1. Chapter 6: Digital Data Communications Techniques
COE 341: Data & Computer Communications (T061) Dr. Radwan E. Abdel-Aal
Chapter 6:
Digital Data Communications Techniques

2. Where are we: Moving from Signal Transmission to Data Communication
Chapter 7: Data Link: Flow and Error control
Data Link
Chapter 8: Improved utilization: Multiplexing
Physical Layer
Chapter 6: Data Communication: Synchronization, Error detection and correction
Chapter 4: Transmission Media
Transmission Medium
Chapter 5: Encoding Signals to represent Data
Chapter 3: Signals and their transmission over media, Impairments

3. Contents Asynchronous and Synchronous Transmission (6.1)
Types of Errors (6.2)
Error Detection (6.3)
Parity Check
Cyclic Redundancy Check (CRC)

4. Asynchronous and Synchronous Transmission:
To communicate meaningful data serially between TX and RX, signal timing should be the same at both
Timing considerations include:
Rate
Duration
Spacing
Need to synchronize RX to TX
Two ways to achieve this:
Asynchronous Transmission
Synchronous Transmission
RX needs to sample the received
data at mid-points
So it needs to establish:
- Bit arrival time
- Bit duration

5. Need for RX and TX Synchronization:
Transmitter
Receiver
TX Clock
RX Clock
Clock drift (example):
If the receiver clock drifts by 1% every sample time,
For Tb = 1 msec, total drift after 50 bit intervals = 50 X 0.01 = 0.5 msec
i.e. instead of sampling at the middle, the receiver will sample bit # 50 at the edge of the bit
RX and TX clocks are out-of-synch  Communication Error!
In general, No. of correctly sampled bits = 0.5 Tb/(n/100)Tb = 50/n,
where n is the % timing error between TX and RX
Two approaches for correct reception:
Send only a few bits ( e.g. a character) at a time (that RX can
sample correctly before losing sync)  Asynchronous Transmission
Keep receiver clock synchronized with the transmitter clock  Synchronous Transmission Tb

6. Asynchronous (?) Transmission: Character-Level Synchronization
Avoids the timing problem by NOT sending long, uninterrupted streams of bits.
So data is transmitted one character at a time:
Has a distinct start bit
Consists of only 5 to 8 bits (so drift will not be a serious problem)
Has a distinct stop bit
Character is delimited (at start & end) by known signal elements: start bit – stop element
Sync needs to be maintained only for the duration of a character
The receiver has a new opportunity to resynchronize at the beginning of next character
 Timing errors do not accumulate from character to character

7. Asynchronous Transmission
(Min)
Binary 1
Binary 1
RX waits for
A character
following the end
of the previous
character
The ‘stop’ bits confirm end
of character  otherwise: Framing Error
Stop bits continue (idling)
till next character
RX “knows” how many bits
To expect in a character,
and keeps counting them
following the ‘start’ bit S1 S2
Receive
Start bit
Stop element
Parity bit: Even or Odd parity?
S1: receiver in idling state
S2: receive in receiving state

8. Asynchronous Transmission
Framing error
Erroneous detection of end/start of a character
Can be caused by:
Noise: (1 is the idling ‘stop’ bit) …
Incorrect timing of bit sampling due to drift of RX clock affects bit count
Erroneous ‘Start’ bit due to noise

9. Asynchronous Transmission
Errors due to lack of sync for an 8-bit system
Let data rate = baud rate = 10 kbps
Bit interval = 1/10k = 100 ms
Assume RX’s clock is faster than TX by 6% (i.e. RX thinks the bit interval is 94 ms)
RX checks mid-bit data at 47 ms and then at 94 ms intervals
Bit 8 is wrongly sampled within bit 7 (bit 7 read twice!)
Possibility of a framing error at the end
100 ms
94 ms
(Timing for
ideal sampling at RX)
47 ms
Half the bit interval
from the ‘start’ rising edge
893

10. Asynchronous Transmission: Efficiency
Uses 1 start bit & 2 stop bits: (i.e. 3 non-data bits)
with 8-bit characters and no parity:
 Efficiency = 8/(8+3) = 72%
 Overhead = 3/(8+3) = 28%

11. Asynchronous Transmission – Pros & Cons
Advantages:
Simple
Cheap
Good for data with large gaps (keyboard)
Snags:
Overhead of 2 or 3 bits per character (~20%)
Timing errors accumulate for large character sizes

12. Synchronous Transmission: Bit-Level Synchronization
Allows transmission of large blocks of data (frames)
TX and RX Clocks must be synchronized to prevent timing drift
Ways to achieve bit-level sync:
Use a separate clock line between TX and RX
Good over short distances
Subject to transmission impairments over long distances
Embed clock signal in data
e.g. Manchester or Differential Manchester encoding
Or carrier frequency for analog signals

13. Synchronous Frame Format
Typical Frame Structure
Preamble bit pattern:
Indicates start of frame
Postamble bit pattern:
Indicates end of frame
Control fields: convey control
information between TX and RX
Data field: data
to be exchanged
data field
8-bit
flag
control field
Preamble/Postamble flags ensure
frame-level synchronization

14. Synchronous Transmission: Efficiency
Example: HDLC scheme uses a total of 48 bits for control, preamble, and postamble fields per frame:
With a data block consisting of 1000 characters (8-bits each),
 Efficiency = 8000/( ) = 99.4%
 Overhead = 48/8048 = 0.6%
Higher efficiency and lower overhead compared to asynchronous

15. Errors in Digital Transmission
Error occurs when a bit is altered between transmission and reception (0  1 or 1  0)
Two types of errors:
Single bit errors
One bit altered
Isolated incidence, adjacent bits not affected
Typically caused by white noise
Burst errors
Contiguous sequence of B bits in which first, last, and any number of intermediate bits are in error
Caused by impulse noise or fading (in wireless)
More common, and more difficult to handle
Effect is greater at higher data rates
What to do about these errors:
Detect them (at least, so we can ask TX to retransmit!)
Correct them (if we can)

16. Error Detection & Correction: Motivation
Assume NO error detection or correction: Number of erroneous frames received would be unacceptable
Hence, for a frame of F bits,
Prob [frame is correct] = (1-BER)F : Decreases with increasing BER & F
Prob [frame is erroneous] = 1 - (1-BER)F = Frame Error Rate (FER)
… 0 0 1
… … F-2 F-1 F
A frame of
F bits
All bits must be
Correct!
Prob [1st bit in error] = BER
Prob [1st bit correct] = 1-BER
Prob [2nd bit in error] = BER
Prob [2nd bit correct] = 1-BER
Prob [Fth bit in error] = BER
Prob [Fth bit correct] = 1-BER
All bits must be
Correct!

17. Motivation for Error Detection & Correction: Example
ISDN specifies a BER = 10-6 for a 64kbps channel
Frame size F = 1000 bits
What is the FER?
FER = 1 – (1 – BER)F = 1 – ( )1000 = 10-3
Assume a continuously used channel
How many frames are transmitted in one day Number of frames/day = (64,000/1000) × 24 × 3600
= × 106
How many erroneous frames/day?
× 106 × 10-3 = × 103
Typical requirement: Maximum of 1 erroneous frame /day!
i.e. frame error rate is too high!
 We definitely need error detection & correction

18. Frame Error Probabilities:
… 0 0 1 P1
1 - P1
1000
Correct
Erroneous
With an
error detection
facility
900
100
Errors,
Undetected
Errors,
Detected P3 P2 20
P1 + P2 + P3 = 1
Without error detection facility: P3 = 0, and:
P2 = 1 – P1 80

19. Error Detection Techniques
Two main error detection techniques:
Parity Check
Cyclic Redundancy Check (CRC)
Both techniques use additional bits that are added to the “payload data” by the transmitter for the purpose of error detection

20. Error Detection: Implementation
Mismatch:
Error Detected

21. Parity Check Simplest error-detection scheme Appending one extra bit:
Even Parity: Will append “1” such that the total number of 1’s is even
Odd Parity: Will append “1” such that the total number of 1’s is odd
Example: If an even-parity is used, RX will check if the total number of 1’s is even
If it is not  error occurred
Note: even number of bit errors go undetected

22. Cyclic Redundancy Check (CRC)
Burst errors will most likely go undetected by a simple parity check scheme
Instead, a more elaborate technique called Cyclic Redundancy Check (CRC) is typically implemented
CRC appends redundant bits to the frame trailer called Frame Check Sequence (FCS)
FCS is later utilized at RX for error detection
In a given frame containing n bits, we define:
k = the number of original data bits
(n – k) = the number of bits in the FCS field
So, that the total frame length is k + (n – k) = n bits
D (k)
FCS (n-k)
T(n)
01110

23. CRC Generation
CRC generation is all about finding the FCS given the data (D) and a divisor (P)
There are three equivalent ways to see how the CRC code is generated:
Modulo-2 Arithmetic Method
Polynomial Method
Digital Logic Method
D (k)
FCS or F (n-k)
T (n)
01110
110101
P (n-k+1)

24. Modulo 2 Arithmetic Binary arithmetic without carry
Equivalent to XOR operation
i.e:
0  0 = 0; 1  0 = 1; 0  1 = 1; 1  1 = 0
1  0 = 0; 0  1 = 0; 1  1 = 1
Examples:
1010
+1010
___________
0000

25. CRC Error Detection Process
Given k-bit data (D), the TX generates an (n – k)-bit FCS field (F) such that the total n-bit frame (T) is exactly divisible by some (n-k+1)-bit predetermined devisor (P) (i.e. gives a zero remainder)
In general, the received frame may or may not be equal to the sent frame
Let the received frame be (T’)
In error-free transmission T’ = T
The RX then divides (T’) by the same known divisor (P) and checks if there is any remainder
If division yields a remainder then the frame is erroneous
If the division yields zero remainder then the frame is error-free unless many erroneous bits in T’ resulted in a new exact division by P
(This is very unlikely but possible, causing an undetected error!)

26. CRC Generation T = 2 (n – k)  D + F (n-k) left shifts
(multiplications by 2)
Data D: ?
LSB
P is 1-bit longer than F
P Must start and end with 1’s

27. CRC Generation
T = 2(n – k)  D + F, What is F that makes T divides P exactly ?
Claim: F is the remainder obtained from dividing {2(n – k)  D} by divisor P
where Q is the quotient and F is the remainder
If this is the correct F, T should now divide P with Zero remainder
Note: For F to be a remainder when dividing by P, it should be 1-bit smaller
(1)

28. CRC Generation: 1. Modulo-2 Arithmetic Method
At TX: CRC Generation (using previous rules):
Multiply: 2(n – k)  D (left shift by (n-k) bits)
Divide: 2(n – k)  D / P
Use the resulting (n – k)-bit remainder as the FCS
At RX: CRC Checking: RX divides the received T (i.e. T’) by the known divisor (P) and checks if there is any remainder

29. Example – Modulo-2 Arithmetic Method
Given
D =
P =
Find the FCS field
Solution:
First we note that:
The size of the data block D is k = 10 bits
The size of P is (n – k + 1) = 6 bits
 Hence the FCS length is n – k = 5
 Total size of the frame T is n = 15 bits

30. Example – Modulo-2 Arith. Method
Solution (continued):
Multiply 2(n – k)  D
2(5)  =
This is a simple shift to the left by five positions
Divide 2(n – k)  D / P (see next slide for details)
÷ yields:
Quotient Q =
Remainder R =
So, FCS = R = : Append it to D to get the full frame T to be transmitted
T =
M FCS

31. Example – Modulo-2 Arith. Method
# of bits < # of bits in P,
 result of division is 0
= FCS = F
Checks you should do (exercise):
- Verify correct operation, i.e. that 2(n-k)D = P*Q + R
- Verify that obtained T ( ) divides P (110101) exactly
(i.e. with zero remainder)

32. Problem 6-12
For P = & D = , find the CRC

33. Chances of missing an error by CRC error detection
Let E be an n-bit number with a bit = 1 at the position of each error bit error occurring in T
Error occurring in T causes bit reversal
Bit reversal is obtained by XORing with 1
So, received Tr = T  E
Error is missed (not detected) if Tr is divisible by P
Since T is made divisible by P, this requires that E is also divisible by P!.
That is a ‘bit’ unlikely!

34. CRC Generation: 2. The Polynomial Method
A k-bit word (D) can be expressed as a polynomial D(x) of degree (k-1) in a dummy variable x, with:
The polynomial coefficients being the bit values
The powers of X being the corresponding powers of 2
bk-1 bk-2 … b2 b1 b0  bk-1Xk-1 + bk-2Xk-2 + … + b1X1 + b0X0
where bi (k-1 ≤ i ≤ 0) is either 1 or 0
Example1: an 8 bit word D = is represented as D(X) = x7+x6+x4+x3+1
We ignore polynomial terms corresponding to 0 bits in the number

35. CRC – Mapping Binary Bits into Polynomials
Example2: What is x4D(x) equal to?
x4D(x) = x4(x7+x6+x4+x3+1) = x11+x10+x8+x7+x4, the equivalent bit pattern is (i.e. four zeros appended to the right of the original D pattern)
Example3: What is x4D(x) + (x3+x+1)?
x4D(x) + (x3+x+1) = x11+x10+x8+x7+x4+ x3+x+1, the equivalent bit pattern is (i.e. pattern 1011 = x3+x+1 appended to the right of the original D pattern)

36. CRC Generation: The Polynomial Way
Binary Arithmetic
Polynomial
T = 2(n – k)  D + F
T(X) = X(n – k)  D(X) + F(X)

37. CRC Calculation - Procedure
Shift pattern D(X), (n-r) bits to the left, i.e. perform the multiplication X(n-r)D(X)
Divide the new pattern by the divisor P(X)
The remainder of the division R(X) (n-r bits) is taken as FCS
The frame to be transmitted T(X) is X(n-r)D(X) + FCS

38. Example of Polynomial Method
D = (k = 10)
P = (n – k + 1 = 6)
n – k = 5  n = 15
Find the FCS field
Solution:
D(X) = X9 + X7 + X3 + X2 + 1
P(X) = X5 + X4 + X2 + 1
X5D(X)/P(X) = (X14 + X12 + X8 + X7 + X5)/(X5 + X4 + X2 + 1)
This yields a remainder R(X) = X3 + X2 + X (details on next slide)
i.e. F = 01110
R is n – k = 5-bit long  Remember to indicate any 0 bits!

39. Example of Polynomial Method
F = R T
01110 P
110101
Insert 0 bits to make 5 bits

40. Choice of P(X)
How should we choose the polynomial P(X) (or equivalently the divisor P)?
The answer depends on the types of errors that are likely to occur in our communication link
As seen before, an error pattern E(X) will be undetectable only if it is divisible by P(X)
It can be shown that all the following error types are detectable :
All single-bit errors, if P(X) has two terms or more
All double-bit errors, if P(X) has three terms or more
Any odd number of errors, if P(X) contains the factor (X+1)
Any burst error whose length is less than the FCS length (n – k)
A fraction (=1-2-(n-k-1) ) of error bursts of length (n-k+1)
A fraction (=1-2-(n-k) ) of error bursts of length > (n-k+1)

41. Choice of P(X)
In addition, if all error-patterns are equally likely, and r = n - k = length of the FCS, then:
For a burst error of length (r + 1), the probability of undetected error is 1/2(r – 1)
For a longer burst error i.e. length > (r + 1), the probability of undetected error is 1/2 r
There are four widely-used versions of P(X)
CRC-12: P(X) = X12 + X11 + X3 + X2 + X + 1
(r = = 12)
CRC-16: P(X) = X16 + X15 + X (r = = 16)
CRC-CCITT: P(X) = X16 + X12 + X5 + 1
CRC-32: P(X) = X32 + X26 + X23 + X22 + X16 + X12 + X X10 + X8 + X7 + X5 + X4 + X2 + X + 1
(r = = 32)
FCS is 1-bit
shorter than P:
FCS Size
P(X) always starts with 1

42. Some CRC Applications CRC-8 and CRC-10 (not shown) are used in ATM
CRC-12 is used for transmission of 6-bit characters. Its FCS length is 12-bits
CRC-16 & CRC-CCITT are used for 8-bit characters in the US and Europe respectively
Used in HDLC
CRC-32 is used for IEEE LAN standard

43. CRC Generation: 3. Digital Logic
k = 10 (size of D) (known data to be TXed)
n – k + 1 = 6 size of P (known divisor) P (X) = X5+X4+X2+1 (110101)
n – k = 5 size of FCS (to be determined at TX)
n = 15
5-element left-shift register
Initially loaded with 0’s
After n left shifts, register will contain the required FCS
Data Block, D
Always
An XOR at C0
P = X5+X4+X2+X0
Divisor is “hardwired” as feedback connections
via XOR gates into the shift register cells
Starting at LSB, for the first (n-k) bits of P, add an XOR only for 1 bits

44. FCS generated in the shift register after n (=15) shift steps
CRC Generation at TX
P = X5+X4+X2+X0
Start with Shift Register
Cleared to 0’s
C4 in
C2 in
C0 in
MSB D
Inputs formed with
Combination Logic
MSB
FCS generated in the shift register after n (=15) shift steps

45. 0’s in the shift register after n (=15) shift steps (if no errors)
CRC Checking at RX
P = X5+X4+X2+X0
Start with Shift Register
Cleared to 0’s
C4 in
C2 in
C0 in
MSB
15 bits D D
Received Frame, T
MSB
FCS
Inputs formed with
Combination Logic
0’s in the shift register after n (=15) shift steps (if no errors)

46. Problem 6-13
A CRC is constructed to generate a 4-bit FCS for an 11-bit message. The generator polynomial is X4+X3+1
Draw the shift register circuit that would perform this task
Encode the data bit sequence (leftmost bit is the LSB) using the generator polynomial and give the code word
Now assume that bit 7 (counting from the LSB) in the code word is in error and show that the detection algorithm detects the error

47. Problem 6-13 – Solution C3  C2 C1 C0  a)
Input data C3  C2 C1 C0 
P(X) = X4+X3+1
b) Data (D) =
D(X) = 1 + X3 + X4 + X6 + X7 + X8
X4D(X) = X12 + X11 + X10 + X8 + X7 + X4
T(X) = X4M(X) + R(X) = X12 + X11 + X10 + X8 + X7 + X4 + X2
Code =
c) Code in error =
yields a nonzero remainder  error is detected
LSB

48. Error Correction
Once an error is detected, what action can the RX take?:
Two alternatives:
RX asks for a retransmission of the erroneous frame
Adopted by data-link protocols such as HDLC and transport protocol such as TCP
A Backward Error Correction (BEC) method
RX attempts to correct the errors if enough redundancy exists in the received data
TX uses Block Coding to allow RX to correct potential errors
A Forward Error Correction (FEC) method
Use in applications that leave no time for retransmission, e.g. VoIP.

49. Error Correction vs. Error Control
Backward error correction by retransmission is not recommended in the following cases:
Error rate is high (e.g. wireless communication)
Will cause too much retransmission traffic network overloading
Transmission distance is long (e.g. satellite, submarine optical fiber cables)
Network becomes very inefficient (Not utilized properly)
Usually:
Error Correction methods: Those that use FEC techniques
Error Control methods: Those that use retransmission