Telecommunications B (EE2T21)

Lecture 3 overview:

Angle modulation
 non-linear modulation techniques for analog signals:
* Phase Modulation (PM)
* Frequency Modulation (FM)

Signal-to-Noise Ratio after detection for analog signals

* Coherent detection of AM, DSB-SC, SSB
* Non-coherent detection of AM

EE2T21 Telecommunications B
Dr.ir. Gerard J.M. Janssen
April 22, 2022

Faculty of Electrical Engineering, Mathematics, and Computer Science

Lectures & Work-instructions Telecommunications B
Telecommunication Techniques

Lectures:
Monday 25-4, 2-5, 9-5 1st+2nd hour
Tuesday 3-5 1st+2nd hour
Wednesday 11-5 3rd+4th hour

Q&A/Working lectures:
Wednesday 4-5 3rd+4th hour
Thursday 28-4, 12-5 5th+6th hour

2
 Bandpass Signals
Mathematical description of bandpass signals:

s (t )  Re{g (t )e jct }
 R (t ) cos[c t   (t )]
 x(t ) cos c t  y (t )sin c t

Complex envelope or complex equivalent baseband signal:

g (t )  f ( m(t ))  x(t )  jy (t )  R(t )e j ( t )

Modulation In-phase and Quadrature-phase Amplitude: AM Phase: PM, FM
function component

3
 Angle modulation
For angle modulation:

g (t )  f ( m(t ))  Ac e j ( t )  - angle modulation

- constant amplitude
 - efficient power generation
using class-C amplifiers

The transmitted signal for angle modulation is:

s (t )  Re{g (t )e jct }  Re{ Ac e j ( t ) e jct }  Ac cos[c t   (t )]

Average signal power:

Ps  s 2 (t )  12 | g (t ) |2  12  R 2 (t )  12 Ac2
Independent of the selected
type of angle modulation!
4
 Phase- and Frequency modulation (1)
Phase Modulation (PM):

 (t )  D p m(t )  so  (t ) is linearly proportional with the

information signal m(t).
 D p = phase deviation constant [rad/V]

Frequency Modulation (FM):

t
 (t )  D f  m( ) d   so  (t ) is proportional with the integral of
 the information signal m(t).
 D f = frequency deviation constant [rad/Vs]
(or sometimes given in [kHz/V]).
1 d (t )
f (t )   D f m(t )  m(t )
2 dt
5
 Phase- and Frequency modulation (2)

For PM and FM:  (t )  L{m(t )} , however, for g(t) we find:

g (t )  Ac e j ( t )  x(t )  jy (t )  Ac cos  (t )  jAc sin  (t )

 Ac cos  L{m(t )}  jAc sin  L{m(t )}
 a non-linear function of m(t )

Therefore, PM and FM are called non-linear modulation techniques

6
 Phase- and Frequency modulation (2)

Let m p (t ) and m f (t ) be the input signals of a PM and FM modulator,

respectively. What is the relation between m p (t ) and m f (t )?

1. Frequency modulation with an PM-modulator:

t
Df
m p (t )  m f ( )d  is the corresponding m p (t ) to obtain FM for
Dp 
m f (t ) with a frequency deviation constant D f using a PM modulator
with phase deviation constant D p .
7
 Phase- and Frequency modulation (3)

2. Phase modulation with an FM-modulator:

D p dm p (t )
m f (t )  is the corresponding m f (t ) to obtain PM for
Df dt
m p (t ) with phase deviation constant D p using an FM modulator with
frequency deviation constant D f .

8
 Direct phase- and frequency modulation

Voltage dependent frequency

determining circuit of an oscillator.

Voltage dependent phase

shifting circuit.

9
 Definition instantaneous frequency (1)

Instantaneous frequency f i (t ) :
Assume:

s (t )  R (t ) cos (t )  R (t ) cos(ct   (t ))
1 1 d (t ) 1 d (t )
 f i (t )  i (t )   fc 
2 2 dt 2 dt
t
So for FM we find with:  (t )  D f  m(  ) d 

1
f i (t )  f c  D f m(t )
2

Therefore, D f is called the frequency deviation constant

and the instantaneous frequency is linearly proportional with m(t).

10
Definition instantaneous frequency (2)

11
 Definition frequency deviation

The frequency deviation is related to the instantaneous frequency:

1 d (t ) FM 1
 f d (t )  f i (t )  f c   D f m(t )
2 dt 2

Peak-frequency deviation (“frequentiezwaai”): F  max{| f d (t ) |}  0

Peak-peak deviation: Fpp  max{ f d (t )}  min{ f d (t )}  2F

1
For FM: F  D f V p [Hz] with V p  max{m(t )}
2
For PM:   max{ (t )}  D pV p (usually < ) [rad].

12
 PM and FM with sine-wave modulation

PM with m p (t )  Am sin mt   (t )  D p Am sin mt

results in: s (t )  Ac cos ct  D p Am sin mt 

For both, PM and FM,
 Ac cos ct   sin mt  we find a similar signal
description.
t

FM with m f (t )  Am cos mt   (t )  D f A m cos m  d 



Am D f
 sin mt
m
 Am D f   F 
results in: s (t )  Ac cos  ct  sin mt   Ac cos  ct  sin mt 
 m   fm 

13
 Definition of the modulation index

For PM, the modulation index is defined as:  p    D pV p

For a simple detector,
with V p  max{m(t )}. Now:  max   , otherwise
ambiguity in phase.

s (t )  Ac cos ct   sin mt   Ac cos ct   p sin mt 

F
For FM, the modulation index is defined as:  f  where B is the
B
bandwidth of the modulation signal m(t): for tone modulation B  f m .

 F 
Now: s (t )  Ac cos  ct  sin mt   Ac cos ct   f sin mt 
 fm 
For PM and FM with sine-modulation we find equal modulation index:
 p   f when   F / f m.
14
 Signal spectrum (1)

The frequency spectra of PM and FM modulated signals are found

according to the definitions:

S ( f )  12 [G ( f  f c )  G * ( f  f c )]

with: G ( f )  F{g (t )}  Ac F{e j ( t ) }  Ac F{e jL{m ( t )}}

Problem: for angle-modulation, g(t) is a non-linear function of m(t).

Therefore, a simple relation between M ( f ) and G ( f ) , like
for the linear modulation schemes: AM, DSB and SSB, does not exist
for PM and FM. Also superposition cannot be applied in PM and FM.

Only for a few cases it is possible to derive the signal spectrum for PM
and FM analytically. One of these cases is tone-modulation.

15
 Signal spectrum (2)

To analyze the signal spectrum with tone-modulation, we take:

1. for PM: m p (t )  Am sin mt
  (t )  D p m p (t )  D p Am sin mt   sin mt
p
2. For FM: m f (t )  Am cos mt
t t

  (t )  D f m f ( )d   D f A m cos m  d 
 

Am D f F
 sin mt  sin mt   sin mt
m fm
f

16
 Signal spectrum (3)

So for both, PM and FM, we can now write: g (t )  Ac e

j ( t )
 Ac e j sin mt

Since g(t) is a periodic function with period Tm  1/ f m , g(t) can be

rewritten as a Fourier series:
 Tm /2
jnmt Ac
g (t )   n
c e with cn   e j sin mt e  jnmt dt
n  Tm Tm /2
uTm Tm
Substitute u  m t  t  2  um  dt  2 du

Ac
 cn   e j (  sin u  nu )
du  Ac J n (  ) J  n (  )  (1) n J n (  )
2 

Bessel function of the 1st kind and order n

17
Signal spectrum (4)

18
 Signal spectrum (3)

Using this, we find for the signal s(t):


jct jnmt jct
s (t )  Re{g (t )e }  Re{ Ac J
n 
n (  ) e e }
 Frequency components at f c  nf m
=Ac J
n 
n (  ) cos(ct  nmt )

For the frequency spectra of g(t) and s(t) we find now:

  
jnmt
G ( f )  F{  cn e }  c  ( f  nf
n m )  Ac J n (  ) ( f  nf m )
n  n  n 

Ac
S( f ) 
2
J
n 
n (  )[ ( f  nf m  f c )   ( f  nf m  f c )]

Transmission bandwidth infinite, in principle.

19
 Signal spectrum (4)

Now we find for the power spectral density:

Ac2  2
Ps ( f )  
4 n
J n (  )[ ( f  nf m  f c )   ( f  nf m  f c )]

20
Signal spectrum (5)

21
Signal spectrum (6)

22
Signal spectrum (7)

23
 Transmission bandwidth

In principle, the absolute transmission bandwidth BT of PM and FM

is infinite!! BT depends on  and B (or f m ).

Carson’s rule: 98% of the transmitted power is contained in the

bandwidth 2(   1)B for arbitrary m(t).

Carson bandwidth: BT  2(   1) B B is the bandwidth of m(t).

For 2    10 a better approximation for BT is: BT  2(   2) B.

For tone modulation with frequency f m : BT  2(   1) f m

24
 Narrowband angle modulation (1)

For small angle variations: |  (t ) | 0.2 rad ez  1  z | z | 1

g(t) can be approximated as: g (t )  Ac e j ( t )  Ac [1  j (t )]

Now we find for s(t): s (t )  Ac cos c t  Ac (t )sin c t

carrier sideband: 90o out of phase

Ac
with: S ( f )   ( f  fc )   ( f  fc )  j ( f  fc )  ( f  fc )
2

 Dp M ( f ) for PM

where: ( f )  F{ (t )}   D f
 2 jf M ( f ) for FM

25
Narrowband angle modulation (2)

26
Detection performance for analogue modulations
Signal quality after detection is determined by:
- received signal power
- noise- and interference power at the receiver input

Receiver

The Signal-to-Noise ratio at the detector input:

received signal power
SNRin 
noise power in BT
determines: - SNRout after detection for analogue signals
- bit error probability (BER) for digital signals
27
 Signal detection quality

Analog modulation: there is a linear relation between

SNRout after detection and SNRin.

- AM, DSB, SSB: SNRout  2 SNRin (BT  2B or B for SSB)

- FM, PM: SNRout >> SNRin is possible

if BT >> 2B, and SNRin > SNRth

A better SNRout can be obtained at the cost of a larger BT .

28
 SNRout for analog modulations (1)
Additive noise in bandpass systems:
BT
f c , BT

fc
N0 rin (t )



/
/
rout (t )

fc
N0

Bandpass filter

r (t )  s (t )  n(t ) n(t) is additive white Gaussian

noise (AWGN), with PSD
 Re{[ g s (t )  g n (t )]e j (ct c ) } Pn(f) = N0 /2 (double sided).

Baseband equivalent complex envelope:

gT (t ) ˆ g s (t )  g n (t )
 [ xs (t )  xn (t )]  j[ ys (t )  yn (t )]
in-phase quadrature-phase
29
 SNRout for analog modulations (2)

Reference  baseband system (without modulation): standard

S Ps Ps
   
 N baseband N 0 B ( N 0 / 2)  2 B

SNR at detector input:

In the following, we assume that
S Ps the bandwidth used for noise
   power calculations refers to the
 N in N 0 BT equivalent noise bandwidth.

S B
 
 N baseband BT

30
 Coherent detection (AM, DSB and SSB)

Coherent detection without phase error (lo   c  0 ) gives:

 (t )  Re{g (t )}  x (t )  x (t )
m T s n

31
 Coherent detection (AM, DSB and SSB)

Let: r(t )  s(t )  n(t )  [ xs (t )  xn (t )]cos ct  [ ys (t )  yn (t )]sin ct

then we find after coherent detection without

phase error and  c  0 :

2r (t ) cos c t  2[ xs (t )  xn (t )]cos 2 c t  2[ ys (t )  yn (t )]sin c t cos c t

 2[ xs (t )  xn (t )]( 12  12 cos 2c t )  [ ys (t )  yn (t )]sin 2c t
LPF
 [ xs (t )  xn (t )]
Check that with phase error  e :
LPF
2r (t ) cos(c t   e )  [ xs (t )  xn (t )]cos  e  [ ys (t )  yn (t )]sin  e
Check again: Product detector circuit § 4.13
32
 SNR: coherent detection AM (1)

AM-modulation: gT (t )  Ac [1  m(t )]  xn (t )  jyn (t )

 (t )  Re{g (t )}  A  A m(t )  x (t )
and m xn2 (t )  yn2 (t )  2 N 0 B
T c c n

Ac2 A 2
Input signal power: Ps  [1  m(t )]2  c [1  m 2 (t )]
2 2
What happens when there is a
Now we find for SNRout: phase error lo   c in the coherent
detector?
Ac2 m 2 (t )Ac2 m 2 (t )
SNRout  
2
xn (t ) 2 N0 B

§6.7, equation 6.133

Stoch. Proc.
33
 SNR: coherent detection AM (2)
Ac m 2 Ac m 2
srms  srms 
^ 2 ^ 2

xn2  2 B  ( N 0 / 2) xn2  2 B  ( N 0 / 2)

After coherent detection: based on sideband power

srms  Ac m 2
^ S Ac2 m 2 (t ) Ac2 m 2 (t )
   
N
  out 2
xn (t ) 2N0 B

Noise components at positive and

negative frequencies are
xn2  2 B  N 0 uncorrelated: their powers add up.
The signal components are
Based on sideband power SNRout  2SNRin . correlated: their amplitudes add up.
34
 SNR: coherent detection AM (3)
Ac2
With input signal power Ps  (1  m 2 (t )) and noise
power Pn  2 N 0 B we find: 2

Ac2
(1  m 2 ) SNRout 2m 2
SNRin  2   1
2 N0 B SNRin 1  m 2

Comparison to baseband:
Ac2 2
(1  m )
Ps SNRout m2
SNRbaseband   2    0.5
N0 B N0 B SNRbaseband 1  m 2

35
 SNR: coherent detection AM (4)

2
For sine wave modulation: m  0.5 and

SNRout 2 SNRout 1
 , 
SNRin 3 SNRbaseband 3

So AM performs at least 4.8 dB worse than baseband (AM

is sub-standard)  loss due to carrier power  3 dB.

36
 SNR: coherent detection DSB-SC

DSB-modulation: gT (t )  Ac m(t )  xn (t )  jyn (t )

 (t )  A m(t )  x (t )
and m c n
Ac2 2
Ac2 2 m (t )
S
Using the signal power: Ps  m (t ) ,    2
2 2 2  N in 2 N0 B
S Ac m (t )
and   
 N out 2 N0 B

SNRout SNRout
we find:  2,  1  standard
SNRin SNRbaseband
3 dB improvement is obtained by coherent addition of the
sidebands.
37
 SNR: coherent detection SSB
USSB
ˆ (t )]
SSB-modulation: g s (t )  Ac [ m(t )  jm LSSB

 gT (t )  [ Ac m(t )  xn (t )]  j[ Ac mˆ (t )  yn (t )]

 (t )  Re{gT (t )}  Ac m(t )  xn (t ),
Coherent detection: m xn2 (t )  N 0 B
S Ac2 m 2 Ac2 m 2
    
 N out xn2 N0 B

Signal- and noise power at detector input:

2 Ac2
Ps  1
2 g s (t )  2 [m 2  mˆ 2 ]  Ac2 m 2 Pn  N 0 B
SNRout SNRout
 1 1
SNRin SNRbaseband
The performance of SSB is identical to baseband: standard.
38
Coherent detection (SSB)

39
Non-coherent AM-detection: envelope detector
s (t )  R (t ) cos(c t   (t )) Vout  KR(t )
Envelope
detector
DC-block

Vout  KR(t )  KAc [1  m(t )]  KAc  KAc m(t )

1 1
   RC 
fc B

Note: the rectifying envelope detector

is a non-linear component:
R (t )  x 2 (t )  y 2 (t ) !

But … , it is cheap!!!
40
 SNR: AM with envelope detection (1)

The output signal of an envelope detector with gain K is:

non-linear

r0 (t )  KRT (t )  K | gT (t ) | K | g s (t )  g n (t ) |
 K | [ Ac (1  m(t ))  xn (t )]  jyn (t ) |
with
g n (t )  xn (t )  jyn (t )  Rn (t )e jn ( t )

The output signal of the non-linear envelope detector is given by:

2
r0 (t )  KRT (t )  K  Ac (1  m(t ))  xn (t )  yn2 (t )

41
 Noise and Rayleigh distribution (1)
In the complex envelope of a noise signal:
g n (t )  xn (t )  jyn (t )  Rn (t )e jn ( t )
xn  yn  0,
xn and yn are normally distributed with
xn2  yn2   2  2 N 0 Bn
and Pxn ( f )  Pyn ( f )  N 0  white Gaussian noise in baseband

The amplitude Rn (t ) | g n (t ) | has a Rayleigh distribution and its phase

 n (t ) is uniformly distributed over [0, 2 ) (pag. 442 - 444).
 Rn  Rn2 /2 2
 2e , Rn  0
f Rn ( Rn )  
0 , Rn  0

met Rn   and  R2n  1  4   2
4
42
Rayleigh distribution (2)

43
 SNR: AM with envelope detection (2)
2 2
For SNRin  1, thus ( xn / Ac ) ,( yn / Ac )  1 the envelope can be
approximated by:
1

r0 (t )  K  A [1  m(t )]  2 Ac [1  m(t )]xn (t )  x (t )  y (t )

2
c
2 2
n
2
n
2

1
2 2 2
 2 xn (t ) x (t )
n y (t )
n

 KAc [1  m(t )] 1   2 2
 2 2
 Ac [1  m ( t )] Ac [1  m (t )] Ac [1  m ( t )] 
 KAc [1  m(t )]  Kxn (t ) x
(1  x)  1  if x  1
2
 r02 (t )  K 2 Ac2  K 2 Ac2 m 2 (t )  K 2 xn2 (t )
Thus for SNRout we find: Signal and noise are
additive, i.e. independent!
Ac2 m 2 Ac2 m 2
SNRout  2 
xn 2 N0 B
identical to the coherent product detector, if SNRin  1.
44
 SNR: AM with envelope detection (3)
For SNRin 1 , the output signal of
the envelope detector (non-linear)
will follow the noise amplitude in
stead of the signal amplitude.
gn (t )

Now, signal and noise

are multiplicative!

r0 (t )  KRT (t )  K | Ac [1  m(t )]  Rn (t )e jn ( t ) |

 K { Ac [1  m(t )]cos  n (t )  Rn (t )}

For small SNRin , SNRout decreases very fast: threshold effect.

45
 Non-coherent detection of AM

Usually, an envelope detector (much cheaper) is used for detecting

AM signals in stead of a coherent product detector:
• for SNRin  1  signal and noise at the envelope detector output
are additive
 performance equal to the coherent detector
• for SNRin 1  noise becomes multiplicative!!
 very bad performance.

Somewhere, a threshold occurs between linear and non-linear

detection behavior. This threshold is found around SNRin = 10 dB.

In practice this threshold is not very relevant for audio/voice signals,

because the SNR for these signals has to be better than 20 – 25 dB
anyway. For AM-data transmission, however, this threshold is
dramatic, and usually a coherent detector has to be used.

46