Chapter 3: Optimum Receivers

for Additive Gaussian Noise

Channel
Goals
 Design & performance characteristics of optimum receiver
 Various modulation techniques
 AWGN channel
 Optimum: Minimize the probability of making errors

2
Overview
 Optimum receivers
 Correlation demodulator
 Matched filter demodulator
 Optimal detector
 Performance of optimum receiver (memoryless modulation)
 Comparison of digital modulation methods

3
 Optimum Receivers for AWGN Channel
 Consider the following receiver configuration
 Assume the channel does not introduce any changes or
disturbances to the modulated signal

Mod. signal Rcvd signal Output

Channel Demodulator Detector

4
Optimum Receivers for AWGN Channel
 The data can be recovered on a component-by-component
basis taking inner product of received signal and M basis
functions, such that

xk   x(t ) f k (t ) dt
T
where x(t ) is the modulated waveform
0

 This is the correlative demodulation

5
Optimum Receivers …
 The above integral can also be implemented by noting that

 x (t ) f
T

k (t ) dt  x (t ) * f k (T  t ) t T
0

 The component of the modulated waveform x(t) along the

kth basis function is the convolution of the waveform x(t)
with a filter whose impulse response is fk(T-t) at the output
sample time T
 This is the matched-filter demodulation

6
Optimum Receivers …
 The above two methods can accurately recover the
original components of the modulating signal if there is no
noise on the channel
 In reality, the received signal at the input of the
demodulator is corrupted by noise
 We consider the case where the channel noise is additive
White Gaussian (AWGN)

7
Optimum Receivers …
 A channel model for the received signal over an AWGN
channel is depicted below

Transmitted signal Received Signal

+
Sm(t) r(t)= Sm(t) + n(t)

Noise n(t)

r(t) = Sm(t) + n(t) 0≤t≤T

 n(t) is the sample function of the AWGN process with
power spectral density
Φnn(ω) = ½ N0 W/Hz

8
Optimum Receivers …
 Objective: Based on the observation of r(t) over the signal
interval, design a receiver that is optimum in the sense that it
minimizes the probability of error
 For the receiver configuration shown in the first block
diagram, the reception process may be divided into two
components; namely, signal demodulation and detection
 Signal demodulation: Converts the received waveform into
an N-dimensional vector
r = [ r1, r2, r3, …… rN]
 Detector: Decides which of the M possible signal
waveforms are transmitted based on the vector r

9
Overview
 Optimum receivers
 Correlation demodulator
 Matched filter demodulator
 Optimal detector
 Performance of optimum receiver (memoryless modulation)
 Comparison of digital modulation methods

10
Optimum Receivers - Correlation Demodulator
 Correlation demodulator: Decomposes r(t) into N-
dimensional vectors
 The signal and noise are expanded into a series of linearly
weighted orthonormal basis functions {fn(t)} which spans the
signal space so that all possible members of the signal set
{Sm(t), 1 ≤ m ≤ M} can be represented
 The basis functions do not span the noise space
 However, the noise terms that fall outside the signal space
are irrelevant to the detection of the required signal (see
text for proof)

11
Correlation Demodulator …

Correlation-type demodulator
12
Correlation Demodulator …
 The N correlators essentially compute the projection of r(t)
onto the N basis functions {fn(t)} such that

 r( t ) f  S ( t )  n ( t )  f k ( t ) dt
T T

k ( t ) dt  m
0 0

rk  s mk  n k k  1 ,2 ,........ N Where

S
T
s mk  m ( t ) f k ( t ) dt k  1 ,2 ,.... N and
0

 n( t ) f
T
nk  k ( t ) dt k  1 ,2 ,........ N
0

 The signal is now a vector smk = [ sm1, sm2, sm3, ……. smN]
whose values depend on which of the M possible signals
was transmitted
13
Correlation Demodulator …
 The components {nk} are random variables that arise from
the additive white Gaussian noise
 In the interval 0 ≤ t ≤ T

r(t)  smk f k (t)  nk (t) f k (t)  n' (t)  rk f k (t)  n' (t)
N N N

k 1 k 1 k 1

Where n (t)  n(t)  nk (t) f k (t)

N
'

k 1

 n’(t) is a zero-mean Gaussian noise process that

represents the difference between original noise process
n(t) and part corresponding to projection of n(t) onto {fn(t)}

14
Correlation Demodulator …
 Recall that n’(t) is irrelevant to the decision as to which
signal is transmitted and thus the decision can be based
on the correlator output signal and noise only
 I.e rk = smk + nk
 Note that the signal components are deterministic and the
noise components are Gaussian with zero-mean such that

E{nk }   E[ n (t )] f k (t ) dt  0
t
for all n
0

 And the covariances are

E{nk nm }    E[ n (t ) n ( )] f k (t ) f m (t ) dtd 
T T

0 0

 N 0    (t   ) f k (t ) f m (t ) dtd   N 0 mk
1 1
T T

2 0 0
2

15
Correlation Demodulator …
 Thus, {nk} are zero-mean, uncorrelated Gaussian random
variables (also independent) with common variance
σn2 = ½ N0
 Further, {rk} is also Gaussian with mean smk and the same
variance σr2 = σn2 = ½ N0
 The output {rk} conditioned on the mth signal being
transmitted are also statistically independent random
variables with probability density function given by

Pr s m    Prk smk 

N
m  1,2,....... N
k 1

16
Correlation Demodulator …
 Where

P rk smk  
1  ( rk  smk ) 2 
exp   ; k  1,2,....... N
N 0  N0 
 So that the joint conditional PDF is

P r s m   exp   
1  N ( rk  smk ) 2 

(N 0 ) N /2
 k 1 N 0 
 Example: Consider an M-ary baseband PAM signal set in
which the basic pulse shape is rectangular
g(t) = a for 0 ≤ t ≤ T and zero otherwise
PAM signal set is one dimensional (N=1) and thus there is
only one basis function, given by
17
Correlation Demodulator …
 1
1  0t T
f (t )  g (t )   T
a 2T  0 otherwise
 The output of the correlator type demodulator is
r   r (t ) f (t ) dt  
1
T T
r (t ) dt
0 T 0


1 N0
 [ s(t )  n(t )] dt  S m  n where E[n]  0 and  n 
T 2

 The probability density function of the sampled output is

thus given by
1  ( r  sm ) 2 
P ( r sm )  exp  
N 0  N 0 

18
Overview
 Optimum receivers
 Correlation demodulator
 Matched filter demodulator
 Optimal detector
 Performance of optimum receiver (memoryless modulation)
 Comparison of digital modulation methods

19
Matched Filter Demodulator
 Here we employ a bank of N-linear filters instead of the N
correlators to generate {rk}
 The impulse response of these filters is matched to the
basis functions such that
 f k (T  t ) 0t T
hk ( t )  
0 otherwise
 The output of these filters are given by

y k ( t )   r (  )hk ( t   ) d    r (  ) f k ( T  t   ) d 
t t
k  1,2 ,..... N
0 0

20
Matched Filter Demodulator …
 Sampling the output of the filters at t = T

y k (T )   r ( t ) f k ( ) d   rk
T
k  1, 2 ,.... N
0

 Hence, the samples outputs of the filter at time t = T are

the set of values {rk} obtained from the N linear correlators
 In general, for a signal S(t) and matched filter h(t)= S(T-t)

y (t )   S ( )S k (T  t   ) d 
t

21
Matched Filter Demodulator …

Matched filter
demodulator
22
Overview
 Optimum Receivers
 Correlation demodulator
 Matched filter demodulator
 Optimal detector
 Performance of optimum receiver (memoryless modulation)
 Comparison of digital modulation methods

23
Optimum Detector
 Output of the correlator or the matched filter demodulator
produces the vector r = [r1, r2,, r3,… rN ]

 This output vector contains all the relevant information

about the received signal waveform

 On the basis of the vector r, we need to make a decision on

what is transmitted such that the decision is optimal, in
some sense, assuming no memory in signals transmitted in
successive intervals

24
Optimum Detector …
 Optimality criterion: Decide on what is transmitted in each
interval based on the observation of the vector r in each
interval such that the probability of correct decision is
maximized (or the probability of error is minimized)
• Define the a posteriori probability as
P{signal Sm was transmitted} = P{sm│r}; m = 1,2,…M

25
Optimum Detector …
 Decision criterion: Select the signal corresponding to the
maximum of the set of a posteriori probabilities {P(sm│r)}
 The criterion maximizes the probability of correct decision and,
hence, minimizes the probability of error

 This decision criterion is called maximum a posteriori

probability (MAP) criterion

26
Optimum Detector …
 Note that:
P(r s m )P(s m )
P(s m r)  (Baye' s Theorem)
P(r)
 P{sm} is the apriori probability of the mth signal being
transmitted

P (r)   P (r sm ) P ( sm )
M
 Note also that:
m 1

 The computation of the a posteriori probability P(sm│r)

requires knowledge of the:
 Apriori probability P(sm) and
 Conditional pdf P(r|sm)

27
Optimum Detector …
 For the special case where M signals are equi-probable
such that the apriori probability P(sm) = 1/M for all M
 Further P(r) is independent of which signal is transmitted
 Then maximizing P(sm│r) is equivalent to finding the
signal that maximizes P(r|sm)

28
Optimum Detector …
 The conditional pdf P(r|sm), or any monotonic function of it,
is called the likelihood function
 The criterion of maximizing P(r|sm) is referred to as the
maximum likelihood (ML) criterion
 Note that the MAP and ML criteria are equivalent when the
apriori probabilities P(sm) are all equal
 I.e {sm} are equiprobable

29
Optimum Detector …
 For the AWGN channel, the likelihood function becomes
 1 
 exp  
  
N
( r s )
P (r s m )  
N 2

 N   m  1,2,.......M
k mk

 0   k 1 N0 

 Taking the logarithms of both sides simplifies the

computation

 k mk
1 1
ln P (r s m )   N ln(N 0 ) 
N
( r  S ) 2

2 N0 k 1

 Finding the maximum of ln (P(r|sm)) over sm is equivalent to

finding the signal sm that minimizes the Euclidean distance

D (r, s m )   ( rk  S mk ) 2
N
( Distance metrics)
k 1

30
Optimum Detector …
 The decision rule on the ML criterion is equivalent to
finding the signal sm that is closest in distance to the
received signal vector r
 Called the minimum distance detection
 Expanding the distance metrics

D(r, sm )  rk2  2rk smk  smk

N N N
2

k 1 k 1 k 1

 r  2 r .sm  sm m  1,2,.......M
2 2

 Since ||r|| 2 is common to all the decision metrics, and

hence it may be ignored in the computation of the metrics

31
Optimum Detector …
 One may use a modified distance metrics

D' (r,sm )  2 r . sm  sm m  1,2,....... M

2

 Note that minimizing D’(r,Sm) is equivalent to maximizing

 D' (r, s m )  C (r, s m )  2r.s m  s m

2

(correlation Metric)

 Note: r.sm represents the projection of the received signal

vector onto each of the M possible transmitted vectors
 Thus, it thus measures the correlation between the
received vector and the mth signal

32
Optimum Detector …
 Note also that ||sm ||2 = εsm can be thought of as a bias
term that serves as compensation for signal sets that have
unequal energies such as PAM
 If all signals have equal energy, it may be ignored in the
computation of the correlation metrics and the distance
metrics D and D’
 The correlation metrics can thus be computed as

C(r, sm )  2 r(t)sm (t) dt   sm

T
m  1,2,.....M
0

33
Optimum Detector …
 These metrics can be generated by a demodulator that
 Cross-correlates the received signal r(t) with each M possible
transmitted signal and
 Adjust each correlator output for the bias
 Select the signal corresponding to the largest correlation metrics

34
Optimum Detector …
 Note that for signals of unequal energies the output of the
correlators are adjusted by ½ εsm
 Alternatively, r(t) could be passed through a bank of N
matched filters and sampled at t=T
 For non-equiprobable signal we apply MAP based on the
probabilities P(sm│r); m = 1,2,…..M
 Or alternatively on the metrics
PM(r, sm )  P(r sm )P(sm )

35
Optimum Detector …
 Example: Consider a binary PAM with s1 = -s2 = √εb where
εb is the energy per bit
 Let the a priori probabilities be P(s1) = p and P(s2) = 1- p
 Question: Determine the metrics for an optimum MAP
detector when the transmitted signal is corrupted by
AWGN?
 Solution:
 The received signal vector (one-dimensional) for binary PAM is
r = √εb + yn(T)
 Where yn(T) is a zero-mean Gaussian random variable with
variance σ2 = 1/2 N0

36
Optimum Detector …
 Consequently, the conditional PDFs of p(rSm) for the two
signals is
1  (r   ) 2 
P(r s1 )  exp  b
;
2  2
 2 2


1  (r   b ) 
2

P ( r s2 )  exp  
2   2 2

 
2

PM (r , s1 )  p P(r s1 ) and PM (r , s2 )  (1  p) P(r s2 )

 If PM(r, S1) > PM(r, S2) then select S1 as the transmitted

signal; otherwise select S2

37
Optimum Detector …
 The decision rule may be expressed as
PM (r , s1 )
s1

1
PM (r , s2 )

s2

 Which upon substitution for PM(r, S1) and PM(r, S2) gives

(r   b ) 2  (r   b ) 2 s
(1  p)
ln or
1


2 2 p

s2

1 2 (1  p) 1 (1  p )
r b  ln  N 0 ln h
s
1


2 p 4 p

s2

38
Optimum Detector …
 For the PAM case, the optimum detector computes the
product r√εb and compares it with the threshold h
 If r√εb > h the s1 is transmitted or
 If r√εb < h then decides on s2

 Note that the threshold depends on p and N0

 For p = ½ h =0
 If p > ½ , then s1 is more probable and hence h < 0

 The above ideas can be generalized for any number of

equiprobable signals when using the ML or MAP criterion

39
Overview
 Optimum Receivers
 Performance of optimum receiver (memoryless modulation)
 Probability of Error for Binary Modulation
 Probability of Error for M-ary PSK
 Probability of Error for QAM
 Comparison of Digital Modulation Methods

40
Probability of Error for Binary Modulation
 Consider PAM signals s1(t) = g(t) and s2(t) = -g(t) where
g(t) is an arbitrary pulse which is non-zero in the interval
0 ≤ t ≤ Tb and zero elsewhere
 Energy of the pulse g(t) be εb and the signals may be
represented geometrically as
s2 s1

- √εb 0 + √εb

 When s1(t) is transmitted, the received signal from the

output of (the matched filter or correlator) demodulator
r  s1  n  Eb  n
 Note that when h = 0 and r > 0 → s1(t) and r < 0 → s2(t)

41
Probability of Error for Binary ….
 Conditional probability density functions of r are given by

1  ( r   b )2 
p ( r s1 )  exp   
N 0  N0 
1  ( r   b )2 
p( r s 2 )  exp   
N 0  N0 

Conditional PDFs of two signals

42
Probability of Error for Binary ….
 Given that s1(t) is transmitted, the probability of error is
given by P{r < 0}, i.e.,

P (e s1 )  P{r  0}
 (r   b ) 2 
exp  N 0  dr
1
0

N 0  
2 b

(r   b )

1 x
N0 x2
 

exp 2
dx where
2  N0 2
 2 b 
 exp
1

dx  Q 
x2




2
2 2 b  N0 
N0

43
Probability of Error for Binary ….
 Similarly, if s2(t) was transmitted, r = -√εb + n and
 2 b 
P(e s 2 )  Q 

 N0 

 When s1(t) & s2(t) are equally probable with probability ½,

the average probability of error will be
1 1  2 b 
Pb  P (e s1 )  P (e s 2 )  Q 

2 2  N0 

 Note that:
 Probability of error depends only on the ratio of εb/N0 and not on
any other characteristics of the signals
 εb/N0 = signal-to-noise ratio (SNR) per bit
 2 εb/N0 = SNRmax from the matched filter or correlator demodulator

44
Probability of Error for Binary ….
 Probability of error may also be expressed in terms of the
distance between the two signals s1 and s2 such that
d12 = 2√εb where εb = ¼ d212
 d2   1 
Pb  Q 12   Q d
 12 2 N


 2N 0 
   0 

45
Probability of Error for Binary ….

Probability of error for binary signals

46
Overview
 Optimum Receivers
 Performance of the optimum receiver (Memoryless
Modulation)
 Probability of Error for Binary Modulation
 Probability of Error for M-ary PSK
 Probability of Error for QAM
 Comparison of Digital Modulation Methods

47
Probability of Error for M-ary PSK
 We had earlier seen that digitally phase modulated signal
waveforms may be expressed as
 2 
sm (t )  g (t ) cos2f c t  (m  1) 1 m  M
 M 
 Which has a vector representation as
 2 2 
sm    s cos (m  1),  s sin (m  1)
 M M 
 Where s= ½ g is the energy in each of the waveforms and
g(t) is the pulse shape of the transmitted signals

48
Probability of Error for M-ary PSK ….
 Note: The signal waveforms have equal energies
 The optimum detector for an AWGN channel simply
computes the correlation metrics
C(r, s m )  r  s m m  1,2,......., M
 Accordingly, the received signal r = [r1, r2 ] is projected
onto each of the M possible signal vectors & the decision
is made in favor of the signal with the highest projection
 Alternatively, a phase detector that computes the phase of
the received signal from r and selects the signal vector sm
whose phase is closest to that of r
 Phase of r is   tan  rr 
1 2

 
r
1

49
Probability of Error for M-ary PSK ….
 Now compute the PDF of Θr from which the probability of
error may be computed
 Assume that the transmitted signal is s1(t) having a phase
of Θr = 0 such that

s1  [  s , 0] and r1  [  s  n1 ]; r2  n2
 n1 and n2 are jointly normal each with zero mean and
variance ½ N0
 r1 and r2 are also jointly normal with means E{r1} = √εs and
E{r2} = 0 and equal variances σr2 = ½ N0

1  (r1   s )2  r22 
 Hence, p(r1, r2 )  exp  
2r  2 
 
2 2
r

50
Probability of Error for M-ary PSK ….
 The PDF of Θr can be obtained by change of variables
from (r1, r2) to
 r2 
V  r12  r22 and  r  tan 1
 
 r1 

 Such that
1  (V   s  2  s V cos  r 
exp   
2
p (V ,  r ) 
2   2 r 
 
2 2
r

 Integrating this joint PDF over the range of values of V

yields the marginal PDF of Θr
2
 V  2  s cos  r 

 
1   s sin 2  r 
  

p(  r )  p( V ,  r )dV 
 
e Ve dV
 2 

0
2 0

 Where s =s/NO is the symbol SNR

51
Probability of Error for M-ary PSK ….
 The figure shows plot of
the probability density
function of Θr for different
values of s
 When s1(t) is transmitted,
a decision error is made if
the noise causes the
phase to fall outside
–π/M ≤ Θr ≤ π/M
 Symbol error probability is


 p( )d
M
PM  1  r r

M

52
Probability of Error for M-ary PSK ….
 In general, the integral p(Θr) does not reduce to a simple
form and evaluated numerically (except for M=2 or 4)
 For M=2, binary phase modulation, the signals s1(t) & s2(t)
are antipodal and hence the probability of error is given by
 2 b 
P2  Q  

 N0 

 For M=4, we have two binary phase modulation signals in

phase quadrature and since there is no cross talk between
the signals on the quadrature carriers, the bit error
probability is identical to the one given for M=2 above

53
Probability of Error for M-ary PSK ….
 On the other hand the symbol error is different
 The probability of correct decision for the 2-bit symbol is
Pc = (1 – P2)2
  2 b 
2

Pc  (1  P2 ) 2  1  Q 

  N0 

 And the probability of symbol error

 2 b   1  2 b 
P4  1  Pc  2Q  1  Q
 


 N0   2  N 0 

54
Probability of Error for M-ary PSK ….
 An approximation to the error probability for large values of
M and large SNR can be made when εs/N0 >>1 and
│Θr│≤1/2π such that
s
p ( r )  cos  r e  s sin r
2


 Substituting this approximation, the symbol error PM can be
expressed as


s

M
PM  1  cos  r e  sin 2  r
d r

s


M

     
 2Q 2 s sin   2Q 2k b sin 
 M  M

 Where k=log2M and s = kb

55
Probability of Error for M-ary PSK ….

Probability of a symbol
error for PSK Signals

56
Overview
 Optimum Receivers
 Performance of the optimum receiver (Memoryless
Modulation)
 Probability of Error for Binary Modulation
 Probability of Error for M-ary PSK
 Probability of Error for QAM
 Comparison of Digital Modulation Methods

57
Probability of Error for QAM
 QAM signal waveforms can in general be expressed as
sm (t )  Amc g (t ) cos 2f c t  Ams g (t ) sin 2f c t
 Where Amc and Ams are amplitudes of the information
bearing quadrature carriers and g(t) is the signal pulse
 The vector representation of these waveforms is

 1 1 
S m   Amc  g , Ams g 
 2 2 

 The probability of error depends on how the signal

constellations are arranged as we will demonstration below

58
Probability of Error for QAM ….
 Consider a QAM signal set for M=4 arranged in the two
different ways as shown below
 Note that the distance between any two signal points is
constrained to be the same such that dmin=2A
 Recall that the probability of error is essentially determined
by the minimum distance between pairs of signals
A2
d=(A12+A22)1/2
√2A d=2A
A1

4-phase modulated signals QAM signals with two amplitude levels

59
Probability of Error for QAM ….
 If the signal points are equally probable, the average
transmitter power for the four-phase modulation case is
1
Pav  x 4 x 2 A 2  2A 2
4
 For the QAM, the points must be placed on circles of radii
A1=A and A2=√3A such that dmin=√(A2 + 3A2) =2A, so that
1
Pav  x[ 2(3A 2 )  2A 2 ]  2A 2
4
 Thus for all practical purposes for the same average
power, the error rate for the two signal sets are the same
 There is no advantage in using the two amplitude QAM
signal set over the M = 4-phase modulated signals

60
Probability of Error for QAM ….
 Consider the case for M = 8 QAM shown below with
 Four different constellations
 All having two amplitudes levels and
 A minimum distance between signal points of 2 (Normalized by A)

C  2 2

Four eight-point QAM

signal constellations

61
Probability of Error for QAM ….
 Assuming the signal points are equiprobable, the average
transmitted signal power is given by

  (a
1 M
A2 M
Pav  2
(A  A ) 
2 2
 a ms
2
)
M M
mc ms mc
m 1 m 1

 Where (amc ,ams) are the coordinates of the signal points

normalized by A
 For the signal constellations shown in Figures (a) and (c)
Pav = A2/8 [4x2 +4x10] = A2/8[4x4 + 4x8} = 6A2 (7.78 dB)
 For those in Figure (b)
Pav = A2/8 [4x2 + 4x(2 + √2)] = 6.83A2 (8.34 dB)
 And for those in Figure (d)
Pav = A2/8 [4x2 + 4(1 + √3)2] = 4.732A2 (6.75 dB)
62
Probability of Error for QAM ….
 The signal set in (d) requires approximately 1dB less that
sets in (a) and (c) and 1.6 dB less than the signal set in (b)
 The signal constellation in (d) is the best QAM constellation
since it requires least power for a given minimum distance
between signal points
 For M ≥ 16, there are many possibilities for selecting the
two dimensional signal space
 However, multi-amplitude circular constellation is not necessarily
the best for AWGN channel

 Rectangular QAM signal constellations are advantageous

as easily being generated as two PAM signals impressed
on phase-quadrature carriers & can be easily demodulated
 Thus, rectangular M-ary QAM signals are most frequently used

63
Probability of Error for QAM ….
 For rectangular constellation in which M = 2k, where k is
even, the QAM signals are equivalent to two PAM signals
on quadrature carriers each having √M =2k/2 signal points
 The probability of error for QAM can be determined from the
probability of error for PAM
 The probability of correct decision for the M-ary QAM
system is thus given by
Pc = (1 - P√M)2
where P√M is probability of error of √M-ary PAM with half the
average power in each quadrature signal of equivalent QAM

1  3 av 
P  2 (1  ) Q  Where  /N = average SNR per symbol

M
M  (M  1) N 0 
av O

64
Probability of Error for QAM ….
 Probability of symbol error for the M-ary QAM is then
PM = 1 – (1 - P√M)2
 For k odd, there is no equivalent expression, but a tight
upper bound can be established as
  3  av 
2

PM  1  1  2 Q  

  ( M  1 )N 0  
 3 k  
 4 Q  bav ; k
 1
 ( M  1 ) N 0 
 Where bav/NO = average SNR per bit

65
Probability of Error for QAM ….

Probability of a symbol error for QAM

66
Overview
 Optimum receivers
 Performance of the optimum receiver (Memoryless
Modulation)
 Comparison of digital modulation methods

67
Modulation – Summary of BER and SER in AWGN
 Summary of probability of bit and symbol error formulas

68
Modulation – Summary …
 Let us compare the performance of QAM with that of PSK
for a given signal size M
 Since both are signal types having two dimensions
 For M-ary PSK
  
PM  2 Q  2 s sin 
 M 
 Since the error probabilities are dominated by the Q-
function it would be sufficient to compare the arguments of
the Q-function for the two signal formats
 Ratio of the arguments give
3
( M  1)
Rm 

2 sin 2

69
Probability of Error for QAM ….
 M-ary QAM yields better performance for M > 4 as the
following values of Rm indicates

M Rm (dB)
4 1.00
8 1.65
16 4.20
32 7.02
64 9.95

70
M-PSK – Performance over AWGN Channel …
 Error rate degrades as
M increases
 Spectral efficiency
increases as M
increases
 Note: at 16dB, error
rate for 16-PSK is 10-4

71
M-QAM – Performance over AWGN Channel
 Power efficiency
decreases with
increasing M, but
not early as fast as
M-PSK
 Note: at 12.5, error
rate for 16-QAM
is 10-4
 Spectral efficiency
increases as M
increases

72
Tradeoffs between BER, power and bandwidth
1. Trade BER performance
for power – fixed data rate
2. Trade data rate for power
– fixed BER
3. Trade BER for data rate –
fixed power

73
Performance Comparison of Various Modulation
 Comparison of Various Digital Modulation Schemes
(BER = 10-6)

74
Modulation Schemes in Current Systems

LTE QPSK, 16 QAM, 64 QAM

75
Modulation Schemes in Current Systems
 Adaptive modulation:
 A modulation technique is selected based on the measured signal
to interference plus noise ratio (SINR)
 Each modulation scheme has a threshold SINR
 Subscribers located farther from an access point or base station
(i.e. with lower SINR values) use a more robust modulation
scheme (lower throughput)
 While subscribers closer to the from an access point or base
station (i.e. with higher SINR values) can use less robust
modulation schemes (higher throughput)
 Hence, there is a need to measure signal quality using the
Reference Signals

76