Modulation and Demodulation

Analog modulation
AM
PM/FM

Digital modulation
Binary Modulation
Quadrature Modulation

Power Efficiency

Noncoherent Detection
1/31/2005
1
 Modulation
• Important fact: a continuous, oscillating signal will
propagate farther than other signals.
• Start with a carrier signal
– usually a sine wave that oscillates continuously.
– Frequency of carrier fixed

1/31/2005
2
 Carrier Signal
• In analog transmission, the sending device
produces a high-frequency signal that acts as
a basis for the information signal.
– This base signal is called the carrier signal or
carrier frequency
• The receiving device is tuned to the frequency
of the carrier that it expects from the sender.

1/31/2005
3
 Modulation
• Signal information is modulated on the carrier
signal by modifying one of its characteristics
(amplitude, frequency, phase).
• This modification is called modulation
• Same idea as in radio, TV transmission
• The information signal is called a modulating
signal.

1/31/2005
4
 Modulation
• Modulation: process of changing a carrier
wave to encode information.
• Modulation used with all types of media
• Why is modulation needed?
– Allows data to be sent at a frequency which is available
– Allows a strong carrier signal to carry a weak data signal
– Reduces effects of noise and interference

1/31/2005
5
 Types of modulation
• Amplitude modulation (used in AM radio) –
strength, or amplitude of carrier is modulated
to encode data
• Frequency modulation (used in FM radio) –
frequency of carrier is modulated to encode
data
• Phase shift modulation (used for data) –
changes in timing, or phase shifts encode data

1/31/2005
6
 Analog modulation

1/31/2005
7
 Amplitude Modulation

1/31/2005
8
 Amplitude modulation

DSB-SC
AM

Signal: m(t)
Modulated signal: A(t)=Acm(t), 1-1 correspondence to m(t)
Carrier: Cos ωct

6carrier
47 4 8
General form : x c ( t ) = A{ (t) × cos ω c t
{
modulated signal fixed
.
1/31/2005
9
 Doubled-Sideband Modulation
(Suppressed Carrier): DSB-SC
6carrier
47 4 8
General form : x c ( t ) = A{ (t) × cos ω c t
{
modulated signal fixed
.
1 1 ωc
X c( f ) = A c 1
M4(4f2+4 4f c3) + A c 1
M4(4f2−4 4f c3) , f c =
2 2 2π
1 4 4 4 4 4 4 442 4 4 4 4 4 4 4 43
tra n sla tio n o f M ( f )

M(f)
m(t)

t f
-W 0 W

AcM(f+f c)/2 AcM(f-f c)/2

t f
- fc fc-W fc fc+W
1/31/2005
phase reversal 10
 Demodulation
Coherent (Synchronous) Demodulator (Detector):
The receiver knows exactly the phase and frequency of the
received signal.
m(t) xc(t) x c(t) d(t) yD(t)
LPF

2cosωct
cosωct

Modulator Demodulator

xc(t) = Acm(t)cosωct
d(t) = [Acm(t)cosωct] 2cosωct
= Acm(t) Message we want!+ 2Ac m(t)cosωct
1/31/2005 y1(t)= Acm(t)
11
 2
1
X(t) d(t)
0.5
1

-4 -2 2 4 -4 -2 2 4
-0.5 -1

-1
-2

Xc(f)

f
0
-fc fc

D(f)

f
0
-2fc 2fc

1/31/2005
12
 Power Analysis
<xc2(t)>=<[A+m(t)]2 (Ac)2 cos2ωc t>
Assume m(t) varies slowly w.r.t. cos2ωct

=<(1/2)(Ac)2 [A +m(t)] 2 >+<(Ac)2 [A+m(t)]2 cos2ωct>

Since <cosX>=0 :
=(1/2)(Ac)2< [A2+2A<m(t) +<m2(t)>]

Assume <m(t)>=0:
= (1/2)(Ac)2 [ A2 + <m(t) 2>]
Carrier dc bias power of m(t),
power power signal power

1/31/2005
13
Efficiency E: the percentage of total power that conveys
information.
Signal Power
E=
Total Power
[1 / 2 ( Ac 2 )] < m ( t ) >
2

EAM = 2 2
(100 %)
[1/2(Ac )][A + < m ( t ) > ]
2

a < mn ( t ) >
2 2

= ( 100 %)
1 + a < mn ( t ) >
2 2

2
< m (t) >
= 2 2
(100 %)
1/31/2005 A + < m (t) >
14
 Frequency Modulation
• Frequency of the carrier signal is modulated to
follow the changing voltage level (amplitude)
of the modulating signal.
• Peak amplitude and phase of the carrier signal
remain constant, but as the amplitude of the
info signal changes, the frequency of the
carrier changes accordingly.

1/31/2005
15
 Frequency Modulation

1/31/2005
16
 Phase-Shift Modulation
• vary phase of carrier
• may use more than simply 180 degree shift
(binary)
• this allows higher bit rate than baud rate
– Eight angles results in 3 bits per signal element. Or 3
bits per baud!

1/31/2005
17
 Angle Modulation: FM and PM
General form：xc(t)=Accos[ωct+φ(t)]
Phase modulation: φ(t)=Kpm(t), Kp: deviation constant
Frequency modulation : d φ(t)/dt = Kf.m(t)
Φ ( t ) = Kf ∫ m(α )dα + Φo
t
Kf: deviation constant; t o

fd: Frequency deviation constant = 2π fd t m(α )dα + Φo

∫to

 PM : x c ( t) = A ccos [ω ct + k pm ( t )]


 FM : x c (t) = A ccos [ω ct + 2πfd ∫to m (α ) d α ]
t

1/31/2005
18
 FM/PM waveforms

1
fc =
2π LC

xBB controls the value of C

1/31/2005
19
 Indirect implementation (Armstrong)
using a mixer and summer

xc (t ) = Ac (cos ωc t − φ (t ) sin ωc t )
If φ(t) very small: xc (t ) = Ac cos(ωc t + φ (t ))

1/31/2005
20
 Direct method using PLL
m(t)

Crystal LPF
∑
oscillator

Frequency Vc(t)
Divider VCO
÷N
xc(t)

1/31/2005
Vc controls frequency, so is m(t)
21
 FM/PM spectrum

Ideally

Reality

1/31/2005
22
 Theorem
• If |Φ(t)|<<1,

X c ( f ) = {[δ ( f − f c ) + δ ( f + f c )] + j[Φ ( f − f c ) − Φ ( f + f c )]}

Ac
2

K f max[m(t )]
• Or if >1 where B is the bandwidth of m
2πB

πAc2  2π 2π 
Xc( f ) = M ( ( f − fc ) + M ( (− f − f c )
2K f  Kf Kf 

1/31/2005
23
 Digital Modulation

1/31/2005
24
 Design Parameters
• Power efficiency : describes the ability of a
modulation technique to preserve the fidelity of the
digital message at low power levels.

– ηP : Eb/N0

• Bandwidth efficiency: describes the ability of a

modulation scheme to accommodate data within a
limited bandwidth.

– ηB : R/B bps/Hz

1/31/2005
25
 Channel capacity formula
• Channel capacity formula
– Rmax <= C = Blog2(1+S/N)
– ηBmax = C/B = log2(1+S/N)
• C is the channel capacity (in bps)
• B is the RF bandwidth
• S/N is the signal-to-noise ratio

1/31/2005
26
 Bandwidth
• Absolute bandwidth : The range of frequencies over
which the signal has a ‘non-zero’ power spectral density.

• Null-to-null bandwidth : equal to the width of main

spectral lobe.

• Half-power ( 3-dB) bandwidth : the interval between

frequencies at which the PSD has dropped to half power,
or 3dB below the peak value.

• 99 percent bandwidth (by Federal Communication

Commission) : occupied 99 percent of signal power.

1/31/2005
27
 General Digital Modulation
• Geometric Representation of Modulation
– set of modulation signal :
S = {s1 ( t ), s 2 ( t ),....... s M ( t )}
– vector representation with orthogonal basis
functions :
N

– si ( t ) = ∑j =1
s ij Φ j ( t )

– ∫ Φ (t )Φ
−∞
i j ( t ) dt = 0 i≠ j
∞
E = ∫ Φ ( t ) dt = 1
2
i
1/31/2005
−∞
28
 Geometric Representation of Modulation
• BPSK: 2 Eb
s1 ( t ) = cos( 2π f c t ) 0 ≤ t ≤ Tb
Tb
2 Eb
s2 ( t ) = − cos( 2π f c t ) 0 ≤ t ≤ Tb
Tb

Q
2
Φ 1 (t ) = cos( 2π f c t ) 0 ≤ t ≤ Tb
Tb

{ } − Eb Eb I
S BPSK = E b Φ 1 ( t ), − E b Φ 1(t )

1/31/2005
29
Geometric Representation of Modulation
• Bit error analysis: in AWGN channel with a
noise spectral density No.
 d ij 
Ps ( ε | s i ) ≤ ∑ Q  
 2N 
j =1, j ≠ i  0 
dij: distance between the ith and jth signal point in
the constellation
∞
1
Q( x) = ∫
x 2π
exp( − x 2 / 2 ) dx

1 M x
Ps (ε ) = Ps (ε | si ) P ( si ) =
M
∑ P (ε | s )
i =1
s i
1/31/2005
30
 Linear Modulation Techniques
• Transmitted signal amp. is proportional to modulated signal
• Good spectral efficiency
• Bad power efficiency since linear AMP is needed.
• E.g. QPSK, OQPSK, π/4QPSK…

s ( t ) = Re[ Am ( t ) exp( j 2 π f c t )
= A [ m R ( t ) cos( 2 π f c t ) − m I ( t ) sin( 2 π f c t )]

1/31/2005
31
 Linear Modulation Techniques
• BPSK Transmitter
Bit 1: A XBPSK(t)
Bit 0: -A xin

cos(2πfct+θc)
2Eb
s BPSK ( t ) = cos( 2 π f c t + θ c ) ( bit 1 )
Tb
2Eb
s BPSK ( t ) = − cos( 2 π f c t + θ c ) ( bit 0 )
Tb
2
 sin π fT b 
PBPSK = P p ( f + f c ) + P p ( f − f c ), P p ( f ) = A c2 T b  
 π fT b 
1/31/2005
32
 Demodulation Techniques
Notch
filter xin
XBPSK(t)
2fc
2cos(2πfct+θc)

xin

1/31/2005
33
 Linear Modulation Techniques
• Waveform of BPSK after adding noise

1/31/2005
34
 Detection with Matched Filter:

Filter impulse response is a square pulse,

1/31/2005
matching the pulse shapes in xin 35
 y (t ) = p ( t ) ∗ h (t ) no (t ) = p (t ) ∗ n ( t )
+∞
= ∫−∞
p (t − τ ) h (τ ) d τ , n *t) =
2
o ∫
+∞

−∞
2
H ( f ) N ( f ) df
+∞
EP = ∫
2
p(t ) dt n = No / 2
2
−∞ o

2E p
SNRmax =
N0
1/31/2005
36
Optimum Detection:

+∞
y(Tb ) = ∫ x(τ )h(Tb − τ )dτ
−∞
+∞
= ∫ x (τ ) p (τ )dτ ,
−∞

τ =Tb
y (Tb ) = ∫ x(τ ) p(τ )dτ
τ =0
1/31/2005
37
 Demodulation
• BPSK Coherent Receiver :
2Eb
s BPSK ( t ) = m ( t ) cos( 2 π f c t + θ c + θ ch )
Tb
2Eb
= m (t )
cos( 2 π f c t + θ )
Tb
Using Carrier Recovery to get carrier frequency:

2 m (t )
2 Eb
cos ( 2πf c t + θ ) = m (t )
2 2 Eb
[1 + 1 cos 2( 2πf c t + θ c )]
Tb Tb

 2Eb 
Bit Error Rate : P e , BPSK = Q  

1/31/2005  N 0 
38
 • BPSK Coherent Receiver Architecture
Symbol/
clock
recovery
Signal
Pulse
Carrier
generator
recovery
Detect
decision
cos(2πfct+φ)

Received Matched
AGC sampler
signal filter

1/31/2005
39
 Linear Modulation Techniques
• From BPSK to QPSK Q
Constellation diagram

2 Eb I
Q Q
Eb

I 2E b I
1/31/2005
40
 Linear Modulation Techniques
• Constellation diagram after adding noise

1/31/2005
41
 QPSK transmitter
2 Eb π
s QPSK (t ) = cos[ 2πf c t + (i − 1) ]
Ts 2
0 ≤ t ≤ Ts i = 1,2,3,4
2 Eb π
s QPSK (t ) = cos[( i − 1) ] cos( 2πf c t )
Ts 2
2 Eb π
− sin[( i − 1) ] sin( 2π f c t )
Ts 2
 π π 
s QPSK (t ) =  E b cos[( i − 1) ]φ1 (t ) − E b sin[( i − 1) ]φ 2 (t ) 
 2 2 
2
 sin 2 π fT b 
PBPSK = P p ( f + f c ) + P p ( f − f c ), P p ( f ) = 2 CT b  
 2 π fT b 
1/31/2005 C is average magnitude squared
42
 QPSK transmitter

Pulse
shaping
an cos(2πfct+φ)
s(t)
bit 2 bits to Map to Local +
stream 4 levels constellation Oscillator
90o
bn -sin(2πfct+φ)
Pulse
shaping

1/31/2005
43
 QPSK Receiver

 2Eb 
Bit Error Rate : Pe , QPSK = Q  

 N0 
1/31/2005
44
 BER & BW efficiency comparison
• BPSK v.s QPSK
– BER is the same !!!

 2Eb 
Bit Error Rate : Pe = Q  

 N0 

But QPSK only need ½ the bandwidth of BPSK !!!

BPSK: symbol rate = bit rate

QPSK: symbol rate = ½ bit rate
1/31/2005
45
 Reducing the maximum phase jump
• Offset QPSK
– To overcome 180° phase shift in QPSK thus
prevent the overhead to overcome signal
envelop distortion caused in QPSK

1/31/2005
46
 Reducing the maximum phase jump
• Offset QPSK Tx

1/31/2005
47
 Reducing the maximum phase jump
• Offset QPSK Rx

1/31/2005
48
 Further reduction of phase jump
• π/4-QPSK
– Alternate between the 2 QPSK constellations
– If the current two bits corresponds to a point in
the left constellation
– Next two bits will be represented by a point on
the right constellation
– Vice versa

– Also called π/4-DQPSK

1/31/2005
49
 Concerns over linear modulation
• On each of the I/Q axis, we are performing AM
modulation
• Transmitted signal amplitude changes with
time
– Cause receiver challenges
– Transmitter power utilization
– Sensitive to additive noise

èuse nonlinear modulation

1/31/2005
50
 Constant Envelope Modulation
• Class C Amp. can be used
– saving power
• Limiter-discriminator detection can be used
– easy and simple architecture
• Good performance against random noise
and signal fluctuation due to Rayleigh
fading
– good performance
• But! BW is larger than linear modulation
1/31/2005
51
 Binary frequency shift keying
• BFSK
– General Form
2 Eb
s FSK ( t ) = υ H ( t ) = cos( 2π f c + 2π ∆ f ) t 0 ≤ t ≤ Tb (bit 1)
Tb
2 Eb
s FSK ( t ) = υ L ( t ) = cos( 2π f c − 2π ∆ f ) t 0 ≤ t ≤ Tb (bit 0)
Tb
– Discontinuous phase FSK
2 Eb
s FSK ( t ) = υ H ( t ) = cos( 2 π f H t + θ 1 ) 0 ≤ t ≤ Tb (bit 1)
Tb
2Eb
s FSK ( t ) = υ L ( t ) = cos( 2 π f L t − θ 1 ) 0 ≤ t ≤ T b (bit 0)
Tb
1/31/2005
52
 Discontinuous phase FSK
• Discontinuous phase FSK
– can be combined by two OOK
– cause spectral spreading and spurious

FSK

OOKH

OOKL
1/31/2005
53
 Continuous phase FSK
• Continuous phase FSK
– similar to FM except that m(t) is binary
2 Eb
s FSK ( t ) = cos( 2π f c t + θ ( t ))
Tb
t
2 Eb
= cos[ 2π f c t + 2π f d ∫ m (η ) d η ]
Tb −∞

=
2 Eb
[cos 2π f c t ⋅ cos θ (t ) − sin 2π f c t ⋅ sin θ ( t ) ]
Tb

1/31/2005
54
 Binary frequency shift keying
• Coherent detection of BFSK

 E 
P e , FSK = Q  b 

 N 0 
1/31/2005
55
 Binary frequency shift keying
• Non coherent detection of BFSK

1  Eb 
Pe , FSK, NC = exp  − 
1/31/2005
2  2N0 
56
 Modulation Index
• Modulation Index of FSK:
(2∆F )
h= ,
Rb
where ∆ F is the peak RF frequency deviation
and R b is the bits rate

Example: 1
∆F = f H − fc = Rb
4
1
2× Rb
⇒ h = 4 = 0 .5
1/31/2005
Rb
57
 CPFSK and modulation index
Q Q
h=¼ h = 1/3

I I

Q Q
h=½ h = 2/3

I I

1/31/2005
58
 Minimum shift keying
– a special type of Continuous phase FSK
– modulation index h = 0 . 5
– peak RF frequency deviation=Rb/4
T
– coherently orthogonal. i.e. υ ( t )υ ( t ) dt = 0
b

∫ H
0
L

– MSK=fast FSK
– MSK=OQPSK with baseband rectangular being
replaced with half-sinusoidal
– MSK = FSK with binary signaling freq. of fc±1/4Tb

1/31/2005
59
 Minimum shift keying

• Advantage of MSK: particularly attractive

for use in mobile radio communication
systems:
– constant envelope
– spectral efficiency ?
– good BER
– self-synchronizing capability

1/31/2005
60
 Minimum shift keying
• MSK as OQPSK:
πt πt
s MSK ( t ) = m I ( t ) cos( ) cos( 2π f c t ) + m Q ( t ) sin( ) sin( 2π f c t )
2Tb 2Tb
Pulse shape: half period cos
• MSK as CPFSK :
 πt 
s MSK ( t ) = cos  2 π f c t − m I ( t ) m Q ( t ) + φk 
 2Tb 
• MSK power spectrum
2
16 A 2
 cos 2 π fT b 
PMSK = P p ( f + f c ) + P p ( f − f c ), Pp ( f ) = c
 
π 2  1 − 16 f 2
T b
2

1/31/2005
61
 Minimum shift keying
• MSK Transceiver

 
xMSK (t ) = 2 Ac cos ωct + ∫ ∑ bm p (t − mTb )dt 
t
1/31/2005
−∞
 m  62
 Minimum shift keying

MSK
1/31/2005
63
 Comparison of MSK, OQPSK, QPSK
90o phase
shift in
each
symbol

1/31/2005
64
 MSK
• Receiver of MSK

1/31/2005
65
 Why is MSK more spectrally efficient?
• Power spectral density of MSK
– 99% BW of MSK = 1.2/Tb
– 99% BW of QPSK or OQPSK = 8/Tb

1/31/2005
66
 Even better spectral efficiency?
Use Gaussian MSK

2
f ln 2
H ( f ) = exp(− 2
B 2

1/31/2005
67
• GMSK & GFSK
– GMSK reduces the sidelobe levels of MSK
– GMSK = Gaussian filter + MSK
– MSK = GMSK with B = ∞
– Important parameter : 3dB-BW – bit duration product
(B3dBTb)

1/31/2005
68
• Transmitter of GMSK :
QUAD architecture
s ( t ) = cos( 2π f c t + φ ( t ))
= cos 2π f c t ⋅ cos φ (t ) − sin 2π f c t ⋅ sin φ (t )

cos φ (t ) ⊗
m(t ) φ (t ) s(t)
k ∫− ∞ ( )
NRZ cos f c t 2π ⊕
t
GLPF 90o

sin φ (t ) ⊗

1/31/2005
69
 • Deciding frequency modulation index Kf
2 E
s FSK (t ) = b
cos( 2 π f c t + φ ( t ))
Tb
t
2 E
=
Tb
b
cos[ 2π fct + K f ∫
− ∞
m (η ) d η ]
∞

∫ ∑
t t
φ (t ) = K f ∫ −∞
m (η ) d η = K f −∞
n = −∞
a n r (η − nT ) d η
∞
⇒ K f ∫ −∞
r ( t ) dt = π / 2
∞
⇒ K f ∫ −∞
Π ( t ) ∗ h G ( t ) dt = π / 2
π /2
∴ K f = ∞
∫ −∞
Π ( t ) ∗ h G ( t ) dt
1/31/2005
70
• Power spectral density of a GMSK signal

1/31/2005
71
• Receiver of GMSK

 2α E  0.68 for GMSK with BT = 0.25

P e = Q  b 
 α≅
 N 0  0.85 for simple MSK ( BT = ∞)
1/31/2005
72
 • Occupied RF Bandwidth

1/31/2005
73
 Simulated power spectral density
• GSM GMSK v.s MSK

1/31/2005
74
 Simulated power spectral density
• Bluetooth v.s GSM

1/31/2005
75
 M-ary PSK
• Two or more bits are grouped to form symbols and
one of the possible M symbols may be sent

2 Es  2π 
Si (t ) = cos  2π f c t + ( i − 1)  , 0 ≤ t ≤ Ts i = 1, 2, K , M
Ts  M 

2Es  2π  2Es  2π 
= cos  ( i − 1)  cos( 2 π f c t) − sin  ( i − 1)  sin( 2 π f c t)
Ts  M  Ts  M 
i = 1, 2, K , M

  2π   2π  
S M - PSK ( t ) =  E s cos  ( i − 1 ) φ
 1 ( t ) − E s sin  ( i − 1 ) φ
 2 ( t ) 
  M   M  
i = 1, 2, K , M
1/31/2005
76
 M-ary PSK
• Constellation diagram of 8PSK

φ2(t)

φ1(t)

1/31/2005
77
 Quadrature amplitude modulation
• Emin : energy of the signal with the lowest amplitude
• ai, bi, index of the signal point

2 E min 2 E min
Si (t ) = a i cos( 2 π f c t) − bi sin( 2 π f c t)
Ts Ts
0≤t≤T i = 1, 2 , K , M

2 2
φ1 ( t ) = cos( 2π f c t ) 0 ≤ t ≤ Ts φ 2 ( t ) = sin( 2π f c t ) 0 ≤ t ≤ Ts
Ts Ts

1/31/2005
78
 Constellation diagram of 16QAM
φ2(t)

φ1(t)

1/31/2005
79
 Digital QAM Modulator
an a*(t)
d[n] Map to 2-D Impulse modulator
1
Serial/Parallel l
bit constellation Impulse modulator
bn
stream
Pulse shaping gT(t)
b*(t)
Matched
s(t) Delay
+ Local
Oscillator
90o

Pulse shaping gT(t)

1/31/2005
Matched delay matches delay through 90o phase shifter
80
 Phase Shift by 90 Degrees
• 90o phase shift performed by Hilbert
transformer 1 1
cos( 2π f 0 t ) ⇒ δ ( f + f0 ) + δ ( f − f0 )
cosine => sine 2 2
j j
sin( 2π f 0 t ) ⇒ δ ( f + f 0 ) − δ ( f − f 0 )
2 2
sine => – cosine
• Frequency response of ideal
 1 if x > 0
Hilbert transformer: 

H ( f ) = − j sgn( f ) sgn( x ) =  0 if x = 0

 − 1 if x < 0
1/31/2005 
81
 Hilbert Transformer
• Magnitude response • Phase response
All pass except at origin Piecewise constant

|H(f )| ∠H ( f )
90o
f f
H ( f ) = − j sgn( f ) -90o

• For fc > 0 • For fc < 0

π π π
cos( 2π f c t + ) = sin( 2π f c t ) cos(2πfct − ) = cos(−(2πfct + ))
2 2 2
π
= cos(2π (− fc )t + ) = sin(2π (− f c )t )
1/31/2005 2
82
 Hilbert Transformer
• Continuous-time ideal • Discrete-time ideal
Hilbert transformer Hilbert transformer
H ( f ) = − j sgn( f ) H (ω ) = − j sgn( ω )
1/(π t) if t ≠ 0 2 sin 2 (π n / 2 ) if n≠0
h(t) = π n
h[n] =
0 if t = 0 0 if n=0

h(t) h[n]

t n
Even-indexed
1/31/2005 samples are zero
83
 QAM Receiver
• Channel has linear distortion, additive noise,
and nonlinear distortion
• Adaptive digital FIR filter used to equalize
linear distortion (magnitude/phase distortion in
channel)
Channel equalizer coefficients adapted during
startup
At startup, transmitter sends known PN training
sequence

1/31/2005
84
 Matched
sampler
filter

Symbol/ cos(2πfct+φ)
clock
recovery
Signal Detect
Received Carrier
AGC Pulse decision
signal recovery
generator
90o
Phase
shift
-sin(2πfct+φ)

Matched
1/31/2005 sampler
filter 85
 In-Phase/Quadrature Demodulation
• QAM transmit signal x (t ) = a (t ) cos( ω c t ) + b (t ) sin( ω c t )
• QAM demodulation by modulation then filtering
– Construct in-phase i(t) and quadrature q(t) signals
– Lowpass filter them to obtain baseband signals a(t) and b(t)

i ( t ) = 2 x (t ) cos( ω c t ) = 2 a (t ) cos 2 (ω c t ) + 2b (t ) sin( ω c t ) cos( ω c t )

= a (t ) + a (t ) cos( 2ω c t ) + b (t ) sin( 2ω c t )
baseband high frequency component centered at 2 ωc
q (t ) = 2 x (t ) sin( ω c t ) = 2 a (t ) cos( ω c t ) sin( ω c t ) + 2 b (t ) sin 2 (ω c t )
= b (t ) + a (t ) sin( 2ω c t ) − b (t ) cos( 2ω c t )
baseband high frequency component centered at 2 ωc
1/31/2005
86
 Performance Analysis of QAM
• Received QAM signal
x ( nT ) = s ( nT ) + v ( nT )
• Information signal s(nT)
s ( nT ) = a n + j bn = ( 2i − 1) d + j ( 2 k − 1) d
where i,k ∈ { -1, 0, 1, 2 } for 16-QAM
• Noise, vI(nT) and vQ(nT) are independent
Gaussian random variables ~ N(0; σ2/T)
v ( nT ) = v I ( nT ) + j vQ ( nT )

1/31/2005
87
 Performance Analysis of QAM
• Type 1 correct detection Q
3 2 2 3

P1 ( c ) = P ( v I ( nT ) < d & vQ ( nT ) < d )

1 1
2 2
I

2 2
1 1

= P ( v I ( nT ) < d ) P ( vQ ( nT ) < d )
3 3
2 2

16-QAM
= (1 − P ( v I ( nT ) > d ))(1 − P ( vQ ( nT ) > d ))
d d
2Q ( T) 2Q ( T)
σ σ
d
= (1 − 2Q ( T )) 2
σ
1/31/2005
88
 Performance Analysis of QAM
• Type 2 correct detection
Q

P2 ( c ) = P ( v I ( nT ) < d & vQ ( nT ) < d )

3 2 2 3

1 1

= P ( v I ( nT ) < d ) P ( vQ ( nT ) < d ) 2 2
I

d d 2 2
= (1 − 2Q ( T ))(1 − Q ( T )) 1 1

σ σ
3 3
2 2

• Type 3 correct detection 16-QAM

P3 ( c ) = P ( v I ( nT ) < d & vQ ( nT ) > − d )
= P ( v I ( nT ) < d ) P ( vQ ( nT ) > − d )
d
= (1 − Q ( T )) 2
1/31/2005 σ
89
 Performance Analysis of QAM
• Probability of correct detection
4 d 4 d
P ( c ) = (1 − 2Q ( T )) + (1 − Q (
2
T )) 2
16 σ 16 σ
8 d d
+ (1 − 2Q ( T ))(1 − Q ( T ))
16 σ σ
d 9 2 d
= 1 − 3Q ( T)+ Q ( T)
σ 4 σ
• Symbol error probability
d 9 2 d
P ( e ) = 1 − P ( c ) = 3Q ( T)− Q ( T)
σ 4 σ
1/31/2005
90
 Average Power Analysis
• PAM and QAM signals are deterministic
• For a deterministic signal p(t), instantaneous power
is |p(t)|2
• 4-PAM constellation points: { -3 d, -d, d, 3 d }
– Total power 9 d2 + d2 + d2 + 9 d2 = 20 d2
– Average power per symbol 5 d2
• 4-QAM constellation points: { d + j d, -d + j d,
d – j d, -d – j d }
– Total power 2 d2 + 2 d2 + 2 d2 + 2 d2 = 8 d2
– Average power per symbol 2 d2

1/31/2005
91
 Summary of QAM.
 ( − L + 1, L − 1) ( − L + 3, L − 1) L ( L − 1, L − 1) 
 ( − L + 1, L − 3 ) ( − L + 3, L − 3 ) L ( L − 1, L − 3 ) 
 
{a i , bi } =  . . . . 
 
 . . . . 
 ( − L + 1, − L + 1) ( − L + 3 , − L + 1) L ( L − 1, − L + 1) 

 ( − 3 , 3) ( − 1 , 3) ( 1 , 3) ( 3 , 3) 
 ( − 3 , 1) ( − 1 , 1) ( 1 , 1) ( 3 , 1) 
{a , b }=  
 ( − 3 , − 1) ( − 1 , − 1) ( 1 , − 1) ( 3 , − 1) 
i i

 
 ( − 3 , − 3) ( − 1 , − 3) ( 1 , − 3) ( 3 , − 3) 

 1   2 Emin   1   3Eav 
Pe ≅ 41 − Q  
 Pe ≅ 41 − 
Q  
 M   N0  M   ( M − 1) N 0 
 

1/31/2005
92
 M-ary FSK
2Es π 
Si (t ) = cos  ( n c + i ) t  0 ≤ t ≤ Ts i = 1, 2, K , M
Ts  Ts 

– error probability under coherent detection

 2 Eb log2 M 
Pe ≤ ( M − 1)Q  

 N 0 
– error probability under non-coherent detection
M −1
 ( − 1) k +1   M − 1   − kE s 
Pe = ∑     exp  
k =1  k + 1   k   ( k + 1) N 0 

1/31/2005
93
 M-ary FSK
• M_ary FSK
– BW of coherent MFSK :
Rb ( M + 3)
B=
2 log2 M

– BW of noncoherent MFSK :
Rb M
B=
2 log2 M
1/31/2005
94