CSC 317 Computer Organization and Architecture

Chapter 2: Data Representation

Spring 2007

Chapter 2 Objectives
Understand the fundamentals of numerical data representation and manipulation in digital computers. Master the skill of converting between various radix systems.

Understand how errors can occur in computations because of overflow and truncation.
Understand the concepts of error detecting and correcting codes.
2

2.1 Introduction
A bit is the most basic unit of information in a computer.
It is a state of on or off in a digital circuit. Or high or low voltage instead of on or off.

A byte is a group of eight bits.

A byte is the smallest possible addressable unit of computer storage.

A word is a contiguous group of bytes

Word sizes of 16, 32, or 64 bits are most common. Usually a word represents a number or instruction.
3

2.2 Positional Numbering Systems

Computers use base 2 to represent numbers
The binary system is also called the base-2 system. Our decimal system is the base-10 system. It uses powers of 10 for each position in a number. Any integer quantity can be represented exactly using any base (or radix).

The decimal number 947 in powers of 10 is: 9 10 2 + 4 10 1 + 7 10 0

The decimal number 5836.47 in powers of 10 is: 5 10 3 + 8 10 2 + 3 10 1 + 6 10 0 + 4 10 -1 + 7 10 -2

2.2 Positional Numbering Systems

An n-bit sequence of binary digits is an unsigned integer A whose value is: A = 2i ai (0 i n-1) The binary number 11001 in powers of 2 is:
1 24+ 1 23 + 0 22 + 0 21 + 1 20 = 16 + 8 + 0 + 0 + 1 = 25

8 bits can represent unsigned numbers from 0 to 255 (28-1)

2.3 Decimal to Binary Conversions

In an earlier slide, we said that every integer value can be represented exactly using any radix system.

You can use either of two methods for radix conversion: the subtraction method and the division remainder method.
The subtraction method is more intuitive, but cumbersome. It does, however reinforce the ideas behind radix mathematics. The division method employs the idea that successive division by a base is equivalent to successive subtraction by powers of the base.
6

2.3 Decimal to Binary Conversions

Suppose we want to convert the decimal number 190 to base 3. We know that 3 5 = 243 so our result will be less than six digits wide. The largest power of 3 that we need is therefore 3 4 = 81, and 81 2 = 162. Write down the 2 and subtract 162 from 190, giving 28.

2.3 Decimal to Binary Conversions

Converting 190 to base 3... The next power of 3 is 3 3 = 27. Well need one of these, so we subtract 27 and write down the numeral 1 in our result. The next power of 3, 3 2 = 9, is too large, but we have to assign a placeholder of zero and carry down the 1.

2.3 Decimal to Binary Conversions

Converting 190 to base 3... 3 1 = 3 is again too large, so we assign a zero placeholder. The last power of 3, 3 0 = 1, is our last choice, and it gives us a difference of zero. Our result, reading from top to bottom is: 19010 = 210013

2.3 Decimal to Binary Conversions

Converting 190 to base 3...

First we take the number that we wish to convert and divide it by the radix in which we want to express our result. In this case, 3 divides 190 63 times, with a remainder of 1. Record the quotient and the remainder.
10

2.3 Decimal to Binary Conversions

Converting 190 to base 3...

63 is evenly divisible by 3. Our remainder is zero, and the quotient is 21.

11

2.3 Decimal to Binary Conversions

Converting 190 to base 3...

Continue in this way until the quotient is zero. In the final calculation, we note that 3 divides 2 zero times with a remainder of 2. Our result, reading from bottom to top is:

19010 = 210013

12

2.3 Decimal to Binary Conversions

Fractional values can be approximated in all base systems. Unlike integer values, fractions do not necessarily have exact representations under all radices. The quantity is exactly representable in the binary and decimal systems, but is not in the ternary (base 3) numbering system.

13

2.3 Decimal to Binary Conversions

Fractional decimal values have nonzero digits to the right of the decimal point. Fractional values of other radix systems have nonzero digits to the right of the radix point. Numerals to the right of a radix point represent negative powers of the radix:
0.4710 = 4 10 -1 + 7 10 -2 0.112 = 1 2 -1 + 1 2 -2 = + = 0.5 + 0.25 = 0.75
14

2.3 Decimal to Binary Conversions

As with whole-number conversions, you can use either of two methods: a subtraction method and an easy multiplication method. The subtraction method for fractions is identical to the subtraction method for whole numbers. Instead of subtracting positive powers of the target radix, we subtract negative powers of the radix. We always start with the largest value first, n -1, where n is our radix, and work our way along using larger negative exponents.
15

2.3 Decimal to Binary Conversions

The calculation to the right is an example of using the subtraction method to convert the decimal 0.8125 to binary.

Our result, reading from top to bottom is: 0.812510 = 0.11012 Of course, this method works with any base, not just binary.
16

2.3 Decimal to Binary Conversions

Using the multiplication method to convert the decimal 0.8125 to binary, we multiply by the radix 2.

The first product carries into the units place.

17

2.3 Decimal to Binary Conversions

Converting 0.8125 to binary . . .

Ignoring the value in the units place at each step, continue multiplying each fractional part by the radix.

18

2.3 Decimal to Binary Conversions

Converting 0.8125 to binary . . .

You are finished when the product is zero, or until you have reached the desired number of binary places. Our result, reading from top to bottom is: 0.812510 = 0.11012 This method also works with any base. Just use the target radix as the multiplier.
19

2.3 Decimal to Binary Conversions

The binary numbering system is the most important radix system for digital computers. However, it is difficult to read long strings of binary numbers-- and even a modestly-sized decimal number becomes a very long binary number.
For example: 110101000110112 = 1359510

For compactness and ease of reading, binary values are usually expressed using the hexadecimal, or base-16, numbering system.
20

2.3 Decimal to Binary Conversions

The hexadecimal numbering system uses the numerals 0 through 9 and the letters A through F.
The decimal number 12 is C16. The decimal number 26 is 1A16.

It is easy to convert between base 16 and base 2, because 16 = 24. Thus, to convert from binary to hexadecimal, all we need to do is group the binary digits into groups of four.
A group of four binary digits is called a hextet
21

2.3 Decimal to Binary Conversions

Using groups of hextets, the binary number 110101000110112 (= 1359510) in hexadecimal is:

Octal (base 8) values are derived from binary by using groups of three bits (8 = 23):

Octal was very useful when computers used six-bit words.

22

2.4 Signed Integer Representation

Several representations exist for negative values:
Sign Magnitude 000 = +0 001 = +1 010 = +2 011 = +3 One's Complement 000 = +0 001 = +1 010 = +2 011 = +3 Two's Complement 000 = +0 001 = +1 010 = +2 011 = +3

100 = -0
101 = -1 110 = -2 111 = -3

100 = -3
101 = -2 110 = -1 111 = -0

100 = -4
101 = -3 110 = -2 111 = -1
23

2.4 Signed Integer Representation

Sign Magnitude
Leftmost bit is sign bit: 0 for positive, 1 for negative Remaining bits are magnitude Drawbacks: Sign bits give problems to addition and subtraction Two representations for 0 Rarely used

24

2.4 Signed Integer Representation

Two's Complement
Easier implementation of addition and subtraction Leftmost bit still indicates sign Positive numbers same as sign magnitude Only one zero (all 0 bits) Negating A: invert (complement) all bits of A and add 1 Example: -55

start with +55: 0000...00110111 complement each bit: 1111...11001000 (1's complement) add 1: 1111...11001001 (2's complement)
25

2.4 Signed Integer Representation

Addition and Subtraction (numbers in 2's complement representation)
Add -55 and +58 (use 8 bits for simplicity) -55: 11001001 +58 00111010 ==> 100000011 underlined leftmost bit is an overflow (ignored here)

Overflow rule:
If two numbers are added and they are both positive or both negative, then overflow occurs if and only if the result has opposite sign
26

2.4 Signed Integer Representation

Subtraction uses addition To subtract B from A, take 2's complement of B and add it to A Digital circuit only need addition and complement circuits

27

2.4 Signed Integer Representation

28

2.4 Signed Integer Representation

Given a full adder (FA), we can use it to add binary digits (up to 3) carry_in
A B carry_out

1-bit Full Adder

A
0 0 0 0 1 1

B carry_in carry_out S
0 0 1 1 0 0 0 1 0 1 0 1 0 0 0 1 0 1 0 1 1 0 1 0
29

2.4 Signed Integer Representation

Several FA's can be used to add binary numbers by feeding out the carry_out one FA to the carry_in of the FA of the left.
add/sub
B31 B2 B1 B0

A31 1-bit FA S31

A2 1-bit FA S2

A1 1-bit FA S1

A0 1-bit FA S0

C32

C31

C3

C2

C1

C0

32-bit Ripple Carry Adder/Subtractor (Better: Carry Lookahead Adder)

Note: add/sub is ON (1) if we want A-B, otherwise is OFF
30

2.4 Signed Integer Representation

Multiplication
More complicate than addition Done by a sequence of shifts and additions Just like the paper-pencil approach. Use an example of a 4-bit multiplication.
1011 1101 1011 0000 1011 Multiplicand (11) Multiplier (13) partial products

1011
10001111

double-precision product (143)

31

2.4 Signed Integer Representation

Paper-pencil approach is inefficient Solution: Do not wait until the end to add partial products. Algorithm (unsigned numbers):
Do n times (e.g, 32 for MIPS) For each 1 on Multiplier, perform an add and a shift right For each 0 in the multiplier, perform only a shift right

Solutions for the multiplication of signed numbers:

1. Convert the numbers to positive numbers, multiple them as above. If signs were different, negate answer. 2. Apply Booth Algorithm

2.4 Signed Integer Representation

Booths algorithm
Multiplier and multiplicand are placed in registers Q & M Q-1, 1-bit register placed to the right of Q0 Initialize A (third register) and Q-1 to 0 Do n times (n is the number of bits in Q): If Q0Q-1 = 01 then A <-- A + M If Q0Q-1 = 10 then A <-- A M Arithmetic shift right A, Q, Q-1

33

2.4 Signed Integer Representation

2.4 Signed Integer Representation

Multiply 3 x 7 = 21 (00010101)
Phase Comments M Initialize registers 0111 1 Q0Q-1 = 10, AA-M ASR A, Q, Q-1 2 Q0Q-1 = 11, ASR A, Q, Q-1 3 Q0Q-1 = 01, AA+M ASR A, Q, Q-1 4 Q0Q-1 = 00, ASR A, Q, Q-1 A 0000 1001 1100 1110 0101 0010 0001 Q Q-1 0011 0 0011 0 1001 1 0100 1 0100 1 1010 0 0101 0
35

2.5 Floating-Point Representation

Integers can be considered as fixed point numbers Decimal point at the far right Floating-point (fp) numbers allow an arbitrary number of decimal places to the right of the decimal point.
For example: 0.5 0.25 = 0.125
A 32-bit fp number: 000000000000000000011101.01001111 is equivalent to:

2 4 + 2 3 + 2 2 + 2 0 + 2 -2 + 2 -5 + 2 -6 + 2 -7 + 2 -8 =
29.30879375
36

2.5 Floating-Point Representation

Very large or very small fp numbers are expressed in scientific notation.
For example: 0.125 = 1.25 10-1 5,000,000 = 5.0 106

Numbers written in scientific notation have three components:

37

2.5 Floating-Point Representation

Computer representation of a fp number:

The one-bit sign field is the sign of the stored value. The size of the exponent field, determines the range of values that can be represented. The size of the significand determines the precision of the representation.
38

2.5 Floating-Point Representation

The IEEE-754 single precision floating point standard uses an 8-bit exponent and a 23-bit significand. The IEEE-754 double precision standard uses an 11-bit exponent and a 52-bit significand.
For illustrative purposes, we will use a 14-bit model with a 5-bit exponent and an 8-bit significand.
39

2.5 Floating-Point Representation

The significand of a floating-point number is always preceded by an implied binary point. Thus, the significand always contains a fractional binary value. The exponent indicates the power of 2 to which the significand is raised. (-1)sign X significand X 2exponent
40

2.5 Floating-Point Representation

Example:
Express 3210 in the simplified 14-bit floating-point model. We know that 32 is 25. So in (binary) scientific notation 32 = 1.0 x 25 = 0.1 x 26. Using this information, we put 110 (= 610) in the exponent field and 1 in the significand as shown.

41

2.5 Floating-Point Representation

The illustrations shown at the right are all equivalent representations for 32 using our simplified model.

Not only do these synonymous representations waste space, but they can also cause confusion.

42

2.5 Floating-Point Representation

Another problem with our system is that we have made no allowances for negative exponents. We have no way to express 0.5 (=2 -1)! (Notice that there is no sign in the exponent field!)
All of these problems can be fixed with no changes to our basic model.
43

2.5 Floating-Point Representation

To resolve the problem of synonymous forms, the first digit of the significand must be 1 (normalized). This results in a unique pattern for each FP number.
In the IEEE-754 standard, this 1 is implied meaning that a 1 is assumed to the left of the binary point. Hidden bit extends the significand by 1 bit Number normalized to the form: 1.bbbbbb

Biased exponents represent negative exponents

They are approximately midway point in range of values. Bias of 127 for single precision,1023 for double precision. Biased exponent: exponent + bias
44

2.5 Floating-Point Representation

Exponent ranges (in theory):
Unbiased: -127 to 128 ( -1023 to 1024) Biased: 0 to 255 ( 0 to 2047)

Example:
Express -12.7510 in the revised single-precision IEEE-754 FP standard Binary: -1100.11 = -1.10011 x 2 3 biased exponent: 3 + 127 = 130 = 1000 0010 significand: 1001 1000 0000 0000 0000 000 (1 to the left of the binary point is hidden) Bit 31 Final FP representation: Bit 0 1 10000010 10011000000000000000000
45

2.5 Floating-Point Representation

2.5 Floating-Point Representation

In IEEE-754 floating point standard:
An exponent of 0 and 255 (2047 for double precision) are used for special values (see next table). Denormalized numbers extend the range of small numbers
Smallest normalized number: 1.0x2-126 Smallest denormalized number: 0.0001 x 2-126 = 2-149

New actual range of exponents:

unbiased exponent range: -126 to 127 (-1022 to 1023)

biased exponent range: -1022 to 1023 (1 to 2046)

47

2.5 Floating-Point Representation

Single Precision E (8) F (23) 0 0 0 nonzero
1-254 anything

255 255

0 nonzero

Double Precision Object Represented E (11) F (52) 0 0 true zero (0) 0 nonzero denormalized number 1-2046 anything floating point number infinity 2047 0 2047 nonzero not a number (NaN)

F: fraction or significand

2.5 Floating-Point Representation

Floating-point arithmetic A FP operation may produce one of the conditions:
Exponent overflow: A positive exponent may exceed the maximum value. Exponent underflow: A negative exponent is less than the minimum value. Significand overflow: May happen during the addition of two significands of same sign. Significand underflow: May happen during significand alignment.
49

2.5 Floating-Point Representation

Addition and subtraction Steps to add 2 FP numbers X and Y:
Step 0: Restore hidden bits: Make explicit the implicit significand bit and change the sign of the subtrahend if it is a subtract operation Step 1: Zero Check: Check if either operand is zero Step 2: Significand alignment: Align significands so that exponents are equal

Step 3: Addition: Add the two significands taking into account their signs. Significand or exponent overflow may exist.
Step 4: Normalization: Normalize the result by shifting left significand digits and decrementing the exponent, which it may cause exponent overflow.
50

2.5 Floating-Point Representation

 Example of addition: Add X and Y

X: 0 10001001 10001000000000000000000 Y: 0 10000110 01101000000000000000000 Step 0: X's significand: 1.10001 Y's significand: 1.01101 Step 1: Passed check Step 2: Align Y's significand: 0.00101101 Step 3: Add significands: 1.10110101 Step 4: result of step 3 is in normalized form Result: 0 10001001 10110101000000000000000

2.5 Floating-Point Representation

Floating-point multiplication and division Simpler processes that addition and subtraction
Step 1: Zero check Step 2: Add (Subtract) exponents and Subtract (Add) bias. The result could be an exponent overflow or underflow. Step 3: Multiply (Divide) significands. Multiplication (Division) is performed in the same way as for integers. Step 4: Normalize the result. It may result in exponent underflow. Step 5: Round the result.
53

2.5 Floating-Point Representation

2.5 Floating-Point Representation

Several alternatives exist for rounding numbers. Round to nearest is the default mode.
Result is rounded to the nearest representable number. Extra bits beyond the 23 bits: 110 --> round up Extra bits beyond the 23 bits: 011 --> round down

Another round mode is round toward zero.

This is equivalent to simply truncation

2.5 Floating-Point Representation

No matter how many bits we use in a floating-point representation, our model must be finite. The real number system is, of course, infinite, so our models can give nothing more than an approximation of a real value. At some point, every model breaks down, introducing errors into our calculations. By using a greater number of bits in our model, we can reduce these errors, but we can never totally eliminate them.
56

2.5 Floating-Point Representation

Our job becomes one of reducing error, or at least being aware of the possible magnitude of error in our calculations. We must also be aware that errors can compound through repetitive arithmetic operations. For example, our 14-bit model (bias = 16) cannot exactly represent the decimal value 128.25. In binary, it is 10 bits wide:
10000000.012 = 128.2510 = 1.000000001 x 27 0 10111 00000000 biased exponent: 10000 + 00111
57

2.5 Floating-Point Representation

When we try to express 128.2510 in our 14-bit model, we lose the low-order bit, giving a relative error of:

128.25 - 128 128.25

0.195%

If we had a procedure that repetitively added 0.25 to 128.25, we would have an error of nearly 1% after only four iterations.

58

2.5 Floating-Point Representation

Floating-point errors can be reduced when we use operands that are similar in magnitude. If we were repetitively adding 0.25 to 128.25, it would have been better to iteratively add 0.25 to itself and then add 128.25 to this sum.

In this example, the error was caused by loss of the low-order bit.

59

2.5 Floating-Point Representation

Floating-point overflow and underflow can cause programs to crash. Overflow occurs when there is no room to store the high-order bits resulting from a calculation. Underflow occurs when a value is too small to store, possibly resulting in division by zero.
Experienced programmers know that its better for a program to crash than to have it produce incorrect, but plausible, results.
60

2.5 Floating-Point Representation

When discussing floating-point numbers, it is important to understand the terms range, precision, and accuracy. The range of a numeric integer format is the difference between the largest and smallest values that is can express. Accuracy refers to how closely a numeric representation approximates a true value.

The precision of a number indicates how much information we have about a value
61

Range of 2s complement and FP numbers

32-bit numbers

2.5 Floating-Point Representation

Most of the time, greater precision leads to better accuracy, but this is not always true.
For example, 3.1333 is a value of pi that is accurate to two digits, but has 5 digits of precision.

There are other problems with floating point numbers. Because of truncated bits, you cannot always assume that a particular floating point operation is commutative or distributive.
63

2.5 Floating-Point Representation

This means that we cannot assume:
(a + b) + c = a + (b + c) or a*(b + c) = ab + ac

Moreover, to test a floating point value for equality to some other number, first figure out how close one number can be to be considered equal. Call this value epsilon and use the statement:
if (abs(x) < epsilon) then ...

64

2.6 Character Codes

Calculations arent useful until their results can be displayed in a manner that is meaningful to people. We also need to store the results of calculations, and provide a means for data input.

Thus, human-understandable characters must be converted to computer-understandable bit patterns using some sort of character encoding scheme.

65

2.6 Character Codes

As computers have evolved, character codes have evolved. Larger computer memories and storage devices permit richer character codes. The earliest computer coding systems used six bits. Binary-coded decimal (BCD) was one of these early codes. It was used by IBM mainframes in the 1950s and 1960s.
66

2.6 Character Codes

In 1964, BCD was extended to an 8-bit code, Extended Binary-Coded Decimal Interchange Code (EBCDIC). EBCDIC was one of the first widely-used computer codes that supported upper and lowercase alphabetic characters, in addition to special characters, such as punctuation and control characters. EBCDIC and BCD are still in use by IBM mainframes today.
67

2.6 Character Codes

Other computer manufacturers chose the 7-bit ASCII (American Standard Code for Information Interchange) as a replacement for 6-bit codes. While BCD and EBCDIC were based upon punched card codes, ASCII was based upon telecommunications (Telex) codes. Until recently, ASCII was the dominant character code outside the IBM mainframe world.

68

2.6 Character Codes

Many of todays systems embrace Unicode, a 16bit system that can encode the characters of every language in the world.
The Java programming language, and some operating systems now use Unicode as their default character code.

The Unicode codespace is divided into six parts. The first part is for Western alphabet codes, including English, Greek, and Russian.

69

2.8 Error Detection and Correction

It is physically impossible for any data recording or transmission medium to be 100% perfect 100% of the time over its entire expected useful life. As more bits are packed onto a square centimeter of disk storage, as communications transmission speeds increase, the likelihood of error increases-sometimes geometrically. Thus, error detection and correction is critical to accurate data transmission, storage and retrieval.

70

2.8 Error Detection and Correction

Check digits, appended to the end of a long number can provide some protection against data input errors.
The last character of UPC barcodes and ISBNs are check digits.

Longer data streams require more economical and sophisticated error detection mechanisms. A checksum is a form of redundancy check to protect data sent or stored.

Cyclic redundancy checking (CRC) codes provide error detection for large blocks of data.
71

2.8 Error Detection and Correction

Checksums and CRCs are examples of systematic error detection. In systematic error detection a group of error control bits is appended to the end of the block of transmitted data.
This group of bits is called a syndrome.

CRCs are polynomials over the modulo 2 arithmetic field.

The mathematical theory behind modulo 2 polynomials is beyond our scope. However, we can easily work with it without knowing its theoretical underpinnings.
72

2.8 Error Detection and Correction

Modulo 2 arithmetic works like clock arithmetic. In clock arithmetic, if we add 2 hours to 11:00, we get 1:00. In modulo 2 arithmetic if we add 1 to 1, we get 0. The addition rules couldnt be simpler:
0+0=0 1+0=1 0+1=1 1+1=0

73

2.8 Error Detection and Correction

Find the quotient and remainder when 1111101 is divided by 1101 in modulo 2 arithmetic.

As with traditional division, we note that the dividend is divisible once by the divisor. We place the divisor under the dividend and perform modulo 2 addition.

74

2.8 Error Detection and Correction

Find the quotient and remainder when 1111101 is divided by 1101 in modulo 2 arithmetic

Now we bring down the next bit of the dividend. We bring down bits from the dividend so that the first 1 of the difference align with the first 1 of the divisor. So we place a zero in the quotient.

75

2.8 Error Detection and Correction

Find the quotient and remainder when 1111101 is divided by 1101 in modulo 2 arithmetic

1010 is divisible by 1101 in modulo 2. We perform the modulo 2 addition.

76

2.8 Error Detection and Correction

Find the quotient and remainder when 1111101 is divided by 1101 in modulo 2 arithmetic

We find the quotient is 1011, and the remainder is 0010.

This procedure is very useful to us in calculating CRC syndromes. Note: The divisor in this example corresponds to a modulo 2 polynomial: X 3 + X 2 + 1.
77

2.8 Error Detection and Correction

Suppose we want to transmit the information string: 1111101. The receiver and sender decide to use the (arbitrary) polynomial pattern, 1101. The information string is shifted left by one position less than the number of positions in the divisor. The remainder is found through modulo 2 division (at right) and added to the information string: 1111101000 + 111 = 1111101111.
78

2.8 Error Detection and Correction

If no bits are lost or corrupted, dividing the received information string by the agreed upon pattern will give a remainder of zero. We see this is so in the calculation at the right. Real applications use longer polynomials to cover larger information strings. Some of the standard polynomials are listed in the text.
79

2.8 Error Detection and Correction

Data transmission errors are easy to fix once an error is detected.
Just ask the sender to transmit the data again.

In computer memory and data storage, however, this cannot be done.

Too often the only copy of something important is in memory or on disk.

Thus, to provide data integrity over the long term, error correcting codes are required.
80

2.8 Error Detection and Correction

Hamming codes and Reed-Soloman codes are two important error correcting codes. Reed-Soloman codes are particularly useful in correcting burst errors that occur when a series of adjacent bits are damaged.
Because CD-ROMs are easily scratched, they employ a type of Reed-Soloman error correction.

Because the mathematics of Hamming codes is much simpler than Reed-Soloman, we discuss Hamming codes in detail.
81

2.8 Error Detection and Correction

Hamming codes are code words formed by adding redundant check bits, or parity bits, to a data word. The Hamming distance between two code words is the number of bits in which two code words differ.
This pair of bytes has a Hamming distance of 3:

The minimum Hamming distance for a code is the smallest Hamming distance between all pairs of words in the code.
82

2.8 Error Detection and Correction

The minimum Hamming distance for a code, D(min), determines its error detecting and error correcting capability. For any code word, X, to be interpreted as a different valid code word, Y, at least D(min) single-bit errors must occur in X. Thus, to detect k (or fewer) single-bit errors, the code must have a Hamming distance of D(min) = k + 1.

83

2.8 Error Detection and Correction

Hamming codes can detect D(min) - 1 errors and

correct

errors

Thus, a Hamming distance of 2k + 1 is required to be able to correct k errors in any data word. Hamming distance is provided by adding a suitable number of parity bits to a data word.

84

2.8 Error Detection and Correction

Suppose we have a set of n-bit code words consisting of m data bits and r (redundant) parity bits. An error could occur in any of the n bits, so each code word can be associated with n erroneous words at a Hamming distance of 1. Therefore,we have n + 1 bit patterns for each code word: one valid code word, and n erroneous words.

85

2.8 Error Detection and Correction

With n-bit code words, we have 2 n possible code words consisting of 2 m data bits (where n = m + r). This gives us the inequality: (n + 1) 2 m 2 n Because n = m + r, we can rewrite the inequality as: (m + r + 1) 2 m 2 m + r or (m + r + 1) 2 r
This inequality gives us a lower limit on the number of check bits that we need in our code words.
86

2.8 Error Detection and Correction

Suppose we have data words of length m = 4. Then: (4 + r + 1) 2 r implies that r must be greater than or equal to 3. This means to build a code with 4-bit data words that will correct single-bit errors, we must add 3 check bits. Finding the number of check bits is the hard part. The rest is easy.
87

2.8 Error Detection and Correction

Suppose we have data words of length m = 8. Then: (8 + r + 1) 2 r implies that r must be greater than or equal to 4. This means to build a code with 8-bit data words that will correct single-bit errors, we must add 4 check bits, creating code words of length 12. So how do we assign values to these check bits?
88

2.8 Error Detection and Correction

With code words of length 12, we observe that each of the digits, 1 though 12, can be expressed in powers of 2. Thus:
1 = 20 2 = 21 3 = 21+ 2 0 4 = 22 5 = 22 + 2 0 6 = 22 + 2 1 7 = 22 + 21 + 2 0 8 = 23 9 = 23 + 2 0 10 = 2 3 + 2 1 11 = 2 3 + 2 1 + 2 0 12 = 2 3 + 2 2

1 (= 20) contributes to all of the odd-numbered digits. 2 (= 21) contributes to the digits, 2, 3, 6, 7, 10, and 11. . . . And so forth . . .

We can use this idea in the creation of our check bits.

89

2.8 Error Detection and Correction

Using our code words of length 12, number each bit position starting with 1 in the low-order bit. Each bit position corresponding to an even power of 2 will be occupied by a check bit. These check bits contain the parity of each bit position for which it participates in the sum.

90

2.8 Error Detection and Correction

Since 2 (= 21) contributes to the digits, 2, 3, 6, 7, 10, and 11. Position 2 will contain the parity for bits 3, 6, 7, 10, and 11. When we use even parity, this is the modulo 2 sum of the participating bit values. For the bit values shown, we have a parity value of 0 in the second bit position.

What are the values for the other parity bits?

91

2.8 Error Detection and Correction

The completed code word is shown above.

Bit 1checks the digits, 3, 5, 7, 9, and 11, so its value is 1. Bit 4 checks the digits, 5, 6, 7, and 12, so its value is 1. Bit 8 checks the digits, 9, 10, 11, and 12, so its value is also 1.

Using the Hamming algorithm, we can not only detect single bit errors in this code word, but also correct them!
92

2.8 Error Detection and Correction

Suppose an error occurs in bit 5, as shown above. Our parity bit values are:

Bit 1 checks digits, 3, 5, 7, 9, and 11. Its value is 1, but should be zero. Bit 2 checks digits 3, 6, 7, 10, and 11. The zero is correct. Bit 4 checks digits, 5, 6, 7, and 12. Its value is 1, but should be zero. Bit 8 checks digits, 9, 10, 11, and 12. This bit is correct.

93

2.8 Error Detection and Correction

We have erroneous bits in positions 1 and 4. With two parity bits that dont check, we know that the error is in the data, and not in a parity bit. Which data bits are in error? We find out by adding the bit positions of the erroneous bits. Simply, 1 + 4 = 5. This tells us that the error is in bit 5. If we change bit 5 to a 1, all parity bits check and our data is restored.
94

Chapter 2 Conclusion
Computers store data in the form of bits, bytes, and words using the binary numbering system. Signed integers can be stored in ones complement, twos complement, or signed magnitude representation. Floating-point numbers are usually coded using the IEEE 754 floating-point standard.

95

Chapter 2 Conclusion
Floating-point operations are not necessarily commutative or distributive. Character data is stored using ASCII, EBCDIC, or Unicode.

Error detecting and correcting codes are necessary because we can expect no transmission or storage medium to be perfect.
CRC, Reed-Soloman, and Hamming codes are three important error control codes.
96