Chapter 1

Chapter 1 EE 322-Digital Communications
Chapter 1
PRINCIPLES OF BASEBAND DIGITAL
DATA TRANSMISSION
1.1 Introduction
Advantages of digital communications
1. Digital signals are more immune to channel noise by using channel coding
(perfect decoding is possible!)
2. Repeaters along the transmission path can detect a digital signal and
retransmit a new noise-free signal (In long distance systems, noise does not
accumulate from repeater to repeater).
3.
4. Digital signals derived from all types of analog sources can be represented
using a uniform format
5. Relatively inexpensive circuits may be used.
6. Privacy is preserved by using data encryption
7. Digital signals are easier to process by using microprocessors and VLSI
(e.g., digital signal processors, FPGA)
8. Data from voice, video, and data sources may be merged and transmitted
over a common digital system.
9. Errors may often be corrected with the use of coding.
10. More and more things are digital e.g., English text, computer data.
Disadvantages of digital communications:

1. Generally, more bandwidth is required that for analog Systems.
2. Synchronization is required
3. Requires A/D conversions at high rate
Fundamental elements of digital communication systems

1. Information source: analog or digital
2. Source encoder: converts the information source to binary form and
removes the redundancy. (source encoding for data compression)
3. Channel encoder: adds redundancy to overcome the effects of noise and
interference introduced by the channel.
4. Digital modulator: maps the binary information sequence into signal
waveforms.
5. Digital demodulator: convert the received waveform to binary information
sequence.
6. Channel decoder: corrects the demodulated errors based on the redundancy
added by the channel encoder.
7. Source decoder: recover the original information source.
Page 1 of 37
Fig. 1 Block diagram of basic digital Communication System
Various unwanted undesirable effect crop up in transmissions

Attenuation
 Reduces signal strength at the receiver.
Distortion
 Waveform perturbation caused by imperfect response of the system to
the desired signal itself.
 Equalizer can be used to reduce the effect.
Interference
 Contamination by extraneous signals from human sources
Noise
 Random and unpredictable electrical signals from internal or external
to the system.
 The term SNR (signal to noise ratio) is used to measure performance
(noise) relative to an information analog signal.
 The term BER (Bit Error Rate) is used in digital system to measure the
deterioration of the signal.
Page 2 of 37
What makes a Communication System GOOD?
We can measure the “GOODNESS” of a communication system in many ways:
1. How close is the estimated (received) signal to the original signal?
o Better estimate = higher quality transmission
o Signal to Noise Ratio (SNR) for analog signals.
o Bit Error Rate (BER) for digital signals.
2. How much power is required to transmit the signal?
o Lower power = longer battery life, less interference
3. How much bandwidth B is required to transmit the signal?
o Less B means more users can share the channel
4. How much information is transmitted?
o In analog systems information is related to Bandwidth of the signal.
o In digital systems information is expressed in bits/sec (bps).
Baseband Digital Data Transmission Systems

Figure 2 shows a block diagram of a baseband digital data transmission system,
which includes several possible signal processing operations.
Fig. 2 Block diagram of a baseband digital data transmission system.
1.2 Line Codes:

Binary data can, however, be transmitted using many other pulse types. The
choice of a particular pair of pulses to represent the symbols 1 and 0 is called line
coding and selection is usually made on the grounds of one or more of the
following considerations:
1. Presence or absence of a DC level.
2. Power spectral density – particularly its value at 0 Hz.
3. Spectral occupancy (i.e. bandwidth).
4. BER performance (i.e. relative immunity from noise).
5. Transparency (i.e. the property that any arbitrary symbol, or bit, pattern can
be transmitted and received).
6. Ease of clock signal recovery for symbol synchronisation.
7. Presence or absence of inherent error detection properties.
Properties of a good line code:

1. Good clock recovery (Self-synchronization) e.g. RZ
2. Zero dc level e.g.(polar, Manchester)
3. Minimum bit-error rate for a given SNR .That is large noise margin e.g. polar
4. Minimum bandwidth e.g. Bipolar NRZ
5. Data transparency(Independent of data) e.g. CMI
6. Inherent error detection. Errors will occur and should be detectable e.g. Bipolar.
Page 3 of 37
Page 4 of 37
Page 5 of 37
Table 1 Comparison of line (baseband) code performance.
In the following discussion, the pulses are spaced Tb seconds apart. Consequently,
the transmission rate is Rb = 1 / Tb pulses per second.
Pulse Duration: There are two classes used here.
– Non return-to-zero (NRZ) where the pulse or symbol duration Ts = the bit period
Tb.
– Return-to-zero (RZ) where the pulse or symbol duration Ts < the bit period Tb.
Usually Ts = 0.5Tb.
• The pulse duration will usually have an effect on the synchronization properties
of the line code (i.e. it determines the presence or absence of a frequency
component at the clock frequency).
1.2.1 Polar line code

In polar signalling, 1 is transmitted by a pulse p(t) and 0 is transmitted by -p(t).
Polar RZ
Page 6 of 37
Polar NRZ
2
 sin  fTb 
PPolar NRZ ( f )  A Tb 
2

  fTb 
Fig.3 Power spectral density of a polar signal
Advantages:
1) Most efficient scheme from the power requirement viewpoint. For a given
power, it can be shown that the detection-error probability for a polar
scheme is the smallest possible.
2) Transparent because there is always some pulse (positive or negative)
regardless of the bit sequence.
Disadvantages:
1) not bandwidth efficient.

2) It has no error-detection or error-correction capability.
3) It has nonzero PSD at dc (ω) = 0). This will rule out the use of ac coupling
during transmission.
4) There is no discrete clock frequency component in the spectrum of the polar
signal. Rectification of the polar signal (RZ), however, yields a periodic
signal of the clock frequency and can readily be used to extract timing.
Page 7 of 37
1.2.2 On-Off line code OOK (Unipolar)
In this case a 1 is transmitted by a pulse p(t) and a 0 is transmitted by no pulse.
Uniploar NRZ
2 2
A2Tb  sin  fTb   1  A2Tb  sin  fTb / 2   1 
n 
PUni. NRZ ( f )  
4   fTb 
 1   ( f )  PUni. RZ ( f )  
16   fTb / 2 
 1    ( f  T )
 Tb   Tb n  b 
Fig. 4 Power spectral density (PSD) of an on-off signal.
Advantages:
1. Easy to generate.
2. Clock component present (For RZ only)
Disadvantages:
1. not transparent
2. excessive transmission bandwidth (Higher BW (2Rb))
3. nonzero power spectrum at dc
4. no error detection (or correction) capability
This code is most used in baseband data transmission and magnetic tape
recording.
Page 8 of 37
1.2.3 Bipolar [Pseudoternary or Alternate Mark Inverted (AMI)] line code:
This is the signalling scheme used in PCM these days. 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). With consecutive pulses alternating,
we can avoid the dc wander and thus cause a dc null in the PSD. Bipolar signalling
actually uses three symbols [p(t), 0, and -p(t)],
and, hence, it is in reality ternary rather than binary signalling.
Bipolar NRZ
2
A2Tb  sin  fTb 
PBipolar RZ ( f )   sin  fTb 
2

4   fTb 
Fig. 5 PSD of bipolar, polar, and split-phase signals normalized for equal powers.
Advantages:
1. Its spectrum has a dc null.
2. It has single-error-detection capability
3. Good error probability and capable of recovering clock information
Page 9 of 37
Disadvantages:
1. The essential bandwidth of the signal is Rb, which is twice the theoretical
minimum bandwidth.
2. Not transparent (Long strings of data causes loss of synchronization).
3. It needs twice much power as unipolar (3 db) and it needs 2 power supplies.
This code has memory and is most used in telephone systems.
High-Density Bipolar (HDB) line code:

Because the AMI is not transparent other methods are used to prevent long strings
of zeros. HDBN also does not have any dc value and have the same data rate.
In this case when a run of N+1 zeros happens, they will be replaced by a code of
length N+1 containing AMI violation.
The most popular form of HDBN is HDB3; which uses to special sequences: 000V
and B00V. The transmit power requirement is a little greater than AMI (about
10%).
B00V is used when there are an even number of ones following the last special
sequence and 000V is used when there are an odd number of ones following the
last special sequence. Consecutive V pulses alternate in sign to avoid dc wander.
Because violation just happens at the fourth bit of the special code, it can be easily
detected and will be replaced by a zero at the receiver.
It is also capable of error detecting because a sign error would make the number
of bipolar pulses between violations even instead of odd.
Fig. 6 (a) HDB3 signal and (b) its PSD.
Page 10 of 37
1.2.4 Manchester coding (split phase or digital phase)
In this line code transitions occur at the middle of each bit. A high to low transition
represents a 1 and a low to high transition represents a 0 or vice versa.
Advantages:
1. There is always a zero dc level regardless of the data sequence.
2. Extracting timing information is easy and the error rate performance is good.
3. The Manchester code is transparent.
Disadvantages:
1. Large bandwidth relative to NRZ type coding and not capable of error
detecting.
• Used on IEEE 802.3 Ethernet LANs.
Differential Manchester
Inversion in the middle of bit interval is used for synchronization – presence or
absence of additional transition at the beginning of next bit interval identifies
the bit 0 = transition, 1 = no transition
• Perfect synchronization
• Fine for long runs of 1s, but wastes bandwidth for long runs of 0-s
• used by IEEE 802.5 (Token Ring)
1 0 0 1 0 0 0 1 1 1 0
Page 11 of 37
Doubinary:
Doubinary signaling: “Dou” means “doubling the transmission capacity of a
straight binary system.”
A zero is represented by no transition and one by alternate transitions according to

this rule:
Positive if the 1 is preceded by and even number of 0s.
Negative: if the 1 is preceded by an odd number of 0s
NRZ-I (Non Return to Zero Inverted)

NRZI maps binary signals to physical signals during transmission. If a data bit is 1,
NRZI transitions at the clock boundary. If a data bit is 0, there is no transition.
NRZI may have long series of 0s or 1s, resulting in clock recovery difficulties.
1.2.5 M-Ary Coding (Signaling)

• In binary coding:
– Data bit ‘1’ has waveform 1
– Data bit ‘0’ has waveform 2
– Data rate = bit rate = symbol rate
• In M-ary coding, take M bits at a time (M = 2k) and create a waveform (or
symbol).
– ‘00’  waveform (symbol) 1
– Symbol rate = bit rate/k
• For M-ary PAM transmission, there are M possible symbols with symbol
duration T.
1/T is referred to as the signaling rate or symbol rate or symbols per second or
baud
Page 12 of 37
 Some equivalences
 Each symbol can be equivalently identified with log2M bits.
 So the baud rate 1/T can be equivalently transformed to bps as:
T  Tb log 2 ( M )
 Gray code: Any dibit differs from an adjacent dibit in a single bit
position.
Advantages:
o M-ary signals reduce required bandwidth
o Required transmission rate is low (bit rate/M) Instead of transmitting
one pulse for each bit (binary PCM), we transmit one multilevel pulse a
group of k-bits (M=2k)
o Bit rate = Rb bits/s  min BW = Rb/2
o Symbol rate = R/k sym/s  min BW = Rb/2k
o Needed bandwidth goes down by k
Disadvantages:
– Low signal to noise ratio (due to multiple amplitude pulses)
• Trade-off is relatively high bit error rate (BER)
Page 13 of 37
1.3 The Intersymbol Interference (ISI) Problem
Intersymbol interference (ISI) occurs when a pulse spreads out in such a way that
it interferes with adjacent pulses at the sample instant. If the rectangular
multilevel pulses are filtered improperly as they pass through a communications
system, they will spread in time, and the pulse for each symbol may be smeared
into adjacent time slots and cause Intersymbol Interference.
 ISI arises when the channel is dispersive

 Frequency limited  time unlimited  ISI
 Time limited  bandwidth unlimited  bandpass channel  time unlimited
 ISI
 We wish to design transmit and receive filters to minimize the ISI.
 When the signal-to-noise ratio is high, as is the case in a telephone system,
the operation of the system is largely limited by ISI rather than noise.
Page 14 of 37
How can we restrict BW and at the same time not introduce ISI?
Three strategies for eliminating ISI:

1) Use a line code that is absolutely bandlimited.
• Would require Sinc pulse shape.
• Can’t actually do this (but can approximate).
2) Use a line code that is zero during adjacent sample instants.
• It’s okay for pulses to overlap somewhat, as long as there is no
overlap at the sample instants.
• Can come up with pulse shapes that don’t overlap during
adjacent sample instants.
 Raised-Cosine Rolloff pulse shaping
3) Use a filter at the receiver to “undo” the distortion introduced by the
channel (Equalizer.
1.3.1 Nyquist’s First Method for Zero ISI: Time domain
Suppose 1/T is the sample rate

The necessary and sufficient condition for p(t) to satisfy
1,  n  0  
 E , for i  0
p  nT    pi  p(iTb )  
0,  n  0  
0, for all i  0
1 1
 Tb  
2 B0 Rb

 i 
p(t )   p R  sin c( Rbt  i)
i   b
E sin( Rbt )
Popt (t )  E sin c( Rbt ) 
 Rbt
Is that its Fourier transform P(f) satisfy

 P f  m T   T
m 
 E R R
 , for  b  f  b
popt ( f )   Rb 2 2
0,
 otherwise
  Rb 
sin  t T  t  T ,  f  
p t    sinc   P  f     2  ,
t T  0,  otherwise 

Page 15 of 37
FIG. 7 (a) Sinc function p(t) as the optimum pulse shape. (b) Optimum pulse spectrum.
1
 When Tb  , rectangular function satisfy Nyquist condition
Rb
 Since pulses are not possible to create due to:

 Infinite time duration.
 Sharp transition band in the frequency domain.
 The time function p(t) decreases as 1/|t| for large |t|, resulting in a
slow rate of decay, Therefore, it is very sensitive to sampler phase
 The Sinc pulse shape can cause significant ISI in the presence of timing
errors.
Page 16 of 37
 If the received signal is not sampled at exactly the bit instant
(Synchronization Errors), then ISI will occur.
 We seek a pulse shape that:
 Has a more gradual transition in the frequency domain.
 Is more robust to timing errors.
 Yet still satisfies Nyquist’s first method for zero ISI.
Raised Cosine-Rolloff Nyquist Filtering
To ensure physical realizability of the overall pulse spectrum P(f), the modified P(f)
decreases toward zero gradually rather than abruptly
1. Flat portion, which occupies the frequency band 0≤|f| ≤f1 for some
parameter f1 to be defined
2. Roll-off portion, which occupies the frequency band f1 ≤|f| ≤2B0-f1
3. One full cycle of the cosine function defined in the frequency domain,
which is raised up by an amount equal to its amplitude
4. The raised-cosine pulse spectrum
 E
 , 0  f  f1
 Rb

 E    ( f  f1 )  
p( f )   1  cos    , f1  f  Rb  f1
 2 Rb   2(0.5Rb  f1 )  
0, Rb  f1  f


2 f1
 The roll-off factor   1
Rb
 cos( Rbt ) 
p(t )  E sin c( Rbt )  2 2 2 
 1  4 Rb t 
The amount of intersymbol interference resulting from a timing error ∆t decreases
as the roll-off factor is increased form zero to unity
For special case of   1

 sin c(2 Rbt ) 
p(t )  E  2 2 
 1  4 Rb t 
Page 17 of 37
FIG. 8 (a) Raised-cosine pulse spectrum for varying roll-off rates. (b) Pulse response p(t) (i.e.,
inverse Fourier transform of P(f ) for varying roll-off rates).
Transmission-Bandwidth Requirement
 The transmission bandwidth required by using the raised-cosine pulse spectrum
R
is BT  b (1   )
2
 Excess channel
 The transmission bandwidth requirement of the raised-cosine spectrum
exceeds that of the optimum Nyquist channel
R
fv   b
2
 When the roll-off factor is zero, the excess bandwidth is reduced to zero
R
 When the roll-off factor is unity, the excess bandwidth is increased to b .
2
Two additional Properties of the Raised-Cosine Pulse Spectrum

Property 1
 The roll-off protion of the spectrum P(f) exhibits odd symmetry about the
R
midpoints f=± b
2
Pv ( f )  Popt ( f )  P( f )
A unique characterization of the roll-off portion of the raised-cosine spectrum
0, for 0  f  f1

 E    ( f  f1 )   Rb
 2 R 1  cos  2(0.5R  f )   , for f1  f  2
 b   b 1 
Pv ( f )  
 E    ( f  f1 )   Rb

 2R 1  cos    , for  f  Rb  f1

  2(0.5Rb  f1 )   2
 b
0, for Rb  f1  f  Rb

Page 18 of 37
Pv ( f ' )   Pv ( f ' )
R
f' f  b
2
1. The baseband raised-cosine pulse spectrum P(f) of Fig. 9 (a) is centered on the
origin at f=0, whereas the vestigial sideband spectrum is centered on the
sinusoidal carrier frequency fc
2. The parameter fv in Fig. 9 (a) refers to the excess bandwidth measured with
respect to the ideal brick-wall solution for zero intersymbol interference,
whereas the parameter fv in refers to the excess bandwidth measured with
respect to the optimum bandwidth attainable with single sideband modulation..
FIG. 9 (a) Nyquist and raised-cosine pulse spectra for positive frequencies.
(b) Residual spectrum P(f ).
Property 2
 The finite summation of replicas of the raised-cosine pulse spectrum, spaced
by Rb hertz, equals a constant

E
 P( f  mR ) 
m
b
Rb

 n   n  
  R   R  b m
P  t   R P( f  mRb )
n   b  b  
1
 Sampling the modified pulse response p(t) at the rate ,
Rb
 n   cos( n ) 
P    E sin c(n)  2 2 
 Rb   1  4n  
sin( n )
1) sin c( n) 
n
1, forn  0

0, forn  1, 2,...
Page 19 of 37
2) cos( n )  1 for n  0,
 Finally, nothing that the Fourier transform of the delta function is unity,
 n    E , for n  0
P   
 Rb   0, for n  1, 2,...

E (t )  Rb  P( f  mR )
m 
b

E
 (t )   P( f  mRb )
Rb m
 Given the modified pulse shape p(t) for transmitting data over an imperfect
channel using discrete pulse-amplitude modulation at the data rate 1/T, the
pulse shape p(t) eliminates intersymbol interference if, and only if, its
spectrum P(f) satisfies the condition

 m Rb
 P  f  T   constant,
m 
f 
2
 Raised Cosine Filter is also called a NYQUIST FILTER.

 NYQUIST FILTERS refer to a general class of filters that satisfy the
NYQUIST’s First Criterion.
1.3.2 Nyquist’s Second Criterion for no ISI
 Values at the pulse edge are distortionless

Consider a pulse specified by
1 n  0,1
p (nTb )  
0 for all other n
Fig. 10 Communication using controlled ISI or Nyquist second criterion pulses.
Page 20 of 37
1.3.3 Nyquist’s Third Criterion for no ISI
 Within each symbol period, the integration of signal (area) is proportional to
the integration of the transmit signal (area)
 ( t ) Rb
 sin( T ) , f  Rb
2 2
( t ) j 2 ft
P( f )   p(t )   e df
0, Rb Rb sin( T )
f  
 2 2
2 n1T
1, n0
A  2 n1 p (t )dt  
2
2
T
0, n0
1.4 Correlative-Level Coding or Partial-response signaling

ISI, when generated in an uncontrolled manner, is an undesirable phenomenon.
However, ISI may become a friend if it is added to the transmitted signal in a
controlled manner.
Known fact: A signal of bandwidth B can be distortionlessly transmitted using its
samples with sampling rate  2B.
Conversely, in a channel with bandwidth B Hz, the theoretical maximum signal
rate is 2B symbols per second.
Why intentionally adding ISI? Answer: To have better bandwidth efficiency.
 Ideal Nyquist pulse shaping is efficient; it cannot be realized.
 Raised consine pulse shaping is realizable; it is bandwidth inefficient.
 By adding ISI to the transmitted symbols in a controlled manner, we can
achieve the Nyqusit rate 2B in a channel bandwidth of B Hertz.
1.4.1 Duobinary signaling (or class I partial response)

– p(nTb)=1, n=0,1
– p(nTb)=1, otherwise
Fig. 11 (a) The minimum bandwidth pulse that satisfies the duobinary pulse criterion and (b) its spectrum
Page 21 of 37
 Interpretation of received signal
2: 11
2: 00
0: 01 or 10 depends on the previous transmission
Fig. 12 Equivalent duobinary signaling
Binary input sequence {bk} : uncorrelated binary symbol 1, 0

1 if symbol bk is 1
ak  
1 if symbol bk is 0
ck  ak  ak 1
Note that ck has three levels (–2,0,2).
Rb 1

Ideal Nyquist channel of bandwidth BT  
2 2Tb
H I ( f )  H opt ( f )[1  exp( j 2 fTb )]
 H opt ( f )[exp( j fTb )  exp( j fTb )]exp( j fTb )
 2 H opt ( f ) cos( fTb ) exp( j fTb )
 R
1, | f | b
H opt ( f )   2
0, otherwise
2 cos( fTb ) exp(  j fTb ), | f | 1/ 2Tb

HI ( f )  
 0, otherwise
 H ( f )  1  exp( j 2 fTb )
sin( t / Tb ) sin[ (t  Tb ) / Tb ]
hI (t )  
 t / Tb  (t  Tb ) / Tb
Tb2 sin( t / Tb )

 t (Tb  t )
Let us ignore the effect of Hopt(f) first in the block diagram in the previous figure.
We directly obtain:
Page 22 of 37
Duobinary signal and Nyquist Criteria

 Nyquist second criteria: but twice the bandwidth
Conclusions
 By adding ISI to the transmitted signal in a controlled (and reversible)
manner, we can reduce the requirement of bandwidth of the transmitted
signal.
 Hence, in the previous example, {ck} can be transmitted in every Tb/2
seconds!
 Doubling the transmission capacity without introducing additional
requirement in bandwidth!
 A larger SNR is required to yield the same error rate because of an increase
in the number of signal levels (from –1, +1 to –2, 0, 2).
 The duobinary signaling is also named class I partial response.
 Full response: The transmission wave at each time instance is fully determined
by a single information symbol.
 Partial response: The transmission wave at each time instance is only
partially determined by one information symbol (i.e., is fully determined by two
or more information symbols).
Decision Feedback for Correlative-Level Coding
 Recovering of {ak} from {ck}

aˆk  ck  aˆk 1
 It requires the previous decision to determine the current symbol.
 So the system should feedback the previous decision.
 Error therefore may propagate!
 Decision feedback : technique of using a stored estimate of the previous
symbol
 Propagate : drawback, once error are made, they tend to propagate through
the output
 Precoding : practical means of avoiding the error propagation phenomenon
before the duobinary coding
How to avoid error propagation? Answer: differential coding (precoding)
Page 23 of 37
Differential Coding
d k  bk  d k 1
 symbol 1 if either symbol bk or d k 1 is 1
dk  
 symbol 0 otherwise
 {dk} is applied to a pulse-amplitude modulator, producing a corresponding
two-level sequence of short pulse {ak}, where +1 or –1 as before
ck  ak  ak 1
 0 if data symbol bk is 1
ck  
2 if data symbol bk is 0
1.4.2 Modified duobinary signaling
 In duobinary signaling, H(f) is nonzero at the origin.

 We can correct this deficiency by using the class IV partial response.
 The PSD of the signal is nonzero at the origin.
 Solution: Class IV partial response or modified duobinary technique.
 Precoding is added to eliminate error propagation in decision system.
 symbol 1 if either symbol bk or d k  2 is 1

d k  bk  d k  2 
 symbol 0 otherwise
ck  ak  ak 2
 |ck|=1 : random guess in favor of symbol 1 or 0
Page 24 of 37
If | ck | 1, say symbol bk is 1
If | ck | 1, say symbol bk is 0
 H ( f )  1  exp( j 4 fTb )
 [exp( j 2 fTb )  exp( j 2 fTb )]exp( j 2 fTb )
H IV ( f )  H Nyquist ( f )[1  exp( j 4 fTb )]
 2 jH Nyquist ( f )sin(2 fTb ) exp( j 2 fTb )
2 j sin(2 fTb ) exp(  j 2 fTb ), | f | 1/ 2Tb
H IV ( f )  
 0, elsewhere
sin( t / Tb ) sin[ (t  2Tb ) / Tb ]
hIV (t )  
 t / Tb  (t  2Tb ) / Tb
2Tb2 sin( t / Tb )

 t (2Tb  t )
SY ( f ) / Tb  sinc2 (2 fTb ), Duobinary


SY ( f ) /(4Tb )  sin (2 fTb )sinc ( fTb ), Modified Duobinary
2 2
 4 cos 2 ( fTb ) I

16 cos 4 ( fTb ) II
| G ( f ) |2 
 SY ( f )   4 cos ( fTb )  8sin (2 fTb ) III
2 2
Tb  4sin 2 (2 fTb ) IV


 16sin (2 fTb )
4
V
Page 25 of 37
1.2
0.8
0.6
0.4
0.2
1.5 Eye Diagram

0
-0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Eye diagram is a means of evaluating the quality of a received “digital waveform”

The eye diagram is created by taking the time domain signal and overlapping the
traces for a certain number of symbols. Eye diagrams reveal the impact of ISI and
noise.
From an experimental perspective, the eye pattern offers two compelling virtues
 The simplicity of generation
 The provision of a great deal of insightful information about the
characteristics of the data transmission system, hence its wide use as a
visual indicator of how well or poorly a data transmission system performs
the task of transporting a data sequence across a physical channel.
Fig. 13 The eye diagram
Fig. 14 4-Ary PAM signaling: (a) four RZ symbols; (b) baseband transmission;
(c) the 4-ary RZ eye diagram.
Page 26 of 37
Fig. 15 (a) Binary data sequence and its waveform. (b) Corresponding eye pattern.
Timing Features
Three timing features pertaining to binary data transmission system,
 Optimum sampling time : The width of the eye opening defines the time
interval over the distorted binary waveform appearing at the output of the
receive-filter
 Zero-crossing jitter : in the receive-filter output, there will always be
irregularities in the zero-crossings, which, give rise to jitter and therefore non-
optimum sampling times
 Timing sensitivity: This sensitivity is determined by the rate at which the eye
pattern is closed as the sampling time is varied.
The Peak Distortion for Intersymbol Interference

 In the absence of channel noise, the eye opening assumes two extreme values
 An eye opening of unity, which corresponds to zero intersymbol
interference
 An eye opening of zero, which corresponds to a completely closed eye
pattern; this second extreme case occurs when the effect of intersymbol
interference is severe enough for some upper traces in the eye pattern to
cross with its lower traces.
Fig.16 Interpretation of the eye pattern for a baseband binary data transmission system.
Page 27 of 37
 Noise margin
 In a noisy environment,
 The extent of eye opening at the optimum sampling time provides a
measure of the operating margin over additive channel noise
(Eye opening)  1  Dpeak
 Eye opening
 Plays an important role in assessing system performance
 Specifies the smallest possible noise margin
 Zero peak distortion , which occurs when the eye opening is unity
 Unity peak distortion, which occurs when the eye pattern is completely
closed.
Fig. 17 Illustrating the relationship between peak distortion and eye opening.
Note: The ideal signal level is scaled to lie inside the range 1 to 1.
 The open part of the signal represents the time that we can safely sample
the signal with fidelity
– By quality is meant the ability to correctly recover symbols and timing
– The received signal could be examined at the input to a digital receiver
or at some stage within the receiver before the decision stage
 Two major issues are 1) sample value variation, and 2) jitter and sensitivity
of sampling instant
 Eye diagram reveals issues of both
 Eye diagram can also give an estimate of achievable BER
Page 28 of 37
Interpretation of Eye Diagram
Vertical and Horizontal Eye Openings
 The vertical eye opening or noise margin is related to the SNR, and thus the
BER
– A large eye opening corresponds to a low BER
 The horizontal eye opening relates the jitter and the sensitivity of the
sampling instant to jitter
– The red brace indicates the range of sample instants with good eye
opening
– At other sample instants, the eye opening is greatly reduced, as
governed by the indicated slope
Page 29 of 37
As shown in the next Figure monitoring of an eye pattern can provide a
qualitative measure of performance regarding the
signal quality, including the following important observations:
1. The width of the eye opening represents the time interval during which the
received signal can be sampled without error from ISI.
2. The best time to sample the received signal is when the eye is open the
widest. When there is no ISI, we have an eye opening of unity, and when
there is a significant amount of ISI, we have an eye opening of zero (i.e.,
the eye is completely closed). With an eye opening of 50% or better (i.e.,
with a signal-to-noise ratio of 6 dB or more), reliable data transmission can
be achieved.
3. The maximum distortion is indicated by the height of the eye opening at
sampling time and it is twice the peak distortion.
4. The noise margin or immunity to noise is defined by the height of the eye
opening at the sampling time.
5. The sensitivity to timing errors is detected by the rate of closure of the eye
as sampling time is varied.
6. Zero (level) crossings can provide clock information, and the amount of
distortion of zero crossings indicates the amount of jitter.
7. The variation of level crossing can be seen from the width of the eye
corners.
8. In a linear system with truly random data, all the eye openings would be
identical.
9. Asymmetries in the eye opening generally indicate nonlinearities in the
transmission channel.
10. When the effect of ISI is quite severe, traces from the upper portion of
the eye pattern cross traces from the lower portion, resulting in the eye
being completely closed.
11. In an M-ary system (as discussed later), the eye pattern contains (M –
1)eye openings stacked up vertically one on the other.
Page 30 of 37
1.6 Equalization
Zero-Forcing Equalization
 Delay line, whose taps are uniformly spaced T second apart; T is the symbol
duration
 Adjustable weights, which are connected to the taps of the delay line
 Summer, which adds successively, delayed versions of the input signal, after
they have been individually weighted.
 Adjustable transversal equalizer (transversal equalizer)
 With channel equalization as the function of interest and the transversal
filter with adjustable coefficients as the structure to perform.
Fig. 18 Transversal filter.
Since the zero-forcing equalizer ignores the effect of additive channel noise, the
equalized system does not always offer the best solution to the intersymbol
interference problem
Page 31 of 37
1.7 Optimum Detection of Binary Signal in Gaussian Noise
Additive White Gaussian Noise (AWGN):

 Thermal noise is described by a zero-mean Gaussian random process, n(t) that
ADDS on to the signal => “additive”
 Its PSD is flat; hence, it is called white noise.
 Autocorrelation is a spike at 0: uncorrelated at any non-zero lag
Probability density function (Gaussian distribution)
• Channel is dispersive
– channel is noisy – control over additive white noise (old problem)
– received signal pulses are affected by adjacent symbols (new problem)
– intersymbol interference (ISI); major source of interference;
– Distorted pulse shape (new problem) - channel requires control over
pulse shape
– Detection of digital pulses corrupted by the effect of the channel
Page 32 of 37
• We know the shape of the transmitted pulse
– Device to be used – matched filter
Matched Filter
• detecting transmitted pulses at the front end of the receiver (corrupted by
noise)
• Receiver model
The filter input x(t) is: x(t )  g t   n(t ), 0 t T

• where
– T is an arbitrary observation interval
– g(t) is a binary symbol 1 or 0
– n(t) is a sample function of white noise, zero mean, psd N0/2
• The function of the receiver is to detect the pulse g  t  in an optimum
manner, providing that the shape of the pulse is known and the distortion is
due to effects of noise =
• To optimize the design of a filter so as to minimize the effects of noise at the
filter output in some statistical sense.
Designing the filter

Since we assume the filter is linear its output can be described as y(t )  go (t )  no (t )
• where
– g o (t ) is the recovered signal
– no (t ) produced noise
• This is equal to maximizing the peak signal-to-noise ratio:
| g (T ) |2 instantaneous power
  SNRout  o 2 
E[n0 (t )] average noise power
• So, we have to define the impulse response of the filter h(t) in such a way
that the signal-to-noise ratio is maximized.
Let us assume that:
- G(f) - FT of the signal g(t);
- H(f) – frequency response of the filter
then: FT of the output signal g0(t)= H(f).G(f), or

go (t )   H ( f )G( f )exp( j 2 ft )df

Page 33 of 37
sampled at time t=T and no noise
 2
| go (T ) |2   
H ( f )G( f ) exp( j 2 ft )df
Next step is to add the noise.
What we know is that the power spectral density of white noise is:
N
S N ( f )  0 | H ( f ) |2
2
Thus the average power of the output noise n(t) is:
 N 
E[n 2 (t )]   S N ( f )df  0  | H ( f ) |2 df
 2 
Substituting,
 2

H ( f )G ( f ) exp( j 2 ft )df

N0 
2 
| H ( f ) |2 df
So, given the function G(f), the problem is reduced to finding such an H(f) that
would maximize η.
We use Schwartz inequality which states that for two complex functions, satisfying
the conditions
 

| 2 ( x) |2 dx   
| 1 ( x) |2 dx 
 2  
the following is true:  
1 ( x)2 ( x)dx   | 1 ( x) |2 dx  | 2 ( x) |2 dx
 
and equality holds iff: 1 ( x)  k ( x) *

2
In our case this inequality will have the form:

 2  
 
H ( f )G( f ) exp( j 2 fT )df   | H ( f ) |2 df
 
| G( f ) |2df
and we can re-write the equation for the peak signal-to-noise ratio as:
2 
 
N0 
| G ( f ) |2 df
Remarks:
• The right hand side of this equation does not depend on H(f).
• It depends only on:
– signal energy
– noise power spectral density
2 
N 0 
• Max value is for:   | G ( f ) |2 df
Let us denote the optimum value of H(f) by Hopt(f).

Hopt ( f )  kG* ( f )exp( j 2 fT )
The result: Except for a scaling coefficient k exp( j 2πfT ) , the frequency response of
the optimum filter is the same as the complex conjugate of the FT of the input
signal.
Page 34 of 37
Definition of the filter functions:
In the frequency domain:
• knowing the input signal we can define the frequency response of the filter
(in the frequency domain) as the FT of its complex conjugate.
In the time domain

Take inverse FT of Hopt(f): hopt (t )  k  G* ( f )exp[ j 2 f (T  t )]df

and keeping in mind that for real signals G* ( f ) = G( f ) :


hopt (t )  k  G ( f ) exp[ j 2 f (T  t )]df


 k  G ( f ) exp[ j 2 f (T  t )]df  kg (T  t )

So,
In the time domain it turns out that the impulse response of the filter,
except for a scaling factor k, is a time-reversed and delayed function of the
input signal.
This means it is “matched” to the input signal, that is why this type of time-
invariant linear filters is known as “matched filter”
NOTE: The only assumption for the channel noise was that it is stationary, white,
with PSD N0/2.
Properties of Matched Filters

Property 1:
• A filter matched to a pulse signal g(t) of duration T is characterized by an
impulse response that is time-reversed and delayed version of the input
g(t):
 k g (T  t ) 0  t  T
Time domain: hopt (t )  kg * (T  t ) or h(t )  
0 else where
Frequency Domain: Hopt ( f )  kG ( f )exp( j 2πfT )
Property 2:
• A matched filter is uniquely defined by the waveform of the pulse but for
the:
- time delay T
- scaling factor k
Property 3:
• The peak signal-to-noise ratio of the matched filter depends only on the ratio
of the signal energy to the power spectral density of the white noise at the
filter input.
Go ( f )  Hopt ( f )G( f )  kG* ( f )G( f )exp( j 2 fT )  k | G( f ) |2 exp( j 2 fT )
using the inverse FT
 
g o (T )   G0 ( f ) exp( j 2 fT )df  k  | G( f ) | df
2
 
• The integral of the squared magnitude spectrum of a pulse signal with

respect to frequency is equal to the signal energy E (Rayleigh Theorem) so
substituting in the previous formula we get:
Page 35 of 37
 
E g (t )dt   G ( f ) |2 df so g 0 (T )  k E
2
 
 N0 
for the average output noise power E[n 2 (t )]   S N ( f )df   | H ( f ) |2 df
 2 
we finally get the following expression:

k 2 N0 
2 
E[n2 (t )]  | G( f ) |2 df  k 2 N0 E / 2
(kE ) 2 2E
and  max  SNRmax  2 
(k N 0 E / 2) N 0
Conclusion:
• We see that the dependence of the peak SNR on the input waveform g(t)
has been completely removed by the matched filter.
• So, in evaluating the ability of a matched-filter receiver to overcome/remove
additive white noise we see that all signals with equal energy are equally
effective.
• We call the ratio E/N0 signal-energy-to-noise ratio (dimensionless)
• The matched filter is the optimum detector of a pulse of known shape in
additive white noise.
Correlation realization of Matched filter

• Theorem: For the case of white noise, the matched filter can be realized by
correlating the input with s(t) where r(t) is the received signal and s(t) is the
known signal wave shape.
ro  to   
to
r (t ) s(t )dt
to T
• Correlation is often used as a matched filter for Band pass signals.
Matched filter realization by correlation processing

t
 r( ) h(t  ) d
y(t )  r (t )  h(t ) 
0
( h(t )  g (T  t ) )
t t
  r ( ) g[ T  (t  ) ] d   r ( ) g ( T  t   ) d
0 0
T
When t=T y(T )   r ( ) g ( ) d
0
Cross – correlation of r (t ) with g (t ) . (Observation) Integration – and – Dump

Filter.
Note: The correlator output and the matched filter output are the same
only at time t = T .
Page 36 of 37
Matched filter Example
h(t )  g (T  t ) y(t )  r (t )  h(t )

2
E 2A T
SNRmax  
N0 / 2 N0
Define the matched filter (MF) concept in an AWGN channel. Give the solution
in both time-domain and frequency-domain forms.
y (t )  si (t )  h opt (t )
si (t ) hopt (t )
A2
A A
T T
T/2 T t T/2 0 T/2 T 3T/2

T t 2T
2 2
 A2
2
A A
T 2 T 2
Page 37 of 37

Chapter 1

Uploaded by

Copyright:

Available Formats

Chapter 1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 1

Uploaded by

Copyright:

Available Formats

Chapter 1 EE 322-Digital Communications

Disadvantages of digital communications:

Fundamental elements of digital communication systems

Fig. 1 Block diagram of basic digital Communication System

Various unwanted undesirable effect crop up in transmissions

Baseband Digital Data Transmission Systems

Fig. 2 Block diagram of a baseband digital data transmission system.

1.2 Line Codes:

Properties of a good line code:

1.2.1 Polar line code

1) not bandwidth efficient.

This code has memory and is most used in telephone systems.

High-Density Bipolar (HDB) line code:

Fig. 6 (a) HDB3 signal and (b) its PSD.

A zero is represented by no transition and one by alternate transitions according to

NRZ-I (Non Return to Zero Inverted)

1.2.5 M-Ary Coding (Signaling)

• Trade-off is relatively high bit error rate (BER)

 ISI arises when the channel is dispersive

Three strategies for eliminating ISI:

1.3.1 Nyquist’s First Method for Zero ISI: Time domain

Suppose 1/T is the sample rate

 Since pulses are not possible to create due to:

Raised Cosine-Rolloff Nyquist Filtering

For special case of   1

Two additional Properties of the Raised-Cosine Pulse Spectrum

 Raised Cosine Filter is also called a NYQUIST FILTER.

1.3.2 Nyquist’s Second Criterion for no ISI

 Values at the pulse edge are distortionless

Fig. 10 Communication using controlled ISI or Nyquist second criterion pulses.

1.4 Correlative-Level Coding or Partial-response signaling

1.4.1 Duobinary signaling (or class I partial response)

Fig. 12 Equivalent duobinary signaling

Binary input sequence {bk} : uncorrelated binary symbol 1, 0

2 cos( fTb ) exp(  j fTb ), | f | 1/ 2Tb

Duobinary signal and Nyquist Criteria

Decision Feedback for Correlative-Level Coding

 Recovering of {ak} from {ck}

How to avoid error propagation? Answer: differential coding (precoding)

1.4.2 Modified duobinary signaling

 In duobinary signaling, H(f) is nonzero at the origin.

 Precoding is added to eliminate error propagation in decision system.

 symbol 1 if either symbol bk or d k  2 is 1

SY ( f ) / Tb  sinc2 (2 fTb ), Duobinary

Tb  4sin 2 (2 fTb ) IV

1.5 Eye Diagram

Eye diagram is a means of evaluating the quality of a received “digital waveform”

Fig. 13 The eye diagram

The Peak Distortion for Intersymbol Interference

Vertical and Horizontal Eye Openings

Fig. 18 Transversal filter.

Additive White Gaussian Noise (AWGN):

– Detection of digital pulses corrupted by the effect of the channel

The filter input x(t) is: x(t )  g t   n(t ), 0 t T

Designing the filter

and equality holds iff: 1 ( x)  k ( x) *

In our case this inequality will have the form:

Let us denote the optimum value of H(f) by Hopt(f).

and keeping in mind that for real signals G* ( f ) = G( f ) :

Properties of Matched Filters