EEE543 – LECTURE 3 DATA TRANSMISSION

Part I -
Line Coding

1
OUTLINE
 On-Off Signaling
 Bipolar Signalling

2
SIGNALING FORMAT
 The digital output of a source encoder is converted (or coded)
into electrical pulses (waveforms) for the purpose of
transmission over the channel
 This process is called line coding or transmission coding
 There are many possible ways of assigning waveforms
(pulses) to the digital data,
 We consider the assigning waveforms to binary symbols in
this section where a pulse p(t) is used to represent 1 or 0

Figure 3.1 3
SIGNALING FORMAT CONT’D
 Unipolar Non-return to Zero signaling
 Symbol 1 is represented by a pulse of constant amplitude for the entire
duration of the bit interval
 Symbol 0 is represented by no pulse

4
SIGNALING FORMAT CONT’D
 Bipolar Non-Return to Zero signaling
 Symbol 1 is represented by a pulse and symbol 0 by a negative pulse
 The assigned amplitude is maintained for the entire bit interval in either
case

5
SIGNALING FORMAT CONT’D
 Unipolar Return to Zero signaling
 Symbol 1 is represented by a positive pulse that returns to zero before the
end of the bit interval
 Symbol 0 is represented by the absence of a pulse

6
SIGNALING FORMAT CONT’D
 Bipolar Return to Zero signaling
 Symbol 1 is represented by a positive half pulse
 Symbol 0 is represented by a negative half pulse of equal amplitude to
symbol 1.
 In either case the pulse return to zero before the end of the bit interval

7
SIGNALING FORMAT CONT’D
 Alternate mark inversion return to zero signaling
 Positive and negative pulses, of equal amplitudes, are use alternatively for
symbol 1 and no pulse is used for symbol 0.
 In either case, the pulse return to zero before the end of the bit interval.

8
SIGNALING FORMAT CONT’D
 Split-Phase (Manchester) signaling
 Symbol 1 is represented by a positive and followed by a negative pulse –
both pulses occupying half of the bit duration
 Symbol 0 is represented by a negative (first) followed by a positive pulse
both half the bit interval

9
SIGNALING FORMAT CONT’D
Example: consider the binary sequence 100110101. Draw the pulse
waveforms of the following signaling formats:
a) Split-Phase (Manchester) signaling
b) Alternate mark inversion return to zero signaling
c) Bipolar Return to Zero signaling

10
PROPERTIES OF LINE CODES
 Digital data can be transmitted by various transmission or
line codes.
 A line code should have the following properties:
 Transmission bandwidth should be as small as possible
 Error detection and correction capability. It is desirable to
detect, and preferably correct, detection errors. In a bipolar
case, a single error will cause bipolar violation and can easily
be detected.
 Favorable power spectral density. It is desirable to have zero
power spectral density (PSD) at f = 0 (dc) because ac coupling
and transformers are often used at the repeaters
 Adequate timing content. It should be possible to extract
timing or clock information from the signal.
 Transparency. It should be possible to correctly transmit a
digital signal regardless of the pattern of 1s and 0s. A code is
transparent if the data are so coded that for every possible
sequence of data, the coded signal is received faithfully. 11
PSD OF LINE CODES
 We wish to find the power spectrum density of a pulse train
generated from using each of the line codes discussed above
 There is a relationship between the PSD and the
autocorrelation function of the baseband modulation signal
 We consider a generic pulse 𝑝 𝑡 whose corresponding Fourier
transform is 𝑃 𝑓
 We can denote the line code symbol at time 𝑘 as 𝑎𝑘

 The transmission rate is 𝑅𝑏 = 1ൗ𝑇 pulses per second, the line

𝑏
code generates a pulse train constructed from the basic pulse
𝑝 𝑡 with amplitude 𝑎𝑘 starting at time 𝑡 = 𝑘𝑇𝑏
 The 𝑘th symbol is transmitted as 𝑎𝑘 𝑝 𝑡 − 𝑘𝑇𝑏

 The baseband signal is a pulse train of the form

𝑦 𝑡 = ෍ 𝑎𝑘 𝑝 𝑡 − 𝑘𝑇𝑏
𝑘
12
PSD OF LINE CODES CONT’D
 Thefigure below illustrate the pulse 𝑝 𝑡 and the generated
pulse train 𝑦 𝑡

Figure 3.2
 Assuming the input 1 for positive pulse of amplitude +𝑎𝑘 and
13
0 a negative pulse of amplitude −𝑎𝑘 : the input that produce
𝑦 𝑡 is 1 0 0 1 1 0 1 0 1
PSD OF LINE CODES CONT’D
𝑦 𝑡 is a pulse-amplitude-modulated (PAM) signal
 The line coder determines the symbol 𝑎𝑘 as the amplitude of
the pulse 𝑝 𝑡 − 𝑘𝑇𝑏
 The on-off, polar, and bipolar line codes are all special cases
of this pulse train 𝑦 𝑡 , where 𝑎𝑘 takes on values 0, 1, or - 1
randomly
 We can analyze many line codes according to the PSD of 𝑦 𝑡
 Unfortunately, the PSD of 𝑦 𝑡 depends on both 𝑎𝑘 and 𝑝 𝑡 , If
the pulse shape 𝑝 𝑡 changes, we may have to derive the PSD
all over again.
 This difficulty can be overcome by the simple artifice of
selecting a PAM signal 𝑥 𝑡 that uses a unit impulse for the
basic pulse 𝑝 𝑡
 The impulses are at the intervals of 𝑇𝑏 and the strength
(area) of the 𝑘th impulse is 𝑎𝑘
 If 𝑥 𝑡 is applied to the input of a filter that has a unit
impulse response ℎ 𝑡 = 𝑝 𝑡 the output will be the pulse 14
train 𝑦 𝑡
PSD OF LINE CODES CONT’D
 The figure below illustrated the signal 𝑥 𝑡 and the filter with ℎ 𝑡 =
𝑝 𝑡

Figure 3.3
 The power spectral density of 𝑦 𝑡 is
𝑆𝑦 𝑓 = 𝑃 𝑓 2 𝑆𝑥 𝑓
 This relationship allows us to determine 𝑆𝑦 𝑓 , the PSD of a line
code corresponding to any pulse shape 𝑝 𝑡 , once we know 𝑆𝑥 𝑓 - 15

this is a generalized approach

PSD OF LINE CODES CONT’D
 We now need to derive 𝑅𝑥 𝜏 the time autocorrelation function
of the impulse train 𝑥 𝑡
 Consider the pulse train 𝑥 𝑡 as a form of a time limiting
rectangular pulse as shown in figure 3.4

Figure 3.4
 The height of the pulse is ℎk and the width is 𝜖, if we let 𝜖 →
𝑎 16
0, the height of the pulse is ℎ𝑘 = 𝑘
𝜖
PSD OF LINE CODES CONT’D
 This way, we guarantee that the strength of the 𝑘th impulse
is 𝑎𝑘
 If we designate the corresponding rectangular pulse train by
𝑥ො 𝑡 the autocorrelation

 The integral is the area under the signal 𝑥ො 𝑡 multiplied by

𝑥ො 𝑡 delayed by 𝜏 for (𝜏 < 𝜖) as shown in figure 3.4 b)
𝑎𝑟𝑒𝑎 = 𝑥ො 𝑡 . ℎ𝑘 𝜖 − 𝜏
= ℎ𝑘 . ℎ𝑘 𝜖 − 𝜏
= ℎ𝑘2 𝜖 − 𝜏

17
PSD OF LINE CODES CONT’D
 The autocorrelation is given by:

 Where

 During the averaging interval 𝑇 (𝑇 → ∞), there are N pulses (𝑁 →

∞), where 𝑁 = 𝑇ൗ𝑇𝑏 then

18
PSD OF LINE CODES CONT’D
 Observe that the summation is over N pulses hence 𝑅0 is the
time average of the square of the pulse amplitudes 𝑎𝑘 then
 𝑅0 can be written as
1
𝑅0 = lim ෍ 𝑎𝑘2 = 𝑎෤𝑘2
𝑁→∞ 𝑁
𝑘
 The autocorrelation is an even function so we can express is
as

𝑅0
 This is a triangular pulse of height and width 2𝜖 centered
𝜖𝑇𝑏
at 𝜏 = 0
 This is expected because as 𝜏 increases beyond 𝜖 , there is no
overlap between the delayed signal 𝑥ො 𝑡 − 𝜏 and 𝑥ො 𝑡 ; hence,
𝑅𝑥ො 𝜏 = 0
19
PSD OF LINE CODES CONT’D
 The figure below illustrate the rectangular pulse

Figure 3.5

 As we increase 𝜏 further, we find that the kth pulse of 𝑥ො 𝑡 − 𝜏 will

start overlapping the (k + 1 )th pulse of 𝑥ො 𝑡 as 𝜏 approaches 𝜏𝑇𝑏

20
PSD OF LINE CODES CONT’D
 The overlapping of the kth pulse with the (k+1)th pulse of x(t)

Figure 3.6
 We see that 𝑅𝑥ො 𝜏 will have another triangular pulse of width
𝑅
2𝜖 centered at 𝜏 = 𝑇𝑏 and of height 1 where
𝜖𝑇𝑏

21
PSD OF LINE CODES CONT’D
𝑅𝑛
 At 𝜏 = 𝑛𝑇𝑏 we have another rectangular pulse of height
𝜖𝑇𝑏
where

 𝑅𝑛 isessentially the discrete autocorrelation function of the

line code symbols 𝑎𝑘
 To find 𝑅𝑥 𝜏 , we let 𝜖 → 0 in 𝑅𝑥ො 𝜏

 As 𝜖 → 0, the width of each triangular pulse → 0 and the

height → ∞ the area is still finite The triangular pulses
become impulses
22
PSD OF LINE CODES CONT’D
 The figure shows the pulses for 𝜖 → 0

Figure 3.7
𝑅𝑛
 For the nth pulse centered at 𝑛𝑇𝑏 , the height is and the
𝜖𝑇𝑏
𝑅𝑛
area is , the autocorrelation is given by:
𝑇𝑏

23
PSD OF LINE CODES CONT’D
 The PSD 𝑆𝑥 𝑓 is the Fourier transform of 𝑅𝑥ො 𝜏 .

 Since 𝑅−𝑛 = 𝑅𝑛 , we have

 Theinput 𝑥 𝑡 to the filter with impulse response ℎ 𝑡 = 𝑝 𝑡

results in the output 𝑦 𝑡 , the PSD of y 𝑡 is given by

24
PSD OF LINE CODES CONT’D
 The PSD of a line code is fully characterized by its 𝑅𝑛 and the pulse-
shaping selection 𝑃 𝑓
 We shall now use this general result to find the PSDs of various
specific line codes by first determining the symbol autocorrelation 𝑅𝑛

25
POLAR SIGNALING
 In polar signaling, 1 is transmitted by a pulse 𝑝(𝑡) and 0 is
represented by −𝑝(𝑡)
 The amplitudes 𝑎𝑘 is equally likely to be 1 or - 1, and a; is
always 1.

 There
are N pulses and 𝑎𝑘2 = 1 for each one, and the
summation on the right-hand side above is N

 Tocompute 𝑅1 we note that both 𝑎𝑘 and 𝑎𝑘+1 are either 1 or -

1. Hence, 𝑎𝑘 𝑎𝑘+1 is either 1 or – 1
𝑁
 Out of N terms the product 𝑎𝑘 𝑎𝑘+1 is equal to 1 for terms
2
𝑁 26
and is equal to -1 for the remaining terms
2
POLAR SIGNALING CONT’D
 𝑅1 is given by

 𝑅1follows the same argument that the product 𝑎𝑘 𝑎𝑘+𝑛 is

also equally likely to be 1 or -1, hence

 The PSD 𝑆𝑦 𝑓 is given by:

27
POLAR SIGNALING CONT’D
 Consider a specific pulse shape 𝑝(𝑡) defined as

 The PSD, 𝑆𝑦 𝑓 , is given by

 The spectrum of 𝑆𝑦 𝑓 is show in the next slide.

 We note the following
 The polar signal has most of its power concentrated in lower
frequencies
 Theoretically, the spectrum becomes very small as frequency
increases but never becomes totally zero above a certain
frequency.
 Weconsider its first non-dc null frequency to be its essential
28
bandwidth
POLAR SIGNALING CONT’D
 The spectrum of the PSD is

Figure 3.8
 The essential bandwidth of the signal is seen to be 2𝑅𝑏 Hz
(where 𝑅𝑏 is the clock frequency)
 This is 4 times the theoretical bandwidth (Nyquist
bandwidth) required to transmit Rb pulses per second 29
POLAR SIGNALING CONT’D
 Increasing the pulse width reduces the bandwidth (expansion
in the time domain results in compression in the frequency
domain)
 A full-width pulse (maximum possible pulse width), the
essential bandwidth is half, that is 𝑅𝑏 Hz.
 Polar signaling has no capability for error detection or error
correction
 Polar signaling also has nonzero PSD at dc (𝑓 = 0 ), this will
rule out the use of ac coupling during transmission

 Polar signaling is the most efficient scheme from the power

requirement viewpoint.
 For a given power, the error detection probability for a polar
scheme is the lowest among all signaling techniques
 Polar signaling is also transparent because there is always
some pulse (positive or negative) regardless of the bit
sequence 30
ON-OFF SIGNALING
 In on-off signaling, a 1 is transmitted by a pulse p(t) and a 0
is transmitted by no pulse
 A pulse strength 𝑎𝑘 is equally likely to be 1 or 0 and
𝑁
 Out of N pulses in the interval of T seconds, 𝑎𝑘 is 1 for
2
𝑁
pulses and is 0 for the remaining pulses on the average.
2

 To compute 𝑅𝑛 we need to consider the product 𝑎𝑘 𝑎𝑘+𝑛

 Since 𝑎𝑘 and 𝑎𝑘+𝑛 are equally likely to be 1 or 0, the product
𝑎𝑘 𝑎𝑘+𝑛 is equally likely to be 1 x 1 , 1 x 0, 0 x 1 or 0 x 0, that
is, 1 , 0, 0, 0.
𝑁
 On the average, the product 𝑎𝑘 𝑎𝑘+𝑛 is equal to 1 for terms
4
3𝑁
and 0 for terms 31
4
ON-OFF SIGNALING CONT’D
 The coefficient 𝑅𝑛 is given by

 Therefore, substituting for 𝑅𝑛 and 𝑅0 in 𝑆𝑥 𝑓 we have

 The last part we splitting the term 1ൗ2𝑇𝑏 corresponding to 𝑅0

into two: 1ൗ4𝑇𝑏 outside the summation and 1ൗ4𝑇𝑏 inside the
summation (corresponding to 𝑛 = 0) 32
ON-OFF SIGNALING CONT’D
 The summation term can be written as

 Substitution of this result in 𝑆𝑥 𝑓 we have

 The PSD of the waveform 𝑦 𝑡 is

𝑆𝑦 𝑓 = 𝑃 𝑓 2𝑆 𝑓
𝑥

33
ON-OFF SIGNALING CONT’D
 The on-off PSD has an additional discrete part.

 Thisdiscrete part may be nullified if the pulse shape is

chosen such that
𝑛
𝑃 , 𝑛 = 0, ±1, ±2, … ,
𝑇𝑏
 For example, if we choose the half-width rectangular pulse:

 Then

34
ON-OFF SIGNALING CONT’D
 The resulting PSD of on-off signaling using the half
triangular pulse is shown below

35
Figure 3.9
ON-OFF SIGNALING CONT’D
𝑇𝑏 𝑛𝑓𝑇𝑏
 The continuous component of the spectrum is sinc 2
16 2
 The discrete component is represented by the product of an impulse
𝑇 𝑛𝑓𝑇
train with the continuous component 16𝑏 sinc 2 2 𝑏
1
 The impulses repeat at the clock frequency 𝑅𝑏 = 𝑇 because its
𝑏
2𝜋 1
fundamental frequency is rad/s, or Hz.
𝑇𝑏 𝑇𝑏
 On-off signaling is less immune to noise interference - This is
because the noise immunity depends on the difference of amplitudes
representing 1 and 0.
 On-off signaling uses pulses of amplitudes 2 and 0, and requires
twice as much power as polar signaling
 If a pulse of amplitude 1 or - 1 has energy E, then the pulse of
amplitude 2 has energy 2 2 𝐸 = 4𝐸
 The average signal power of on-off is

36
BIPOLAR SIGNALING
A 0 is transmitted by no pulse, and a 1 is transmitted by a
pulse 𝑝(𝑡) or −𝑝(𝑡) depending on whether the previous 1 was
transmitted by 𝑝(𝑡) or −𝑝(𝑡)
 Bipolar signaling therefore uses three symbols 𝑝 𝑡 , 0, −𝑝 𝑡

 To calculate the PSD, we have

 On the average, half of the 𝑎𝑘 s are 0, and the remaining half

are either 1 or -1, with 𝑎𝑘2 = 1
 𝑅0 is therefore given by

37
BIPOLAR SIGNALING CONT’D
 To compute 𝑅1 , we consider the pulse strength product 𝑎𝑘 𝑎𝑘+1
 There are four equally likely sequences of two bits: 11, 10, 01,
00
 Since bit 0 is encoded by no pulse (𝑎𝑘 = 0), the product
𝑎𝑘 𝑎𝑘+1 is zero for the last three of these sequences.
3𝑁
 On the average, there are combinations have 𝑎𝑘 𝑎𝑘+1 = 0
4
𝑁
and only combinations have nonzero 𝑎𝑘 𝑎𝑘+1
4
 Due to the bipolar rule, the bit sequence 1 1 can be encoded
only by two consecutive pulses of opposite polarities (i.e.
𝑁
𝑎𝑘 𝑎𝑘+1 = −1) for the combinations, therefore
4

38
BIPOLAR SIGNALING CONT’D
 To compute 𝑅2 we need to observe the product 𝑎𝑘 𝑎𝑘+2 , in this
case we consider all possible combinations of three-bit
sequences: 111, 101, 110, 100, 011, 010, 001, 000
 Six combinations have either the first and/or the last bit 0
hence 𝑎𝑘 𝑎𝑘+2 = 0 for all these six combinations
 The first two combinations are the only ones that yield
nonzero 𝑎𝑘 𝑎𝑘+2
 For the sequence 111 𝑎𝑘 𝑎𝑘+2 = 1, and for 101 𝑎𝑘 𝑎𝑘+2 = −1,
using the bipolar rule
N N
 On average, 𝑎𝑘 𝑎𝑘+2 = 1 for terms, 𝑎𝑘 𝑎𝑘+2 = −1 for terms
8 8
3N
and 0 for terms
4

39
BIPOLAR SIGNALING CONT’D
 Ingeneral, for n > 2, the product 𝑎𝑘 𝑎𝑘+2 can be 1, -1, or 0
 An equal number of combinations have values 1 and - 1. This
causes 𝑅𝑛 = 0, therefore
𝑅𝑛 = 0, 𝑛>2
 The PSD of Bipolar signaling is given by:

 Notethat 𝑆𝑦 𝑓 = 0 for 𝑓 = 0 (dc), regardless of 𝑃 𝑓 . Hence,

the PSD has a de null, which is desirable for ac coupling
1 1
 sin
2
𝜋𝑓𝑇𝑏 = 0 at 𝑓 = , that is, at 𝑓 = = 𝑅𝑏 Hz. Thus,
𝑇𝑏 𝑇𝑏
regardless of 𝑃 𝑓 we are assured of the first non-de null 40
bandwidth 𝑅𝑏
BIPOLAR SIGNALING CONT’D
 For the half-width pulse

 The figure below illustrate

41

Figure 3.10
BIPOLAR SIGNALING CONT’D
1
 The essential bandwidth of the signal is 𝑅𝑏 = 𝑇 which is half that of
𝑏
polar using the same half-width pulse or on-off signaling and twice
the theoretical minimum bandwidth.
 Bipolar signaling has several advantages:
 its spectrum has a de null;
 its bandwidth is not excessive;
 it has single-error-detection capability

42
EXERCISES
1. Consider a full-width rectangular pulse shape
𝑡
𝑝 𝑡 =Π
𝑇𝑏

(a) Find PSDs for the polar, on-off, and bipolar signaling.
(b) Sketch roughly the PSDs and find their bandwidths. For each case,
compare the bandwidth to the case where p(t) is a half-width rectangular
pulse.
2. (a) A random binary data sequence 110100101- • • is transmitted
by using a Manchester (splitphase) line code with the pulse 𝑝 𝑡
shown

Sketch the waveform y(t)

2. (b)Derive 𝑆𝑦 𝑓 , the PSD of a Manchester (split-phase) signal in
43
part (a) assuming 1 and 0 equally likely. Roughly sketch this PSD
and find its bandwidth.