ADC

PCM signal
Sample
Quantize
Analog Input
Signal
Encode
Line Code
x
s
(t)
x
Q
(t)
x
k
(t)
x(t)
Analog to Digital Conversion
1
1. Sampling of Continuous Signal
First step of analog to digital conversion is the periodic sampling
of continuous analog signal. A sampled signal is considered at some
distinct instant t = nT; n = 0, 1, 2, 3, for duration .
If some one asks what is the amplitude of the sampled signal
between two consecutive sample, the answer is undefined same is
true for digital signal.

x
s
(t)
t
Sampled signal
2
Let an analog signal x(t), band limited to B Hz shown in fig.1, is
sampled at a rate fc = 1/T samples/sec. If the width of each sample is
then sampled signal x
s
(t) can be expressed as the product of x(t) and a
unit amplitude pulse train s(t) of period T and width of each pulse of
shown in fig.2. Sampled form x
s
(t) of x(t) is shown in fig.3.
T

s(t)
t

x
s
(t)
t
x(t)
-B B
f
X(f)
Fig. 1 (a) Continuous base band signal
(b) Spectrum of x(t)
Fig.2 Unit amplitude pulse train Fig. 3 Sampled signal
3
Lt us express s(t) in terms of Fourier co-efficient,

=
(

|
.
|

\
|
+
|
.
|

\
|
+ =
1
0
2
sin
2
cos
2
s(t)
n
n n
T
nt
b
T
nt
a
T T
a t t
( ) nd
T
n
c
T
n
T
n
T
nt
n
T
t d
T
nt
t d
T
nt
t s a
T
T
n
sin sin
/ sin
2
sin
2
2
cos . 1
2
cos ) (
2 /
2 /
2 /
2 /
2 /
2 /
t
t
t
t t t t
t
t
t
t
t
t
t
t
t
=
|
.
|

\
|
=
|
.
|

\
|
=
(

|
.
|

\
|
=
|
.
|

\
|
=
|
.
|

\
|
=

}
}
Where
;Where d = /T, known as duty cycle.
T

s(t)
t
t t
t
t = =
|
.
|

\
|
= =

1 . sin
0 0 0
T
n
c Lt a Lt a
n n n
4
0
2
sin . 1
2
sin ) (
2 /
2 /
2 /
2 /
=
|
.
|

\
|
=
|
.
|

\
|
=
}
}

t d
T
nt
t d
T
nt
t s b
T
T
n
t
t
t
t
( )
( )

=
|
.
|

\
|
+ =
|
.
|

\
|
+ =
1
1
2
cos sin 2
2
cos sin
2
s(t )
n
n
T
nt
nd c d d
T
nt
nd c
T T
t
t
t
t
T

s(t)
t
5
( ) ( )

=
+ = =
1
2 cos sin ) ( 2 ) ( ) ( ). ( ) (
n
c s
t nf nd c t dx t dx t s t x t x t
( ) ( )

=
+ =
1
cos sin ) ( 2 ) (
n
c
t n nd c t x d t dx e
Now
) ( ) ( e X t x Let
| | ) ( ) (
2
1
) cos( ) (
c c c
n X n X t n t x e e e e e + +
| | ( )
( ) ( )
( ) ( )

=
+ + =
+ + + =
+ + + =
1 1
1 1
1
sin ) ( sin ) ( ) (
sin ) ( sin ) ( ) (
sin ) ( ) (
2
1
2 ) ( ) ( ) (
n
c
n
c
n
c
n
c
n
c c s s
nd c n X d nd c n X d dX
nd c n X d nd c n X d dX
nd c n X n X d dX X t x
e e e e e
e e e e e
e e e e e e
( ) ( )


=

= =
= + =
n
c
n n
c
nd c n X d nd c n X d dX sin ) ( sin ) ( ) (
0 ,
e e e e e
6
( ) ( ) ( ) ... sin ) ( ) ( sin ) ( 2 sin ) 2 ( ... ) ( + + + + + + = d c dX dX d c dX d c dX X
c c c s
e e e e e e e e
( ) ( )


=

= =
= + =
n
c
n n
c
nd c n X d nd c n X d dX sin ) ( sin ) ( ) (
0 ,
e e e e e
f
f
c
+B
f
c
-B
-f
c
-B B
f
c
Xs(f)
7
Plot of above equation is shown in fig.4.
To extract x(t)X() from x
s
(t)X
s
() a low pass filter with cutoff
frequency between B and f
c
-B can be used. There must not any
overlapping between X() and X(-
c
) for proper filtering, therefore
f
c
-BB f
c
2B because of practical characteristics of low pass filter.
Therefore sampling rate of a signal must be greater than or equal to the
maximum frequency of the original/ base band signal.
f
f
c
+B
f
c
-B
-f
c
-B B
f
c
Xs(f)
Fig. 4 Spectrum of Sampled signal
The condition f
c
2B is called Nyquist sampling rate.
8
Example:
An analog signal of highest frequency of 3.4 KHz
sampled maintaining a guard band of 300Hz, find
sampling rate.
fc = G.B+2B = 0.3+23.4 =7.1KHz
9
2. Quantization of Sampled Signal
Second step of digitations of an analog signal is quantization of
sampled signal into some known discrete amplitude levels. Now
both the axis of the signal becomes discrete hence a quantized
signal is actually a digital signal.
Due to quantization the signal losses its originality and an error
is introduced with the signal known as quantization error. Actually
amplitude of a sample is adjusted with its nearest quantization
level. The spacing between adjacent levels are called quantization
interval.
10
If a be the quantization interval then mean square error of a sample
of amplitude S
j
+ in the vicinity of jth level S
j
, in the interval, of
fig. 5 is expressed like,
S
j
S
j
+a/2
S
j
+a/2
S
j
+
Fig. 5 Quantization error
12
1
) (
2
2 /
2 /
2 2
a
d
a
E
a
a
= e e = e
}

2 2
a a
ses
Where
11
Let the signal at transmitting end is quantized into total N discrete levels with
maximum plus-minus signal excursion of P volts, maximum excursion in positive or
negative direction is V and the range of quantized samples is A .
Fig.6 shows a quantization technique of 6 levels. Here quantized amplitude would be
a/2, 3a/, 5a/2 and 7a/2. Now V = 4a , A = 7a and P = 8a. Therefore for N levels
quantization, V = Na/2 , A = (N-1)a and P = Na.
A
P
V
0
a/2
3a/2
5a/2
-5a/2
-3a/2
-a/2
4a
-4a
7a/2
-7a/2
Now peak signal to quantization ratio
(SQR) of quantization system will be,
( ) ( ) N N
a
a N
E
V
SQR
log 20 8 . 4 3 log 20
12 /
4 /
log 10
) (
log 10
2
2 2
2
2
+ = =
|
|
.
|

\
|
=
|
|
.
|

\
|
e
=
Fig. 6 quantization technique of 8 levels 12
Example-2
A sinusoidal wave of peak amplitude A has mean square amplitude of
A
2
/2. According to equation of previous slide SQR of the sine wave will
be, SQR(dB)=10log{(A
2
/2)/(a
2
/12)} =7.78+20log(A/a)
0 2 4 6 8 10 12
5
10
15
20
25
30
A/a
S
Q
R
(
d
B
)
Fig. 7 SQR of a sinusoidal wave
13
SQR increases with increase in amplitude (loudness of voice signal) of
baseband signal. Our aim is to keep SQR constant irrespective of amplitude of
baseband signal
3. Pulse Code Modulation
Pulse Code Modulation (PCM) is essentially final step of analog to
digital conversion where each quantized sample in converted to
equivalent binary pulses, resembles to decimal to binary
conversion.
Fig.8 show 8 levels PCM technique. In PCM gray code is
preferable compare to ordinary binary code to combat variable
number of bit error for adjacent levels like 3 and 4, 7 and 8 etc
when sampled pulses are transmitted directly. Because in gray
code there is only one bit difference between adjacent decimal
numbers hence there is only possibility of single bit error instead
of variable number of bit error during reception of each sample.
14
T T
6
3
4
t
Fig. 8(a) Quantized sample of 8 levels quantization
1 1 0
T T
t
1 0 0 0 1 1
Fig.8(b) PCM data of fig. 8(a)
15
Exerceise-1
A voice signal of highest frequency of 3.4KHz sampled
maintaining a guard band of 200Hz. After sampling the signal is
quantized into 256 levels. Determine sampling rate and bit rate of
PCM.
16
Companding
From the analysis of previous section we can say, if the quantization
interval is kept directly proportional to amplitude of signal then SQR
will remain constant over the entire dynamic range.
Let a = kA; where A is the amplitude a signal at a particular instant
and k is a constant.
k
k A k
A
a
A
dB SQR log 20 79 . 10
12
log 10
12 /
log 10
12 /
log 10 ) (
2 2 2
2
2
2
= = = =
17
Therefore SQR(dB) is independent of signal amplitude. Above
phenomenon can be implemented by compressing the signal
nonlinearly (larger the sample amplitude, the more it is
compressed) then providing uniform quantization is provided on it
like fig.9.
Compressed
output signal
Input signal
Fig. 9 Non linear compression of signal
18
At receiving end signal is recovered by reverse operation
(nonlinear expansion). The process of compression and expansion
of signal is called companding shown in fig10.
Input
Output
Compression
PCM
D/A Expansion
Fig. 10 Companding technique
Input signal x is compressed in North America and Japan according
to following relation known as -law companding.
) 1 ln(
) 1 ln(
) sgn( ) (

+
+
=
x
x x F ;-1 x 1
19
where sgn(x) is the polarity of x and is a parameter used to define
amount of compression. Practical compander uses = 255.
| | 1 ) 1 (
1
) sgn( ) (
1
+ =

y
y y F

The reverse operation or expansion of signal is expressed as,

; where y is the compressed signal.
20
Another companding technique prevalent in Europe and most of
the region of the world is known as A-law companding
expressed as,
) A ln( 1
x A
) x sgn( ) x ( F
A
+
= 0 |x| 1/A
Inverse or expansion of the signal is expressed as,
| |
A
) A ln( 1 y
) y sgn( ) y ( F
1
A
+
=

) A ln( 1
1
y 0
+
s s
Where
Practical compander uses A = 100.
21
22
PAM, PWM and PPM
In all of above pulse modulation techniques unit amplitude pulse
train is used as carrier wave.
In pulse amplitude modulation (PAM) the amplitude of the
carrier pulses are proportional to instantaneous amplitude of
continuous base band signal x(t) shown in fig.11.
In PWM the width of a pulse is proportional to the instantaneous
amplitude of continuous base band signal. In PPM the position of
the pulse i.e. distance from a reference instant is proportional to the
instantaneous amplitude of continuous base band signal.
23
Fig. 11 PAM, PWM and PPMwave
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
x(t),x
p
(t)
t

T
s

s(t)
t
t
x
po
(t)
PWM
PPM
x
d
(t)
t
PAM
24
PAM wave x
p
(t) can be generated using simple electronic circuit of fig.12 where
pulses at G
1
and G
2
of FETs are shown in fig.13 and fig.14 respectively. PAM signal
actually resembles to flat top sampled signal of period T
s
. Since the width of each PAM
pulse is therefore pulses at G
2
stars at the position of unit delayed from kT
s
to
discharge the capacitor C.
C
G
2
G
1
x(t)
x
p
(t)
G
1
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
t
G
2
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
t
Fig. 12 Circuit to convert continuous signal to PAM wave
Fig. 13 Pulses at G
1
of above circuit
Fig. 14 Pulses at G
2
of above circuit
PAM (Pulse Amplitude Modulation)
25
25
C
G
2
G
1
x(t)
x
p
(t)
G
1
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
t
G
2
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
t
0 T
s
2T
s
3T
s
4T
s
5T
s
6T
s
x(t),x
p
(t)
t
PAM

26
In pulse width modulation (PWM) width of a carrier pulse is proportional to instantaneous
amplitude of the baseband signal of course the amplitude of each pulse is equal.
PWM can be generated simply from a comparator circuit of fig.15 where input at non-
inverting terminal is the baseband signal x(t) and a sawtooth wave at inverting terminal.
Resulting wave is shown in fig.16 where starting position of each pulse is at kT
s
but terminating
points depend on width of each pulse, hence amplitude of baseband signal.
Fig. 15 PWM and PPM generator
Sawtooth
wave
x(t)
PWM wave
Monostable
Multivibrator
PPM
wave
Comparator
PWM (Pulse Width Modulation)
x
d
(t)
t
27
x
d
(t)
t
t
x
po
(t)
PWM
PPM
Fig. 16 Output waves of PWM and PPM generator
In pulse position modulation (PPM),
amplitude and width of each pulse is
kept same but starting position of each
pulse i.e. the distance between instant,
t = kT
s
and stating point of kth pulse
depends on instantaneous amplitude of
base band signal.
PPM wave can be generated simply
applying PWM pulses at the input of a
monostable multivibrator.
Sawtooth
wave
x(t)
PWM wave
Monostable
Multivibrator
PPM
wave
Comparator
PPM (Pulse Position Modulation)
28
Let us now consider the case of conversion of PWM signal to
PAM. PWM pulses of fig. 17 (a) is passed through a ramp generator
with large capacitor at output to hold the final amplitude of the
signal, the resulting wave is shown in fig. 17(b).
A pulse train of period T
s
is locally generated and added with the
signal of fig 17(b) such that pulses fall on the constant part of the
wave shown in fig. 17(c). Finally pulses of fig.17(c) is passed
through a clipper circuit of threshold voltage shown by doted line of
fig. 17.(c).
Clipper circuit actually clipped the signal below the threshold
level, therefore the resulting wave would be like fig.17(d) i.e. the
equivalent PAM signal. Baseband signal can be regenerated by
passing the PAM wave through a low pass filter.
Conversion of PWM to PAM
29
PWM
Fig. 17(a) PWM wave
Fig.17 (b) Output of Ramp Generator
Fig. 17(c) Locally generated pulses are added with ramp output
Fig. 17(d) Out put of clipper circuit
30
It is found that average amplitude of sample to sample difference
is less than that of original sampled wave of voice signal therefore
less number bit is required for PCM, maintaining same SQR. If
PCM is done on consecutive pulse difference instead of individual
pulse known as differential pulse code modulation (DPCM).
When sample to sample difference is expressed by a single bit
then the modulation scheme is called delta modulation is
considered as an special case of DPCM.
Differential Pulse Code Modulation (DPCM)
31
Major components of transmitter and receiver of delta modulation are sampler,
predictor, quantizer, adder and smoothing filter shown in fig. 18. In transmitter the
analog baseband signal x(t) is sampled and a difference signal, is generated. Here
is jth sampled pulse of x(t) is x
j
and g
j
is jth predicted pulse.
x(t)
g
j
x
j
+
_
Sampler
Predictor
Quantizer
k
+
g
j
k +
_
Predictor
Smoothing
filter
Fig. 18(a) Delta modulator
Fig. 18(b) Delta demodulator
j
e
j
e
32
Predicted pulse is determined from linear combination of some previous sampled
pulses shown in fig. 19 is expressed as,

=

=
k
1 s
s j s j
x

h g
where is the jth estimated sample determined as, g k
Based on difference signal a pulse of amplitude k is generated by a quantizer
like,
j
x

j
e

e
'

e
'
=
negative is if k
positive is if k
P
j
j
j
g
j
......
Z
-1
Z
-1
Z
-1
Z
-1
h
1
h
2
h
k

Fig. 19 Predictor circuit

33
Let us demonstrate delta modulation transmitter and receiver with
the graph of fig.20. Here predicted sample g
1
at 1
st
sampling instant is
determined from linear combination of some fixed previous samples.
Now g
1
has to be compared with 1
st
sample x
1
and it is found from
fig.20 that x
1
> g
1
, therefore transmitter will generate g
1
+ k and 1
st
estimated sample, = g
1
+ k will be detected at receiver.
At 2
nd
sampling instant g
2
is determined from the similar relation
and compared with x
2
where x
2
> g
2
, therefore transmitter will
generate g
2
+ k and 2
nd
estimated sample = g
2
+ k will be detected
at receiver. Same condition is also found at 3
rd
sampling instant.
1
x
2
x
34
At 4
th
sampling instant x
4
< g
4
therefore transmitter will generate
-k and 4
th
estimated sample, = g
4
- k will be detected at
receiver and so on.
Estimated samples are discrete points at the vicinity of the
baseband wave but some times points are above the curve and
some times fall below the curve depending on the profile of the
baseband curve. These discrete points are made a continuous using
a smoothing filter.
4
x
35
Fig. 20 Comparison of and x
j
) ( t x
j
x(t)
g
j
x
j
+
_
Sampler
Predictor
Quantizer
k
+
k
36
Since the estimated sampled points are little away from the baseband wave
hence a quantization noise is incorporated with delta modulation like fig. 21. If the
step size k is lowered then the separation between and x
j
will be reduced and
the zigzag curve of points would be smoother.
1
x

Fig. 21 Output of the demodulator for larger step size

t
x(t)
37
When step size k is very small then the zigzag curve of estimated points are simply
unable to follow the baseband curve when the curve changes rapidly like fig.22, known as
slope overload distortion.
Maximum slope supported by the zigzag curve is k/T
s
; where T
s
is the sampling period
Therefore a tradeoff has to be made with quantization noise and slope overload distortion
taking optimum step size k
opt
. Output of the demodulator for optimum step size is shown in
fig. 23.
K
T
s
Fig. 22 Output of the demodulator for
smaller step size
Fig. 23 Output of the demodulator for
optimum step size
38
Example-1
Determine optimum step size for the signal x(t)= 5sin(220t) considering
sampling frequency of 8KHz.
Ans. The slope of the signal x(t),
The maximum slope is obtained taking =1.
The maximum slope of the signal m
max
= = 200 . From the trajectory of zigzag curve of fig. 23, the maximum slope
supported by it,
= kf
s
=k8000
Therefore 200 = 8000k
k
opt
= 200/8000 = /40 volt
) 20 2 cos( t t
20 2 5 t
) 20 2 cos( 20 2 5
) (
t
dt
t dx
t t =
s
T
k
= ) tan( u
39
Line coding
Binary data (logic 0 or 1 of PCM) can be transmitted using a number
of different types of serial pulses. The choice of a particular pair of
pulses to represent the symbols 1 and 0 is called Line Coding. Line
coding can be represented by the diagram of fig. below.
ADC
PCM signal
Sample
Quantize
Analog Input
Signal
Encode
Line Code
x
s
(t)
x
Q
(t)
x
k
(t)
x(t)
Fig. Line coding
40
Each line code is described by a symbol mapping function a
k
and a
pulse shape p(t):

=
=
k
b k
kT t p a t x ) ( ) (
Line code can be classified based on symbol mapping functions (a
k
) like,
Unipolar: In unipolar signalling binary symbol 0 is represented by the absence
of a pulse called space and the other binary symbol 1 is represented by the
presence of a pulse called mark. It is also called on-off keying.
Polar: In polar signalling a binary 1 is represented by a pulse p(t) and a binary
0 by the opposite (or antipodal) pulse p(t).
Bipolar: Bipolar Signalling is also called alternate mark inversion which uses
three voltage levels (+V, 0, -V) to represent binary symbols. Zeros, as in
unipolar, are represented by the absence of a pulse and ones (or marks) are
represented by alternating voltage levels of +V and V.
41
Line code can again be classified based on pulse shapes p(t) like,
NRZ (Nonreturn-to-zero): The pulse occupies the full duration of
a symbol.
RZ (Return to Zero): The pulse occupies the half of duration of a
symbol.
Manchester (split phase): In Manchester encoding, the duration of
the bit is divided into two halves. The voltage remains at one level
during the first half and moves to the other level during the second
half. Binary logic 1 is +ve in 1st half and -ve in 2nd half. Binary
logic 0 is -ve in 1st half and +ve in 2nd half.
42
BINARY DATA
A
-A
0
(b) Polar NRZ
A
0
(c) Unipolar RZ
A
-A
0
(d) Bipolar RZ
1 1 0 1 0 0 1
Mark
(hole)
Mark
(hole)
Mark
(hole)
Mark
(hole)
space space space
A
-A
0
(e) Manchester NRZ
Binary Signaling Formats
Volts
A
Time
0
(a) Unipolar NRZ
T
b
According to above classification different types of line coding is shown
in fig. below.
Fig. Line coding
Multiplexing Techniques
43
44
45
46
47
Time division Multiplexing
Fig. below shows the frame of 30-channel system where out of 32 time slot (TS) the
0th TS is for frame alignment, 16
th
TS is for signaling and the rest 30 TS (1 to 15 and
17 to 31) for speech. Each TS contains 8 PCM bits, equivalent to each quantized
pulse of 256 level system. Such frame is called E1 frame flowed by European.
0 1 to 15 16 17 to 30 31
Frame Alignment
Signaling
b
1
b
2
b
3
b
4
b
5
b
6
b
7
b
8
Fig.E1 frame of 30 Channels
48
The bit rate of the 30-channel PCM frame is, 8 kHz (sampling rate) 8 (bits per sample
or TS) 32 (number of TS/frame) = 2.048 Mbps. The total number of bite/frame =
832 = 256bits. We know the sampling rate of voice signal is 8000 samples/sec,
therefore sampling period, Ts = (1/8000) sec. The 8 PCM bits of each sample is placed
in a TS, therefore the length of a frame, T
f
= (1/8000) sec 125 s. The width of each bit,
d = 125/256 = 0.488 s.
Multiplexing hierarchy of European system is shown in fig.3.
E1
M
U
X
30 Voice
channels,
each of 64
kbps
.
..
E1 = 2.048 Mbps
( 30 voice channel)
E2
M
U
X
E1
E2 = 8.448 Mbps
( 120 voice channels)
E3
M
U
X
E3 = 34.368 Mbps
( 480 voice channels)
E4
M
U
X
E2
E3
E4 = 139.264 Mbps
( 1920 voice channels)
E5
M
U
X
E5= 564.992 Mbps
( 7680 voice channels)
E4
Fig.3 European synchronous digital hierarchy
49
North American TDM frame contains 24 TS where each TS contains 8
bits of a user (equivalent of 8 PCM bits/sample) and the 0
th
TS
contains a single bit for frame alignment called T1 frame like fig.4.
0 Channel-1
Frame Alignment
b
1
b
2
b
3
b
4
b
5
b
6
b
7
b
8
Channel-2
Channel-24
Fig.4 T1 frame of 30 Channels
The total number of bits/frame = 824+1 = 193 bits. The 8 PCM bits of each sample is
placed in a TS, therefore the length of a frame, T
f
= (1/8000) sec 125 s. The bit rate of T1
frame = 193/125 =1.544 Mbps.
50
The objective of digital modulation is to convert the rectangular
digital pulses to smooth sinusoidal wave hence considerable
reduction in bandwidth is achieved. The bandwidth reduction is
essential to cope the transmitted wave with the transmission
medium of lower bandwidth.
For example MODEM is connected between PC and telephone
line to convert the rectangular data pulse from the computer
(infinite bandwidth) into continuous FSK wave of lower bandwidth
to cope with the allowed bandwidth of the telephone line
(transmission medium).
Digital Modulation Techniques
51
In wired communication the digital modulation is necessary for,
To reduce the bandwidth of the baseband signal according the
capacity of the transmission medium and receiver circuits.
If the wavelength of the signal is considerable with the length of
the wire then the wire acts an antenna and radiates most of the
signal energy. The remedial measure of the situation is to increase
the frequency of the signal i.e. modulation with high frequency
carrier.
52
In wireless communications the reasons of digital modulation are,
In wireless communication digital modulation is necessary to avoid
interferences of surrounding users with choose of appropriate carrier.
Noisy immunity is increased for a high frequency modulated wave.
Size of an antenna depends on wavelength of a signal. After
modulation (using high frequency carrier) wavelength of the
modulated wave is reduced hence size of the antenna is reduced.
A modulated wave can propagate far longer way than the baseband
signal.
53
Frequency Shift Keying (FSK)
In binary FSK two sinusoidal waves of frequency f
1
and f
2
over a
period [0, T] are used to represent binary logic 1 and 0 respectively.
The waves are like,
for logic 1
for logic 0
; where is the initial phase of sinusoidal wave i.e. at the stating
point symbol wave.
) 2 cos( ) (
1 1
u t + = t f A t S T k t kT ) 1 ( + s s
;
) 2 cos( ) (
2 2
u t + = t f A t S
;
T k t kT ) 1 ( + s s
54
Input binary bits
as a selector
FSK wave
Oscillator-1
Oscillator-2
M
U
X
Fig. 1 FSK modulator
FSK wave can be generated using two oscillators of frequency f
1
and f
2
connected to a multiplexer like fig.1. The multiplexer is
switched between the oscillators by binary input bits.
) 2 cos( ) (
1 1
u t + = t f A t S
) 2 cos( ) (
2 2
u t + = t f A t S
55
Frequencies f
1
and f
2
are chosen such that S
1
(t) and S
2
(t) are mutually orthogonal.
In that case at receiving end S
1
(t) and S
2
(t) can be detected using cross
correlation called coherent demodulation.
}
+
=
T k
kT
dt t S t S
) 1 (
2 1
0 ) ( ) (
To satisfy above condition,
and ; where m and n are integers. The difference
between frequencies,
For m = 2 the separation becomes 1/T which ensure phase
continuity at bit transitions called Sundes FSK shown in fig.2. In
case of discontinuity of phase at each bit transition point like fig. 3,
spectral width becomes wider.
t t n T f f 2 ) ( 2
2 1
= +
t t m T f f 2 ) ( 2
2 1
=
T
m
f f
2
) (
2 1
=
56
S(t)
1 0 1 0
t
S(t)
1 0 1 0
t
Fig.2 Continuous FSK
Fig.3 Discontinuous FSK
57
The received signal contaminated by awgn is expressed as,
r(t)=S
i
(t)+n(t); where i=1,2 and n(t) is the additive white
Gaussian noise (awgn). The coherent detector of orthogonal FSK
is shown in fig.4.
Decision
r(t)
1
0
}
+ T k
kT
dt
) 1 (
( ) ( ) t f t f
2 1
2 cos 2 cos t t
Fig. 4 Coherent demodulator of FSK
58
Binary Phase Shift Keying (BPSK)
In BPSK binary logic 1 and 0 are presented by two sinusoidal waves
of same frequency and peak amplitude over the period [0, T] but
initial phases are 0 and respectively. BPSK wave is expressed like,
) 2 cos( ) 0 2 cos( ) (
1
t f A t f A t S
c c
t t = + =
T k t kT ) 1 ( + s s for logic 1
) 2 cos( ) 2 cos( ) (
2
t f A t f A t S
c c
t t t = + = T k t kT ) 1 ( + s s for logic 0
Wave shape of BSK is shown in fig.5.
1 0 1 0
Fig. 5 BPSK waveform
;
;
59
PSK signals are represented graphically in two dimensional co-ordinate
system called signal constellation. Here x axis is represented by a
function,
t f
T
t
c
t 2 cos
2
) (
1
=
T t s s 0
and y axis by,
t f
T
t
c
t 2 sin
2
) (
2
=
T t s s 0
;
;
Orthogonal basis function Unit vector
and and
Correlation Dot product
and . = 0 and . = 1
Component along x-axis from
is given as,
Component along x-axis
from is given as,
t f
T
t
c
t 2 cos
2
) (
1
=
t f
T
t
c
t 2 sin
2
) (
2
=
0 ) ( ) (
0
2 1
=
}
T
t t 1 ) ( ) (
0
1 1
=
}
T
t t
j b i a S

+ =
j

) ( ) (
2 1
t b t a S + =
a t S
T
=
}
0
1
) (
i

a i S =

.
60
Let us represent S
1
(t) and S
2
(t) in terms of and
) (
1
t ) (
2
t
( )
) ( ) (
) ( sin
2
) ( cos
2
2
2 sin
2
sin 2 cos
2
2
cos
2 sin sin 2 cos cos
) 2 cos( ) (
2 1 1 1
2 1
t S t S
t
T
A t
T
A
T
t f
T
A t f
T
T
A
t f A t f A
t f A t S
i i
c i c i
c i c i
i c i

u u
t u t u
t u t u
u t
+ =
+ =
|
|
.
|

\
|
+ =
=
+ =
( ) 2 / f 2 cos ) (
2
0
1
2 2
0
2
1
T A dt t A dt t S E
T T
= = =
} }
t
( ) 2 / f 2 s ) (
2
0
1
2 2
0
2
2
T A dt t in A dt t S E
T T
= = =
} }
t
Energy of S
1
(t) or S
2
(t)
) ( sin ) ( cos
) ( ) ( ) (
2 1
2 1 1 1
t E t E
t S t S t S
i i
i
u u

+ =
+ =
61
The abscissa of the constellation point,
The ordinate of the constellation point,
In BPSK
i
= 0 or
( )
i
T
T
i i
E S
dt t t S t S
dt t t S S
u

cos
) ( ) ( ) (
) ( ) (
1
0
1 2 1 1 1
0
1 1
= =
+ =
=
}
}
( )
i
T
T
i i
E S
dt t t S t S
dt t t S S
u

sin
) ( ) ( ) (
) ( ) (
2
0
2 2 1 1 1
0
2 2
= =
+ =
=
}
}
E E E S
i
= = or cos
1
u
0 sin
2
= =
i
E S u
62
( )= = 0 , ) , (
1 2 1
S S S
The co-ordinate of the constellation points,
In Cartesian co-ordinate system shown in fig.6.
Fig.6 BPSK Constellation
S
2
=-E
S
1
=E

1
(t)

2
(t)
( ) 0 , cos
i
E u ) 0 , ( E = ) 0 , ( E or
63
Now BPSK modulator and its coherent demodulator can be
implemented like fig. 3.
Local
Oscillator

Polar NRZ data pulse

sequence b(t)
Ab(t)cos(2f
c
t)
Acos(2f
c
t)
Fig.3 BPSK (a) Modulator (b) Demodulator
CR

r(t)
Acos(2f
c
t)
1
0
1 or 0
L
}
+ T k
kT
dt
) 1 (

=
=
k
k
kT t p b t b ) ( ) (
Where b
k
{1, -1} and
[
|
.
|

\
|

=
T
T t
t p
2 /
) (
64
Here output of the correlator of demodulator assuming S
1
(t)
transmitted is,
( )
( )
{ }
{ }
2
) 2 cos( ) ( 0
2
) 2 cos( ) ( ) 4 cos( 1
2
) 2 cos( ) ( ) 2 ( cos
) 2 cos( ) ( ) 2 cos(
) 2 cos( ) ( ) (
) 2 cos( ) (
) 1 (
) 1 ( ) 1 (
) 1 (
2
) 1 (
) 1 (
1
) 1 (
AT
dt t f t n
AT
dt t f t n dt t f
A
dt t f t n t f A
dt t f t n t f A
dt t f t n t S
dt t f t r L
T k
kT
c
T k
kT
c
T k
kT
c
T k
kT
c c
T k
kT
c c
T k
kT
c
T k
kT
c
= + + =
+ + =
+ =
+ =
+ =
=
}
} }
}
}
}
}
+
+ +
+
+
+
+
t
t t
t t
t t
t
t
If S
2
(t) is transmitted then,
2
AT
L =
CR

r(t)
Acos(2f
c
t)
1
0
1 or 0
L
}
+ T k
kT
dt
) 1 (
65
M-ARYPSK
In BPSK bit rate and modulation symbol rate are equal. In MPSK n = log
2
M bits are
represented by a modulation symbol hence bit rate is n times higher than symbol rate.
Therefore n times more information can be transmitted by MPSK (for the same
baud/symbol rate) scheme compared to that of BPSK at the expense of bit error rate.
MPSK wave is expressed as,
t f A t f A
t f A t S
c i c i
i c i
t u t u
u t
2 sin sin 2 cos cos
) 2 cos( ) (
=
+ =
) ( ) (
2 2 1 1
t S t S
i i
+ =
T t s s 0
Where i = 1, 2, 3,M, and
are orthogonal basis functions.
M
i
i
t
u
) 1 2 (
=
) ( ), (
2 1
t t
66
The abscissa of the constellation points,
i
T
i i
E dt t t S S u cos ) ( ) (
0
1 1
}
= =
The ordinate of the constellation point,
i
T
i i
E dt t t S S u sin ) ( ) (
0
2 2
}
= =
The co-ordinates of the constellation points in Cartesian co-ordinate
system are,
( )
i i i i
E E S S u u sin , cos ) , (
2 1
=
In polar co-ordinate system amplitude is,
( ) ( ) E E E
i i
= +
2 2
sin cos u u
The phase angle,
2
2 1
tan
i
i
i
S
S

= u
67
Therefore the polar co-ordinates of the signal constellation are
) , (
i
E u
For M = 8, initial phases are 0, /4, /2, 3/4, , 5/4, 3/2 and 7/4
taking i = 0, 1, 2 ,..., 7. Constellation of 8-ARY PSK is shown in fig.1
below with its decision range.

1
(t)

2
(t)
Fig.1 8-ARY PSK Constellation
68
QPSK
Quadrature PSK is a special case of M-ary PSK (MPSK) where M = 4; Therefore the
initial phases, ; i = 1, 2, 3, 4 provides /4, 3/4, 5/4 and 7/4. Bit rate of
QPSK scheme is lower than that of M-ARY PSK for M > 4 (for same baud rate) but
provide better performance in context of bit error rate. Each modulation symbol
carries two binary bits correspond to a point on constellation shown in fig. 2. In any
constellation gray code is used to ensues single bit discrepancy between adjacent
points.
4
) 1 2 ( t
u

=
i
i
01 11
00 10

1
(t)

2
(t)
Fig. 2 QPSK Constellation