Chap 7 Principles of Digital Data Transmission

․ TX digital data over a channel.

‥ Considering binary case (1, 0) with distinct waveforms (pulses).
․ At RX, two pulse are detected and converted to binary data (1, 0).

7.1 Digital Communication Systems

․ Fundamental building blocks of digital communication systems.

7.1.1 Source
․ Input to a digital system: a sequence of digits.
․ Input = output (a data set).
․ Binary communication, M-ary communication.

7.1.2 Line Coder

․ Digital output of source encoder is converted to pulse (waveforms) over
transmission channel, which is “line coding“ or “transmission coding“ .

1
․ Line code example：

‥ On-off ( RZ )： 1  p  t  , 0  0 .

‥ Polar ( RZ )： 1  p  t  , 0   p  t  (most power efficient , better noise

immunity ) .

‥ Bipolar ( RZ ) ( alternate mark inversion , AMI )： 0  0 , previous 1   p  t  

current 1   p  t 

( Perr , multiple bit << Perr , single bit )

‥ Duobinary (modified duobinary) line code is better than bipolar in BW
efficiency. (Used in Hard disk drive)
‥ On-off NRZ ( non-return-to-zero )： 1  1 , 0  0 .
‥ Polar NRZ： 1  1 , 0  1 .

7.1.3 Multiplexer
․ Increasing capacity of physical channel.
‥ Using digital multiplexer, e.g., TDMA, FDMA, CDMA, to combine several

2
 sources. ( time division , frequency division , code division multiple access )
‥ A physical channel is shared by several source messages simultaneously.

7.1.4 Regenerative Repeater

․ Regenerative Repeater in digital TX line：
‥ Detect the incoming digital signal and regenerate “new clean“ pulse for next
line transmission.
‥ Eliminate noise and combat signal distortion (digital comm. advantage )

․ At repeater, “clock signal“ can sample the incoming pulse.

‥ TX Rb pulse rate, it can be acquired the periodic clock timing information.

‥ RZ polar signal can contain clock frequency Rb Hz. (periodic timing)

(transparent line code)

‥ On-off signal (a)  polar signal (b) + periodic clock signal (c)  clock freq.
can be extracted. (Resonant circuit tuned to the clock frequency)

‥ Bipolar signal, after rectified (abs), get “on-off “ signal.  get clock
frequency.
․ Transparent line code：polar signal . (A line code in which the bit pattern does not

3
 affect the accuracy of the timing information is said to be a transparent line code.)
․ Nontransparent line code：on-off and bipolar signal . (affect the timing information,
e.g. , all 0 , all 1 )
7.2 Line Coding
․ Line code should have the properties.
‥ TX BW：small.

‥ Power efficiency：given BW and specified Perr , small TX power .

‥ Error detection and correction capability ： desirable detection, preferably

correct . ( bipolar , single error detected easily )
‥ Favorable power spectral density：” 0 “ PSD at f  0 . (No DC) , AC
coupling used at repeaters .
‥ Adequate timing content：possible to extract timing clock from signal.
‥ Transparency：correctly TX a digital signal regardless pattern 1 or 0. (long “0”,
on-off signal failed timing extract)

7.2.1 PSD of Various Line Codes

․ Recall , relation of autocorrelation and PSD：( Chap 3.8.2 )
T /2
1
autocorrelation： Rg    lim g  t g  t    dt
T  T 
T /2

1
PSD ( FT of Rg   )： S g  f   FT Rg    lim f
2
GT
T  T

After LTI h( t ) system , get PSD (output)： S y  f   H  f  S g  f 

2

․ Consider the line code baseband signal with generic pulse p  t  train：

y  t    ak p  t  kTb 

where ak : line coder amplitude , kth symbol pulse p  t  kTb 

y  t  : PAM signal

4
․ Using convolution , y  t  is rewritten by p  t  and line code impulses：

y  t   p  t    ak   t  kTb 

x t 

․ Get PSD of y  t  depending on ak and p  t  ：

S y  f   FT  Ry   , Ry   : y  t  autocorrelation

= P  f   Sx  f  , S x  f   FT  Rx   , Rx   : x  t  autocorrelation
2

P  f  =FT  p  t 

5
․ In the impulse train case    0 , pulse width   , get Rx   , S x  f  , S y  f 


1
Rx   
Tb
 R    nT  , R
n 
n b n : ak , ak  n autocorrelation

S x  f   FT of Rx  

1
=
Tb
Re
n 
n
 jn 2 fTb
, R n  Rn  even function of  

1  


Tb
R
 0

 2 
n 1
Rn cos n2 fTb 

S y  f   P  f  Sx  f 
2

P f   
2

  
Tb  n 
Rn e jn 2 fTb 

P f  
2



Tb   R0  2 
n 1
Rn cos n 2 fTb 


 relative Rn and P  f  for different line codes .

* Note：study Rx   ：

․ Assume rectangular pulse train x̂  t  with  small pulse.

6
 ak
hk 


7
‥ Get Rxˆ   :    case 

T /2
1
Rxˆ    lim xˆ  t  xˆ  t   dt
T  T 
T / 2


1
= lim
T 

T k
h 2 k      (有效面積)

1    
= lim
T 

T k
a2k  2 
  
R0   
= 1   ,    , even function
 Tb   
R 1
 0    ,   0 ,    1
Tb 

where
Tb
R0  lim
T  T
a
k
2
k (有效面積)

 T  NTb
1
= lim
N  N
a
k
2
k

‥ At   Tb , get R1 ( kth pulse and ( k+1 )th pulse overlapping )：

1
R1  lim
T  N
k ak ak 1
R
Rxˆ    1    Tb  ,   0
Tb

‥ At   nTb , get Rn

1
Rn  lim
N  N
k ak ak n
R
Rxˆ    n    Tb 
Tb

1
 All  , Rx   
Tb
 R    nT 
n 
n b

8
 7.2.2 Polar Signaling
․ Get the PSD of polar signaling：

P f 
2

Sy  f    R0
Tb
P f 
2

=
Tb
Tb   fTb 
 sinc 2  
4  2 

where polar signaling： 1  p  t  , 0   p  t  , 1& 0 equal probability  R1  0

1 1
R0  lim
N 

N k
a 2 k  lim N  1
N  N
1 1 N N 
R1  lim
N 

N k
ak ak 1  lim  1   1   0
N  N2 2 

N N 
 ak ak 1 =1 , ak ak 1 = -1
2 2 

Rn =0 , n 1

 t   2t 
p t        
 Tb / 2   Tb 
T   fTb 
P  f   b sinc  
2  2 

9
․ Polar signaling properties：

‥ For a given power, Pe is the lowest among all signalings.

‥ It is transparent. ( always  pulse )

‥ Yielding periodic clock signal. ( using rectification (abs) of RZ polar signal )
‥ Polar signaling is not the most BW efficient.

) first non  dc null frequency

Essential BW ( EBW ：
EBW：2 Rb  4 times Nyquist theoretical BW ( Rb / 2) 
Full  width pulse , EBW： Rb

‥ Polar signaling has no capability for error detection or correction.

‥ Polar signaling with DC in PSD. ( Disadvantage )

10
 7.2.3 Constructing a DC Null in PSD by Pulse Shaping

․ Line code PSD S y  f  contains P  f  . If P  f  f 0  0 , S y  f 

2
f 0 0


P f    p t   e
 j 2 ft
dt


P 0   p  t dt  area of p  t  


․ Split phase signal ( Manchester or twinned-binary signal )

 P  0   0 , Null DC

7.2.4 On-Off Signaling

․ Get R0 and Rn of on-off signaling：

‥ 1  p  t  , 0  0 , equally probability.

1 N 2 N
1   0  
2 1
R0  lim 
N  N2 2  2
N
‥ For ak ak  n  11 , 1 0 , 0 1 , 0  0 ,  probability
4
1 N
1   0  
3N 1
Rn  lim   n  1
N  N4 4  4

11
․ Get S x  f  and S y  f  ：

 

1  j 2 fTb 
Sx  f   R0   Rn e
Tb  n 

 n0 
1 1   j 2 fTb
= 
2Tb 4Tb n  
e
n0

1 1
= 
4Tb 4Tb
e
n 
 j 2 fTb

1  n
 e  j 2 fTb

Tb
  f  T   proof F .S .
 b 

1  1 
 n 
=
4Tb
1     f T 
 Tb n   b 

Tb   fTb   1 
 n 
Sy  f  
16
Sin c 2   1 
 2   Tb
   f T 
n   b 

( Similar to polar signaling )

12
․ On-off signaling properties：

‥ on-off   polar  periodic clock signal  1 / 2

1
 on-off PSD = polar PSD
4
 on-off Energy 4E , polar Energy 2E ( for same amplitude ( 2 , 0 ) on-off 
( 1 , -1 ) polar)

on-off power 4E / Tb , polar power 2E / Tb

‥ on-off ( 1 , 0 ) less noise immunity to polar ( 1 , -1 )

‥ Not transparent
‥ No timing extraction
‥ BW efficiency  polar
no error correction detection  polar
with DC PSD  polar

7.2.5 Bipolar Signaling

․ Bipolar signaling is used for telephone network PCM

0  no pulse , 1  p  t  or  p  t  depending on the previous 1,  p  t  or p  t 

 Not 2  level , reality 3  level  p t  , 0 ,  p t 

․ Calculate PSD：

‥ Get R0 ： ak 2  
half ak = 0 , remaining half ak =1 or -1 .  ak 2 =1

1 N
 1   0   
2 N 2 1
 R0  lim 
N  N2 2  2

‥ Get R1 ：  ak ak 1 

four equally like 2 bits：11 , 10 , 01 , 00  ak ak 1 

 ak ak 1  1 , 0 , 0 , 0

1 N
 -1   0  -
3N 1
 R1  lim 
N  N4 4  4

13
‥ Get R2 ：  ak ak  2 

all possible combinations of 3 bits：111, 101, 110, 100, 011, 010, 001, 000

111  ak ak  2 =1

101  ak ak 2 =-1

others  ak ak 2 =0

N
 0  0
1 N 6N
 R2  lim  8 1   -1 +
N  N 8 8 

‥ Get Rn ：  ak ak  n 

equally number of product ak ak n = +1 , -1 , 0

 Rn  0 , n  2

‥ Get PSD：

P f 
2

Sy  f   1  cos 2 fTb 
2Tb
P f 
2

 sin 2  fTb   f  0 (dc), S y  0   0 , dc null

Tb
1  1
first non  dc null BW   Rb  f  
Tb  Tb 
For half-width pulse：
Tb   fTb  2
Sy  f   sinc2   sin  fTb 
4  2 

1
 Essential BW = Rb ,    polar  2 Rb 
 Tb 

(Twice theoretical min. BW)

 For bipolar case, BW is Rb Hz (Not depending on half-width or full-width

pulse due to sin  fTb  f  1  0 )

Tb

14
․ Advantages of Bipolar Signaling：
(1) PSD with dc null
(2) BW is not excessive
(3) With single-error-detection capability

․ Disadvantages of Bipolar Signaling：

(1) Bipolar with rectified  on-off signal  clock period signal
 twice as much power (3dB) as a polar signal
(2) Not transparent
(3) Long “zeros“ causes clock extracting failed

15
 7.3 Pulse Shaping
․ PSD S y  f  is controlled by line code S x  f  and pulse shape P  f 
2

‥ Pulse shape is a more strong influence to S y  f  .

7.3.1 Intersymbol Interference ( ISI ) and Effect

․ Time-limited (TL) signal, e.g., rectangular pulse  p  t    Not Band-limited
FT

(No BL) P  f  after

BL Channel
 Distortion P̂  f  ( errors )

․ Band-limited (BL) signal, e.g., rectangular BW P  f  

IFT
Not Time-limited

(No TL) signal p  t  , cause ISI after

BL Channel
 No Distortion P  f 

‥ Intersymbol interference (ISI)： p  t  ( No TL ) is spreading a pulse beyond

interval Tb , and causing to interfere with neighboring pulses .

‥ To resolve ISI, pulse amplitudes ak can be detected correctly in the pulse

overlapping.
 Properly shaped BL signal 
IFT
Not TL signal, no ISI at the decision-
making instants ( zero interference to all
other pulse at decision instants ) ( Nyquist
criterion )

7.3.2 Nyquist’s First Criterion for Zero ISI

․ Nyquist’s criterion：zero ISI at t   nTb  n  1 , 2 , ......

1 , t  0

p t     1 
0 , t   nTb  Tb  
Rb 
 

16
․ TX Rb bps rate with the theoretical min. BW Rb / 2 Hz. (satisfying Nyquist’s

criterion) What ‘ s pulse？

1 , t  0
p  t   sinc  Rbt   
0 , t=  nTb
 ( No ISI )
1  f 
P f     
Rb  Rb 

․ The sinc pulse is impractical：

‥ Start pulse at   , after truncation , BW > Rb / 2

‥ Undesirable feature , i.e. , decaying too slowly at rate 1 / t

 sampling instants with offset , “ rate 1 / t “ causes the summation of all
1
remaining pulses interference being very large . ( n )

․ Another Nyquist pulse with a BW kR0 / 2 1  k  2  and p  t  decaying faster.

‥ The desired pulse p  t   P ( f ) ：BW in  Rb / 2 , Rb  , with zero ISI

The sample signal p  t  ：

p  t   p  t   Tb  t     t 

17
  1
FT of p  t  ：(Repeated P  f  )  Rb  
 Tb 

1
Tb
 P  f  nR   1
n 
b

 P  f  nR   T
n 
b b  sum spectrum  constant 

‥ Considering 0  f  Rb , sum of P  f  and P  f  Rb  ：

P  f   P  f  Rb   Tb , 0  f  Rb

‥ Let x  f  Rb / 2 , get

P  x  Rb / 2   P  x  Rb / 2   Tb , x  0.5 Rb
 Conjugate symm.
P  Rb / 2  x   P  Rb / 2  x   Tb , x  0.5Rb

18
  at f  Rb / 2 , odd symm.
P  0.5 Rb   0.5 P  0 

Rb
․ BW of P  f  ：0.5 Rb + f x , f x ：excess BW of BW , r：ratio BW
2
(roll-off factor)
excess bandwidth
r
theoretical minimum bandwidth
fx
=
0.5  Rb
=2 f xTb

0  r 1  f x  Rb / 2 

․ Total P  f  BW：

Rb rRb R
BT    1  r  b
2 2 2

< Raised-Cosine Shaping filter P  f  >

․ P f  is a Nyquist first criterion spectrum with excess BW ( r ) . (0<r<1)

19
‥ For r = 1 , full-cosine roll-off signal ,

p  t   IFT  P  f  
cos  Rbt
 Rb sinc  Rbt 
1  4 R 2b t 2 raised cosine

T
 No ISI at t  nTb , No ISI at t  nTb  b ( r = 1 case )
2
It decays rapidly as 1/ t 3 .
(Insensitive to sampling, jitter variation)

20
 * Note：Considering channel effect H c  f  , TX pulse pi  t  can be preprocessing：

Pi  f  Hc  f  = P  f   get new Pi  f  overcome channel effect.

EX 7.1 ：Given BT & r, find pulse TX rate？ (satisfying Nyquist‘s first criterion)

2  1
 Rb  BT  decay faster as 3 
1 r  t 
 1 R 
 If r  1 , BT  Rb  satisfying Nyquist criterion , decary 3 , but BW  b 
 t 2 

7.3.3 Controlled ISI or Partial Response Signaling

․ Nyquist second criterion：

‥ Pulse broadening in time domain  2Tb pulse width   reducing pulse

 R 
bandwidth.  BW  b 
 2 
‥ A known and controlled ISI with compensating few interference patterns ( 2
pulse ) “ Correlative or partial-response scheme “  duobinary pulse

 1 , n  0,1
p  nTb   
0 , for all other n

 Pulse causes zero ISI with all pulses except the succeeding pulse.

 Consider two successive pulses at 0 and Tb .

21
 both pulses “ + “  sample value = 2 at Tb  sample “ + “ , current

1, previous 1.

both pulses “ – “  sample value = -2 at Tb  sample “ – “ , current

0, previous 0.

both pulses “ 0 “  sample value = 0 at Tb  sample “ 0 “ , current

is opposite to previous. (knowledge the previous bit)


 No noise 
 2 
even " 0 "
2
   If one sample is detected wrong , 
 -2 
odd " 0 "
 2  
 2    the rule is violated. Error is detect ed ！

odd " 0 "
 -2 
Rb
‥ Satisfying zero ISI and BW=
2

7.3.4 Example of a Duobinary Pulse

․ Duobinary pulse is satisfying pulse BW to Rb / 2 and Nyquist 2nd criterion.


p  t   sinc   Rb t   sinc  Rb  t  Tb   
sin  Rb t 

 Rbt 1  Rbt 

P  f   FT of p  t   Hint：1+e  j 2
 e  j  e j  e j  
2   f   f   j f / Rb
 cos     e
Rb  b   Rb 
R

22
Binary data rate = Rb bps

min. BW (theoretical) = Rb / 2 Hz

pulse decaying fast 1/ t 2

p  t  noncausal, infinite duration  1/ t decay can approaching .

2

23
 7.3.5 Pulse Relationship between Zero-ISI, Duobinary, and
Modified Duobinary
․ Consider three pulses
Rb
‥ Zero ISI pulse , pa  t  ：satisfying Nyquist 1st rule , BW： ~ Rb , PSD with
2
DC, e.g., raised cosine pulse.
Rb
‥ Duobinary pulse pb  t ：satisfying Nyquist 2nd rule , BW= , PSD with DC.
2
Rb
‥ Modified Duobinary pulse pc  t ：satisfying Nyquist 2nd rule , BW= , PSD
2
without DC.

․ Duobinary pulse pb  t  and zero ISI pulse pa  t  ：

pb  t   pa  t   pa  t  Tb 
Pb  f   Pa  f   1  e j 2 fTb 

Pb  t   Pa  f  2 1  cos  2 fTb   2 cos  fTb   f  0.5 / Tb , frequency null , DC  0

․ Modified duobinary pulse： ( DC = 0 )

 1 , n  1

pc  nTb    1 , n = 1
0 , o.w. n


․ pc  t  can be generated by pa  t  satisfying 1st rule：

pc  t   pa  t  Tb   pa  t  Tb 
Pc  f   FT of pc  t 
 2 j Pa  f  sin  2 fTb   DC null

24
 7.3.6 Detection of Duobinary Signaling and Differential Encoding
< Basic Duobinary Signaling TX/RX >
․ For the controlled ISI method of duobinary signaling：TX block

‥ Binary message bits I k  0,1 , polar signals：

ak  2I k 1

‥ Under controlled ISI, samples of TX signal y  t  ：

y  kTb   bk  ak  ak 1

‥ For RX, Detect I k from y  kTb  or bk . Consider all values bk .

bk  2  ak  1 , I k  1
bk  2  ak  1 , I k  0
bk  0  ak   ak 1 , I k  1  I k 1  ak  2 I k  1

(Dependent on previous bit. If I k 1 is error, I k will be error)

(error propagation problem)

< Differential Encoded Duobinary Signaling, No error propagation>

․ TX block：Duobinary generator with differential encoder.

25
 ‥ Differential encoding generates a new binary sequence

pk  I k  pk 1 , mod-2, initial state p0  0 or p0  1

‥ Polar line coder generates

ak  2 pk  1

‥ After duobinary signaling bk  ak  ak 1 zero-ISI pulse generator, the samples

of the received signal y  t  ：

y  kTb   bb  ak  ak 1
 2  pk  pk 1   2
 2  pk 1  I k  pk 1  1
2 1  I k  , pk 1  1, 1 Ik  1  Ik

 2  I k  1 , pk 1  0, 0  Ik  Ik
 summarize
 0 , Ik  1
y  kTb   
2 , I k  0

‥ Decision algorithm is based on the current sample y  kTb  , get I k ：

2  y  kTb 
Ik 
2


 No propagation error 
* Note： I k can be M-ary signal , mod-2 
changed
mod-M

26
 7.3.7 Pulse Generation
․ Using Tapped Delay Line (TDL) filter to generate p  t  pulse with the sufficiently

small sampling interval Ts . Filter tap gains  P  kTs  sample values.

‥ Narrow rectangular pulse with width Ts , I / P TDL filter .

O / P TDL filter passed through LPF (smoothed out).

( Ts  Improvement approaching continuous p  t  )

‥ TX pulse shape  channel  RX pulse shape + Equalizer

27
 7.4 Scrambling
․ Scrambler
‥ Make data more random and remove long strings of 1s or 0s.
‥ Helpful in timing extraction due to removing long strings of 0s.

Fig 7.20

‥ Scrambler consists of “feedback shift register“ (Scrambler I :S )

P

(Scrambler O :T )
P
‥ Descrambler has “feedforward shift register“
‥ After scrambler output , get (D：Delay)
S  D3T  D5T  T (Fig. 7.20 (a))

 Adding D T  D T 
3 5
to both sides, get  D T  D T  0
n n

S  T   D3T  D5T   1  F  T , F  D3  D5

‥ After Descrambler output , get

R  T  D3T  D5T  S (regenerate input sequence S)

* Note：one single detection error in T, affect three output bits in R

(disadvantage of causing multiple errors)

28
 EX 7.2 Data stream 101010100000111 is fed to scrambler, find the scrambler output
(assume initial registers  0)

Sol：○
1 initial T = S

1st feedback D 3
 D5  S  FS  S  FS  O / P 

2nd feedback D 3
 D5  D3  D5  S  F 2 S  S  FS  F 2 S  O / P 


․ Output：
T  S  FS  F 2 S  F 3 S  F 4 S 
where

F  D 3  D5
F 2   D 3  D 5  D 3  D5   D 6  D10 , D 8
 D8  0 
F 3   D 6  D10  D3  D 5   D9  D11  D13  D15
F 4   D 9  D11  D13  D15  D 3  D5 

12
 D14  D14  0 , D16  D16  0 , D18  D18  0 ,
=D  20 
 D  shift all "0" 15 values  
․ Get output：

T  1  D3  D5  D6  D9  D10  D11  D12  D13  D15  .... S

S  101010100000111
D S  000101010100000111
3

D 5 S  00000101010100000111
D 6 S  000000101010100000111
D 9 S  000000000101010100000111
D11S  00000000000101010100000111
D12 S  000000000000101010100000111
D13 S  0000000000000101010100000111
D15 S  000000000000000101010100000111
T  101110001101001

29
 7.5 Digital Receivers and Regenerative Repeaters
․ Receiver or regenerative repeater performs three functions：
(1) Reshaping incoming pulse via equalizer.
(2) Extracting the timing information to sample incoming pulses at optimum instants.
(3) Making symbol detection decisions based on the pulse samples.

(Separation of dc power from ac signals via transformer-coupling signal or

blocking DC.)

7.5.1. Equalizers
․ Pulse train is attenuated and distorted by transmission medium.
‥ Attenuation  Compensated by preamplifier
‥ Distortion  Compensated (inversion) by equalizer, reduce ISI.
 
channel frequency boosting amplication
attenuated noise
․ For simple modulation, complete equalization is not necessary.
‥ Pulse dispersion can be tolerated.
‥ Pulse is positive or negative. ( > 0 or < 0 )
 Simple equalization with reducing noise amplication and reducing ISI.

< Zero-Forcing Equalizer >

․ Eliminate or minimize ISI (interference) .
‥ Not for all t pulses ( neighboring )
‥ Only for sampling instant pulses ( neighboring )
 RX decision is based on sample values only
‥ Equalizer is designed by FIR filter.

30
 ‥ FIR filter design goal：equalize output pulse with zero ISI at sampling instants
 Satisfying Nyquist or controlled ISI criterion.

‥ I : Pr  t  with non-zero pulse amplitudes at Tb , 2Tb .....

P

‥ If C0  1, Ck  0, O : P0  t   Pr  t  NTb  ( delay NTb )

P

‥ Adjusting taps Ck , get Nyquist pulse O/P.

․ Derive the solution of Ck to get output pulse (No ISI).

‥ The output P0  t  : ( ignoring delay NTb )

N
P0  t    C P  t  nT 
n  N
n r b

‥ Sample at t  kTb , Pr  kTb   Pr  k  , P0  kTb   P0  k  , get

N
P0  kTb    C P  kT
n  N
n r b  nTb  , k  0, 1, 2.....

31
 N
P0  k    C P k  n
n  N
n r

 Satisfying zero ISI.

1, k  0 
P0  k     2N 1
0, k  1, 2,.....  N 

 Get 2 N  1 equations and matrix form ：

where Pr : Toeplitz matrix  2 N  1   2 N  1

‥ Taking inverse of matrix Pr , get solution of C：

C  Pr 1 P0

Ex7.3
For the RX pulse Pr  t  :

Pr  0   1
Pr 1  0.3 Pr  2   0.1
Pr  1  0.2 Pr  2   0.05

Design three taps equalizer (N=1)

Sol：

0   1 -0.2 0.05 C1 

1  = -0.3 1 -0.2   C0 
  
0   0.1 -0.3 1   C1 

32
  
P0 Pr

C1  0.21, C0  1.13, C1  0.318

get P0  0   1, P0  1  P0 1  0

*Note：
․ Zero ISI for k  0, 1,....  N

․ Outside this range, P0  kTb   0 . ( residual ISI )

․ EX7.3 , three-tap ZF equalizer ,

N
P0  k    C P k  n
n  N
n r , N =1

 get P0  0   1, P0  1  P0 1  0

But P0  3  0.01, P0  2   0.0145, P0  2   0.0176
P0  3  0.0318

 N is large, residual nonzero sample values will be small

< Minimum Mean Square Error (MMSE) Method >

․ P0  k   approach zero ISI   k  

․ Find equalizer taps to minimize MSE over large window   K , K 

K
1
  P0  k     k  
2
MSE 
2 K  1 k  K
(kronecker delta)
1, k  0
where   k   
0, k  0
  
․ Solution of MMSE , get  MSE  0 
 C 

C  Pr  P0

where Pr  : pseudo-inverse of nonsquare matrix Pr  2 K  1   2 N  1

․ MMSE with more robust equalizer for the reduction of ISI

33
*Note：
Chap 12 with adaptive equalization and other general equalizers.

34
 7.5.2 Timing Extraction
․ RX digital signal needs to be sampled at precise instants.
‥ RX clock needs to synchronize to TX clock (symbol or bit synchronization)
(Delayed by channel response)
․ Three general synchronization method：
‥ 1. Derivation from a primary or a secondary standard
(TX/RX slaved to master timing source)
 High cost, large data/high speed communication systems
‥ 2. TX an extra synchronizing signal ( pilot clock )
 Occupy the TX resource and use the additional TX power
(available capacity is large in comparison to data rate)
‥ 3. Self-synchronization, timing information is extracted from
the RX signal itself.
 A very efficient method of timing extraction or clock recovery.
(derived from RX signal itself)
․ For on-off signal, it contains a discrete clock frequency. When the on-off signal is
applied to “resonant circuit“ tuned to the clock frequency. Output signal is “clock
“ signal.
․ For bipolar signal, it doesn’t contain any clock frequency.
Using “Nonlinear device“ to get timing clock. “simple rectification“ converts a
“bipolar signal“ to an “on-off signal“ , it can be used to extract clock signal.

 variations of sample instants (or pulse positions) cause timing jitter

․ Timing jitter： 
 small random deviations of the incoming pulse .

 High Q of tuned circuit can suppress jitter.

long "1"  output amplitude increased  new jitter

 
long "0"  output amplitude decreased  generation

35
․ Complete timing extractor and time pulse generator for polar signal：

‥ Timing extractor and phase shifter  clock recovery

(extract the pulse maximum points)
‥ Novel highly stable “phase-locked loop”  clock recovery
(overcome jitter)

7.5.3 Detection Error

․ RX signal passed through equalizer. Then, the detector samples the pulse signal at
clock timing can detect the signal.

․ EX polar signal p  t  with AP peak amplitude.

polar signal + noise p  t   n  t  with the AP  n error peak amplitude

 Symmetric polar signal, detection threshold is “ 0 “.

36
  Calculate the average number of detection errors, Prob. of detection error.
(or likelihood of detection error)

․ n  t  : Gaussian noise (from thermal effects)  Chap 10, Optimum RX .

impulse noise (from switching , transients , strikes)  Chap 14, burst error
correcting codes .

37
 7.6 Eye Diagrams：An Important Tool

․ Eye diagram can show the effect of noise and channel ISI.
‥ Equalizer design to compensate the ISI channel distortion.
Timing extraction process for signal detection.
․ Eye diagram can examine：
‥ Severity of ISI
‥ Accuracy of timing extraction
‥ Noise immunity
․ Given a BB signal at channel output:

y  t    ak p  t  kTb 

Fig 7.25

‥ EX ：binary signal by polar NRZ ( open eye )

(1) Channel infinite BW  pass
 NRZ pulse ( no channel distortion )
(Fig. 7.25 (a), (b))
(2) Consider channel output using polar code RZ pulse  eye open at
the midpoint of pulse ,
midpoint is the best sampling point
(Fig. 7.25 (c), (d))
38
 (3) Channel with finite BW and distortion  NRZ polar signal with ISI
partially closed eye pattern
(Fig. 7.25 (e), (f))

․ Eye diagram is used to determine “optimum tap setting“ for equalizer .

Taps are adjusted to obtain the “maximum“ vertical and horizontal eye opening .
․ Eye diagram measurement regarding the following signal quality：

‥ (1) Maximum opening point.  y 

 Noise margin：eye opening amount at the sampling instants
Maximum eye opening indicates the optimum
sampling instants.

‥ (2) Sensitivity to timing jitter.  x 

 Width of eye indicates the time interval of correct decision.
(Maximum horizontal opening)
Decision-making instant deviates (timing jitter)  vertical noise margin
reduces slope of eye reduced by the
noise tolerance higher Prob. error.
‥ (3) Level-crossing ( timing ) jitter
 Practical RX extracts “pulse rate“ and “sampling clock“ from
“zero-level crossing of RX signal“.
Variation of level crossing shows “timing jitter“.

39
․ Practical eye diagram example of polar signal with raised-cosine pulse.

( zero ISI , max . eye open )

(Nyquist 1st criterion, no requirement of zero crossing pulse.)

40
 7.7 PAM：M-ARY Baseband Signaling for Higher Data Rata

․ For binary BB modulation, each line code uses 1 bit / Tb sec.

․ For high data rate , increased by M times：

 1 
‥ 1 bit /  Tb   M times BW
M 

‥ M bits / Tb  same as 1 BW

(4-ary PAM)  Pulse Amplitude Modulation (PAM)

 2bits / one symbol T  

S (2 times Data rate)

(11 , 01 , 10 , 00)

(3bits  8-ary , 4bits  16-ary)

I M  log 2 M bits (data information bits)

‥ M-ary PAM
 amplitude increased by M , TX power increased by M 2

 BW no change 1 / Tb 

EX7-4 (Good)
Determine PSD of 4-ary PAM , 1 and 0 same probability .

Sol：○
1 4-ary line code  4 distinct symbols

3 message bits 00

 1 message bits 01

ak  
 1 message bits 10
 +3 message bits 11

41
  1
○2 Get R0 , Rn ,  ak : Prob. 
 4
1
R0  lim  ak 2
N  N
k

1 N
 3   1  1   3   5
2 N 2 N 2 N 2
 lim 
N  N4 4 4 4 

Rn : a a
k k n ,  3, 1   3, 1 all case , Find average value 

1
Rn  lim
N  N
k ak ak  N

Possible values, all cases (n>0)

 1, 3, 9   
1 1 1
, ,  Prob.
8 4 8
 1 1
  1.   2  
 16 8
1 N N N N N N 
Rn  lim   9    9    1   1   3   3   0
N  N
8 8 8 8 4 4 
1 5
 S x  f   R0 
Tb Tb
5
S y  f   P  f  Sx  f   P  f  
2 2

Tb
1
( original , PSD of binary polar signaling S y  f   P  t 
2
,
Tb
( 4-ary PSD is 5 times of binary polar PSD )
 Same PSD shape ！！

42
 < Pulse Shaping and Eye Diagrams in PAM >
․ Use the Nyquist 1st criterion pulse , i.e. , raised cosine pulse ( r = 0.5 ) ,
to generate 4-level PAM .
‥ It can use Nyquist 2nd criterion pulse, i.e. , controlled ISI for M-ary signaling .

(HW)

‥ The eye diagrams show four equally separated signal values without ISI at the
optimum sampling points.

43
 7.8 Digital Carrier Systems
․ PAM , polar , …… signals are baseband signals , which use “ low frequencies “ for
wires and coaxial cables modem communication .
․ For radio link or satellites communication, the baseband signals need to be shifted
to “Higher frequencies“, i.e. , modulating a high-frequency sinusoid (carrier) .
․ Modem for two-way ( duplex ) communications：
‥ In TX, Modulator transmits data.
‥ In RX, Demodulator receives data.

7.8.1 Basic Binary Carrier Modulations

․ Three basic forms of carrier modulation
‥ Amplitude modulation (on-off keying (OOK) or amplitude shift keying (ASK))
‥ Phase modulation ( phase shift keying (PSK) )
‥ Frequency modulation ( frequency shift keying (FSK)) 
angle modulation

․ OOK or ASK：
‥ Carrier modulated signal

 ASK  t   m  t  cos C t

where
 t  Tb / 2 
m  t    ak p  t  kTb  , p  t     
 Tb 
ak  0,1  on  off  , m  t  : on  off baseband signal
1  p  t  cos C t

0  0

44
․ PSK：

․ Carrier modulated signal  PSK  t   m  t  cos C t

where
 t  Tb / 2 
m  t    ak p  t  kTb  , p  t     
 Tb 
ak  1  polar  , m  t  : polar baseband signal
1  p  t  cos C t
0   p cos C t  p cos C t   
․ FSK：
․ Carrier modulated signal

 FSK  t    ak p  t  kTb  cos C1t   1  ak  p  t  kTb  cos C 0t

where FSK signal is a superposition two AM signals with different carrier

frequencies and complementary different amplitudes.

ak  0,1 on  off 
1  p  t  cos C1t 

0  p  t  cos C 0t 
sum of two interleaved ASK signals
with two modulating frequency wc 0 , wc1

45
 * Note：(Examples)
ASK used in optical fiber comm. (laser-intensity mod)
PSK used in digital satellite comm.
FSK used in earliest telephone-line modems.

7.8.2 PSD of Digital Carrier Modulation

․ Binary carrier modulation of ASK , PSK , FSK：

  t   m  t  cos C t

․ Find PSD of   t  ：

T  f 
2

S  f   lim
T  T

where  T  f   FT of T  t 

‥ T  t  ：( truncated signal )

  T  T 
T  t     t   u  t    u  t   
  2  2

   T   T  
 m  t   u  t    u  t      cos C t
   2   2  
 mT  t   cos C t

MT  f 
2

 M T  f   FT of mT  t  , S M  f   lim
T  T
1
 T  f    M T  f  fC   M T  f  fC  
2

46
․ Get S  f  :

1 M T  f  f C   M T  f  fC 
2

S  f   lim
T  4 T
  f C  BW of M  f 
zero overlap  fC M  f 
1  M T  f  fC  M T  f  fC  
2 2

= lim   
T  4  T T 
 
1 1
 S M  f  fC   S M  f  fC 
4 4

 Modulation causes a shift in the baseband signal PSD

47
 * Note：(NRZ)

․ ASK PSD with discrete component at C ( on-off PSD with DC) .

․ PSK is same shape a ASK without discrete component at C .

․ FSK with higher BW than ASK and PSK. No discrete component in FSK.

( Phase continuity during C 0 C1 frequency switching)

Pe  SNR  3dB   Pe  SNR 

․
 PSK   ASK , FSK 

7.8.3 Connections between Analog and Digital Carrier Modulations

․ ASK (with nonnegative amplitude) is an AM signal with modulation index   1 .
․ FSK is simply an FM signal with only limited number of instantaneous frequencies.
․ PSK is similar to analog modulation (for only real signal, No image signal).
‥ For PSK ：

 PSK  t   A cos C t   k  , kTb  t  kTb  Tb

 A cos  k cos C t  A sin  k sin C t
 ak cos C t  bk sin C t
where ak  A cos  k , bk   A sin  k  QAM 

 For binary PSK ,   0,  ,  ak   A , bk  0 

 BPSK  t  =  AcosC t  PAM , 2  level 

(multiple bits , M-ary PAM)

 Digital representation of DSB-SC amplitude modulation.
 PSK is more power efficient than ASK.

7.8.4 Demodulation
․ ASK Detection
‥ ASK coherent demodulation (synchronous detection, low SNR, superior
performance)
‥ ASK noncoherent ( envelope detection ) ( high SNR , noncoherent 
coherent , simply detection )
․ FSK Detection

‥ FSK noncoherent detection ( 2 interleaved ASK with C 0 and C1 )

48
  a pair of filters C 0 , C1  , envelope detector, sampled and compared data .

( simple detector )
‥ FSK coherent detection

 generating 2 reference frequencies C 0 , C1 for two demodulators.

demodulated and compared data.

․ PSK Detection
‥ After envelope detection , “ 1 and 0 “ stay constant ( cannot compare and
demodulate )
‥ Coherent detector of BPSK ( OK )

49
․ Differential PSK ( DPSK )：
‥ DPSK uses “noncoherent“ method to detect data . (No carrier synchronization)

Figure 7.35

‥ Differential encoding：( before modulation )

The addition is modulo-2.
0  Current pulse and previous pulse (same)
1  Current pulse and previous pulse (different polarity)

 Modulated Signal：  A cos C t

 Encoded signal in Fig. 7.35 (b)

‥ Demodulation of DPSK：

 A2 
 RX pulse  previous pulse  detected bit 0  z  
 2 

 A2 
RX pulse  previous pulse  detected bit 1  z   
 2 

50
 * Note：
․ PSK uses coherent detection (need synchronization carrier) without self noise,
multiplication interference.
․ PSK has small BW than FSK.
․ Noncoherent detection methods with self noise multiplication interference.

51
 7.9 M-ary Digital Carrier Modulation

․ Binary ASK, FSK, PSK are 1 bit / Tb  bps  transmission rate.

‥ Reduce Tb to improve data rate  BW increased

‥ Increase bits to improve data rate  More power TX

․ To conserve BW and use M-ary signaling, M-level ASK, M-frequency FSK, and M-
phase PSK modulations increase transmission data rate.

< M-ary ASK and Noncoherent Detection>

․ M-ary ASK TX

‥ log 2 M  bits  / Tb , TX Data Rate  , TX waveform :

  t   0 , A cos C t , 2 A cos C t , ......  M  1 A cos C t

   
00 01 10 11  EX : M  4 

‥ Same as AM with M different amplitudes and modulation index   1 .

‥ TX power increased by M 2 times. BW is the same as binary ASK.
․ M-ary ASK RX
‥ Envelope noncoherent detection ( simple )
‥ Coherent Detection

< M-ary FSK and Orthogonal Signaling >

․ M-ary FSK TX waveform：( TX log 2 M  bits  / Tb )

A cos 2 fit , i  1, 2, ......M

where f m  f1   m  1  f , m  1, 2,  , M ( equal spacing )

f M  f1 1
f  frequency deviation     M  1  f
2 2

 f  , too much BW

 f  , noise and interference error

min.  f  orthogonal FSK sets
52
․ Find the min. f to achieve the orthogonal set of FSK signals：
Tb

 A cos  2 f t  A cos  2 f t dt =0,

0
m n mn


Tb Tb
A2
0 A cos  2 f m t  A cos  2 f n t dt   cos 2  f m  f n  t  cos 2  f m  f n  t dt
2 0

A2 sin 2  f m  f n  Tb A2 sin 2  f m  f n  Tb
 Tb  Tb
2 2  f m  f n  Tb 2 2  f m  f n  Tb

   f m  f n  Tb  103   first term ignore 

A2 sin  2  f m  f n  Tb 
0
2 2  f m  f n 

  f m  f1   m  1  f

sin  2  m  n   fTb   0 , m  n

 smallest  f , orthogonality

1
f  Hz  min . spacing , min . shift FSK 
2Tb
1 1
(Ex :  1KHz ,  f min
  500 Hz )
Tb 2Tb
․ Minimum shift FSK signals ( orthonormal basis functions ):

2  i 1 
 i t   cos 2  f1  t i  1, 2, ..., M
Tb  2 Tb 

Tb
 1, mn
where    t    t  dt  0 , m  n
0
m n

‥ Each FSK symbol： ( FSK Basic Waveform )

Tb
A cos 2 f mt  A  m  t  , m  1, 2, ..., M (M=2, two FSK symbols)
2

53
․ M-ary FSK Demodulation：
‥ Band of M coherent or noncoherent detectors of M-ary FSK signal
‥ Making decision based on the strongest detector branch.
‥ For M-ary FSK BW ( Carson’s rule )
1 1 M 3
B , f   M  1  f   M  1 / 4Tb , BW  2  f  B  
Tb 2 2Tb
* Note：
M-FSK no power enhancement
M-FSK BW increases linearity with M
Orthogonal M-ary scheme：

< M-ary PSK , PAM , and QAM >

․ M-ary PSK signals：

2 2
 PSK  t   am cos C t  bm sin C t , 0  t  Tb
Tb Tb
= am 1  t   bm  2  t 

‥ Two-dimensional signal space with orthogonal basis functions：

2 2
1  t   cos C t ,  2  t   sin C t
Tb Tb
Tb Tb

  1  t  2  t  dt  0 .    t dt  1
i
2
, i  1, 2
0 0

‥ Signal points  am , bm  with constant on a circle of radius A.

am  A cos  m , bm   A sin  m
am 2  bm 2  A2  cons tan t

54
 ‥ M-ary PSK signaling, all signal points with uniformly equally spaced angle
obtain the best noise immunity.
2
 m  0   m  1 , m  1, 2, .....M
M

․ EX M = 4 , 4 PSK ( or quadrature PSK , QPSK ) modulation

‥ Two binary PSK signals, one used in-phase carrier of cos C t , other used

quadrature carrier of sin C t .

‥ TX/RX on same frequency and same channel to double the transmission rate
․ EX QAM modulation ( quadrature amplitude modulation , not constant magnitude,
flexible amplitude )
M-ary QAM signal：

pi  t   ai p  t   cos C t  bi p  t   sin C t
=ri p  t   cos C t  i  , i  1, 2, ...M

ri  ai 2  bi 2 ,  i  tan 1  bi / ai 
where 2
p t   u  t   u  t  Tb  
Tb 

(pulse shape, simple rectangular pulse)

(or Nyquist pulse BW limited)

55
․ QAM modulator and demodulator

‥ Two signals m1  t  and m2  t  with each M -ray pulse sequence are

modulated by cos C t and sin C t .

m1  t   ai p  t  
 two PAM signals having M levels
m2  t   bi p  t  

(EX , 16  4 levels, +3 , +1 , -1 , -3)

(carries 4 bits, log2 16  4 )

(16  ai bi  , 16 constellations)

‥ 16 QAM：without increasing BW and 4 times data rate of Binary PSK

(4 bits) (1 bits)

․ If m2  t   0 , PAM signal：( Digital version of DSB-SC signal )

56
 pi  t   ai p  t   cos C t , t   0, Tb 

 Increase M times data rate.

7.10 Matlab Exercise

57
 RZ eye-diagram
1 NRZ eye-diagram
1

0.5
0.5

0
0

-0.5
-0.5

-1 -1
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Time Time

Raised-cosine eye-diagram
1.5

0.5
Amplitude

-0.5

-1

-1.5
-30 -20 -10 0 10 20 30
Time

RZ eye-diagram
3 NRZ eye-diagram
3

2
2

1
1

0 0

-1 -1

-2 -2

-3 -3
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Time Time

58
 Raised-cosine eye-diagram
5

-5
-30 -20 -10 0 10 20 30
Time

59