AWGN

1.
2.
3.
4.
4.1
4.2
4.3
5. Regenerative repeater
1.
Sec. 5.1, 5.2, pp. 309-317, Haykin

e jn0t , n = 0,1,2,... , Walsh functions, Legendre polynomials, Laguerre

functionseternal functionsWalsh functions
M N ( M )
Gram-Schmidt orthogonalization process
1-1(C-12-6)

Two signals which might be used in a digital communication system are

sketched below
Signal #1
1
-1

0 .25 .5 .75

Signal #2
1
1
0
t
-1
.25 .5 .75 1

For these two signals, determine

(a) each signal's energy
(b) a single set of orthonormal basis functions which can be used to
represent each signal
(c) the distance between the signal in signal space.
1-2(Prob. 5.1, p. 338, Haykin)

In Section 3.7 we described line codes for pulse-code modulation. Referring to

the material presented therein, formulate the signal constellations for the following
line codes:
(a) Unipolar nonreturn-to-zero code
6-1

(b) Polar nonreturn-to-zero code

(c) Unipolar return-to-zero code
(d) Manchester code
1-3(Prob. 5.2, p. 338, Haykin)

An 8-level PAM signal is defined by

T
t
2
si (t ) = Ai rect
T

where Ai = 1 , 3 , 5 , 7 . Formulate the signal constellation of {s i (t )}i =1 .

8

1-4(Prob. 5.3, p. 338, Haykin)

The following figure displays the waveform of four signals s1 (t ) , s2 (t ) , s3 (t ) and
s4 (t ) .

(1) Using the Gram-Schmidt orthogonalization procedure , find an orthonormal

basis for this set of signals.
(2) Construct the corresponding signal-space diagram.
s1 (t )

s3 (t )

s2 (t )

t
0

T
3

s4 (t )

1
t

t
0

2T
3

T
3

t
0

1-5(Prob. 5.4, p. 338, Haykin)

(a) Using the Gram-Schmidt orthogonalization procedure, find a set of orthonormal

basis functions to represent the three signals s1(t), s2(t), and s3(t) shown in Figure
P5.4.
(b) Express each of these signals in terms of the set of basis functions found in part
(a).

FIGUR
6-2


E P5.4

1-6(Prob. 5.6, p. 339, Haykin)

A source of information emits a set of symbols denoted by {mi }i =1 . Two

M

candidate modulation schemes, namely, pulse-duration modulation (PDM) and

pulse-position modulation (PPM), are considered for the electrical representation of
this set of symbols. In PDM, the ith symbol is represented by a pulse of unit
amplitude and duration (i/M)T. On the other hand, in PPM, the ith symbol is
represented by a short pulse of unit amplitude and fixed duration, which is transmitted
at time t = (i/M)T. Show that PPM is the only one of the two that can produce an
orthogonal set of signals over the interval 0 t T.
1-7(Prob. 5.7, p. 339, Haykin)

A set of 2M biorthogonal signals is obtained from a set of M orthogonal signals

by augmenting it with the negative of each signal in the set.
(a) The extension of orthogonal to biorthogonal signals leaves the dimensionality of
the signal space unchanged. Why?
(b) Construct the signal constellation for the biorthogonal signals corresponding to
the pair of orthogonal signals shown in Figure P5.5.

FIGURE P5.5
1-8(Prob. 5.8, p. 339, Haykin)

(a) A pair of signals si(t) and sk(t) have a common duration T. Show that the inner
product of this pair of signals is given by

s (t )s (t )dt = s
T

T
i

sk

where si and sk are the vector representation of si(t) and sk(t), respectively.
(b) As a followup to part (a), show that

(s (t ) s (t )) dt = || s
T

s k || 2

1-9(Prob. 5.9, p. 339, Haykin)

Consider a pair of complex-valued signals s1(t) and s2(t) that are respectively
6-3

represented by
s1(t) = a111(t) + a122(t),
s2(t) = a211(t) + a222(t),

t
t

where the basis functions 1(t) and 2(t) are both real valued, but the coefficients a11,
a12, a21, a22 are complex valued. Prove the complex form of the Schwarz inequality:

s1 (t )s (t )dt
*
2

s1 (t ) dt s 2 (t ) dt

where the asterisk denotes complex conjugation. When is this relation satisfied with
the equality sign?

2.
x(t)

E x x (t ) x * (t )dt

(2-1)

x(t) h(t)

R xh x (t )h * (t )dt = x (t + )h * (t )dt

(2-2)

R x R xx ( ) = x (t ) x * (t )dt = x (t + ) x * (t )dt

(2-3)

E x x 2 (t )dt

(2-4)

R xh x (t )h (t )dt = x (t + )h (t )dt

R x R xx ( ) = x (t ) x (t )dt = x (t + ) x (t )dt

(2-5)
(2-6)

x(t)(finite duration) x(t)=0 for t<0

t>T x(t)
Rx(t-T) h(t)=x(T-t)

y (t ) = x (t ) * h (t ) = x ( )h(t )d = x ( ) x (t T + )dt = R x (t T ) (2-7)

x(t)(matched filter)

2-1(Prob. 2.67, p. 170, Oppenheim & Willsky, Signals & Systems, 2nd ed.,
1997)
6-4

(a) LTI h(t)() h(t)=0 for t<0 and t>Ty(t)

T

x(t) h(t)=ax(T-t)(a ) y(T) h 2 (t )dt = M

0

: y (t ) = x ( )h (t )d

y (T ) = x ( )h (T )d x 2 (t )dt

T
= x 2 (t )dt
0

1/ 2

T h 2 (t )dt
0

1/ 2

1/ 2

T x 2 (t )dt
0

h 2 (T t )dt

T
= M 1 / 2 x 2 (t )dt
0

T
y (T ) = M 1 / 2 x 2 (t )dt
0

T
y (T ) = a 2 x 2 (t )dt
0

1/ 2

1/ 2

(Schwarz Inequality)

1/ 2

1/ 2

1/ 2

h(t)=ax(T-t)

= a x 2 (t )dt a =
0

x 2 (t )dt

x0 (t )
1 x1 (t ) x0 (t ) x1 (t )
x0 (t )x1 (t )

(b) Lo x0 (t ) L1 x1 (t )

(i) Lo x0 (t ) x1 (t ) L1 x0 (t ) x1 (t )

6-5

(ii) t=4 (i) x0 (t )

x0 (t ) x1 (t ) L0 x1 (t ) L1 x0 (t ) t=4

: x0 * L0 t =4 = 4 , x0 * L1 t =4 = 2 , x1 * L0 t =4 = 2 , x1 * L1 t =4 = 4

x0 (t ) :

(c)

(i) p(t) R p (0) = max R p ( )

6-6

x(t)=p(t-to) R xp (t0 ) = max Rxp ( )

: R p ( ) = p (t + ) p (t )dt p 2 (t )dt

p 2 (t )dt

1/ 2

p 2 (t + )dt

1/ 2

R p ( ) R p (0)

R p (0) = max R p ( )

R xp ( ) = p (t + t0 ) p (t )dt = R p ( t0 )

R xp (t0 ) = R p (0) = max Rxp ( )

p(t)

(ii) p(t) x(t) Rxp ( )

C = 3 108 m / sec ,

Rxp ( )

6-7

10 sec
2 distance = 3 108 m / sec 10 10 6 sec = 3 10 3 m
distance = 1500 meter

(iii) p(t)
(sharply peaked)
?

p2 (t ) sharp peak !

2-2(Prob. 2.66, p. 169, Oppenheim & Willsky, Signals & Systems, 2nd ed.,
1997) LTI Walsh functions

6-8

(a) hi (t ), i = 1,2,3 xi (t ), i = 1,2,3

(i) xi (t )
(ii) xi (t ) = 0, t < 0
(iii) xi (t ) 1 t 0
(iv) yi (t ) = xi (t ) *h i (t ) t=4
:

(b) yij (t ) = xi (t ) *h j (t ), i j t=4 yij (t ) = ?, i, j = 1,2,3

: t=4 x1 * h2 = x2 * h3 = x1 * h3 = x2 * h1 = 0 hi (t ) LTI
xi (t ) i=1, 2, 3

(c) (tapped-delay-line) hi (t ) , i=1, 2, 3

:

6-9

g k (t k ) h(t) LTI
k =0

rectangular pulse of amplitude A and duration T matched filter

(Ex. 4.1, p. 252, Haykin)
T

t
2
g (t ) = h (t ) =
T

t T
g o (t ) = kA2T

T
t=T Fig. 4.3, p. 253, Haykin integrate-and-dump
t=T RC
S/N
2-1(Prob. 4.4, p. 301, Haykin)
If the ideal integrator is replaced by the simple resistance-capacitance (RC) low-pass

filter, determine the output SNR as a function of the time constant RC.
Determine the value of RC that maximizes the output SNR. The frequency
response of this filter is
1
1
H( f ) =
where f 0 =
f
2RC
1+ j
f0
The requirement is to optimize the selection of the 3-dB cutoff frequency f 0 of the
filter so that the peak pulse signal-to-noise ratio at the filter output is maximized. With
this objective in mind, show that the optimum value of f 0 is 0.2/T, for which the
loss in signal-to-noise ratio compared to the matched filter is about 1 dB.
3.1-2-2 (Prob. 7.17, p. 459, Proakis)
2-2(Prob. 7.31, p. 463, Proakis)
In the case when n is a power of 2, an n n Hadamard matrix is constructed by
means of the recursion

1 1
H2 =

1 1

H n
H 2n =
H n
6-10

Hn
H n

(a) Let Ci denote the ith row of an n n Hadamard matrix as defined above. Show
that the waveforms constructed as
n

s i (t ) = cik p(t kTc ) ,

k =1

i = 1, 2, K, n

are orthogonal, where p(t) is an arbitrary pulse confined to the time interval
0 t Tc .
(b) Show that the matched filters (or crosscorrelators) for the n waveforms {si(t)} can
be realized by a single filter (or correlator) matched to the pulse p(t) followed by a
set of n crosscorrelators using the code words {Ci}.
2-3(D-6-4)

For the pairs of binary signaling waveforms sketched below, determine whether
they are antipodal, orthogonal, and determine the matched filter for each.
(a)

-1

-1

T
0

"0"

(b)

"1"

"0"

"1"

(c)

T
0

"0"

"1"

0
T t
0

"0"

T t
"1"

3.
Eq. (5.89), p. 334, Haykin
Eq.(7.6-10),p.406, Proakis Appendix C, Carlson s1 s 2
AWGN
6-11

Dis tan ce

2 N o = Q

2 N o

p[error ] = Q s1 s 0

No 0
"0"

E{ |"0"} = E{R1 R0 |"0"}

{
= E { R( , t )[s (t ) s (t )]dt }
= E { [s (t ) + N ( , t )][s (t ) s (t )]dt }
= s (t )[s (t ) s (t )]dt + E { N ( , t )[s (t ) s (t )]dt }
144424443 14444244443

= E{R1 |"0"} E{R0 |"0"} = E R( , t )s1 (t )dt R( , t )s0 (t )dt

TB

TB

TB

TB

TB

TB

0 (QE {N ( , t )}= 0 )

= s0 (t ), s1 (t ) s0 (t )

= E1E0 E0

Var{ |"0"} = E [(R1 R0 ) E (R1 R0 )] |"0"

2

= E [(R1 E{R1}) (R0 E{R0 })] |"0"

2

= Var{R1 |"0"} + Var{R0 |"0"} 2 cov{R1 , R0 |"0"} = 4

Var{R0 |"0"} = E (R0 E{R0 }) |"0"

2

2
TB

= E R( , t )s0 (t )dt E0 |"0"

6-12

= E{ R( , )R( , )s0 ( )s0 ( )dd 2 E0 R( , t )s0 (t )dt |"0"} + E02

0
144
40 4444
4244444444
3 1404424443
TB

TB

TB

1 = E

TB

[s ( ) + N ( , )][s ( ) + N ( , )]s ( )s ( )dd }

TB

= s02 ( )s02 ( )dd + 2 s0 ( )E{N ( , )}s0 ( )s0 ( )dd

0
0
1
40 44
424444
3 1044
44444244444443
TB

TB

TB

TB

E 02

+ E{N ( , )N ( , )}s0 ( )s0 ( )dd

0
1
40 444444244444443
TB

TB

3 =

TB

TB

TB

E{N ( , )N ( , )}s0 ( )s0 ( )dd

N0
( )s0 ( )s0 ( )dd
2

TB

N0
2

N0
E0
2

TB

s02 ( )d

2 = 2 E0

TB

[s0 (t ) + E{N ( , t )}]s0 (t )dt = 2 E02

EN
EN

Var{R0 |"0"} = E02 + 0 0 2 E02 + E02 = 0 0

2
2

Var{R1 |"0"} = E [R1 E{R1}] |"0"

2

2
TB

= E R( , t )s1 (t )dt E1E0 |"0"

= E{ [R( , )R( , )]s1 ( )s1 ( )dd

0
14
4044444244444443
TB

TB

2 E1E0 R( , t )s1 (t )dt |"0"} + 2 E1E0

TB

TB TB
4 = E s0 ( ) + N ( , )[s0 ( ) + N ( , )]s1 ( )s1 ( )dd

0 0

= 2 E1E0 +

N0
E0
2

6-13

EE
N

Var [R1 |"0"] = 2 E1E0 + 1 0 2 2 E1E0 = 0 E1

2
2

Cov[R1 , R0 |"0"] = E{[R1 E{R1}][R0 E{R0 }] |"0"}

= E{R1R0 R0 E{R1} E{R0 }R1 E{R1}E{R0 } |"0"}
= E{R1R0 "0"} E{R1 "0"}E{R0 "0"}
14243 144
42444
3
5

5 = E
0

TB

= E 0 E 0 E1

[s ( ) + N ( , )][s ( ) + N ( , )]s ( )s ( )dd

TB

= s02 ( )d s0 ( )s1 ( )d +
TB

TB

TB

= E0 E1E0 +

TB

s0 ( )s1 ( )E{N ( , )N ( , )}d d

N0
E1E0
2

Cov = {R1 , R0 "0"} = E0 E1E0 + 0 E1E0 E0 E1E0

2

N0
E1E0
2

Var{ "0"} =
=

E1 N 0 E0 N 0
N

+
2 0 E1E0
2
2
2

N0
E0 + E1 2 E1E0
2

"1"

E { "1"} = E1 E1E0
Var { "1"} =

N0
E0 + E1 2 E1E0
2

(PDF)
T
6-14

T=
=

E{ "1"} + E{ "0"}
2

( p["0"]=p["1"])

E1 E0
2

p[error "0"] = E1 E0 e

1 x E { "0"}

2 Var { "0"}

dx

2 var{ "0"}

E1 E0

E1E0 E0

= Q
2
= Q

N0

E0 + E1 2 E1E0

)
)

= Q

E0 + E1 2 E1E0
2

N0
E0 + E1 2 E1E0
2

s1 (t ) s0 (t )
E 0 + E1 2 E1 E 0

=
Q

2N0
2N0

p[error"1"]=p[error"0"]

s (t ) s0 (t )
p[error ] = Q 1

2N 0

s1 (t ) s0 (t ) s1 (t ) s0 (t ) 3

3-1(C-5-3, C-6-1, C-6-2)

Three pulses sketched below are used to communicate three messages in a digital
communication system.
(a) Determine an orthonormal basis which spans this signal set, and determine the
orthonormal series expansion for each signal.
(b) Determine the norm for each signal.

(c) Determine the distance between each pair of signals.

(d) Determine the cross-correlation and correlation coefficient between each pair of
signals.
(e) If only two messages are to be communicated, which pair of signals is the best
choice?
(f) For each of these signals, determine a signal which is antipodal and a signal which
is orthogonal.
(g) Determine the impulse response h(t) for the filter matched to the pulse if the
sampling time is 0.75 sec, 1.0 sec, or 1.5 sec.
(h) For your answer in Part (g), sketch the output of the matched filter for all time
when the filters input signal is the signal to which the filter is matched.
6-15

(i) Sketch the response for all time of the filters matched to signal #2 and #3 designed
for a 1 second sampling time when signal #1 is used as input.
(j) Design matched filters with a one-second sampling time for signals which are
orthogonal and antipodal to signal #1 determined in Part (f). Sketch their
output for all time when signal #1 is used as the input. How do your sketches
compare at the sampling time to your results in Part (i)? Why?
(k) Determine the bit error probability if the signals are #1 and #2.
(l) Determine the bit error probability if the signals are #1 and #3.
(m) Determine the bit error probability if the signals are #2 and #3.
(n) Determine the bit error probability if the signals are #1 and its orthogonal signal.
(o) Determine the bit error probability if the signals are #1 and its antipodal signal.
#1 S1 (t )

#2

#3

S 2 (t )

S 3 (t )

2
1

1.58

1
1

0.5

0 0.5
-1.58

0.5

-2
message

1. Signal space computations

(a)spanorthonormal basis

orthonormal series expansion

(b)norm
(c)
(d)cross-correlation
correlation coefficient
(e)
(f)antipodal signalorthogonal signal

6-16

(a)

1
2
(b) s (t ) = s 2 (t )dt
0

12

s1 (t )

12
1
= 4dt + 1dt
0

12

s2 (t )

12
1
2
2
= (1.58) dt + (1.58) dt
0

12

= 2.5
12

= 2.5

s3 (t ) = s1 (t ) = 2.5

(c) si (t ) s j (t ) = si (t ) + si (t ) 2 si (t ), s j (t )
2

s1 (t ), s2 (t ) = 2 1.58dt + 1 ( 1.58)dt = 1.58 2

12

12

s1 (t ), s3 (t ) = 2 ( 2)dt + 1 1dt = 2 + 1 2 = 1.5

12

12

6-17

12

s2 (t ), s3 (t ) = 1.58 ( 2)dt +
12

12

( 1.58) 1dt = 3( 1.58) 2

d12 s1 (t ), s2 (t ) = [2.5 + 2.5 2(1.58 2 )] = 1.849

12

d13 s1 (t ), s3 (t ) = [2.5 + 2.5 2( 1.5)] = 2.828

12

d 23 s2 (t ), s3 (t ) = [2.5 + 2.5 2 3 ( 1.58 2 )] = 3.121

12

(d)cross-correlation
Rij = si (t ), s j (t )

R12 = 1.58 2 = 0.79

R13 = 1.5
R23 = 3( 1.58 2) = 2.37
Correlation coefficient
ij =

Rij
Ei E j

12 = 0.79 2.5 = 0.312

13 = 1.5 2.5 = 0.6

23 = 2.37 2.5 = 0.948

(e) s2 (t ) s3 (t ) s2 (t ) s3 (t )
(f)Antipodal ("opposite")

Orthogonal (Inner product equals zero)

6-18

()
2. matched filter
(a) h(t)
0.75 sec1.0 sec 1.5sec
(b) (a)matching signal

(c) #1 #2 #3
TS = 1 sec
(d) (f)antipodal(orthogonal
TS = 1 sec#1 (c)
(a)run backwards
in time h(t ) = s(TS t )

(a)

6-19

(b)
TS = 1 sec
TS = 0.75 sec 1.5 sec TS = 1 sec
0.25 sec 0.5sec

6-20

(c)

6-21

TS = 1 #2 #3

2.5 -1.5

2.5 ( 1.5) > 2.5 #1 #3 #1 #2

(d) orthogonal:
s1 (t ), sorthog (t ) = 0 = s1 (t ), sorthog (t )dt
1

Antipodal:
s1 (t ), santip (t ) = E = s1 (t ), santip (t )dt = s1 (t )s1 (t )dt
1

3. Bit Error Computations

N 0 2 = 1 Watt/Hz AWGN
PAM
(a) #1 #2bit error probability
(b) #1 #3
(c) #2 #3
(d) #1
(e) #1 antipodal

s s
p[error ] = Q 1 0
2N
0

= Q dis tan t

2N
0

= s0 s1 = [E1 + E0 2 s0 , s1

6-22

12

E0 = E1 = 2.5

inner product

(a)

)]

s1 , s2 = s1 (t )s2 (t )dt = 1 2 2 2.5 + 1 2.5 = 0.5 2.5

1

s1 s2 = 2.5 + 2.5 2.5

12

1.849
p[error ] = Q
= Q(0.9245) 1 0.8238 0.18
4
(b)

s1 , s3 = 1 2 [2( 2 ) + 1( 1)] = 1.5

s1 s3 = [2.5 + 2.5 2( 1.5)] = 8 = 2.828

12

p[error ] = Q(2.828 2 ) = Q(1.414 ) 1 0.9207 0.08

(c) s2 , s3 = 1 2 2 2.5 2.5 = 3 2.5 2

s2 s3 = 2.5 + 2.5 + 3 2.5

12

= 3.12

p[error ] = Q(1.56 ) 1 0.9406 0.06

(d)

s1 , sorthog = 0
s1 sorthog = 5

p[error ] = Q 5 2 = Q(1.118) 1 0.8686 0.13

(e)

s1 , santip = E = 2.5
s1 santip = 10

p[error ] = Q 10 2 = Q(1.581) 1 0.9429 0.06

antipodal signal pair p[error]
Q(x)
3-1(Prob. 4, mid1989, Pawlowski)
A baseband digital communication system uses the two symbols sketched below
to transmit information:

6-23

1) Sketch the impulse response h0 (t ) and h1 (t ) , each with a sampling time of

1 sec, for filters matched to s 0 (t ) and s1 (t ) respectively.
2) Sketch the optimum receiver for this system. Clearly label and define each
function, and specify all parameter values
3) .Assume E = 19.22, and that zero-mean, WSS, AWGN with N0 = -60dBW/Hz
is added to the transmitted signal. Determine the symbol error probability at
the receiver output.
4) If the system organizes its information in words which are eight symbols in
length, determine the word error probability at the receiver output.
5) Is the signal s1(t) the best possible choice when s0(t) is used for binary
baseband signaling? If your answer is yes, explain why. If your answer is no,
also explain why not and sketch a better choice of s1(t).

4.

multidimensional signal
AWGNAdditive White Gaussian Noise
performanceerror probability
SNR
4.1
orthogonal
antipodal signalOn-off
4.1-1Orthogonal Signals
1 01 s 0 (t )
s1 (t ) R bpsbits per second
Tb = 1 / R

0 s 0 (t ) = 1 0 t Tb

1 s 1 ( t ) = 1

(1)
0t<

Tb
2

Tb
< t Tb
2
T
t = b
2

s0 s1
6-24

(2)

Tb

s0 (t ) s1 (t )dt = 0

(3)

AWGN n(t ) N 0 / 2 Watt/Hz

r (t ) = si (t ) + n(t )
i = 0,1
0 t Tb
(4)
r (t ) 0 t Tb 01
optimum
receiversignal correlator
matched filter detector 2
s0 (t )
r (t ) = s0 + n(t )
0 t Tb
2 r0 r1 t = Tb
Tb

Tb

Tb

r0 = r (t ) s0 (t )dt = s0 (t )dt + n(t ) s0 (t )dt

2

(6)

= + n0
S 0 (t )

S1 (t )
A
t

Tb
2

Tb

Tb
-A

1 s0 (t ) s1 (t ) (Fig. p5.5 , p339, Haykin )

r0

()d

S0 (t )

Detector

S1(t )
r1

Output
data

()d
0

Sample
at t=Tb

Tb

Tb

Tb

r1 = r (t ) s1 (t )dt = s0 s1 (t )dt + n(t ) s1 (t )dt

= n1

(7)

n0 n1 correlator
Tb

n0 = n(t ) s0 (t )dt

(8)

6-25

Tb

n1 = n(t ) s1 (t )dt

(9)

E = A 2Tb s 0 (t ) s1 (t ) s1 (t )

0 t Tb

r (t ) = s1 + n(t )

r0 = n0

(10)

(11)

r1 = E + n1
n(t )

N0

Zero mean
Tb

E (n0 ) = s0 (t ) E[n(t )]dt = 0

(12)

E (n1 ) = 0

(13)

variance i2

i2 = E (ni2 ) =

Tb

Tb

si (t ) si ( ) E[n(t )n( )]dtd

N 0 Tb
si (t ) si ( ) (t )dtd
2 0
N Tb
= 0 si2 (t )dt
2 0
N 0
=
i = 0, 1
2

(14)

(15)

s0 (t )

P (r0 | s0 (t )) =

2
2
1
e ( r0 ) / 2
2

(16)

P(r1 | s0 (t )) =

2
2
1
e r1 / 2
2

(17)

(probability density functionPDF) 3

3 s0 (t ) p(ro | 0) p (r1 | 0)

6-26

s0 (t ) r1 > r0
Pe = P(r1 > r0 ) = P(n1 > + n0 ) = P(n1 n0 > )
n1 n2 Zero-mean Gaussian random variable x
x = n1 n0 Zero-mean Gaussian x

E ( x 2 ) = E[(n1 n0 ) 2 ] = E (n1 ) 2 + E (n2 ) 2 2 E (n1n0 )

(18)

E (n1n0 ) = 0
E ( x 2 ) = 2(

(19)

N 0
) = N 0 = x2
2

(20)

Pe =

1
2 x

e x

/ 2 x2

= erfc E
dx = Q

2N
0
N0 2

(21)

(impulse response)
0 t Tb
(22)
h(t ) = s(Tb t )
y(t)(convolution integral) s(t)
t

y (t ) = s ( )h(t )d

(23)

(23) h(t ) (22)

t

y (t ) = s ( )h(Tb t + )d

(24)

y(t) t = Tb
Tb

y (Tb ) = s 2 ( )d = E

(25)

E s(t) t= Tb

4
Monte Carlo Pe SNR
(0,1)(0,0.5)
(binary source output) 0 1 0 r0 = E + n0
r1 = n1 1 r0 = n0 r1 = E + n1 n0 n1
zero-mean 2 = EN 0 / 2
E 1(Signal-to-Noise Ratio, SNR) E / N 0
1/2 2 N=10000 SNR
10 N=10000
6-27

Pe = 10 3 5 (21)
Monte Carlo

5
4.1-1-1 1 AWGN
10 / Tb s0 (t ) 10
(A,A,,A) s1 (t ) (A,A,A,A,A,-A,-A,-A,-A,-A) s0 (t )

rk = A + nk

k = 1,2,L,10

s1 (t )
A + nk
rk =
A + n k

1 k 5
6 k 10
6-28

{nk } zero-mean 2
MATLAB {nk } 2 = 0 2 = 0.1

2 = 1.0 2 = 2.0
4.1-1-2 1 6
S0 (t )

S1 (t )
A

Tb

Tb
2

-A

Tb

s0 (t ) s1 (t )

4.1-1-3
4.1-1-4 (Prob.5.5, p. 338, Haykin)
An orhtogonal set of signals is characterized by the property that the inner product of
any pair of signals in the set is zero. The following figure shows a pair of signals s0(t)
and s1(t) that satisfy this condition.
1) Construct the signal constellation for s0(t) and s1(t).

s0(t)

s1(t)

A
Tb/2 Tb
0

Tb

-A
Further problems:
2) Determine the correlator outputs at the sampling instants.
3) Consider the use of matched filters for the demodulation of the signals shown,
determine the outputs.
4) Consider the detector for the signals shown, which are equally probable and
have equal energies. The optimum detector for these signals compares r0 and
r1 and decides a 0 was transmitted when r0 > r1 and that a 1 was transmitted
when r1 > r0. Determine the probability of error.
4.1-1-5 (Prob. 5.12, p. 340, Haykin)
6-29

Figure P5.12 shows a pair of signals s1(t) and s2(t) that are orthogonal to each
other over the observation interval 0 t 3T. The received signal is defined
by
x(t) = sk(t) + w(t), 0 t 3T
k = 1, 2
where w(t) is white Gaussian noise of zero mean and power spectral density N0/2.
1) Design a receiver that decides in favor of signals s1(t) or s2(t), assuming that
these two signals are equiprobable.
2) Calculate the average probability of symbol error incurred by this receiver for
E/N0 = 4, where E is the signal energy.

FIGURE P5.12
4.1-2 Antipodal Signals

(antipodal signal) s 0
s1 s 0 (t ) = s(t ) s1 (t ) = s(t ) 7
Sec. 4.3, pp. 253-258, Haykin AWGN

r (t ) = s(t ) + n(t )
0 t Tb
(26)
S0 (t )

S1 (t )

A
0

Tb

Tb
-A

S0 (t )

S1 (t )

A
0

Tb
2

Tb

0
-A

Tb Tb
2

6-30

7 antipodal signals

8 antipodal signal (a) (b)

8 s0 (t )

r (t ) = s(t ) + n(t )
r = Tb

(27)

r =+n

(28)

Tb

n = n(t ) s (t )dt

(29)

n(t) Zero-mean E(n)=0

2 = E (n 2 ) =

Tb Tb

s (t ) s ( ) E [n(t )n( )]dtd

N 0 Tb Tb
s (t ) s ( ) (t )dtd
2 0 0
N Tb 2
= 0
s (t )dt
2 0
N 0
=
2
=

(30)

(31)

s(t ) r
2
2
1
e ( r ) / 2
2
s (t ) r

P(r | s(t )) P(r | 0) =

P(r | s(t )) P(r | 1) =

2
2
1
e ( r + ) / 2
2

6-31

(32)

(33)

9 0 1

equiprobable

1
0 1 P(0)=P(1)= 9
2
s(t ) r < 0
1

Peo = P ( r < 0 0 ) =
=

1
2

e ( r )

/ 2 2

e r

= Q = Q

/2

2
N0

dr

(34)

dr
1

= erfc
2

E
N0

Pe1 = P (r > 0 1 ) = Pe0

2E
Pe = P (0)Pe 0 + P (1)Pe1 = Q
N
0

= erfc E
2
N
0

(35)

eq. (4.38)~(4.41) Fig. 4.6, p.258, Haykin 0 1

eq. (4.28)~(4.37) Fig. 4.5, pp. 255-2572134
antipodal signal performance
antipodal signal
antipodal signal 3dB
10
Monte Carlo Pe SNR
(01)(0,0.5)
(binary source output) 0 1 0 1 (mapping)
E n zero-mean
2 = EN 0 / 2 E 1SNR 1/2 2
0 r > 0 I r<0
0
N=10000 SNR 10
N=10000 Pe = 10 3
11 (35) Monte
6-32

Carlo 5 performance 3dB

10 antipodal signal

11
4.1-2-1
2 = 0 2 = 0.1 2 = 1.0 2 = 2.0
4.1-2-2(Prob. 7.17, p. 459, Proakis, 7.26)

The demodulation of the binary antipodal signals

Eb

s1 (t ) = s2 (t ) = T
0

0t T
otherwise

can be accomplished by use of a single integrator, as shown in the following

figure , which is sampled periodically at t= kT, k=0, 1 , 2 , .. The additive
noise is zero-mean Gaussian with power-spectral density of
1) Determine the output SNR of the demodulator at t = T
6-33

N0
W/Hz.
2

2) If the ideal integrator is replaced by the RC filter shown in the figure, determine
the output SNR as a function of the time constant RC.
3) Determine the value of RC that maximizes the output SNR.
r (t ) = si (t ) + n(t )

Output decision

t
() dt
0

Detector

t=kT

R
t = kT +

<< T

4.1-2-3(Prob. 7.23, p. 460, Proakis, 7.32)

Consider a signal detector with an input

r = A+ n

where +A and A occur with equal probability and the noise variable n is
characterized by the (Laplacian) p.d.f. shown in the following figure .
1 ) Determine the probability of error as a function of the parameters A and .
2 ) Determine the SNR required to achieve an error probability of 10 5 .
How dose the SNR compare with the result for a Gaussian p.d.f.?

p ( n) =

1
2

2 /

n
4.1-2-4(Prob. 5.13, pp. 339-340, Haykin , Prob. 7.24, p. 461, Proakis, 7.33)

A Manchester encoder maps an information 1 into 10 and a 0 into 01. If the output
of the encoder is transmitted by use of NRZ, the signal waveforms corresponding
to the Manchester code are shown in the following figure. Determine the
probability of error if the two signals are equally probable.
s1 (t )

s 2 (t )

T
t

-A

0
-A

6-34

4.1-3On-off
antipodal signal
on-off 0 I s(t)

n(t )
r (t ) =
s (t ) + n(t )

0 is transmitted
1 is transmitted

(36)

n(t ) AWGN

n
r (t ) =
+ n

(37)

0 is transmitted
1 is transmitted

P(r | 0) =

P(r | 1) =

2
2
1
e r / 2
2

(38)

0 is transmitted

2
2
1
e( r ) / 2
2

(39)

1 is transmitted

0 1 equiprobable 12

12 on-off
0

Pe 0 ( ) = P(r > ) =

1
2

e r

/ 2 2

dr

(40)

threshold 1

Pe1 ( ) = P(r < ) =

e ( r )

/ 2 2

dr

(41)

Pe ( ) =

1
1
Pe 0 ( ) + Pe1 ( )
2
2

(42)

6-35

2
(40)(41)(42)

opt =

Pe (

opt

) = Q

2N

(43)

On-off 10
Monte Carlo Pe SNR
(0,1)(0,0.5)
(binary source output) 0 1 0 1 (mapping) 0 E
n zero-mean

2 = EN 0 / 2 E 1SNR 1/2
E/2 r>l/2 1 r<l/2
0
N=10000 SNR 10
N=10000 Pe = 10 3
13(43) Monte
Carlo (24)(34)(43) 5 11 13
On-off antipodal signal 6dB 3dB
3dB

13
4.1-3-1
2 = 0 2 = 0.1 2 = 1.0 2 = 2.0
4.1-3-2(Prob. 4-21, p. 221, Couch)
6-36

Assume a typical binary sequence and show that if the corresponding polar NRZ
signal and unipolar NRZ signal have the same peak-to-peak amplitude, the polar
signal has less power (an advantage) than the unipolar signal. If noise is added to
these signals, how do the probabilities of error compare for these two signaling
techniques?
4.1-4
antipodal on-off
(signal space)
signal constellation diagram 14
antipodal on-off
( ,0) (0, )

(r0 , r1 )
(r0 , r1 ) = ( + n0 , n1 )

(r0 , r1 ) = (n0 , + n1 )

= 0.1 = 0.3 = 0.5 100 x

oy *
15

14 Aantipodal signals Bon-off signals Corthogonal signals

6-37

15 Monte Carlo

4.2 Multi-amplitude Signal Transmission

level

16

6-38

0t T
sm (t ) = Am g (t )
Am m g (t )

g (t ) =

(45)

0 t T
otherwise

1/T
0

(44)

4 { Am } = {3d , d , d ,3d }
Am = (2m 3)d
m = 0,1,2,3

(46)

Euclidean Distance 2d 4 16
pulse-amplitude-modulatedPAM

Fig. 4.20,
p. 276, Haykin IV 16
00 s0 (t )
01 s1 (t )

11 s2 (t )

10 s3 (t )

symbol
symbol interval
R = 1 / Tb symbol interval T = 2Tb PAM
17 PAM AWGN

0t T
(47)
r (t ) = si (t ) + n(t )
i = 0,1,2,3

correlatormatched filter

r = r (t ) g (t )dt
0

= Ai g 2 (t )dt + g (t )n(t )dt

(48)

= Ai + n

17 4 PAM
n

2 = E (n 2 ) =

Tb

N0
2
N0
=
2
N0
=
2
=

Tb

Tb

Tb

Tb

g (t ) g ( ) E[n(t )n( )]dtd

(49)

g ( t ) g ( ) ( t ) dtd

(50)

g 2 ( t ) dt

r
6-39


2
2
1
e ( r Ai ) / 2
2
Ai 4

P(r | si (t )) =

(51)

Di = r Ai , i = 0,1,2,3 n > d

+ 3d 3d 4
symbol
P4 =
=

3
P ( r Am
4
3
2

d /

1
e
2

> d ) =
x

/2

dx

3
2

1
e
2

3
Q

/ 2

(51)

dx
=

3
Q

2d 2
N 0

(5.95), p. 335, Haykin 4

symbol

E av =

1 4 T 2
1
s k (t )dt = d 2 (9 + 1 + 1 + 9) = 5d 2

0
4 k =1
4

(52)

d 2 = Eav / 5 Eav / 2 = Eavb

P4 =

3 2 Eav
Q
2 5 N 0

3 4 E avb
= Q
2 5N
0

(53)

SNR 10 log10 ( Eavb / N 0 ) M = 4 PAM

18 Monte Carlo Pe SNR
(0,1)
(0,0.25)(0.25,0.5)(0.5,0.75)(0.75,1.0) 0001
1110{-3d,-d,d,3d}
r = Am + n r A m Am
d 1SNR E avb / N 0 = 5 / 4(d 2 / 2 )
N=10000 SNR 10
N=10000 Pe = 10 3
19(53) Monte
Carlo

6-40

18 Monte Carlo PAM

19
4 M = 2 k

s m (t ) = Am g (t )
m = 0,1,2,L, M 1
(54)
0t T
2d
Am = (2m M + 1)d
m = 0,1,2, L, M 1

(55)

M PAM

PM =

2( M 1) 6(log 2 M ) avb
Q
( M 2 1) N
M
0

(56)

M=16 PAM
18 Monte Carlo Pe SNR N=10000
SNR 10 N=10000
Pe = 10 3 20
56 * Monte Carlo

6-41

21
4.2-1
2 = 0 2 = 0.1 2 = 1.0 2 = 2.0
4.2-2(Prob. 7.1, p. 453, Proakis, 7.1)
Determine the average energy of a set of M PAM signals of the form

s m (t ) = s m (t ),
where s m = E g Am ,

m = 1, 2,K, M

0t T

m = 1, 2,K , M . The signals are equally probable with

amplitudes that are symmetric about zero and are uniformly spaced with distance
d

0
d between adjacent amplitudes as shown in the above figure. The decision is made
in favor of the amplitude level that corresponds to the smallest distance.
Determine the probability of error for the optimum detector.
4.3 Multidimensional Signal Transmission
M = 2k
k = log 2 M 16
one-dimensional signal

M = 2k ab
s i (t ), i = 0,1,..., M 1

si (t ) sk (t )dt = E ik

i, k = 0,1,L, M 1

(57)

E ik Kronecker delta

ik

1
=
0

i = k

(58)

i k

22 4 (0, T )
6-42

E =

s i2 ( t ) dt =

A 2T
M

i = 0 ,1 , 2 , L , M 1

(59)

22 4
set M
s0 = ( E ,0,0,LL,0)
s1 = (0, E ,0,LL,0)
M
sM = (0,0,LL,0, E )

23 M = 2 M = 3

23 M=2 M=3

6-43

24 (Fig. 5.9, p. 327, Haykin)

s i (t ) AWGN
r (t ) = si (t ) + n(t )
i = 0,1,L, M 1
0t T
(60)
24 r (t )
M s0 (t )

r0 = s02 dt + n(t ) s0 (t )dt = E + n0

(61)

ri = s 2 (t ) si (t )dt + n(t ) si (t )dt

T

n(t ) si (t )dt = ni
T

i = 1,2,3,L, M 1

(62)

i2 = E ( ni2 ) =
N0
2
N0
=
2
=

Tb Tb

2
i

(63)

si (t ) si ( ) E[n(t )n( )]dtd

s (t )s ( ) (t )dtd
N
s (t )dt = 2
0

(64)

P ( r0 | s0 (t ) ) =
P ( ri | s0 (t ) ) =

1
2
1
2

e ( r0 E )
e ri / 2

/ 2 2

(65)
i = 1,2,L, M 1

(66)

s0 (t ) r0 > ri i = 1,2,L, M 1

(67)

Pc = P( r0 > r1 , r0 > r2 ,L r0 > rM 1 )

6-44

(68)

Pc|s0 f r0 |s0 ( r0 )dr0

(69)

Pc |s 0 = P ( r0 > n1 , r0 > n2 ,..., r0 > n M 1 | s0 )

(70)

= P ( r0 > n1 | s0 ) P ( r0 > n2 | s0 ) L P( r0 > n M 1 | s0 )

P( r0 > ni | s0 ) =

r0

(71)

f ri ( ri )dri , i = 1,2,..., M 1

r
= 1 Q 0

2r 2

0
= 1 Q
N

(72)

symbol

(73)

PM = 1 Pc = 1 P( r0 > r1 , r0 > r2 , L r0 > rM 1 )

r
= 1 1 Q 0

r
= 1 1 Q 0

y=

r0

( r0 E ) 2
2 2

PM =

1
= y
2

2E
N0

f r0

(74)

s0 ( r0 ) dr0

M 1

1
e

( r0 E ) 2

2 2 dr
0

(75)

{1 [1 Q( y )] }e
2
1

M 1

M 1

( y 2 E / N 0 ) 2 / 2

(76)

dy

M = 2 76
Eb

P2 = Q
N
0

(77)

76 symbol symbol
symbol

PM
P
= M
M 1 2k 1

(78)

M = 2 k , k k n

( ) k
k
n

symbol
k

k PM
2 k 1
=
k
PM
k
1
2k 1

n n 2
n =1

(79)

6-45

Pb =

2 k 1
PM
2k 1

(80)

(5.100)(5.101), p. 336, Haykin M=4

25 Monte Carlo Pe SNR
(0,1)
(0,0.25)(0.25,0.5)(0.5,0.75)(0.75,1.0) 0001
1110
00 s 0 = ( E ,0,0,0)

01 s1 = (0, E ,0,0)
10 s 2 = (0,0, E ,0)
11 s3 = (0,0,0, E )

si
si E 1 E = 2 Eb
Eb = 1 / 2 N=10000 SNR
10 N=10000

Pe = 10 3 27(76) (77)(80)
* Monte Carlo

25 Monte Carlo M=4

6-46

26
4.3-1
2 = 0.1 2 = 1.0 2 = 2.0
4.3-2 27

27
4.3-1: Bi-orthogonal signals
Bi-orthogonal signal
28 M=4

s 0 = ( E ,0,0, LL ,0)
s1 = (0, E ,0, LL ,0)
M
s M / 21 = (0,0, LL ,0, E )
s M / 2 = ( ,0,0, LL ,0)
M
s M 1 = (0,0, LL ,0, )
s i (t ) AWGN
0t T
r (t ) = si (t ) + n(t )
n(t ) N 0 / 2
r (t )
6-47

ri =

r(t )s (t )dt

i = 0,1, L ,

M
1
2

(81)

s 0 (t )
ri =

s (t )dt

i = 0,1, L ,

E + n0
=
ni
ni =

(82)

i=0
i0

n(t )s (t )dt

i = 0,1, L ,

M
1
2

M
1
2

(83)

2 = EN 0 / 2

| ri |

(84)

| r j |= max{| ri |}
i

28

M=4

r j > 0 s j (t ) r j < 0 s j (t ) s 0 (t )

Pc =

r0 / N 0 / 2

r0 / N 0 / 2

x2 / 2

dx

M 1

P (r0 )dr0

(85)

P(r0 ) =

e ( r )

/ 2 2

(86)

2
symbol
PM = 1 Pc

(87)

M=4 29 Monte
Carlo Pe SNR
(0,1)(0,0.25)(0.25,0.5)(0.5,0.75)
(0.75,1.0) 00011110
6-48

00 s0 = ( E ,0)

01 s1 = (0, E )
10 s 2 = (0, E )
11 s3 = ( E ,0)

s 2 = s1 s 3 = s 0
si
si E 1 E = 2 Eb
Eb = 1 / 2 N=10000 SNR
10 N=10000
Pe = 10 3 30(85) (86)(87)
* Monte Carlo

29 Monte Carlo M=4

30
4.3-3(Prob. 7.27, p. 462, Proakis, 7.37)

Consider an M-ary digital communication system where M = 2 N , and N is the

dimension of the signal space. Suppose that the M signal vectors lie on the
6-49

vertices of a hypercube that is centered at the origin, as illustrated in the following

figure. Determine the average probability of a symbol error as a function of
E s / N 0 where E s is the energy per symbol, N 0 / 2 is the power-spectral density
of the AWGN, and all signal points are equally probable.

2 (t )
s2

s1

N=2

1 (t )
s3

s4

2 (t )
s5

s3

s6

s4

N=3
s7

3 (t ) s
8

1 (t )
s2

s1

4.3-4 (Prob. 7.19, p. 460, Proakis, 7.28)

Three equally probable messages m1 , m 2 , and m3 are to be transmitted over an

AWGN channel with noise power-spectral density

N0
. The message are
2

1 0 t T
s1 (t ) =
0 otherwise

1 0t 2

T
s 2 ( t ) = s3 ( t ) = 1
0T
2

0 otherwise

1 ) What is the dimensionality of the signal space?

2 ) Find an appropriate basis for the signal space (Hint: You can find the basis
without using the Gram-Schmidt procedure).
3 ) Draw the signal constellation for this problem.
4 ) Derive and sketch the optimal decision regions R1 , R 2 , and R3 .
5 ) Which of the three messages is more vulnerable to errors and why? In other
words, which of p(Error m i transmitted), i=1, 2, 3 is larger?
6-50

4.3-5 (Prob. 7.20, p. 460, Proakis, 7.29)

An optimal demodulator can be realized as:

z A correlation-type demodulator
z A matched-filter-type demodulator
Where in both cases j (t ) , 1 j N were used for correlating r(t) or designing
the match filters. Show that an optimal demodulator for a general M-ary
communication system can also be designed based on correlating r(t) with s i (t ) ,
1 i M , or designing filters that are matched to s i (t )' s , 1 i M . Precisely
describe the structure of such receivers by giving their block diagram and all
relevant design parameters.
4.3-6 (Prob. 7.22, p. 460, Proakis, 7.31)

In an additive white Gaussian noise channel with a noise power-spectral density of

N0
, two equiprobable messages are transmitted by
2
At

s1 (t ) = T
0

0t T
otherwise

At

A
s 2 (t ) =
T
0

0t T
otherwise

1) Determine the structure of the optimal receiver.

2) Determine the probability of error.

6-51

5. Regenerative repeater (Fig 3.18, p.208, Haykin)

Repeaters for analog communication systems
Wireline Channel repeaters

repeaters AWGN
(bit error rate) Pb Pb SNR E b N o
channel attenuation
2.6 s (t ) 0< <1

r (t ) = s (t ) + n(t )
Transmitted
signal

Received
signal

channel

s(t)

r ( t ) = s( t ) + n ( t )

Noise
n(t)

Attenuation

Figure 2.6 Mathematical model of channel with attenuation and additive noise.

SNR

2 Eb

No

repeaters
(regenerate)(noise-free) regenerative repeaters
Wireline filter-optic communication channels

Regenerative repeater repeater

repeater
repeater repeater
repeaters
Binary PAM (hop, repeater repeater)

Pb = Q(

2 Eb
No

K regenerative repeaters
Pb KQ(

Eb
No

K analog repeaters

6-52

(2.53)

Pb Q(

Eb
KN o

(2.54)

On Board Processing(OBP) Wireline

Channel regenerative repeater

Ex. 7 (Ex.7.7.1, p.438, Proakis)

A binary digital communication system transmits data over a wireline channel of
length 1000km. Repeaters are used every 10km to offset the effect of channel
attenuation. Let us determine the

Eb
No

that is required to achieve a probability of a bit

error of 10 5 if (a) analog repeaters are employed , and (b) regenerative repeaters are
employed. The number of repeaters used in the system is K=100. If regenerative
repeaters are used, the

Eb
No

10 5 = 100Q (

obtained from (2.54) is

2 Eb
No

) 10 7 = Q(

2 Eb
No

which yields approximately 11.3 dB. If analog repeaters are used, the
from (2.53) is 10 5 = Q(

2 Eb
100 N o

) , which yields

Eb
No

Eb
No

obtained

29.6 dB. Hence, the difference in

the required SNR is about 18.3dB, or approximately 70 times the transmitter power of
the digital communication system.

(Prob. 7.55, p. 471, Proakis, 7.44)

Consider a transmission line channel that employs n-1 regenerative repeaters plus the
terminal receiver in the transmission of binary information. We assume that the
probability of error at the detector of each receiver is p and that errors among
repeaters are statistically independent.
(1) Show that the binary error probability at the terminal receiver is
1
Pn = 1 (1 p ) n
2

(2) IF p = 10 6 and n=100, determine an approximate value of Pn.

(Prob. 7.56, p. 471, Proakis, 7.45)
A digital communication system consists of a transmission line with 100 digital
(regenerative) repeaters. Binary antipodal signals are used for transmitting the
information. If the overall end-to-end error probability is p = 10 6 , determine the
probability of error for each repeater and the required Eb / N 0 to achieve this
performance in AWGN.
6-53