[MODULE - 2 |
Signaling Over AWGN Channels
Syllabus
Introduction, Geometric representation
AWGN channel into a vector ci
Of signals, Gram-Schmidt Orthogonalization procedure, Conversion of the continuous
*hannel, Optimum receivers using coherent detection: ML. Decoding, Correlation receiver, matched
2.2 Geometric Representation of Signals
2.3. Gram-Schmidt Orthogonalization Procedure .
[ERENCES ETI
OSCE
2.4 Conversion of the Continuous AWGN Channel into a Vector Channal
2.5 Optimum Receivers using Coherent Detection........sssseesssecseeeeseeeeeeeenees 2-16
2.6 1 Matched Fikersnecc cece siecle steers: clleinies eeindenioreieiesiteie siete tere
2.7. Properties of Matched Filter.
LETT
2.8 Correlation Receiver«When the signal is transmitted over a channel, it is
corrupted due to noise.
«The receiver then observes the received signal and
determines about which symbol was transmitted.
This decision is taken with the help of some set of
rules. This is called detection of the signal.
+The receiver can also use an information in the
received signal to extract estimates of physical
parameters or waveforms of interest. This is called
estimation of the signal.
Detection of the signal is done with the help of
correlators, optimum filter, matched filter etc.
+ Estimation can be done with the help of maximum
likelihood estimation, wiener filter, linear prediction,
adaptive filters etc.
+ Fig. 2.1.1 shows the model of digital communication
system.
“There is a message source, that emits mesages,
My, MyM. There are total 'M' number of
messages which are equally likely. Hence
probability of these messages is,
Hm) = 3
¢ The vector trasmitter clubs 'N' number of messages,
so that they can be transmitted together. Thus the
vector 5; is given as,
«The modulator constructs the waveform 5;(t) for the
symbol s,. The signal s;(t) has the finite energy ie,
T
E, = fo2(t)d, i-12..M
0
«The signal is transmitted over a noisy channel. j
adds additive white Gaussian noise of zero mean
(AWGN). It is denoted by N(t). ie,
X(t) = 5()+N() for O a(-sahtt) G8
Balt) = s2(8)-San% acneesten
‘The function is orthogonal to 4,(t) over the internal 0 to T. The second basis function
Salt) w+ (23.6)
a =
Vin
Here Exp = [7 93(t)dt is the energy of g(t)
1) To prove that $2(t) has unit energy
Energy of d9(#) will be Bea - fl (23.7)
‘Thus 6,(t) has unit energy.
Ii) To prove that ¢4(¢) and ¢2(¢) are orthonormal
Consider the intergration J, 63(1)¢2(H)a. By putting for moms ao,
KO) tina = Ce Pie fe EE fa 82(#)dt
Putting for go(t) from — (235),
Tox(t) 4 (t)dt = 754(0)[52(#) sary (t)]at = sy (8) 89 (t).dt— [spy 5,(t)0(#)at
lo a 0 aE 0 0
Putting for s2,(t) from equation (23.4) in above equation,
(dentdat—fF (oatenbattre ioe]
Bonioa - =fjn above equation observe that there is a Product of
two terms 8,(t) and so(t} We know that the two
bols are not present at a time. Hence the product
a(t) -82(#)=0. Therefore the intergration terms will be
zero and hence complete RHS will be zero. ie,
pao do (t)dt = 0 ves (2.3.8)
Thus the two basis functions are orthonormal.
tii) Generalized equations for orthonormal basis
functions
From equation (2.3.6). we can write the generalized
equation for orthonormal basis function $;(t) as,
od) = 82 i=1,2...N —... (2.3.9)
VE:
where g,(t) is given by generalizing equation (2.3.5)
ie,
i-1
gilt) = 5;(t) -— D 54 9; (4) sss (2.3.10)
jel
And the coefficients sy are given by equation 2.2.2
ie.
sj = fosilt) O)(0) dE f= DZone id
4
wes (23.11)Conversion of the Continuous AWGN
Channel into a Vector Channal
«Let us consider that the noisy received signal is
expressed by a random process X(t) which is given
as,
X(t) = 8,(t)+N(t) #=1,2)..,,M_ (24.1)
Here s;(t) is the transmitted signal and N(t) is white
Gaussian noise process of zero mean and power
spectral density of *
This received signal is applied to a bank of
correlators as shown in Fig. 2.4.1.
T
J a }—x,
0
oy) Ti
x(t) J dt xX,
O
Galt) 7
f a Pn
°
on(t)
Fig. 2.4.1 Response of correlators to noisy input
The output of each correlator is a random
variable which is given as,
T
Xj = FROM) dt, j=L2..N
0
T
= Jisi(t)+ NO] 4;(tdt
0
from equation 2.4.1T 2
= Jott e+] NO oat
0 a
way tp 12h 2 N24)
jom process be defined as,
x) (0) (24.3)
« Let the new rand
x(t) = X=
Putting equation 2.4.1 and equation 2.42 in
above equation,
X(t) = 540+N-¥ 55 +Nj)Oj(4)
jt
N N
= 54(0+ NO- D559" DN (0)
ja ja
= ne 3Nj 400)
jt
since $54 ¢)(0)" 548)
1
‘Thus X(t) depends only on the noise process N(#)
and not on the signal s,(#). Hence let X'(#) = N'(}
From equation (2.4.3) we can write,
y
Xi) = YX; (0+ XO
jt
NX
= Lx OHNO
jt
since x(t) = NO) (2.4.4)
Here N’(t)is totally due to noise process ‘N(t) at the
input of correlators.
«The correlator outputs X,,Xz,..-Xy are Gaussian
random variables. They are characterized completely
by mean and variance.
a TThe received signal of equation (24.4) can be
represented by the vector form as follows :
oe)
X(t) = [Xp Xqroe Xl 2). we
on (#),[2.5 | Optimum Receivers using Coherent
Detection VTU : July-13, 17
e The transmitted signal s,(t) is often corrupted by
noise when it is received. Such a received signal is
represented as,
x(t) = 5(t)+n(t) for 0 70. -¥0 (1) (2.62)
«Similarly, let x,(t) is transmitted, but x,)(T) is
greater than x9,(T). Then incorrect decision will be
taken if noise n(T) is less than
Flin T+ 202(T)+201(7) ie, error will be
generated if,
ma s aO+¥9 0, gy
1g (T) s ta 7x) ww (2.6.3)
The above two error conditions are summarized in
Table 2.6.1 :tal 07) ‘Probability. of error can
ee be obtained by
evaluating the i
probability that
2(T) — Xox(T)
t)| 2.8 | Correlation Receiver
VTU : Dec.-12, 14, 15, 16, July-
In this section we will study a little different type of
receiver which is called correlator. Fig. 2.8.1 shows the
block diagram of this correlator.
CC eT
Locally generated
signal
x(t)
| Integrator teT
marl bao
signal ° x
f(=x(t)+n(t) f(x(t) r(t)
Fig. -2.8.1 Block diagram of the correlatorjn the adjacent figure f(#)
ie, S=x()+n(0,
iiplied to the locally gener,
signal x(t) This result of multiplication F(t): x(t) is
integrated. The output of the integrator is sampled at
yeT (ie. end of one symbol period). Then baeed a
this sampled value, decision is made, This is how the
correlator works. It is called correlator since it correlates
the received signal f(t) with a stored replica of the known
signa 2() In Fig. 281, the product f(1)x(t) is
integrated over one symbol period, ie T. Hence
output r(#) can be written as,
"Presents input noisy
The signal f(t) is
‘ated replica of input
T
rt) = J fex(ae
0
Att=T, the above equation will be,
T
Output of correlator : (I) = ff (t) x(t) dt
0
w= (28.1)
Now let us consider the matched filter as shown in
Fig. 282 below.
* [Matcher] teT
Matched
or —— ler ae
AD hit)
Fig. 2.8.2 Block diagram of a matched fitier receiver
In the-above block diagram observe that the matched
filter does not need locally generated replica of input
filter can
signal x(1) The output of the matched Ker
cbiained by convolution of input f(t) and its impulse
Tesponse h (t)-ie.,
re) = fh)
= j f(a): h(t-1) at ww (2.8.2)
From equation (2.6.30) we know that impulse
Tesponse h(t) of the matched filter is given a5
we (2.8.3)
A(t) = Here) (283)h(a) = FE x(t-t+4)
0
ee this value of /(f-1) in equation (28.2) we
r= f Foyer H+ de
Since the integration is performed over one bit
period, we can change integration limits from 0 to T
ie,
T
r(t) = Ke) f@x(T-t+ oat
a
At t=T, the above equation will be,
r(T)
1
T
al f(a)x(T-T+2)at
0
_ okt
= wo} fx(oae
just for convenience of notation,
Let us put 7
——
{ oT
| Output of matched filter : r= | fOx eae
| Nog
|
o (28.4)
This equation gives the output of matched filter
Observe that this equation and equation 2.8.1 (which
gives output or correlator) are identical. In above
equation the constant X is present which can be
normalized to 1. The similarity between
equation (2.8.1) and equation (2.84) shows that the
matched filter and correlator gives same output.
‘Therefore we can state,
‘and correlator are two distinct,
1s which give the same result. These
ed to synthesize the optimum filter.
The matched filter
independent technique
two techniques are usQuadrature Receiver Using Correlation
* Necessity : When the signal is transmitted over
AWGN channel, there is uncertainty due to noise
over the channel. Additionally there is uncertaintydue to phase of the received signal. This uncertainty
arises due to signals arriving from different paths.
Such a received signal will have the form,
x(t) =
[FE cos(2n ft+0)+w(t) ~ (285)
Here 6 is the phase shift in received signal m(t) is
white Gaussian noise process of zero mean and
PSD of ao
*Due to uncertainty of phase, the conventional
noncoherent systems becomes useless because
output of correlator becomes function of phase 6
+ Principle involved in quadrature receiver : Two
correlation receivers are used which are supplied
with quadrature carriers. Their outputs are then
squared and added. This cancels the phase function
due to trigonometric identity.
+ Block diagram and operation : Fig. 28.3 shows the
block diagram of a correlation receiver using
correlators.
«The output of upper correlator becomes JEcos®
and that of lower correlator becomes VE sin @.
The above two outputs are squared and added as
shown in Fig. 2.8.3. Hence outpitt will be,
YE cos 0
T=]
ea
@~s
{feos xin
x]
‘Squarer
~E sine
[Fen cexsp
Fig, 2.8.3 Quadrature receiver
wt) = Ecos? 0+ Esin? 6
= Ecos? 0+sin? 6)=E
«Thus the dependence of unknown phase 6 is not
there in the output y(f) It depends only on energy
of the signal.
Review Questions
1. What is correlator ? What is the difference and
similarity between correlator and matched filter ?
VU: Dec.-12, Marks 5
2. Explain in brief about correlation receive.
ps sors
‘July-16, Dec.-16 Marks 8, May-17, Marks 5
3. Explain the coherent quadrature receiver using
correlators. CTT
000