

Underflow
Underflow is the condition of producing a result that should be non-zero that is
closer to zero than can be represented by the word using the selected code.
Underflow can occur by dividing a small number by a large number.
Arithmetic Examples
Strategies for performing arithmetic with finite length of computer words has
been to work with integers (fixed point), an exponential notation (floating point)
analogous to scientific notation, or significance arithmetic. 56 Fixed point arithmetic
places the burden on the programmer of keeping track of which part of the word
represents whole numbers, and which part represents fractions. Further coding and work
is done if the range of numbers includes negative numbers. Significance Arithmetic
explicitly includes quantifying accuracy and precision of an operation.
The purpose of the examples below is to get you thinking in terms of what has to
be considered when working with only a fixed number of digits. These examples assume
a fixed point notation using decimal arithmetic.
Suppose we are allowed a 5 digit number, with three digits to the left of the
decimal point and two digits to the right of the decimal point. We consider addition,
multiplication, and division. Traditional subtraction is not affected. (Think about why
this is true.)

56
Weisstein, Eric W. "Significance Arithmetic.", MathWorld--A Wolfram Web Resource (10 February
2008). http://mathworld.wolfram.com/SignificanceArithmetic.html visited 12 February 2008.
Number Systems and Data Representation 121

Addition example: 555.55 + 555.55

Throw Col 2 Col 1 Col 0 Col –1 Col –2
Away
5 5 5 .5 5
5 5 5 .5 5
Carry 0 0 0 .0 0
Intermediate
1 1 1 1 .1
Sum
Answer 1 1 1 1 .1 0

Multiplication example: 689.76 x 3.2

Col Col Col Col Col Col Col Col –
5 4 3 2 1 0 –1 2
A 6 8 9 .7 6
times B 3 .2
Carry 1 1 1 1 1 1
Multiply by 2 6 8 4 2
Column (-1)
Intermediate 1 1 3 7 9. 5 2
Sum
Carry 2 1 2 2 2 1
Multiply by 8 4 7 1 8
Column (0)
Intermediate 2 1 1 1 1
Partial Carry A
Intermediate 2 1 1 9 6. 1 3 2
Partial Sum A
Intermediate 2 1
Partial Carry B
Intermediate 2 2 1 0 7 2 3 2
Partial Sum B
Answer 2 2 0 7 2 3 2
The assumed rules define which 5 digits are retained in the answer. If you
retained Column 0 as the baseline, the reported answer would be 207.20. The usual
practice is to retain the highest order digits, which gives us 2207.2 for this problem. If
you are limited to only 5 digits, you will lose some information no matter what you do.
Notice that the number of additions needed depend upon the number of times we
need to consider a carry. Designers of multiplication circuits for computers think
carefully about how to reduce the number of steps involved in multiplication. Good
designs require thoughtful insight into the basic steps of multiplication, and attention to
common special cases.
Number Systems and Data Representation 122

Division example: Some processors place the dividend (numerator) initially into
a register combination that is twice the length of the word being used for storage in main
memory. For our 5 digit example, the dividend would be placed into a 10 digit register
pair. The result consists of a quotient and a remainder, each in a storage- length word.
The notation below is similar to the format for doing division by hand in
elementary schools in Italy. 57 It is a convenient notation for making explicit which
quotient digit is associated with which multiplication. This results in fewer mistakes
when done by hand.
Col Col Col Col Col Col Col Col Col Col
2 1 0 –1 –2 –3 –4 –5 –6 –7
Divisor Divi-
dend
127.26) 8 1 7 .4 5 0 0 0 0 0 Quotient
Multiply 8 9 0 .6 2 (7
Subtract 2 6 .8 3 0 0 0 0 0 (.
Multiply 2 5 4 5 2 (2
Subtract 1 .3 7 8 0 0 0 0
Multiply 1 2 7 2 6 (1
Subtract .1 0 5 4 0 0 0
Multiply 0 0 0 0 0 (0
Subtract .1 0 5 4 0 0 0
Multiply 8 9 7 4 5 (7
Subtract 1 5 6 5 5 0

The answer is approximately 7.2107

International Notation Systems for Arithmetic
Notation for writing arithmetic varies around the world, both the symbols used for
arithmetic operators, and the layout of the problem when done by hand. For example, it
is common that the symbol for multiplication is a dot ( · ) rather than ( x ). Computer
programming languages often use the asterisk (*). The notation used in America for the
division problem 13/7 is expressed in Germany as 13 : 7.
Notation can suggest novel or more efficient methods of solving problems. The
following are from Germany. 58

Manual Multiplication in Germany

Compared to multiplication taught in America, the German notation places the two
numbers to be multiplied on the same line, separated by a centered dot used as the
multiplication symbol. Multiplication begins with the left-most digit of the multiplier,
and proceeds to the right. An example is presented twice. The first example is the
57
Student from Italy (c. 2001)
58
Karla Gaudet, example with instructions. (05 January 2008)
Number Systems and Data Representation 123

compact version as it would normally be done. The second version goes step by
miniature step.

Multiplication example. 324 · 961

3 2 4 · 9 6 1
2 9 1 6
1 9 4 4
3 2 1
3 1 1 3 6 4
The multiplier is 961.
The “carry” operation is normally done mentally without writing it down. The set
of tables below show the process in individual steps, with the “carry” recorded explicitly.
Step 1
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 3 3
Multiply 4·9 6
Step 2
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 1 1 3
Multiply by 2·9 8 6
Step 3
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 2 1 3
Multiply by 3·9 7 8 6
Step 4
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Number Systems and Data Representation 124

Step 5
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 2 2
Multiply 4·6 4
Step 6
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 2·6 2 4
Step 7
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Longer Variation A:
Step 8A
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Subtotal 2 1 9 4 4
Step 9A
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 8 2 4
Subtotal 2 1 9 4 4
Carry
Multiply 4·1 4
Number Systems and Data Representation 125

Step 10A
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 8 2 4
Subtotal 2 1 9 4 4
Carry
Multiply 2·1 2 4
Step 11A
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 8 2 4
Subtotal 2 1 9 4 4
Carry
Multiply 3·1 3 2 4
Step 12A
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 8 2 4
Subtotal 2 1 9 4 4
Carry
Multiply 3·1 3 2 4
Subtotal 3 3 2 4
Number Systems and Data Representation 126

Step 13A. Add the subtotals.

Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 8 2 4
Subtotal 2 1 9 4 4
Carry
Multiply 3·1 3 2 4
Subtotal 3 3 2 4
TOTAL 3 1 1 3 6 4
End of problem.
Shorter Variation B:
Step 8B. In this version, the next total includes the subtotal immediately above it.

Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Subtotal 2 3 1 1 0 4
Carry
Multiply 4·1 4
Step 9B.
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Subtotal 2 3 1 1 0 4
Carry
Multiply 2·1 2 4
Number Systems and Data Representation 127

Step 10B.
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Subtotal 2 3 1 1 0 4
Carry
Multiply 3·1 3 2 4
Step 11B. In this version, the next total includes the subtotal immediately above
it.
Multiplicand.multiplier 3 2 4 · 9 6 1
Carry 2 1 3
Multiply 7 8 6
Subtotal 1 2 9 1 6
Carry 1 1 2
Multiply 3·6 8 2 4
Subtotal 2 3 1 1 0 4
Carry
Multiply 3·1 3 2 4
TOTAL 3 1 1 3 6 4

Manual Division in Germany

In Germany, the symbol used for division is the colon (:) rather than the slash (/) or the
division symbol ( ÷ ). The format of multiply and subtract is the same as done in
America. The notation differs by
• Placing the divisor to the right of the numerator
• Placing the quotient to the right rather than on top of the numerator.
• No distinguishing division symbol to the left (v) and line above the numerator
Number Systems and Data Representation 128

2 4 5 : 2 3 = 10.652173
2 3
1 5 0
1 3 8
1 2 0
1 1 5
5 0
4 6
4 0
2 3
1 7 0
1 6 1
9 0
6 9
Number Systems and Data Representation 129

Complement Arithmetic
Introduction to Complement Arithmetic
Complement arithmetic is an efficient way to perform subtraction involving
numbers represented by a finite length word. A major benefit is that it permits subtraction
without having to think about borrowing from adjacent columns when you use finite
length arithmetic. Binary digital computers use “One’s Complement” or “Two’s
Complement” arithmetic internally. It is very useful for computer professionals to be able
to do this in hexadecimal or octal for tracing through memory dumps to examine the
performance of a program for errors.
From a computer’s point of view, the advantage is that subtraction can be
implemented as a combination of complement and addition. No extra subtraction
hardware is required. The logical complement operation is very quick to do.
If you declare that numbers having a one in the most significant bit are negative,
and those with zero in the most significant bit are positive, then you can use two’s
complement notation to signify negative numbers in binary. With this convention, radix
complement (two’s complement in binary) provides a unique representation of the
number zero. Without this convention, you cannot tell a positive number from a negative
number by just looking at the bit pattern.
Base 10 Complement Arithmetic
We will begin with an illustration in base ten. We learned base 10 arithmetic
because most humans have 10 digits (8 fingers, 2 thumbs)… the original digital
calculator. If we want to perform the subtraction 724 - 538, we would do the following.

After 6 11 14
Borrow

7 2 4

- 5 3 8

Result 1 8 6

The problem with this approach is that you must determine if you need to borrow
from the column immediately to the left. If so, you add the amount borrowed from the
current column and subtract it from the left-adjacent column. After completing the
borrow operation, you then perform the subtraction. Humans can do this without much
additional attention when we work with base 10 because we have practiced it for years.
It is not quite as efficient as we would like for implementation on a computer. We can
make this easier.
Suppose we are only allowed to use 3-place numbers, 000 through 999. We are
required to represent all positive and negative numbers we work with by using these
Number Systems and Data Representation 130

numbers. To represent a number using complement arithmetic using base 10, do the
following.

9 9 9

Subtract 5 3 8

Intermediate 4 6 1
Result

Add 0 0 1

Complement 4 6 2

The top line consists of a finite length word filled with our highest digit in our
base 10 number system, which is 9. Recall, the digits in base 10 run from 0 through 9.
The next line is our traditional negative number. If our negative number were only 2
digits, it would be treated as having leading zeros to the left. The number in the second
row is subtracted from the top row. Notice that no borrows are required from adjacent
columns. The third row is the result of this subtraction. In base 10, this number is the
“Nine’s Complement”.
The final operation is to add 1 to the nine’s complement. This produces our
“Radix Complement”, where the radix of the number system is 10. This is our negative
number. To complete the subtraction, add this radix complement to the original number.
Throw away any digits that carry beyond the third column.

“Throw
Away”
Column

Original
7 2 4
Number

Radix Add à
4 6 2
Complement

Final
1 1 8 6
Result

The result is 186.

Try some examples. Using 4 digits, compute: 3928 - 489 = 3439.
Radix Complement of -489 is 9511. 3928 + 9511 = 1|3439 = 3439. The vertical
bar separates the 4 digit number (which we keep) from the fifth digit (which we throw
away).
Try: -239 - 398 using 4 digits.
Number Systems and Data Representation 131

Base 16 (Hexadecimal) Complement Arithmetic

The principles with base 16 are the same as for base 10, except that we have 16
digits (0 through F) to use. Instead of subtracting from 9999, we subtract from FFFF.
The problem is to compute:

7 D 4

- 5 C E

Result 2 0 6

Begin by finding the 15’s Complement of 5CE.

F F F

- 5 C E

Result A 3 1

Find the Radix Complement by adding 1 to the 15’s Complement.

A 3 1

Add 1

Radix A 3 2
Complement

To perform the original subtraction, we now do addition. Refer to the Hexadecimal

Addition Table.
Number Systems and Data Representation 132

“Throw
Away”
Column

Carry 1 1

Original 7 D 4
Number

Radix A 3 2
Complement

Result 1 2 0 6

Drop the digit carried into position 4.

The answer is 206H, where the suffix H identifies the number as a hexadecimal number.
Try another one!

1 F 3

Subtract - 4 6 7

Result

Hmm… How do we do this? The negative number is bigger than the positive number!
Let’s try.
Find the radix complement of 467H.

F F F

- 4 6 7

B 9 8

+ 1

Radix B 9 9
Complement
Number Systems and Data Representation 133

Add the radix complement to 1F3H.

Carry 1

1 F 3

+ B 9 9

D 8 C

The difference between 1F3H and 467H is D8CH.

Convert this back to minus a positive number.

F F F

- D 8 C

2 7 3

2 7 4

Append the minus sign. The answer in signed-magnitude format is -274H.

Base 2 (Binary) Complement Arithmetic
Consider the subtraction problem: 11002 - 01112 . The first step is to convert the
negative number to radix complement.

1 1 1 1

- 0 1 1 1

1 0 0 0

+ 1

Radix 1 0 0 1
Complement

There is an historical anomaly in the name of this representation. The number

“1000” is called the “one’s complement of 0111”. The number “1001” is called the
“two’s complement of 0111”. This is the concept that people nation-wide think of when
you mention “two’s complement”. The problem occurs when you work with base 3. It
leads to an ambiguity of terminology. The term “radix complement” is unambiguous, but
many people do not understand the term “radix”. The next step is to add the radix
complement.
Number Systems and Data Representation 134

“Throw
Away”
Column

Carry

Original 1 1 0 0
Number

Add 1 0 0 1

Final 1 0 1 0 1
Result

The answer is 01012 .

Try the problem in decimal to see if the answer makes sense.
11002 = 1210 .
01112 = 710 .

12 - 7 = 510 = 01012 .

Oops… it worked! J

Complement Arithmetic Problems

Problem 1. 3-digit Base 10: Compute 33 – 15.

9 9 9 “Throw
Away”
Column

Subtract 1 5 Original 3 3
Number

Intermediate Radix Add à

Result Complement

Add 0 0 1 Final
Result

Radix
Complement
Number Systems and Data Representation 135

Problem 2. 3-digit Base 10: Compute 33 – 45.

9 9 9 “Throw
Away”
Column

Subtract 4 5 Original 3 3
Number

Intermediate Radix Add à

Result Complement

Add 0 0 1 Final
Result

Radix
Complement

Convert this back to minus a positive number.

9 9 9

Subtract

Intermediate
Result

Add 0 0 1

Final Result

Append the minus sign to get the signed- magnitude representation.

Number Systems and Data Representation 136

Problem 3. 3-digit Hexadecimal. Compute FAD16 – DA116

Use the Hexadecimal Addition Table to perform hexadecimal addition.

F F F “Throw
Away”
Column

Subtract D A 1 Original F A D
Number

Intermediate Radix Add à

Result Complement

Add 0 0 1 Final
Result

Complement

Problem 4. 3-digit Hexadecimal. Compute FAD16 – FEA16

F F F “Throw
Away”
Column

Subtract F E A Original F A D
Number

Intermediate Radix Add à

Result Complement

Add 0 0 1 Result

Radix
Complement
Number Systems and Data Representation 137

Convert this back to minus a positive number.

“Throw F F F
Away”
Column

Subtract

Intermediate
Result

Add 0 0 1

Final Result

Notice that all the numbers used in computation are positive numbers. The
purpose of complement arithmetic is to handle negative numbers without using signed
arithmetic. It is possible since we are using finite sized words to represent numbers.
We attach the minus sign to the result as a notation that the number is a negative
number. Notice that FFF (as a positive number) is greater than FC3 (as a positive
number), so FFF – F3C > 0. Therefore 03C + 1 > 0. We attach a minus sign to 03D
because we started with FAD – FEA. We note that FAD < FEA, so FAD – FEA < 0 .
Problem 5. 8-digit Binary. Subtract 11 01102 – 1 11012

“Throw 1 1 1 1 1 1 1 1
Away”
Column

Subtract 1 1 1 0 1

Intermediate
Result

Add 0 0 0 0 0 0 0 1

Radix
Complement
Number Systems and Data Representation 138

“Throw
Away”
Column

Original
1 1 0 1 1 0
Number

Radix Add à
Complement

Final
Result

Problem 6. 8-digit Binary. Subtract 11 01102 – 1011 11012

1 1 1 1 1 1 1 1

Subtract 1 0 1 1 1 1 0 1

Intermediate
Result

Add 0 0 0 0 0 0 0 1

Radix Complement

“Throw
Away”
Column

Original
1 1 0 1 1 0
Number

Radix Add à
Complement

Result
Number Systems and Data Representation 139

Convert this back to minus a positive number.

1 1 1 1 1 1 1 1

Subtract

Intermediate
Result

Add 0 0 0 0 0 0 0 1

Final Result

Again, notice that all the numbers used in computations are positive numbers.
We attach a minus sign to the result to tag it as a negative number since our original
problem is 11 01102 – 1011 11012 where 11 01102 < 1011 11012 .
See if your answers are correct.

Fast Two’s Complement

There is a fast method for doing two’s complement without doing an addition.
Locate the position of the right- most 1-bit. For example, in the word below, the right-
most bit is in position 3.

Bit Position 7 6 5 4 3 2 1 0

Example: 1 1 0 0 1 0 0 0

To take the two’s complement of this example, leave bits in positions 3, 2, 1, and
0 alone. Take the one’s complement of bits in positions 7, 6, 5, and 4. The answer for
this example is:

Bit Position 7 6 5 4 3 2 1 0

Example: 0 0 1 1 1 0 0 0

In general, find the position of the right-most 1-bit. Take the one’s complement
of all the higher order bits. This approach for finding two’s complement is good for
hardware design.

Fast Two’s Complement Problems

Problem 1. Consider the binary number 001101002 .

a. What bit position is the right- most 1-bit in?
b. What bit positions are left alone when writing the fast two’s complement?
c. Write the two’s complement of this binary number.
Number Systems and Data Representation 140

Problem 2. Consider the binary number 110100002 .

a. What bit position is the right- most 1-bit in?
b. What bit positions are left alone when writing the fast two’s complement?
c. Write the two’s complement of this binary number.
Number Systems and Data Representation 141

Binary Coded Decimal Excess-3 Notation

George Robert Stibitz of Bell Labs in 1939 proposed representing decimal
numbers by adding 3 to the Binary Coded Decimal representation. 59 It is called “Excess-
3” because the code is greater than the reference value by 3. Instead of using the word
“excess”, the shorter notation “XS” is used.
This simplified the hardware implementation. It is much easier to take the
complement for subtraction. It is easier to detect a “carry” during addition. In the table
below, the binary code is given for BCD and BCD XS-3 for comparison.
Table 51. Complements in Decimal, BCD, and BCD Excess-3

Decimal 0 1 2 3 4 5 6 7 8 9
Complement
of Decimal 9 8 7 6 5 4 3 2 1 0
(9 – x)

BCD 0000 0001 0010 0011 0100 0101 0110 0111 1000 1001
Complement of
1111 1110 1101 1100 1011 1010 1001 1000 0111 0110
BCD

BCD XS-3 0011 0100 0101 0110 0111 1000 1001 1010 1011 1100
Complement of
1100 1011 1010 1001 1000 0111 0110 0101 0100 0011
BCD XS-3
Notice that the complement of BCD XS-3 produces the BCD XS-3 code for
the decimal complement. For example, the BCD XS-3 code for decimal 5 is
binary 1000. The complement is 0111, which is the BCD XS-3 code for
decimal 4. The complement of decimal 5 is 4 (ie, 9 – 5 = 4 ). The BCD XS-3
code produces a useful result.
On the other hand, the BCD code (not BCD XS-3 code) for decimal 5 is
binary 0101. The complement is 1010, which is not a code for any decimal
digit in the BCD (not XS-3) code.
Addition with BCD XS-3
The concept of addition using BCD Excess-3 is simple. Implementation is
easier to handle in hardware than BCD without Excess-3 coding because of
the symmetric placement of the plain text within the code.
The details that need attention are how to handle situa tions that result in a
sum exceeding the code. There are two ways for the sum to overflow: code
overflow and word overflow. In both cases, a carry bit C is set to 1. The
value of C is added to the next higher significant digit.

59
Paul E. Ceruzzi , “Building the Complex Number Computer”, Reckoners: The Prehistory of the Digital
Computer, from Relays to the Stored Program Concept, 1935-1945, Chapter 4: Number, Please: Computers
at Bell Labs, pages 80 – 84, Greenwood Press (1983). http://ed-thelen.org/comp -hist/Reckoners-ch-4.html
visited 05 February 2008.
Number Systems and Data Representation 142

• A code overflow exists when the sum falls outside of the

boundaries of the code. In the table below, valid codes for decimal
numbers are from 00112 to 11002. Any answer that falls outside
this range produces a code overflow.
• A word overflow exists when the sum produces a value that
exceeds the largest value a word is able to represent. For the 4-bit
word, this value is 11112.
Table 52. Binary Coded Decimal Excess-3

X=4
D=3 P=7
Decimal 0 1 2 3 4
Code 0 1 2 3 4 5 6 7
Binary 0000 0001 0010 0011 0100 0101 0110 0111
Code Overflow
Y=5
Q=8 V=13
Decimal 5 6 7 8 9
Code 8 9 10 11 12 13 14 15
Binary 1000 1001 1010 1011 1100 1101 1110 1111
Code Overflow

Let X and Y be two single digit decimal numbers. Let D be the Displaceme nt of
the Excess code, which is 3 for this system. The BCD XS-3 codes are
• P = X + 3 = X + Excess
• Q = Y + 3 = Y + Excess
To get the coded sum, S = (P + Q) – Excess = (P + Q) – D = (P + Q) – 3 = (X + Y) +
Excess. If the value of S exceeds the valid code, a carry value C must be set to 1 and the
value of S adjusted. There are several special cases as illustrated in the table below.
Table 53. Code and Word Overflow Carry Zones for BCD Excess-3 Addition

0 1 2 3 4 5 6 7 8 9
0 00 01 02 03 04 05 06 07 08 09
1 01 02 03 04 05 06 07 08 09 10
2 02 03 04 05 06 07 08 09 10 11
3 03 04 05 06 07 08 09 10 11 12
4 04 05 06 07 08 09 10 11 12 13
5 05 06 07 08 09 10 11 12 13 14
6 06 07 08 09 10 11 12 13 14 15
7 07 08 09 10 11 12 13 14 15 16
8 08 09 10 11 12 13 14 15 16 17
9 09 10 11 12 13 14 15 16 17 18
Number Systems and Data Representation 143

In the addition table, the right digit is the coded sum, the left digit is the carry
digit to be added to the next higher significant digit. Since the arithmetic is being done in
binary, the carry digit is either a zero or one, and is a single binary digit (bit).
The sum of the BCD XS-3 values ((P – D) + Q) may result in an overflow
condition which produces a wrong answer prior to adjustment. There is no code or word
overflow for the upper left corner. The band that includes (5 + 5) has code overflow but
no word overflow. The band that includes (6 + 7) has both word and code overflow. The
bottom right corner that includes (8 + 8) has word overflow, but no code overflow. When
either overflow condition occurs, the carry bit C is set to 1.
In the (Nassi-Schneiderman) flow chart for BCD Excess-3, the final Sum is S.
Variables E and R are intermediate temporary variables. The value of E incorporates the
value of the carry bit, which allows this addition procedure to apply for higher order
positions of a number. V is the first code position past the decimal table. The quantity
(D – V) adjusts the intermediate sum Result R to wrap around the fixed length table to
obtain the correct BCD XS-3 Sum.

Figure 7. BCD Excess-3 Addition Flow Chart

BCD XS-3 Addition Examples with Input Carry = 0

When you look at examples given in tables below, also look at the table of Binary
Coded Decimal Excess-3. Beginning with the code ent ry for the first number of an
addition problem, count to the right in the table for the number of columns given by the
second number of the addition problem. Note what happens if the result falls outside the
range of permitted codes.
Number Systems and Data Representation 144

Table 54. Example Set 1 of BCD XS-3 Addition with Input Carry C = 0

Row 5+2 5+4 5+7 5+9

1 5
1000 5 1000 5
1000 5 1000
2 -D
1101 -D 1101 -D
1101 -D 1101
3 10101 10101 10101 10101
4 + 2 0101 +4 0111 +7 1010 + 9 1100
5 NWO NWO NWO WO
NCO 1010 NCO 1100 CO 1111 CO 10001
Cß1 Cß1
6 +D - V 0110 0110
7 10101 0111
8 C=0 S=7 C=0 S=9 C=1 S=2 C=1 S=4
9 Sum 7 9 12 14
In Row 2, D = Displacement of code = 3. Thus, - D = - 3, which is coded in two’s
complement in Row 2. In Row 5, WO = Word Overflow, NWO = No Word Overflow,
CO = Code Overflow, NCO = No Code Overflow. In Row 8, C = Carry, S = Sum.
The overflow bit (left- most bit) in Row 3 is discarded and not used in the addition
of the value shown in Row 4.
Problem (5+2) illustrates the case of no word overflow and no code overflow. It
is a case that is not near a boundary of the code overflow identification table. Other
examples are just before a boundary (5 + 4), (5 + 7), and (5 + 9). A later example will
repeat these problems, but with an input carry C = 1, which will push the result over the
boundary.
Problem (5 + 4), like (5 + 2), has no overflow.
Problem (5 + 7) shows the case of a code overflow, but no word overflow. Notice
that the carry bit is set, C ß 1, in row 5. The overflow in Row 7 is ignored and does not
affect setting the carry bit.
Problem (5 + 9) illustrates the case of both a word overflow and a code overflow.
Anytime either overflow condition occurs, the carry bit is set, C ß 1.
Another example set includes the case of word overflow, but no carry overflow
(9 + 9).
Number Systems and Data Representation 145

Table 55. Example Set 2 of BCD XS-3 Addition with Input Carry C = 0

Row 9+0 9+3 9+6 9+9

1 9 1100 9 1100 9 1100 9 1100
2 -D 1101 -D 1101 -D 1101 -D 1101
3 11001 11001 11001 11001
4 + 0 0011 +3 0110 +6 1001 +9 1100
5 NWO NWO WO WO
NCO 1100 CO 1111 CO 10010 NCO 10101
Cß1 Cß1 Cß1
6 +D - V 0110 +D - V 0110 +D - V 0110
7 10101 1000 1011
8 C=0 S=9 C=1 S=2 C=1 S=5 C=1 S=8
9 Sum 9 12 15 18
Problem (9 + 9) illustrates the case of word overflow, but no code overflow.

BCD XS-3 Addition Examples with Input Carry = 1

The difference between this example and the case of Carry = 0 is that the
adjustment in Row 2 has 1 added to it. The value of (1 – D = – D + 1) in binary is 11102 .
The effect of this change ripples through the examples for cases on the change boundary
shown in the table for Code and Word Overflow Carry Zones for BCD Excess-3
Addition.
Table 56. Example Set 1 of BCD XS-3 Addition with Input Carry C = 1

Row 5+2 5+4 5+7 5+9

+1 +1 +1 +1
1 5 1000 5 1000 5 1000 5 1000
2 1 - D 1110 1-D 1110 1-D 1110 1-D 1110
3 10110 10110 10110 10110
4 + 2 0101 +4 0111 + 7 1010 +9 1100
5 NWO NWO WO WO
NCO 1011 CO 1101 CO 10000 CO 10010
Cß1 Cß1 Cß1
6 +D - V 0110 + D - V 0110 +D - V 0110
7 10011 0110 1000
8 C=0 S=8 C=1 S=0 C=1 S=3 C=1 S=5
9 Sum 8 10 13 15
Compared to the same example with Carry C = 0, the Sum for this example with
Carry = 1 is greater by 1.
The interesting case of the next example is the sum of (9 + 9) with input Carry
C = 1. This is the problem of (9 + 10).
Number Systems and Data Representation 146

Table 57. Example Set 2 of BCD XS-3 Addition with Input Carry C = 1

Row 9+0 9+3 9+6 9+9

+1 +1 +1 +1
1 9 1100 9 1100 9 1100 9 1100
2 1-D 1110 1-D 1110 1-D 1110 1-D 1110
3 11010 11010 11010 11010
4 +0 0011 + 3 0110 +6 1001 +9 1100
5 NWO WO WO WO
CO 1101 CO 10000 NCO 10011 NCO 10110
Cß1 Cß1 Cß1 Cß1
6 +D - V 0110 + D - V 0110 +D - V 0110 +D - V 0110
7 10011 0110 1001 1100
8 C=1 S=0 C=1 S=3 C=1 S=6 C=1 S=9
9 Sum 10 13 16 19

Subtraction with BCD XS-3

BCD subtraction (X – Y) is done by adding the radix complement of the smaller
magnitude number to the larger magnitude number, and applying the correct sign. The
radix complement is taken of the whole number, not of individual digits! To do this, take
the binary complement of each BCD XS-3 digit. Then, add 1 to the number, doing a
carry to the next significant digit if necessary.
Table 58. Radix Complement of XS-3 Without Code Overflow

Decimal Excess-3 Remarks

x 302 0110 0011 0101
x 697 1001 1100 1010 1’s Complement
x +1 698 1001 1100 1011 Radix Complement

Table 59. Radix Complement of XS-3 With Code Overflow

Decimal Excess-3 Remarks

x 290 0101 1100 0011
x 709 1010 0011 1100 1’s Complement
Add 1
x +1 1010 0011 1101
Code Overflow.
0000 0000 0110 XS-3 adjust: add 6
Note carry to next
710 1010 0100 0011
significant digit

BCD numbers are stored in signed magnitude format rather than storing negative
numbers in radix complement format. This approach is useful for variable length
numbers. This requires determination of the sign of the result.
Even if negative BCD numbers are stored in radix complement format, a sign is
required. Recall for negative binary numbers in complement form, the range of numbers
Number Systems and Data Representation 147

is almost evenly split between positive and negative numbers. The most significant bit
serves as a sign bit. The problem for BCD is that 10 numbers are coded into a 16 number
code. At least a 19 number code is required to represent 9 positive and 9 negative
numbers, and a signless zero. A sign or flag is needed to identify a number
Complement arithmetic can still be used for subtraction. One approach is to
always prepare both BCD numbers in radix complement format before performing the
operation. A hybrid approach tests signs and magnitudes, and take radix complements
only if necessary. This is shown in the modified Nassi-Schneiderman chart below.

Figure 8. BCD Excess-3 Subtraction Flow Chart

If signs are fetched first, it is sometimes possible to avoid having to compare

magnitudes.
Magnitude comparisons can be done while fetching digits. If lengths of numbers
are known before scanning, it may be possible to determine the relative magnitudes
before completely scanning both numbers by starting comparisons with the most
significant digits. This requires storing the length of the number. This is an advantage if
comparisons, without subsequent arithmetic, are a common need.
Using a finite length word for storing length of a BCD number necessarily places
a limit on the number of digits the BCD number can have. For most business
applications, this is not a practical problem. It is a problem for tasks that produce a
number with many digits, such as computing pi.
Number Systems and Data Representation 148

BCD XS-3 Subtraction Example Same Sign and |X| = |Y|

Let X = 37 and Y = 37. Find Z = X – Y.
Then set Z to zero, and set the sign of Z to “positive”. Problem finished.

BCD XS-3 Subtraction Example Same Sign and |X| < |Y|
Let X = 302 and Y = 608. Find Z = X – Y = – 306. Let the finite word length be
3 decimal digits long.
Let the internal BCD XS-3 code be P = X + 3, Q = Y + 3. Let the BCD XS-3 code
for the sum be S = Z + 3. Let T = Radix Complement, and W = T in BCD XS-3 Code.
Find the radix complement of X.
Decimal BCD XS-3 Binary
X 302 P= 0110 0011 0101BCD XS-3
X 697 1001 1100 1010BCD XS-3
T= X + 1 698 W= 1001 1100 1011BCD XS-3

Add [Radix Complement(X)] + Y.

Decimal W2 W1 W0
T= X + 1 698 W 1001 1100 1011
Q2 Q1 Q0
Y 608 Q 1001 0011 1011
Using a finite length word of 3 decimal digits, discard the overflow digit for Z to
produce the result of addition as Z = 306.
Perform the binary computation according to the BCD Addition flow chart,
working from right to left, one BCD XS-3 digit at a time. To avoid ambiguity, use the
same subscript of digit also for a Carry In to belong to that digit. The subscript for a
Carry Out is for the next higher significant digit.
Table 60. BCD XS-3 Subtraction for Same Sign and |X|<|Y|

698 T2 +Y2 6+6 T1 +Y1 9+0 T0 +Y0 8+8

+608 +C2 +1 +C1 +1 +C0 +0
1 W W2 1001 W1 1100 W0 1011
2 1-D 1110 1-D 1110 -D 1101
3 T T2 10111 T1 11010 T0 11000
4 Q Q2 1001 Q1 0011 Q0 1011
5 WO NWO WO
R R1 : CO 10000 R1 : CO 1101 R0 : NCO 10011
C3 ß1 C2 ß1 C1 ß1
6 +D - V 0110 +D - V 0110 +D - V 0110
7 S 0110 10011 1001
8 C 3 = 1 S2 = 3 C 2 = 1 S1 = 0 C 1 = 1 S0 = 6
9 Sum 13 10 16
Number Systems and Data Representation 149

Finally, the sign of the answer is the complement of the sign of X.

Sign(Z) = Complement[Sign(X)] = negative sign: – .
The final answer is Z = : – 306.

BCD XS-3 Subtraction Example Same Sign and |X| > |Y|
Let X = – 461 and Y = – 320. Find Z = X – Y = – 141. Let the finite word
length be 3 decimal digits long.
Let the internal BCD XS-3 code be P = |X| + 3, Q = |Y| + 3. Let the BCD XS-3
code for the sum be S = Z + 3. Let T = Radix Complement, and W = T in BCD XS-3
Code.
Find the radix complement of Y.
Decimal BCD XS-3 Binary
|Y| 320 Q= 0110 0101 0011BCD XS-3
Y 679 Q = 1001 1010 1100BCD XS-3

T= Y + 1 680 W’= 1001 1010 1101BCD XS-3 Notes:

Code Overflow in T0 .
Generate C1 to next digit.

Concentrate on the right- most digit. Apply the rules for BCD XS-3 Addition for the case
of Code Overflow to complete the process of obtaining the radix complement of Y.

1001 1010 1101 W’ from above

0110 Add Code Overflow adjustment of + D – V
1 0011 C1 and W0 . Add carry C1 to next significant BCD XS-3
digit.
1001 1011 0011 W

Add X + [Radix Complement(Y)]. Work on right column completely, then move

to the next column to the left, applying the proper displacement –D or 1–D if a carry C=1
was generated from the previous column.
Number Systems and Data Representation 150

Table 61. BCD XS-3 Subtraction for Same Sign and |X|>|Y|
Decimal
|X| 461 P2 0111 P1 1001 P0 0100
Displacement 1– D 1110 –D 1101 –D 1101
|X| |X|2 10101 |X|1 10110 |X|0 10001
T= Y +1 680 W W2 1001 W1 1011 W0 0011

Rß|X| + W R R2 1110 R1 10001 R0 0100

Code Word & Code No
Overflow Overflow Overflow
C3 ß1 C2 ß1 C1 ß0
D– V 0110 D– V 0110
R2 +D-V 10100 R1 +D-V 0111 R0 0100
S S2 4 S1 7 S0 4
|Z| 141 |Z|2 1 |Z|1 4 |Z|0 1

Using a finite length word of 3 decimal digits, discard the overflow digit for Z.
Sign(Z) ßSign(X) = negative = –.
Z = – 141.
Number Systems and Data Representation 151

Binary Multiplication
Multiplication in binary follows the same procedural rules as multiplication in
base ten. It is simpler because the multiplication table is much smaller. Remember how
long it took to memorize the multiplication table? The binary table is much easier! The
inputs are X and Y.

Multiply X
0 1
0 0 0
Y
1 0 1
Example 1: Multiply 110102 x 101112 = 26 x 23 = 598. In the example
below, reserve the row immediately below the multiplier to hold the carry when adding
the columns. Label the columns with the position number, beginning from the right hand
side. The row labeled “Column 0” is the result of multiplying X by the digit found in
Column 0 of Y. The row labeled “Column 1” is the result of multiplying X by the digit
found in Column 1 of Y. After multiplying X by each digit of Y, add the resulting
rows. Be sure to account for any carries from the previous column. The number of digits
in the answer will be the total number of digits in the individual multipliers.

Column
4 3 2 1 0
Position Number
X 1 1 0 1 0
Y 1 0 1 1 1
Carry 1 1 1 10 1 1 0 0 0
Column 0 1 1 0 1 0
Column 1 1 1 0 1 0
Column 2 1 1 0 1 0
Column 3 0 0 0 0 0
Column 4 1 1 0 1 0
Sum 1 0 0 1 0 1 0 1 1 0
Check the answer by converting 10010101102 to decimal.
Place Value 512 256 128 64 32 16 8 4 2 1
Sum 1 0 0 1 0 1 0 1 1 0
The answer is 598.
Number Systems and Data Representation 152

Example 2. Multiply 11011.0112 by 101.1102 = 27.375 x 5.75 = 157.40625.

Column
Position 4 3 2 1 0 -1 -2 -3
Number
X 1 1 0 1 1. 0 1 1
Y 1 0 1. 1 1 0
Carry 1 1 1 10 10 10 10 10 1 1 1 0 0
Column -3 0 0 0 0 0 0 0 0
Column -2 1 1 0 1 1 0 1 1
Column -1 1 1 0 1 1 0 1 1
Column 0 1 1 0 1 1 0 1 1
Column 1 0 0 0 0 0 0 0 0
Column 2 1 1 0 1 1 0 1 1
Sum 1 0 0 1 1 1 0 1. 0 1 1 0 1 0
Check the answer by converting 1001 1101.0110 102 to decimal.
Place 128 64 32 16 8 4 2 1 1/2 1/4 1/8 1/16 1/32 1/64
Value
Sum 1 0 0 1 1 1 0 1 0 1 1 0 1 0
The answer is 157.40625 .
Just as with decimal multiplication, the number of digits to the right of the radix
point is the same as the number of digits to the right of the radix point in both original
numbers. In this example, X has 3 digits to the right of the radix point (which you can
call the binary point), and Y has 3 digits to the right of the radix point. The answer has 6
digits to the right of the radix point.
Similarly, the number of digits to the left of the radix point is the same as the
number of digits to the le ft of the radix point in both original numbers. In this example,
X has 5 digits to the left of the radix point, and Y has 3 digits to the left of the radix point.
The answer has 8 digits to the left of the radix point.
Some methods of multiplication are faster than others. One example is the Booth
Algorithm, which is used in some designs of arithmetic logic units. 60 You can view a
simulation of multiplication using this method at the U. Massachusetts web site. 61

60
Yap Siong Chua, CDA 4101 class notes, U. North Florida, Jacksonville, FL (1981).
61
Israel Koren, “Booth’s Algorithm”, (09 December 2005).
http://www.ecs.umass.edu/ece/koren/arith/simulator/Booth/ Visited 04 January 2008.
Number Systems and Data Representation 153

Binary Multiplication
Multiply 111012 by 101102 . = 29 x 22. Show all your work. The answer is 638.
Column
4 3 2 1 0
Position Number
X 1 1 1 0 1
Y 1 0 1 1 0
Carry
Column 0
Column 1
Column 2
Column 3
Column 4
Sum
Check the answer by converting it to decimal.
Place Value 512 256 128 64 32 16 8 4 2 1
Sum
Answer =
Number Systems and Data Representation 154

Hexadecimal Multiplication
Remember memorizing the times-ten table? You can do the same for
hexadecimal. The multiplication table for hexadecimal is given below. Use this for
hexadecimal multiplication just as you use the decimal multiplication table for decimal
multiplication.
Table 62. Hexadecimal Multiplication Table
0 1 2 3 4 5 6 7 8 9 A B C D E F
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 1 2 3 4 5 6 7 8 9 A B C D E F
2 0 2 4 6 8 A C E 10 12 14 16 18 1A 1C 1E
3 0 3 6 9 C F 12 15 18 1B 1E 21 24 27 2A 2D
4 0 4 8 C 10 14 18 1C 20 24 28 2C 30 34 38 3C
5 0 5 A F 14 19 1E 23 28 2D 32 37 3C 41 46 4B
6 0 6 C 12 18 1E 24 2A 30 36 3C 42 48 4E 54 5A
7 0 7 E 15 1C 23 2A 31 38 3F 46 4D 54 5B 62 69
8 0 8 10 18 20 28 30 38 40 48 50 58 60 68 70 78
9 0 9 12 1B 24 2D 36 3F 48 51 5A 63 6C 75 7E 87
A 0 A 14 1E 28 32 3C 46 50 5A 64 6E 78 82 8C 96
B 0 B 16 21 2C 37 42 4D 58 63 6E 79 84 8F 9A A5
C 0 C 18 24 30 3C 48 54 60 6C 78 84 90 9C A8 B4
D 0 D 1A 27 34 41 4E 5B 68 75 82 8F 9C A9 B6 C3
E 0 E 1C 2A 38 46 54 62 70 7E 8C 9A A8 B6 C4 D2
F 0 F 1E 2D 3C 4B 5A 69 78 87 96 A5 B4 C3 D2 E1
Example: Multiply X = OxA4B2 times Y = C316 .
Sum Carry 1 1
Carry term 0 1 2
Carry term 1 7 3 8 1
X A 4 B 2
times Y C 3
E C 1 6
8 0 4 8
Answer 7 D 7 3 9 6
The answer is 7D7396H.
In the above example, “Carry term 0” contains the carry terms from multiplying
by 3. “Carry term 1” contains the carry terms from multiplying by C. “Sum Carry”
contains the carry terms from addition in the adjacent right column.
Number Systems and Data Representation 155

Hexadecimal Multiplication Problem

Problem 1. Multiply X = FADH times EB16 .
Sum Carry
Carry term 0
Carry term 1
X F A D
times Y E B

Answer
Answer =
Number Systems and Data Representation 156

Fixed Point Binary Division

Division in binary is similar to division in decimal.
Example: Divide 10011101.011012 by 11011.0112 . The first step is to
normalize the division problem to make the divisor an integer. This is a nice, but not
necessary, step. This transforms the problem into dividing
10011101011.012 by 110110112

1 0 1. 1 1
11011011 1 0 0 1 1 1 0 1 0 1 1. 0 1
1 1 0 1 1 0 1 1
1 0 1 1 1 1 1 1 1
1 1 0 1 1 0 1 1
1 0 1 0 0 1 0 0 0
1 1 0 1 1 0 1 1
1 1 0 1 1 0 1 1
1 1 0 1 1 0 1 1
0 0 0 0 0 0 0 0
The answer is 101.112 . This result is consistent with Example 2 of fixed point
multiplication.

Fixed Point Binary Division Problems

Problem 1. Divide 101.1100112 by 11.01012 .

.
110101 1 0 1 1 1 0 0 .1 1
Number Systems and Data Representation 157

Common Fractions in Different Bases

Fractions cannot be exactly represented in a finite number of digits in all number
systems. For some kinds of work, one number system may be better than another number
system. For example, the fraction 1/3 cannot be represented in base 10 using a finite
number of digits. Some fractions can be represented by repeating decimals, and a bar is
placed on top of that group of digits that repeats infinitely. We therefore represent 1/3 in
decimal as
1
= 0.333333333K = 0. 3 = 0.3
3
Not all numbers with fractional parts can be represented in this way because the
fractional part is not a repeating decimal. The numbers e and π are famous examples.
When you look at the stock market, you see the fractional part of prices listed in
multiples of negative powers of two for the high, low, last, and change, such as 5/8 =
0.101, 3/4 = 0.11, 1/2 = 0.1. Use of these fractions permitted stock price data to be stored
in binary on computers exactly, before the invention of Binary Coded Decimal. Stock
markets in the United Stated converted from fractional to decimal systems in early 2001.
You can learn more about the stock market on the Internet at http://www.nyse.com.
Fraction 1/2 1/4 1/8 1/16 1/32
Power of 2 2 –1 2 –2 2 –3 2 –4 2 –5
Binary Fraction 0.10000 0.01000 0.00100 0.00010 0.00001
Here are some examples of fractions expressed in different bases. Fractions with exact
representation with a finite number of digits are indicated in bold. Underlined groups of
digits are repeating digits. The proper notation is to have a solid over-bar. Neither MS
Equation 3.0, nor Corel Equation 2.0, can do that for groups of selected digits.
Table 63. Ordinal Fractions in Different Bases
Ordinal Fractions
Fraction Base 2 Base 3 Base 5 Base 7 Base 10 Base 12 Base 16
1/2 0.1 0.1 0.2 0.3 0.5 0.6 0.8
1/3 0.01 0.1 0.13 0.2 0.3 0.4 0.5
1/4 0.01 0.02 0.1 0.15 0.25 0.3 0.4
1/5 0.0011 0.0121 0.1 0.1254 0.2 0.2497 0.3
1/6 0.001 0.01 0.04 0.1 0.6 0.2 0.2A

1/7 0.001 0.012 0.032412 0.1 0.1428571 0.186A35 0.249

1/8 0.001 0.01 0.03 0.06 0.125 0.16 0.2
1/9 0.000111 0.01 0.021 0.053 0.1 0.14 0.15
1/10 0.00011 0.0022 0.02 0.0462 0.1 0.12497 0.19
Number Systems and Data Representation 158

Table 64. Decimal Fractions in Different Bases

Decimal Fractions
Fraction Base 2 Base 3 Base 5 Base 7 Base 10 Base 12 Base 16
1/10 0.00011 0.0022 0.02 0.0462 0.1 0.12497 0.19
2/10 0.0011 0.0121 0.1 0.1254 0.2 0.2497 0.3

3/10 0.01001 0.0220 0.12 0.2046 0.3 0.37249 0.4C

4/10 0.0110 0.1012 0.2 0.2543 0.4 0.4972 0.6

5/10 0.1 0.1 0.2 0.3 0.5 0.6 0.8

6/10 0.1001 0.1210 0.3 0.43 0.6 0.7249 0.9

7/10 0.10110 0.2002 0.32 0.4620 0.7 0.84972 0.B3

8/10 0.1100 0.2101 0.4 0.543 0.8 0.9724 0.C

9/10 0.11100 0.2200 0.42 0.6204 0.9 0.A9724 0.E6

Failure to account for the inability to exactly represent decimal fractions using
fixed length binary words in the design of the Patriot missile software resulted in the loss
of 28 lives in a Scud attack on 25 FEB 1991 on the American barracks in Dharan, Saudi
Arabia during the Gulf War. See the article by Arnold. 62
Beware of thinking that base 10 is necessarily the best based upon observing that
so many of the fractions above can be expressed exactly with a finite length number. If
God had given us 7 digits, instead of 10, at the end of our upper appendages, we would
be more inclined to prefer the fractions 1/7, 2/7. 3/7, 4/7, 5/7, 6/7, and 1 = 7/7.

62
Douglas N. Arnold, “The Patriot Missile Failure” (23 AUG 2000).
http://www.ima.umn.edu/~arnold/disasters/patriot.html Visited 04 JUN 2002.
Number Systems and Data Representation 159

Prime Numbers and Number System Base

One integer is considered “divisible” by another integer only if the remainder
from the division is zero. A positive integer is called “prime” if it is divisible only by
itself and one. One is declared to not be a prime number.
The properties of the set of integers are independent of the base of the
number system used to express them. The base of the number system in use does not
affect the set of prime numbers. It does not alter which numbers are prime. It does not
alter the number of prime numbers.
There is no “largest” prime number. For every number, there is a larger prime
number. The “infinitude of primes” was proven by Euclid in 300 B.C., and is called
"Euclid's Second Theorem". There is an infinite number of prime numbers. These facts
keep the National Security Agency in business, and keep cryptologists busy and happy
(or unhappy). A statement and proof of Euclid’s Second Theorem can be found on the
Internet. 63
Historical context of important discoveries are interesting. Euclid lived
approximately 325 BC to 265 BC. Euclid of Alexandria was born 2 years before
Alexander the Great died in Babylon (325 B.C.), and 3 years after Aristotle died (322
B.C.). Aristotle had been the personal tutor of Alexander the Great, and was a student of
Plato. 64 There is a strong conjecture that Euclid studied at Plato’s Academy in Athens. 65
Euclid was alive during the reigns of Ptolemy I and Ptolemy II in Egypt. Alexandria,
Egypt, was the intellectual capital of the world. During this period, Ptolemy II ordered
the translation of Jewish Law, producing the Septuagint, one of the most influential
translations ever to be accomplished. Euclid died in Alexandria. Elsewhere, it was
during this period that the Roman Aqueduct was built.
Here are the first five prime numbers, expressed in different bases.

63
Chris K. Caldwell, “Euclid's Proof of the Infinitude of Primes (c. 300 BC)”, The University of Tennessee
at Martin. http://primes.utm.edu/notes/proofs/infinite/euclids.html Visited 04 January 2008.
64
Aine Donovan, David E. Johnson, George R. Lucas, Jr., and Paul E. Roush (eds.), Ethics for Military
Leaders, Simon and Schuster (1998).
65
J J O'Connor and E F Robertson, “Euclid of Alexandria”, School of Mathematics and Statistics,
University of St. Andrews, Scotland (January 1999).
http://www-groups.dcs.st-and.ac.uk/~history/Mathematicians/Euclid.html, 13 September 2002.
Number Systems and Data Representation 160

Table 65. First 12 Prime Numbers in Different Bases

First Dozen Prime Numbers in Different Bases

Base 10 Base 2 Base 3 Base 5 Base 7 Base 12 Base 16
2 10 2 2 2 2 2
3 11 10 3 3 3 3
5 101 12 10 5 5 5
7 111 21 12 10 7 7
11 1011 102 21 14 B B
13 1101 111 23 16 11 D
17 10001 122 32 23 15 11
19 10011 201 34 25 17 13
23 10111 212 43 32 1B 17
29 11101 1002 104 41 25 19
31 11111 1011 111 43 27 1F
37 100101 1101 122 41 31 25

If an integer is not a prime number, it can be divided by a smaller prime number.

Consider the number 19. Try the division in different bases.
Base 2 Demonstration that 19 is prime.
Divide by
2 3
1 0 0 1 1 1 0
10 1 0 0 1 1 11 1 0 0 1 1
1 0 1 1
0 0 1 1 1 1
1 0 1 1
R= 1 0 1
0
R= 1

Divide by
5 7
1 1 1 0
101 1 0 0 1 1 111 1 0 0 1 1
1 0 1 1 1 1
1 0 0 1 1 0 1
1 0 1 0
R2= 1 0 0 R2= 1 0 1
R10 =4 R10 =5
Number Systems and Data Representation 161

11 13
1 1
1011 1 0 0 1 1 1101 1 0 0 1 1
1 0 1 1 1 1 0 1
R2= 1 0 0 0 R2= 1 1 0
R10 =8 R10 =6

17 19
1 1
10001 1 0 0 1 1 10011 1 0 0 1 1
1 0 0 0 1 1 0 0 1 1
R2= 1 0 R2= 0
R10 =2 R10 =0

The first divisor not equal to one that results in a remainder of zero is
100112 = 1910 . Therefore, this is a prime number.
Base 3 Demonstration that 19 is prime.
Divide by
2 3
1 0 0 2 0
2 2 0 1 10 2 0 1
2 2 0
1 1
0 0
R3= 1 R3= 1
R10 =1 R10 =1

Divide by
5 7
1 0 2
12 2 0 1 21 2 0 1
1 2 1 0 2
1 1 R3= 1 2
0 R10 =5
R3= 1 1
R10 =4
Number Systems and Data Representation 162

11 13
1
102 2 0 1 111 2 0 1
1 0 2 1 1 1
R3= 2 2 R2= 2 0
R10 =8 R10 =6

17 19
1 1
122 2 0 1 201 2 0 1
1 2 2 2 0 1
R2= 2 R2= 0
R10 =2 R10 =0
The first divisor not equal to one that results in a remainder of zero is 2013 = 1910 .
Therefore, this is a prime number.
Base 5 Demonstration that 19 is prime.
Divide by
2 3 5
1 4 1 1 3
2 3 4 3 3 4 5 3 4
2 3 3 0
1 4 0 4 R3= 4
1 3 3 R10 =4
R5= 1 R3= 1
R10 =1 R10 =1

Divide by
7 11
2 1
12 3 4 21 3 4
2 4 2 1
R3= 1 0 R3= 1 3
R10 =5 R10 =8
Number Systems and Data Representation 163

Divide by
13 17 19
1 1 1
23 3 4 32 3 4 34 3 4
2 3 3 2 3 4
R3= 1 1 R3= 2 R3= 0
R10 =6 R10 =2 R10 =0

The first divisor not equal to one that results in a remainder of zero is 345 = 1910 .
Therefore, this is a prime number.
Prime Number Problems
Problem 1. a. Show that 7 is a prime number using base 2.
b. Show that 7 is a prime number using base 3.
c. Compare the results.
Number Systems and Data Representation 164

Floating Point Arithmetic

The general concept is called “exponential notation”. There is an extensive
collection of links to web sites on the subject of exponential notation. 66
Fixed Point Numbers with Fractions
Define a fixed point number as a number written in positional notation that
contains a radix point (a period in the United States, or a comma in Europe). The
number immediately to the left of the radix point is in the units position. In the decimal
system, the special name of the radix point is “decimal point”. When the fixed point
number has no fractional part, we call it an integer. Fixed point notation is the notation
normally used by most people in their every-day life, and in business and finance.
Examples of some approximated numbers written in fixed point notation are:
14.5939 kg = 1 slug
π ≈ 3.1415 92653 5897 32384 62643
e ≈ 2.7 1828 1828 45 90 45 2353 60287
γ ≈ 0.5772 15664 90153 28606 06512
Some useful numbers are very large or very small, and writing such numbers in
fixed point notation is inefficient and inconvenient. There are other ways of writing very
large and very small numbers: scientific notation, and floating point notation.
Business data processing often uses fixed point numbers with fractions for
currency. The internal representation for fixed point numbers on computers usually does
not provide for recording the position of a radix point. The program must keep track of
the proper position, and use formatting statements to position the radix point properly.
Scientific Notation
Scientific notation permits us to write very large numbers and very small numbers
using a compact notation. A number written in scientific notation consists of several
parts. An example of a number written in scientific notation is
6.02216 9 x 1023 mol –1
This number is known as Avogadro’s number. 67 It is the number of elementary entities
(molecules) in one mole of a substance (referenced to the number of atoms in 12 grams of
Carbon-12). 68
In the example above, the “6.022169” portion of the number is called the
mantissa, significand, or coefficient. At least one author called it the “base”, but that
name can cause confusion with some other concepts. Except for the number zero, the

66
Mitchell N. Charity, “An Exponential Notation Meta Page” (10 August 2003).
http://www.vendian.org/mncharity/export1/exponential_notation_meta/ Visited 04 January 2008.
67
Handbook of Mathematical Functions With Formulas, Graphs, and Mathematical Tables, Applied
Mathematics Series, AMS-55, Milton Abramowitz and Irene A. Stegun (eds.), National Bureau of
Standards, U.S. Department of Commerce, U.S. Government Printing Office (June 1964).
68
“Unit of Amount of Substance (Mole)”, The International System of Units (SI), Section 2.1.1.6:, Page 8,
NIST Special Publication 330, 2001 Edition, Barry N. Taylor (editor), National Institute of Standards and
Technology (July 2001).
Number Systems and Data Representation 165

mantissa of a number written in scientific notation has one non-zero number to the left of
the decimal point.
The “23” portion of the number is called the characteristic or exponent. The
exponent can be a positive or negative integer. When the exponent is zero, the number is
written without the exponent term. The number “10” is the exponent base.
Avagadro’s number is
6022 16900 00000 00000 00000 = 6.02216 9 x 10 23
Very small numbers written in scientific notation have negative powers of the
base, such as the gravitational constant,
G = 6.6732 x 10 –11 N m2 / kg2
The number
– q = - 1.60219 17 x 10 –19 C
is the elementary charge of an electron, measured in Coulombs. The mantissa is
“-1.6021917”. The characteristic is “-19”. The charge of an electron is negative, as
indicated by the minus sign of the mantissa. The value of the charge of a single electron
is very much less than one, as indicated by minus sign in the characteristic. Written in
fixed point notation, this number is
– 0. 00000 00000 00000 00016 02191 7

Scientific Notation Conversion Problems

Problem 1. Write h = 0.00000 00000 00000 00000 00000 00000 00066 26196 J/s in
scientific notation. This is the Plank constant.
Problem 2. Write c = 2997 92500 m/s in scientific notation. This is the speed of
light in a vacuum.

Scientific Notation Multiplication

Multiplication and division using scientific notation is easy. The charge of
electrons in one mole of hydrogen is the product of the charge per electron times the
number of electrons in a mole of hydrogen. The mantissas are multiplied, and the
characteristics are algebraically added.
(6.02216 9 x 1023 mol –1 ) x (– 1.60219 17 x 10 -19 C)
= (6.02216 9) x (– 1.60219 17) x 10 23 – 19 C/mol
= – 9. 64866 91879 797 x 10 4 C/mol
That was much easier than multiplying
6022 16900 00000 00000 00000 x 0. 00000 00000 00000 00016 02191 7
Here is another example for multiplication.
(3.0 x 10 3 ) x (6.0 x 10 2 ) = (3.0 x 6.0) x 10 3 + 2 = 18.0 x 10 5
Number Systems and Data Representation 166

Scientific Notation Multiplication Problems

Problem 1. Approximate the distance of one light- year. 69
( 2.998 x 108 m/s ) x (3.147 x 107 s/yr ) =
Problem 2. Naively estimate the kinetic energy of the earth revolving around the Sun,
assuming a circular orbit (ignoring Kepler). 70
(1/2) x (5.98 x 1024 kg ) x ( 2.99 x 104 m/s )2 =

Scientific Notation Normalization

Notice that the form of 18.0 x 10 5 is not quite in scientific notation. The
mantissa has more than one digit to the left of the decimal point. The process of
reformatting the number into scientific notation is called normalization.
18.0 x 10 5 = (1.8 x 10 1 ) x 10 5 = 1.8 x (10 1 + 5 ) = 1.8 x 10 6

Scientific Notation Division

Division is almost as easy. Consider dividing 1.5 x 10 4 by 3.0 x 10 5 . The
procedure is to divide the mantissas, algebraically add the characteristics [(+4) + (-5) in
the example below], and normalize the result.
1.5 • 10 4  1.5 
= 4 −5 1 −1 −1
(
−1 −1
)
 • 10 =   • 10 = 0.5 • 10 = 5.0 • 10 • 10 = 5.0 • 10
−2

3.0 • 10 5
 3.0   2

Scientific Notation Addition and Subtraction

Addition and subtraction using scientific notation require more effort. Numbers
must be written in a form having the same exponent before performing addition or
subtraction. For example:
( 2.1 x 10 3 ) + ( 1.4 x 10 2 ) = ( 21.0 x 10 2 ) + ( 1.4 x 10 2 )
= 22.4 x 10 2 = 2.24 x 10 3
It is also correct to write
( 2.1 x 10 3 ) + ( 1.4 x 10 2 ) = ( 2.1 x 10 3 ) + ( 0.14 x 10 3 ) = 2.24 x 10 3
Here is another example:
( 3.2 x 10 3 ) – ( 1.1 x 10 4 ) = ( 0.32 x 10 4 ) – ( 1.10 x 10 4 )
= – 0.78 x 10 4 = – 7.8 x 10 3
It is easy to lose precision when adding or subtracting numbers with very different
exponents. For example:
2 x 10 10 – 0.5 x 10 –10 . = 2 00000 00000 – 0.00000 00000 5
= 1 99999 99999 .99999 99999 5

69
Encyclopedia.com, “Light Year”, http://www.encyclopedia.com. 01 July 2002.
70
George Goth, “The Magnitudes of Physics”, Skyline College, San Bruno CA (08 Aug 2001)
http://www.smccd.net/accounts/goth/MainPages/magphys.htm, 01 July 2002.
Number Systems and Data Representation 167

In physics, it is common to be interested in very small differences of very large

numbers. Consider the pressure of the atmosphere on the surface of the earth. When we
hear sound, we are sensing very small variations in that pressure.
Engineering Notation
Engineering notation is similar to scientific notation, except that the characteristic
is always a multiple of 3. Because of this, the mantissa may have one to three digits to
the left of the decimal point.
Examples:
Fixed Point Engineering Scientific
13,200,000,000.0 13.2 x 109 1.32 x 1010
0.000 000 325 325 x 10 – 9 3.25 x 10 – 7
102,200,000,000,000,000,000,000 102.2 x 1021 1.022 x 1023
The rule for choosing multiples of three is a normalization rule. Notice that
engineering notation is naturally related to the practice in the United States to form
groups of three digits to aid in quickly determining the value of a number. This is
reinforced both in engineering and the SI units of measure by the language adopted for
number prefixes.
Table 66. SI Units of Measure Prefixes

tera giga mega kilo milli micro nano pico

x 10 12 x 10 9 x 10 6 x 10 3 x 10 0
x 10 – 3 x 10 – 6 x 10 – 9 x 10 – 12
Floating Point Notation
Floating point numbers are like numbers in scientific notation. It is all in binary.
The radix point (the period) is either immediately to the right or to the left of the most
significant digit, depending on the specific processor. An example of a floating point
number is:
0.11 x 2 110
The binary fraction 0.112 is the mantissa. Convert its value to decimal as follows:

1 1 3
0.112 = 2 −1 + 2 −2 = + = = 0.75
2 4 4
The binary exponent 1102 equals decimal 6. The value of the exponent term is
2 6 = 6410
The decimal value of the original number is
0.75 x 64 = 48
On a different processor, this same number might have a floating point
representation as
1.1 x 2 109
Number Systems and Data Representation 168

Floating Point Multiplication

Multiplication with floating point numbers is similar to multiplication with
scientific notation. Multiply the mantissas and algebraically add the characteristics.
Example:
( 0.11 x 2 110 ) x ( 0.1001 x 2 100 ) = ( 0.11 x 0.1001 ) x ( 2 110 + 100 )
Perform binary addition for the exponents using the binary addition table.
Carry 1 0 0
1 1 0
+ 1 0 0
Answer 1 0 1 0
When arithmetic is done with floating point numbers, the answer is placed back
into standard form after the operation is finished. Just before normalization, the answer is
0.011011 x 21010 . The process of normalization adjusts the characteristic in order to place
the most significant digit immediately to the right of the radix point.
0.011011 x 21010 = ( 0.11011 x 2 –1 ) x 21010 = 0.11011 x 21001

Floating Point Division

Floating point division is similar to scientific notation division. The procedure is
to divide the mantissas, algebraically add the characteristics, and normalize the result.
Number Systems and Data Representation 169

Example 1. Divide ( 0.1001 x 21001 ) by ( 0.1101 x 2101 ).

= 
0.1001  1001−101
• 2
 0.1101 

Adding the characteristics algebraically is the easy part.

2 1001−101 = 2100
The division is more tedious.

0.1101 0. 1 0 0 1 0 0 0 0 0 0 0 0
The first step is to shift the radix point in both terms to the right by the same number of
places until the divisor is an integer. In this example, shift the radix point by 4 places. In
fractional notation, this is
0.1001 1001
=
0.1101 1101

.1 0 1 1 0 0 0 1 0 0 1 0 0 1
1101 1 0 0 1 .0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 0 1
1 0 1 0 0
1 1 0 1
1 1 1 0
1 1 0 1
1 0 0 0 0
1 1 0 1
1 1 0 0 0
1 1 0 1
1 1 0 0 0

The answer is
0.1011 0001 0010 01 x 2100
= 0.1011 0001 x 2100
= 1011.0001
Notes:
0.1001 x 21001 = 1001 x 2101 = 9 x 25 = 9 x 32 = 288
0.1101 x 2101 = 1101 x 21 = 13 x 2 = 26
288/26 = 11.076923
Number Systems and Data Representation 170

Example 2. Divide ( 0.11101 x 2110 ) by ( 0.101 x 21101 ),

 0.11101 • 2 110   0.10111  110−1101
  =
1101  •2
 0.101 • 2   0.101 
The characteristic is 110 – 1101 = – 111.
For the mantissa:

0.101 0. 1 1 1 0 1 0 0 0 0 0 0 0

Shift the radix point in both terms to the right by the same number of places until
the divisor is an integer. In this example, shift the radix point by 3 places. Perform the
division.

1 .0 1 1 1 0 0 1
101 1 1 1 .0 1 0 0 0 0 0
1 0 1
1 0 0 1
1 0 1
1 0 0 0
1 0 1
1 1 0
1 0 1
1 0 0 0
1 0 1
1 1 1
The result of the division is 1.0111 00 x 2 –101 . Normalize the resulting floating
point notation:
1.0111 00 x 2 –101 = ( 0.1011 100 x 2 1 ) x 2 –101
= 0.1011 100 x ( 2 1 x 2 –101 )
= 0.1011 100 x 2 – 100
= 0.0000 1011 100
Number Systems and Data Representation 171

Example 3. Divide ( 0.10111 x 21011 ) by ( 0.11 x 210 ).

 0.10111 • 2 1011   0.10111  1011−10
  =  •2
 0.11 • 210
  0.11 
The characteristic is 1011 – 10 = 1001.
The mantissa requires more work.

0.11 0 .1 1 1 0 0 0 0 0 0 0 0 0
Shift the radix point in both terms to the right by the same number of places until
the divisor is an integer. In this example, shift the radix point by 2 places. Perform the
division.

0 .1 1 1 1 0 1 0 1
11 1 0 .1 1 1 0 0 0 0 0
1 1
1 0 1
1 1
1 0 1
1 1
1 0 0
1 1
1 0 0
1 1
1 0 0
The answer in binary is
0.1111 01 x 21001
= 1 1110 1010.10
= 256 + 128 + 64 + 32 + 8 + 2 + (2/3) = 490.6

Representation of Floating Point Numbers

Representing numbers in a computer is a design problem. Computers do not have
infinite length words. Words in a computer are finite in length. This means that not all
numbers can be represented exactly. Rhetorical questions:
How should you represent zero?
How should you represent a negative number?
Should you have a representation of infinity? If so, how?
How should you represent a mantissa?
How should you represent a characteristic?
What other representations are desirable? Error conditions? Status
indicators?
Number Systems and Data Representation 172

Several technical references for the following discussion are valuable for future
computer scientists and computer engineers. 71 72 73 74

Negative Numbers
In the discussion that follows, mention will occasionally be made that some
methods of representing numbers result in two representations for zero, +0 and –0. For
testing equality, you usually want a test that +0 = -0 to be true, and it would seem that a
unique value for zero would always be desirable. Alas, such is not always the case.
Distinguishing between these is useful when dealing with something called “branch
cuts”, which is a topic in complex variables. 75 Scientists, engineers, and mathematicians
worry about such things. Computer scientists that do not work with engineers and
scientists, and people involved in business data processing, do not need to consider this.
More will be said about this when discussing IEEE formats for floating point numbers.
How do you provide for negative numbers? There are three major ways of doing
this.
• signed magnitude
• complement
• bias or Excess (XS)
For purpose of discussion, pretend that we are restricted to a 4-bit word. It is
important to start numbering bits from the right hand side, starting with zero. The left-
most bit is called the Most Significant Bit (MSB), and the right-most bit is called the
Least Significant Bit (LSB). For example:
3 2 1 0
MSB LSB
Signed Magnitude
The sign bit approach is to reserve one bit for use to indicate the sign of the
number. When we compare two numbers, a positive number should appear to be larger
than a negative number. It makes sense to assign the most significant bit to be the sign
bit. If a sign bit equal to zero represents a negative sign, and a sign bit equal to one
represents a positive sign, we get the right outcome when we do a comparison.

71
P. H. Abbott, D. G. Brush, C. W. Clark III, C. J. Crone, J. R. Ehrman, G. W. Ewart, C. A. Goodrich, M.
Hack, J. S. Kapernick, B. J. Minchau, W. C. Shepard, R. M. Smith, Sr., R. Tallman, S. Walkowiak, A.
Watanabe, and W. R. White, “Architecture and software support in IBM S/390 Parallel Enterprise Servers
for IEEE Floating-Point arithmetic” IBM Journal of Research and Development, Volume 43, Numbers 5/6
(1999), http://www.ibm.com, 08 June 2002
72
ESA/390 Principles of Operation, International Business Machines, Document Number SA22-7201-06
(1999), http://publib.boulder.ibm.com/cgi-bin/bookmgr/BOOKS/DZ9AR006/CCONTENTS , 10 Jun 2002.
73
Intel® ItaniumTM Architecture Software Developer’s Manual, Volume 1, Application Architecture,
Revision 2.0 (December 2001).
74
David Goldberg, “What Every Computer Scientist Should Know About Floating-Point Arithmetic”,
Computing Surveys, Association for Computing Machinery (March 1991).
75
W. Kahan, “Branch Cuts for Complex Elementary Functions, or Much Ado About Nothing's Sign Bit”;
Chapter 7 in The State of the Art in Numerical Analysis ed. by M. Powell and A. Iserles (1987) Oxford.
Number Systems and Data Representation 173

± 2 1 0
MSB LSB
A complication to using the sign bit approach is that there are two possible
representations of the number zero: a plus- zero, and a minus-zero. The trichotomy test
classifies a number as being
• less than zero
• equal to zero
• greater than zero
When there are two possible representations for zero, the test becomes more
complicated. You must test for both a positive and a negative zero. Two representations
for zero means that one combination of bits is being wasted that could be used to
represent another number. For our example signed nibble, we can represent only 14
unique numbers.
Complement
Negative numbers can be represented in binary using either one’s complement or
two’s complement. Complement arithmetic has the advantage of using positive integer
addition for adding positive and negative numbers. Like the signed magnitude approach,
one’s complement representation has the disadvantage of having two representations for
the number zero.
Two’s complement (radix complement) arithmetic does not have the disadvantage
of ambiguous representations of zero. If you declare that numbers having a one in the
most significant bit is a negative number, you can use two’s complement to
unambiguously represent negative numbers. If you do not agree to that convention, you
have ambiguity. Consider the bit pattern in an example nibble:
0 1 1 0
Is the intended number plus six, or minus ten?
Bias or Excess
Rather than interpreting the most significant bit as an indicator of sign, the bias
approach is to declare that the B smallest numbers are negative numbers. Let X be the
value we want to represent. The number Y is the stored value, which is computed as
Y = X + B. In our 4-bit word, suppose we declare a bias of 4. We have 16 possible
numbers, with the range [-4, 11].
0 0 1 0 is – 2
while the pattern
0 1 1 0 is + 2
and
1 1 1 1 is + 11
This approach allows you to partition your set of numbers in an arbitrary fashion.
Bias representation imposes additional procedural requirements on arithmetic operations.
Number Systems and Data Representation 174

Suppose you want to add -1 plus -1 using a bias of 4. Naïve addition gives us
-1+4=3 0 0 1 1
+ 0 0 1 1
Sum = 6 0 1 1 0
The sum of 6 corresponds to +2, which is not what we want. We want to compute
X1 + X2 = (-1) + (-1) = -2. To get the desired answer, add the two numbers and
subtract the bias. The required computation is
Yanswer = Y1 + Y2 - B
Addition is commutative; you can exchange order of addition without changing
the result. Subtracting the bias before doing the addition can preserve precision in the
result at intermediate steps when using a finite length word for arithmetic, such as when
using a real computer.
Yanswer = Y1 - B + Y2
Bias representation of negative numbers turns out to be very useful when using
exponents, with a desire to limit the number of places reserved for fractions (negative
exponents). The amount of bias is designated in the name. The notation above would be
called XS-4 notation, where “XS” means “Excess”.
It is common on binary computers to allocate half the range of numbers that can
be represented to be negative numbers, though you could design a system differently.
The bias is usually set to be some power of two, or just one less than a power of two,
depending on the processor.
IEEE excess notation declares that when the most significant bit is on, the
represented number is a positive integer. The bit pattern of the excess is all ones, except
for the most significant bit which is zero. When this is done, the most significant bit can
be interpreted as a sign bit, with a negative number having the sign bit turned off. For an
8-bit number, this is called XS-127. The range of possible numbers for this situatio n is
[ -127, 128 ]. A zero characteristic is represented by 0111 11112 .
IBM hexadecimal floating point excess notation declares that when the most
significant bit is turned on, the represented number is a positive integer or zero. The bit
pattern of the excess is all zeros, except for the most significant bit which is a one. For a
7-bit number (which IBM uses), this is called XS-64. The range of possible numbers for
this situation is [ -64, 63 ]. A zero characteristic is represented by 100 00002 .
IEEE IBM
# Bits Decimal Hex Decimal Hex Equivalent
Excess Excess Excess Excess Binary XS
7 64 80 256
8 127 7F
11 1023 7FF
15 16383 7FFF
17 65535 1FFFF
Number Systems and Data Representation 175

Mantissa Normalization and the Characteristic

The details of representing the mantissa and characteristic depend on the
normalization scheme chosen. We saw a precursor to the variation in rules when we
considered the differences between scientific notation and engineering notation.
In a fixed length word computer, you are limited to how many different numbers
you can represent. You can represent a greater range of numbers if you are willing to
sacrifice how densely packed the representable numbers are. Suppose you have only 8
bits to represent all numbers, with allocation of bits as shown below. Because of using
one bit for a sign, only 255 different numbers can be represented using these 8 bits, rather
than 256, because +0 = -0.
7 6 5 4 3 2 1 0
± Mantissa Characteristic
With binary normalization, the characteristic tells you how many places the radix
point of the fixed point binary number was shifted to satisfy the normalization rule. For
example, using two’s complement in the characteristic, the number
7 6 5 4 3 2 1 0
0 1 1 0 0 0 1 1
± Mantissa Characteristic
has a characteristic of 0112 = 3. The value in binary of 0.11 x 23 = 110.02 . The
number
7 6 5 4 3 2 1 0
0 1 1 0 0 1 0 1
± Mantissa Characteristic
has a two’s complement characteristic of 1012 . In signed- magnitude, this number is
-0112 = -3 . The value of the floating point number is
0.11 x 2 – 3 = 0.0001 12 = (1/16) + (1/32) = 0.09375.
Floating Point Implicit Binary Normalization
Effective use of every bit is zealously sought. Recall that the normalization
scheme required shifting until a 1 arrived in the most significant position in the mantissa.
If that bit is always 1, then there is no need to physically store it. Just pretend it is there,
and use the most significant bit for another digit that you do not know a priori.
This is called implicit normalization, and it is the usual method of binary
normalization. For a fixed number of bits, this extends the range of numbers that can be
represented.
Number Systems and Data Representation 176

Block Normalization
Consider a different normalization rule, Block-2. For each increment of the
characteristic, shift the mantissa 2 places instead of 1. Apply this new rule to the
example bit patterns above. The first one,
7 6 5 4 3 2 1 0
0 1 1 0 0 0 1 1
± Mantissa Characteristic
now has a characteristic of 2 x 3 = 6. The value in binary of 0.11 x 2 6 = 110000.02 .
The number
7 6 5 4 3 2 1 0
0 1 1 0 0 1 0 1
± Mantissa Characteristic
has a two’s complement characteristic of 2 x ( -3 ) = -6 . The value of the floating point
number is 0.11 x 2 – 6 = 0.0000 00112 = 0.0117 1875.
Notice that the range of numbers you can represent has dramatically increased.
You can represent numbers much closer to one, and much farther from one. This also
reduces the frequency of having to do normalization. However, you can still only
represent 255 different numbers. For our 8-bit floating point number, here are some of
the properties for different decisions for normalization. The binary number closest to one
that you can represent is given by
7 6 5 4 3 2 1 0
0 See table below. 1 0 0
± Mantissa Characteristic
The positive binary number farthest from one that you can represent is given by
7 6 5 4 3 2 1 0
0 1 1 1 1 0 1 1
± Mantissa Characteristic
Number Systems and Data Representation 177

Number of bits Binary mantissa of Positive normalized Positive number

shifted for each smallest number number closest to farthest from zero
increment of the you can represent zero you can you can represent
characteristic represent
Block-1 =
2 – 1 x 2 – 4 = 1/32 (2 4 – 1) x 2 – 1
Binary 0.1000
= 0.3125 = 7.5
Normalization
2 –2 x 2 – 8 =
(2 4 – 1) x 2 2
Block-2 0.0100 1/1024
= 48
≈ 9.7656 x 10 – 4
2 – 3 x 2 –12
(2 4 – 1) x 2 5
Block-3 0.0010 = 2 – 15 = 1/32768
= 480
≈ 3.05176 x 10 – 5
Block-4 = 2 – 4 x 2 – 16
(2 4 – 1) x 2 8
Hexadecimal 0.0001 = 2 – 20 = 1/1048576
= 3840
Normalization ≈ 9.53674 x 10 – 7
As the number of bits shifted for each increment of the characteristic increased,
the range of numbers also increased. In going from a shift of 1 bit to a shift of 4 bits for
normalization, the range of values changed from A = [ 0.3125, 7.5 ] to approximately
B = [ 9.54 x 10 – 7 , 3840 ], with only 255 different values on each interval. The holes
between large numbers in set B are bigger than the holes between numbers in set A.
Floating Point Overflow and Underflow
Floating Point Overflow is the condition of attempting to compute a characteristic
that is larger than the largest characteristic that a floating point word can store. This
condition often is the result of attempting to divide by zero, or dividing a very large
number by a very small number. Often, the floating point overflow is the result of a
mistake in program logic or a bad data input.
Floating Point Underflow is the condition of attempting to compute a
characteristic that is less than zero. This is equivalent to computing a number that is very
close to zero. The importance of floating point underflow depends upon the application.
For many applications, assuming the result to be zero is an adequate assumption. This
can be catastrophic for applications that investigate the importance of small differences
between resulting computations, as is common in physics and engineering.

Subnormal Floating Point Numbers

A subnormal (or denormalized) floating point number is a number with a zero
characteristic, but a non-zero mantissa. A subnormal floating point number does not
have an implicit bit. Ability to accommodate subnormal floating point numbers extends
the range of numbers closer to zero that can be represented, after all the shifting that can
be counted by the characteristic has been done. Permitting subnormal floating point
numbers delays the onset of underflow. Not all floating point processors support
subnormal floating point numbers.
Number Systems and Data Representation 178

Introduction to IEEE Standards and Implementation

For a fixed length representation (a fixed number of bits), there is a tradeoff
between the number of bits used for the characteristic and the number of bits used for the
mantissa. What is the best allocation of bits for the selected purpose?
Floating point arithmetic can be implemented in either software or hardware.
Floating point hardware is useful if you must do a lot of floating point operations.
Floating point hardware increases the cost of the processor, so this capability is a waste if
you do not use it. For floating point hardware, the tradeoffs are implemented by the
engineer and cannot be altered.
When floating point arithmetic is done in software, the programmer has the option
of performing these tradeoffs to design a floating point format that is optimum for the
application. Most programmers do not think about this. Instead, they usually mimic the
structures implemented in hardware for other processors that include floating point
hardware. IEEE Standards 754 and 854 specify floating point formats.
These standards were constructed so that a person who does floating point
arithmetic using one compiler and processor combination will get about the same results
as a person doing floating point arithmetic using another compiler and processor
combination. While standards have helped, they still do not guarantee performance
independent of compiler and platform. The standards only specify format for storage in
memory. The standards do not specify the layout of special purpose registers used during
computation, nor the algorithms for performing arithmetic. For critical applications,
identical precision and accuracy cannot be taken for granted for applications ported to
other hardware environments or compilers. Trust, but verify. Special situations can
benefit from custom designed floating point software.
Sometimes programmers need to convert floating point numbers to fixed point
numbers for use in real-time automatic control systems. When this is not carefully done,
it can result in disaster. Lack of attention to detail in conversion of numbers from one
representation to another led to the explosion of the European Space Agency’s Ariane 5
rocket on 04 June 1996. The problem was traced to a conversion from a 64-bit floating
point number to a 16-bit fixed point number. 76
Zucker has a nice interactive web site for converting decimal to floating point. 77

IEEE Standards 754 (1985) and 854 for Floating Point Arithmetic
The Institute of Electrical and Electronics Engineers (IEEE) standard for floating
point arithmetic is the industry standard for design. 78 It defines the format for floating
point numbers and the methods for doing computation.

76
Douglas N. Arnold, “The Explosion of the Ariane 5” (23 AUG 2000).
http://www.ima.umn.edu/~arnold/disasters/ariane.html visited 04 JUN 2002.
77
Ron Zucker, “Floating Point Conversion Using Excess Notation”, University of North Florida.
http://www.unf.edu/~rzucker/cot3100dir/excess.html visited 07 January 2008.
78
P. H. Abbott, D. G. Brush, C. W. Clark III, C. J. Crone, J. R. Ehrman, G. W. Ewart, C. A. Goodrich, M.
Hack, J. S. Kapernick, B. J. Minchau, W. C. Shepard, R. M. Smith, Sr., R. Tallman, S. Walkowiak, A.
Watanabe, and W. R. White, “Architecture and software support in IBM S/390 Parallel Enterprise Servers
Number Systems and Data Representation 179

IEEE Zero, Subnormals, Infinity, and Not A Number (NAN)

IEEE floating point zero is defined by setting all bits in both the characteristic and
the mantissa to zero. The value is considered zero regardless of the state of the sign bit.
IEEE identifies a number as subnormal when all bits of the characteristic are zero, but the
mantissa is nonzero. IEEE identifies a number as infinity when all bits of the
characteristic are one, but the mantissa is zero.
In addition to numerical values, the IEEE floating point specification defines two
other objects, “not a number” (NaN) or “not a value”. There are two types of NaNs:
Signaling NAN (SnaN), and Quiet NAN (QnaN). NANs have all bits of the
characteristic set to one, but the mantissa is non- zero. The Quiet NAN mantissa has a
one in the most significant bit. The Signaling NAN mantissa has a zero in the most
significant bit.
IEEE Single Precision Floating Point

0 Sign 1 Characteristic 8 9 Mantissa 31

IEEE single precision floating point has an 8-bit characteristic using Excess-127
7
( 2 – 1 ) notation, and a 23-bit mantissa in signed magnitude form, with implicit-bit
binary normalization (24-bit precision), where the implicit bit is in the 20 position (just to
the left of the binary point). The sign bit is the sign of the mantissa, with 0 indicating
positive and 1 indicating negative.

X = ±2 (Characteristic−Excess ) • (1 + Mantissa )

The maximum number that can be represented with this format is approximately
2 x 2127 ≈ 3.4 x 1038 . The smallest normalized number that can be represented is
approximately 2 x 2 – 127 ≈ 1.2 x 10 – 38 . The smallest subnormal number that can be
represented is approximately 2 x 2 – 127 x 2 – 23 = 2 x 2 – 150 ≈ 1.4 x 10 – 45 .
IEEE Double Precision Floating Point

0 Sign 1 Characteristic 11 12 Mantissa 63

IEEE double precision floating point has an 11-bit characteristic using Excess-
1023 ( 210 – 1 ) notation, and 52-bit mantissa in signed magnitude form, with implicit-bit
binary normalization, where the implicit bit is in the 20 position. The sign bit is the sign
of the mantissa, with 0 indicating positive and 1 indicating negative. Therefore, the
mantissa has 53-bit precision.
X = ±2 (Characteristic−Excess ) • (1 + Mantissa )
The maximum number that can be represented is approximately
2x2 1023
≈ 1.8 x 10308 . The smallest normalized number that can be represented is

for IEEE Floating-Point arithmetic” IBM Journal of Research and Development, Volume 43, Numbers 5/6
(1999), http://www.ibm.com, 08 June 2002.
Number Systems and Data Representation 180

approximately 2 x 2 – 1023 ≈ 2.2 x 10 – 308 . The smallest subnormal number that can be
represented is approximately
2 x 2 – 1023 x 2 – 52 = 2 x 2 – 1075 ≈ 4.9 x 10 – 324 .

IEEE Double Extended Precision Floating Point

79 Sign 78 Characteristic 64 63 Mantissa 0

IEEE extended precision floating point has a 15-bit characteristic using Excess-
16383 ( 214 – 1 ) notation, and 64-bit mantissa, with implicit-bit binary normalization,
where the implicit bit is in the 20 position. The sign bit is the sign of the mantissa, with 0
indicating positive and 1 indicating negative. Therefore, the mantissa has 65-bit
precision.
X = ±2 (Characteristic−Excess ) • (1 + Mantissa )
The maximum number that can be represented is approximately
2x2 16383
≈ 1.2 x 104932 . This is more than the number of molecules in the universe
(assuming a finite universe and a big bang). The smallest no rmalized number that can be
represented is approximately 2 x 2 – 16383 ≈ 3.4 x 10 – 4932 . The smallest subnormal
number that can be represented is approximately
2 x 2 – 16383 x 2 – 64 = 2 x 2 – 16447 ≈ 1.8 x 10 – 4951 .

Intel Floating Point Unit Formats

The Intel Floating Point Unit (FPU) uses the IEEE format for Single Real and
Double Real floating point numbers.
Intel Extended Real Floating Point

79 Sign 78 Characteristic 64 63 Mantissa 0

The 80-bit floating point register format has a 15-bit characteristic and 64-bit
mantissa, with explicit-bit binary normalization (64-bit precision) where the most
significant bit is in the 20 position. The sign bit is the sign of the mantissa, with 0
indicating positive and 1 indicating negative. The 15-bit characteristic is biased by
214 – 1 = 16383.
X = ±2 (Characteristic−Excess ) • (Mantissa )
The largest number that can be represented with this format is approximately
2 x 216383 – 1 ≈ 1.18 x 104932 . The smallest normalized number that can be represented
is approximately 2 x 2 – 16383 ≈ 3.4 x 10 – 4932 . The smallest subnormal number that
can be represented is approximately 2 x 2 – 16383 x 2 – 64 = 2 x 2 – 16447 ≈ 1.8 x 10 – 4951 .
Number Systems and Data Representation 181

Itanium and Pentium 4 Extended Double Precision Floating Point Register

The Itanium IA-64 processor by Intel fully complies with ANSI/IEEE Standard
754 for floating point arithmetic. 79 In addition, the Intel processors have an Extended
Double Precision floating point register format.
81 Sign 80 Characteristic 64 63 Mantissa 0
The 82-bit floating point register format has a 17-bit characteristic and 64-bit
mantissa, with explicit-bit binary normalization (64-bit precision) where the most
significant bit is in the 20 position. The sign bit is the sign of the mantissa, with 0
indicating positive and 1 indicating negative. The 17-bit characteristic is biased by
216 – 1 = 65,535.
X = ±2 (Characteristic−Excess ) • (Mantissa )
The largest number that can be represented with this format is approximately
2 x 265536 – 1 ≈ 4 x 1019728 .

IBM Floating Point Arithmetic

The IBM System 390 supports the IEEE Floating Point Arithmetic standard,
which IBM calls “Binary Floating Point (BFP)”. In addition, S/390 continues to support
the IBM proprietary hexadecimal normalization architecture. IBM calls this
“Hexadecimal Floating Point (HFP)”. 80
Hexadecimal Floating Point means that the characteristic is a power of 16 rather
than a power of 2. Each increment by 1 of the characteristic represents a shift of 4 bits of
the mantissa when it is being normalized. Excess notation for IBM HFP also differs; the
excess for IBM HFP is a power of 2, the excess for IEEE and IBM BFP is one less than a
power of 2. Excess for IBM HFP is 64 = 26 for all sizes of Hexadecimal Floating Point
words. Excesses for IEEE and IBM short and long Binary Floating Point are 127 = 27 – 1
and 1023 = 210 – 1. The IA-64 processor Extended Double Precision excess is
65535 = 232 – 1, and the IBM Binary Extended Floating Point excess is 16383 = 214 – 1.
Hexadecimal Short Floating Point

0 Sign 1 Characteristic 7 8 Mantissa 31

The 32-bit short floating point word has a 7-bit characteristic using XS-64
notation. The 24-bit mantissa uses explicit hexadecimal normalization to extend the
range of numbers and reduce the frequency of shifting. The IBM hexadecimal format
does not have an implicit bit to the left of the radix point. The sign bit is zero for a plus
sign, and it is one for a negative sign.
X = 16(Characteristic− Excess) • (Mantissa )

79
Intel® ItaniumTM Architecture Software Developer’s Manual, Volume 1, Application Architecture,
Revision 2.0 (December 2001).
80
ESA/390 Principles of Operation, International Business Machines, Document Number SA22-7201-06
(1999). http://publib.boulder.ibm.com/cgi-bin/bookmgr/BOOKS/DZ9AR006/CCONTENTS , Visited 10
Jun 2002.
Number Systems and Data Representation 182

Hexadecimal Long Floating Point

0 Sign 1 Characteristic 7 8 Mantissa 63

The 64-bit short floating point word has a 7-bit characteristic using XS-64
notation. The 56-bit mantissa uses explicit hexadecimal normalization to extend the
range of numbers and reduce the frequency of shifting. The IBM hexadecimal format
does not have an implicit bit to the left of the radix point. The sign bit is zero for a plus
sign, and it is one for a negative sign.
Hexadecimal Extended Floating Point
Extended floating point is represented with two 64-bit words. The characteristic
uses Excess-64 (26 ) notation.
0 Sign 1 High Order 7 8 Left 14 Hex Digits of 63
Characteristic Mantissa

64 Sign 65 Low Order 71 72 Right 14 Hex Digits of 127

Characteristic Mantissa
The mantissa has a total of 112 bits. The high order characteristic belongs with
the high order word. The low order characteristic belongs to the low order word. The
sign bit of both words are the same, which is the sign of the mantissa.
Let X be the value in the high order word. Let Y be the value in the low order
word. Let the value of the extended floating point variable be
Z = X+Y
Let the sign of the high order word be Xsign. Let the characteristic be
Xcharacteristic. Let the mantissa be Xmantissa. Then
X = 16( Xcharacteristic−Excess ) • ( Xmantissa )
Let the sign of the low order word be Ysign. Note that Ysign = Xsign. Let the
characteristic be Ycharacteristic. Let the mantissa be Ymantissa.
Y = 16 (Ycharacteristic−Excess) • (Ymantissa )
Note that the low order mantissa is shifted by 14 hexadecimal digits in position
from the high order mantissa. That means the low order characteristic is always 14 less
than the high order characteristic.
Ycharacteristic = Xcharacteristic – 14
Therefore,
Y = 16 ( Xcharacteristic−14− Excess) • (Ymantissa )
Evaluating,
(
Z = 16 ( Xcharacteristic− Excess) Xmantissa + 16 −14 • (Ymantissa ))
Number Systems and Data Representation 183

Because the minimum value that can be stored in the low order characteristic field
is zero, the minimum characteristic that can be used in the high order characteristic is 14.
The range of values that can be used in the high order characteristic is [-50, 63].
Remember that these are powers of 16. IBM Hexadecimal Extended Floating Point
therefore cannot represent numbers as close to zero as IBM Hexadecimal Long Floating
Point. To help keep this in perspective, the moment of a Proton is approximately
1.4 x 10 – 26 J/T (Joules / Tesla).
Binary Short Floating Point
The IBM Binary Short Floating Point word conforms to the IEEE Single
Precision Floating Point word specifications.
Binary Long Floating Point
The IBM Binary Long Floating Point word conforms to the IEEE Double
Precision Floating Point word specifications.
Binary Extended Floating Point

0 Sign 1 Characteristic 15 16 Mantissa 127

The 128-bit binary extended floating point has a 15-bit characteristic using
XS-16383 (214 –1 ) notation. The 112-bit mantissa (113-bit precision) uses implicit
normalization. Note that IEEE does not have a specification for a floating point word
format greater than Double Precision. The IBM Binary Extended Floating Point word
gives greater precision (more bits in the mantissa) than the Intel IA-64 (Pentium 4 and
Itanium) Extended Double Precision word, but with reduced range (2 fewer bits in the
characteristic). The IBM Hexadecimal Floating Point characteristic, for all its formats,
produce a greater range of numbers that can be represented ((1/16) x 16 – 64 , 1663 ), or
approximately
( 5.4 x 10 –79 , 7.2 x 10 75 ). Chris Impey reported that one estimate of the number of
photons in the universe is 1080 . 81
The maximum number that can be represented is approximately
2 x 216383 ≈ 1.2 x 104932 . The smallest normalized number that can be represented is
approximately 2 x 2 – 16383 ≈ 3.4 x 10 – 4932 . The smallest subnormal number that can
be represented is approximately 2 x 2 – 16383 x 2 – 112 = 2 x 2 – 16495 ≈ 6.5 x 10 – 4966 .

81
Chris Impey, “Reacting to the Size and Shape of the Universe”, Mercury Magazine Vol. 30 No. 1
(January/February 2001). http://www.aspsky.org/mercury/mercury/30_01/universe.html Visited 05 July
2002.
Number Systems and Data Representation 184

Floating Point Conversions

Convert IEEE Single Precision Floating Point to IBM Hexadecimal Short Floating
Point
Procedure:
Step 1: Let M = (IEEE Single Precision Floating Point characteristic) – (IEEE Single
Precision Floating Point Excess).
Note: The IEEE format uses implicit normalization. The radix point is immediately
to the left of the IEEE mantissa most significant bit.
Step 2: Insert the prefix bit pattern 0001 immediately to the left of the radix point.
Step 3: Let MM = M MOD 4. This is the remainder obtained by dividing M by 4.
Step 4: Compute the number of bits (J) to right-shift the mantissa and prefix bit
pattern.
• If MM = 0, then J = 4
• If MM < 0, then J = |MM|. |MM| is the absolute value of MM.
• If MM > 0, then J = 4 – MM.
The number of bits of precision in the mantissa lost is J – 1 .
Step 5: Copy the shifted bit pattern into the IBM Short HFP mantissa.
Step 6: Compute the IBM HFP characteristic, without excess.
M + 4
• K=  . The brackets with feet at the bottom, and no roof, is
 4 
the FLOOR function. The FLOOR of an exact integer is that integer.
The FLOOR of a number with a fraction is the next LOWER integer.
Examples: 3.4 = 3 , − 3.4 = −4 .
Step 7: Add the IBM HFP excess to K and store it in the HFP characteristic field.
Step 8: Copy the sign bit from the IEEE Single Precision FP word to the sign bit of
the IBM HFP word.
Example: Convert the following Convert IEEE Single Precision Floating Point to
IBM Hexadecimal Short Floating Point
Number Systems and Data Representation 185

IEEE Single Precision Floating Point

S Characteristic Mantissa
0 1 4 8 9 12 15
0 1 0 0 1 0 0 0 1 0 0 1 1 0 0 1

Mantissa (Continued)
16 20 24 28 31
0 0 1 1 1 1 0 1 0 0 0 1 0 0 0 1

Step 1: Let M = (IEEE Single Precision Floating Point characteristic) – (IEEE Single
Precision Floating Point Excess). M = 145 – 127 = 18.
Note: The IEEE format uses implicit normalization. The radix point is immediately
to the left of the IEEE mantissa most significant bit.
Step 2: Insert the prefix bit pattern 0001 immediately to the left of the radix point.
The appended mantissa is 0001.001 1001 0011 1101 0001 0001
Step 3: Let MM = M MOD 4. This is the remainder obtained by dividing M by 4.
MM = 18 MOD 4 = 2
Step 4: Compute the number of bits (J) to right-shift the mantissa and prefix bit
pattern.
• If MM = 0, then J = 4
• If MM < 0, then J = |MM|. |MM| is the absolute value of MM.
• If MM > 0, then J = 4 – MM = 4 – 2 = 2.
The number of bits of precision in the mantissa lost is J – 1 = 2 – 1 = 1.
Step 5: Copy the shifted bit pattern into the IBM Short HFP mantissa. The pattern
copied is: 01001 1001 0011 1101 0001 0001
Step 6: Compute the IBM HFP characteristic, without excess.
 M + 4  18 + 4   22 
• K= =  =   = 5.5 = 5 .
 4   4  4
Step 7: Add the IBM HFP excess to K and store it in the HFP characteristic field. K
+ 64 = 5 + 64 = 69 = 4516 = 100 01012 .
Step 8: Copy the sign bit from the IEEE Single Precision FP word to the sign bit of
the IBM HFP word.

IBM Hexadecimal Short Floating Point

S Characteristic Mantissa
0 1 4 7 8 12 15
0 1 0 0 0 1 0 1 0 1 0 0 1 1 0 0

Mantissa (Continued)
16 20 24 28 31
1 0 0 1 1 1 1 0 1 0 0 0 1 0 0 0
Number Systems and Data Representation 186

Convert IBM Hexadecimal Extended Floating Point to IEEE Double Extended

Precision Floating Point
Procedure:
Step 1: Subtract the IBM Hexadecimal Floating Point Excess from the high order
characteristic.
Step 2: Let K = the numerical value of the high order characteristic.
Step 3: Let N = 4 K. Each positive increment of the IBM Hexadecimal Floating Point
characteristic represents a shift of the radix by 4 bit positions to the left to
perform the normalization.
Step 4: Let J = the number of bit positions from the left end of the mantissa occupied
by the first one-bit.
Step 5: Let M = N – J .
Step 6: Convert M to binary.
Step 7: Add the IEEE Double Extended Precision Floating Point Excess to M.
Step 8: Record the result in the characteristic field of the IEEE Double Extended
Precision Floating Point word.
Step 9: The most significant bit of the mantissa of the IBM Hexadecimal Floating
Point becomes the implicit bit in the IEEE Double Extended Precision
Floating Point word, and therefore does not explicitly get recorded.
Step 10: Beginning with the bit immediately to the right of the most significant bit,
copy the remaining bits from the IBM Hexadecimal Floating Point mantissa
into the IEEE Double Extended Precision Floating Point mantissa.
Step 11: Stop after the IEEE Double Extended Precision Floating Point mantissa is
filled. Note that 37 to 40 least-significant-bits of precision in the mantissa
were lost in the conve rsion.
Convert IEEE Double Extended Precision Floating Point to IBM Hexadecimal
Extended Floating Point
Procedure:
Step 1: Set all IBM Hexadecimal Extended Floating Point bits to zero.
Step 2: Copy the IEEE Double Extended Precision Floating Point sign bit into the
IBM Hexadecimal Extended Floating Point High Order sign bit and Low
Order sign bit.
Step 3: Evaluate the IEEE Double Extended Precision Floating Point Excess from the
characteristic. Call this G.
Step 4: Subtract the IEEE Double Extended Precision Floating Point Excess from the
characteristic. Call this P = G – XSIEEE DE FP .
Step 5: Compute the IBM Hexadecimal Floating Point Characteristic without the IBM
P + 4
HFP Excess. K =  
 4 
Step 6: Compute the IBM Hexadecimal Floating Point Characteristic with the IBM
HFP Excess. N = K + 64.
Step 7: Write N into the IBM HFP High Order Characteristic.
Step 8: Append the implicit most significant bit to the left of the radix point for the
IEEE DEP FP mantissa.
Number Systems and Data Representation 187

Step 9: Let MM = P MOD 4. This is the remainder obtained by dividing P by 4.

Step 10: Compute the number of places to shift the appended mantissa to obtain the
IBM HFP mantissa.
• If MM = 0, then J = 4
• If MM > 0, then J = 4 – MM
• If MM < 0, then J = | MM | where the vertical bars identify the absolute
value function.
Step 11: Beginning with the most significant bit of the shifted mantissa, copy the
remaining bits from the IEEE Double Extended Precision Floating Point
mantissa into the IBM Hexadecimal Floating Point Mantissa.
Step 12: Compute the IBM HFP Low Order Characteristic. Call this Q.
Step 13: Write Q into the IBM HFP Low Order Characteristic.

Floating Point Conversion Problems

Problem 1. Convert the following IEEE Single Precision Floating Point word to an
IBM Hexadecimal Short Floating Point word.
IEEE Single Precision Floating Point
S Characteristic Mantissa
0 1 4 8 9 12 15
0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0

Mantissa (Continued)
16 20 24 28 31
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1

IBM Hexadecimal Short Floating Point

S Characteristic Mantissa
0 1 4 7 8 12 15

Mantissa (Continued)
16 20 24 28 31

Problem 2. Convert the following IEEE Single Precision Floating Point word to a
IBM Hexadecimal Short Floating Point word.
Number Systems and Data Representation 188

IEEE Single Precision Floating Point

S Characteristic Mantissa
0 1 4 8 9 12 15
0 1 0 0 0 0 1 0 0 0 1 0 0 0 0 0

Mantissa (Continued)
16 20 24 28 31
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

IBM Hexadecimal Short Floating Point

S Characteristic Mantissa
0 1 4 7 8 12 15

Mantissa (Continued)
16 20 24 28 31

Problem 3. Convert the following IEEE Single Precision Floating Point word to a
IBM Hexadecimal Short Floating Point word.
IEEE Single Precision Floating Point
S Characteristic Mantissa
0 1 4 8 9 12 15
1 0 1 1 1 1 1 0 1 0 0 1 0 0 0 0

Mantissa (Continued)
16 20 24 28 31
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

IBM Hexadecimal Short Floating Point

S Characteristic Mantissa
0 1 4 7 8 12 15

Mantissa (Continued)
16 20 24 28 31

Problem 4. Convert the following IBM Hexadecimal Extended Floating Point word to
an IEEE Double Extended Precision Floating Point word.
Number Systems and Data Representation 189

IBM Hexadecimal Extended Floating Point

High Order Word
S High Order Characteristic High Order Mantissa
0 1 4 7 8 12 15
0 1 0 0 1 0 1 0 0 0 0 1 0 1 1 0

High Order Mantissa (Continued 1)

16 20 24 28 31
0 1 0 1 1 1 0 0 1 1 1 1 1 0 0 1

High Order Mantissa (Continued 2)

32 36 40 44 47
0 0 0 1 0 0 0 0 0 1 0 1 1 0 0 1

High Order Mantissa (Continued 3)

48 52 56 60 63
1 1 1 0 1 1 1 0 0 1 0 1 0 0 1 0

Low Order Word

S Low Order Characteristic Low Order Mantissa
64 65 68 71 72 76 79
0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0

Low Order Mantissa (Continued 1)

80 84 88 92 95
0 0 0 0 1 0 1 1 0 1 1 1 0 0 1 0

Low Order Mantissa (Continued 2)

96 100 104 108 111
1 1 0 0 0 1 0 1 1 0 0 0 1 0 1 0

Low Order Mantissa (Continued 3)

112 116 120 124 127
1 1 1 0 0 1 0 0 0 1 0 0 0 1 0 1

IEEE Double Extended Precision Floating Point

S Characteristic
79 78 75 72 71 67 64

Mantissa
63 60 56 52 48
Number Systems and Data Representation 190

Mantissa (Continuation 1)
47 44 40 36 32

Mantissa (Continuation 2)
31 28 24 20 16

Mantissa (Continuation 3)
15 12 8 4 0

Problem 5. Convert the following IEEE Double Extended Precision Floating Point word
to an IBM Hexadecimal Extended Floating Point word.
IEEE Double Extended Precision Floating Point
S Characteristic
79 78 75 72 71 67 64
1 0 1 1 1 1 1 1 0 1 0 0 1 0 1 0

Mantissa
63 60 56 52 48
0 1 1 1 0 0 1 0 1 1 1 1 1 0 1 1

Mantissa (Continuation 1)
47 44 40 36 32
1 1 0 0 0 1 1 1 0 0 1 1 1 1 0 0

Mantissa (Continuation 2)
31 28 24 20 16
1 1 1 1 1 0 0 0 0 1 1 0 1 1 0 0

Mantissa (Continuation 3)
15 12 8 4 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1
Number Systems and Data Representation 191

IBM Hexadecimal Extended Floating Point

High Order Word
S High Order Characteristic High Order Mantissa
0 1 4 7 8 12 15

High Order Mantissa (Continued 1)

16 20 24 28 31

High Order Mantissa (Continued 2)

32 36 40 44 47

High Order Mantissa (Continued 3)

48 52 56 60 63

Low Order Word

S Low Order Characteristic Low Order Mantissa
64 65 68 71 72 76 79

Low Order Mantissa (Continued 1)

80 84 88 92 95

Low Order Mantissa (Continued 2)

96 100 104 108 111

Low Order Mantissa (Continued 3)

112 116 120 124 127
Number Systems and Data Representation 192

Problem 6. Convert the following IEEE Double Extended Precision Floating Point word
to an IBM Hexadecimal Extended Floating Point word.
IEEE Double Extended Precision Floating Point
S Characteristic
79 78 75 72 71 67 64
1 0 1 1 1 1 1 1 0 0 0 0 1 0 1 1

Mantissa
63 60 56 52 48
0 1 1 1 0 0 1 0 1 1 1 1 1 0 1 1

Mantissa (Continuation 1)
47 44 40 36 32
1 1 0 0 0 1 1 1 0 0 1 1 1 1 0 0

Mantissa (Continuation 2)
31 28 24 20 16
1 1 1 1 1 0 0 0 0 1 1 0 1 1 0 0

Mantissa (Continuation 3)
15 12 8 4 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1

IBM Hexadecimal Extended Floating Point

High Order Word
S High Order Characteristic High Order Mantissa
0 1 4 7 8 12 15

High Order Mantissa (Continued 1)

16 20 24 28 31

High Order Mantissa (Continued 2)

32 36 40 44 47

High Order Mantissa (Continued 3)

48 52 56 60 63
Number Systems and Data Representation 193

Low Order Word

S Low Order Characteristic Low Order Mantissa
64 65 68 71 72 76 79

Low Order Mantissa (Continued 1)

80 84 88 92 95

Low Order Mantissa (Continued 2)

96 100 104 108 111

Low Order Mantissa (Continued 3)

112 116 120 124 127

Problem 7. Convert the following IEEE Double Extended Precision Floating Point word
to an IBM Hexadecimal Extended Floating Point word.
IEEE Double Extended Precision Floating Point
S Characteristic
79 78 75 72 71 67 64
1 0 1 1 1 0 1 1 1 0 0 1 0 1 1 0

Mantissa
63 60 56 52 48
0 1 1 1 0 0 1 0 1 1 1 1 1 0 1 1

Mantissa (Continuation 1)
47 44 40 36 32
1 1 0 0 0 1 1 1 0 0 1 1 1 1 0 0

Mantissa (Continuation 2)
31 28 24 20 16
1 1 1 1 1 0 0 0 0 1 1 0 1 1 0 0

Mantissa (Continuation 3)
15 12 8 4 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1
Number Systems and Data Representation 194

IBM Hexadecimal Extended Floating Point

High Order Word
S High Order Characteristic High Order Mantissa
0 1 4 7 8 12 15

High Order Mantissa (Continued 1)

16 20 24 28 31

High Order Mantissa (Continued 2)

32 36 40 44 47

High Order Mantissa (Continued 3)

48 52 56 60 63

Low Order Word

S Low Order Characteristic Low Order Mantissa
64 65 68 71 72 76 79

Low Order Mantissa (Continued 1)

80 84 88 92 95

Low Order Mantissa (Continued 2)

96 100 104 108 111

Low Order Mantissa (Continued 3)

112 116 120 124 127

Numerical Methods
Numerical methods are procedures for solving mathematical problems
numerically, including problems for which no closed form solutions exist. Attention is
given to designing procedures that make efficient use of a computer while controlling
computational errors. Analysis includes quantifying errors, approximation or
convergence to solutions, and sensitivity of solutions to variations of inputs.
Number Systems and Data Representation 195

Numerical methods have been applied to algebra, calculus, differential equations,

and matrix algebra. Computational fluid dynamics, weather forecasting, underwater
sound propagation, and analysis of vibration of complex structures (cars, aircraft, space
station) all require use of numerical methods.
The accumulation of approximation errors can render a solution useless. The
study of such errors and methods to minimize such errors are introductory topics in
numerical analysis. Some of the approaches used to minimize these errors include
• Thoughtful selection of the order of performing arithmetic operations. For
example, doing arithmetic between pairs of small quantities before
combining those results with larger quantities. Another strategy is to seek
to alternate multiplications and divisions to keep results as close to 1 as
possible.
• Rearranging the computation of a quantity to reduce the number of
operations that produce large errors. For example, evaluating
f(x) = a2 x2 + a1 x + a0 as f(x) = ( a2 x + a1 ) x + a0 which reduces the
number of multiplications from 4 to 2.
• Selection of numerical methods that require fewer operations that produce
large errors.
Numerical Methods include Numerical Analysis and Approximation Theory.
• Selection of mathematical approaches that converge to an adequate
solution (not necessarily ever capable of reaching an ideal exact solution)
with fewer operations.
• Application of asymptotic approximations. 82
You must first understand the problem you want to solve! A perfectly
computed solution to the wrong problem is still a wrong solution.
• Identify and solve the right problem.
• Express the problem in simplest forms.
Approximation theory83 is the mathematical study of how given quantities can be
approximated by other (usually simpler) ones under appropriate conditions.
Approximation theory also studies the size and properties of the error introduced by
approximation. Approximations are often obtained by power series expansions in which
the higher order terms are dropped.

82
James P. Keener, Principles of applied mathematics: Transformation and approximation, Pages 424 –
428. Addison-Wesley (1988).
83
Weisstein, Eric W. "Approximation Theory.", MathWorld--A Wolfram Web Resource (10 February
2008). http://mathworld.wolfram.com/ApproximationTheory.html visited 12 February 2008.
Number Systems and Data Representation 196

More Computer Arithmetic

Basic arithmetic is the foundation for much sophisticated work in science and
engineering. Doing a good job with arithmetic is prerequisite to doing a good job with
other computations. What constitutes “good” depends upon the application. Digital
Engineering Institute has a nice web site. 84 Some of the considerations that lead to
different designs include: required accuracy, required precision, attention to circuit
propagation delay, importance of speed, number of processors, control of fan-out, chip
area required for circuit design, physical mass and volume of resulting circuit, and
importance of power consumption.

84
Richard B. Katz, “Arithmetic”, NASA Office of Logic Design (15 January 2002).
http://www.klabs.org/DEI/Arithmetic/ Visited 14 January 2008.
Number Systems and Data Representation 197

Sound, Smell and Color

Other kinds of data exist besides generic numbers. Some standards are emerging
through common use.
Sound
Sounds may be represented either as processed signals or by abstract
representations. Digitized waveforms often require large files that use much storage
space and take much time to transmit over a network. Compression techniques have been
developed to extract parameters that describe and permit an approximate reconstruction
of the original sound. Encryption methods are also used to protect royalties from sale of
copyrighted materials. These methods are used by the entertainment industry.
Another method for representing sounds is to describe the method used to
generate sounds. An example of such an abstract method is a musical score that a
performer uses when playing music. This specifies the sequence of musical notes, along
with directions of how they are to be played. This is a compact notation that can be
edited easily by a composer, and used by automated musical instruments to produce
music. This approach is used in synthesized music.
The MIDI Manufacturers Association has a tutorial, links to specifications, and
other information related to Musical Instrument Digital Interface (MIDI). 85 Codes of the
MIDI specification are not brief enough to include in this tutorial. Another approach is to
use MusiXTeX. 86
Smell
How can you implement a computer version of “scratch and sniff”? The
detection, analysis, and information display has been automated. One exa mple is
Cyranose. It measures the resistance of 32 polymer composite sensors when exposed to a
vapor. The data is digitized and analyzed. Parameters that describe the vapor (smell)
constitute an encoding of the data. 87 For an entrepreneurial and light- hearted exploration,
see Charles Platt’s article on Digiscent. 88 Even Amazon is getting into the act. 89
Color
Consider codes for color used on web pages on the internet. This assumes that a
particular code generates the same color on different monitors. While that assumption is
not strictly true, we expect it to be close.

85
Jim Heckroth (Crystal Semiconductor Corp), “Tutorial on MIDI and Music Synthesis”, MIDI
Manufacturers Association (2001). http://www.midi.org/about-midi/tutorial/tutor.shtml Visited 14 January
2008.
86
Daniel Taupin, MusiXTeX, “Directory: CTAN home / tex-archive/ macros/ musixtex/ taupin “ CTAN:
The Comprehensive TeX Archive Network. http://www.ctan.org/tex-archive/macros/musixtex/taupin/
Visited 14 January 2008.
87
http://cyranosciences.com/technology/onboard.html Visited 2001. Not available on 14 January 2008.
88
Charles Platt, “You’ve Got Smell!”, Wired News (November 1999).
http://www.wired.com/wired/archive/7.11/digiscent_pr.html Visited 14 January 2008.
89
“Amazon To Market Online Smells” (1998). http://www.rayno.com/amasmell.htm Visited 14 January
2008.
Number Systems and Data Representation 198

RGB color monitors represent colors by using combinations of three primary

colors: red, green, and blue. It is common today for the intensity of each primary color to
be represented by one byte, which allows 256 different levels of each primary color. The
total number of colors represented by this scheme using three bytes is 256 x 256 x 256 =
16,777,216. Colors are coded as three pairs of hexadecimal digits such as OxF040FF, or
as a triplet of decimal numbers such as (240, 64, 255). Internally, these codes are the
same. The format used for input depends upon the software.

Hexadecimal Color Codes

The color table below shows 85 of the 16,777,216 colors that can be represented
using 3 bytes per color.
OxFF0000 Ox00FF00 Ox0000FF
OxF00000 Ox00F000 Ox0000F0
OxC00000 Ox00C000 Ox0000C0
Ox800000 Ox008000 Ox000080
Ox400000 Ox004000 Ox000040
Ox0F0000 Ox000F00 Ox00000F
OxFF00FF OxFFFF00 Ox00FFFF
OxF000F0 OxF0F000 Ox00F0F0
OxC000C0 OxC0C000 Ox00C0C0
Ox800080 Ox808000 Ox008080
Ox400040 Ox404000 Ox004040
OxFFFFFF OxFFFFFF OxFFFFFF
OxFFF0FF OxFFFFF0 OxF0FFFF
OxFFC0FF OxFFFFC0 OxC0FFFF
OxFF80FF OxFFFF80 Ox80FFFF
OxFF40FF OxFFFF40 Ox40FFFF
OxFF00F0 OxFFF000 OxFFF0F0
OxFF00C0 OxFFC000 OxFFC0C0
OxFF0080 OxFF8000 OxFF8080
OxFF0040 OxFF4000 OxFF4040
OxF0FFF0 OxF0FF00 Ox00FFF0
OxC0FFC0 OxC0FF00 Ox00FFC0
Ox80FF80 Ox80FF00 Ox00FF80
Ox40FF40 Ox40FF00 Ox00FF40
OxF000FF Ox00F0FF OxF0F0FF
OxC000FF Ox00C0FF OxC0C0FF
Ox8000FF Ox0080FF Ox8080FF
Ox4000FF Ox0040FF Ox4040FF

Obviously, if you are looking at a black and white copy of this, you are not going
to see the colors. You can see the colors by viewing an electronic copy. There are
several very nice web sites with color code charts for use with HTML.
Number Systems and Data Representation 199

Analog Computers
The term “analog” refers to devices that are not binary digital computers. Analog
computers rely on electromechanical, electronic, and other components to represent,
retain, and transform continuous values. These values may be used for subsequent
computation, displayed, or used as control signals to other devices. Analog computers
tend to be special purpose devices, sometimes quite sophisticated.
Digital electronic computers are all analog devices at the most basic level.
Computers that are part digital and part analog are called hybrid computers.
While digital computers have made a tremendous contribution, they are still
limited by the speed at which inputs can be digitized and functions computed. This is a
significant limitation in some applications of data acquisition and signal processing.
Critical issues for analog devices include the accuracy of converting data,
repeatability, and reliability. Such devices are sensitive to its environment: heat,
electromagnetic fields, vibration, humidity, gravity, etc.
We continue to push for extreme performance in computation and storage speed,
small size, low power consumption, and other properties. Scientists and engineers are
pushing the boundaries of physics, materials science and biological sciences to find new
and reliable ways to compute.