S-72.

1140 Transmission Methods in

Telecommunication Systems (5 cr)

Noise in analog carrier wave (CW) modulation

systems
 Noise in analog CW modulation systems

 Understanding noise
– Lowpass presentation of bandpass noise and its conversion
to baseband noise
– Noise statistics of quadrature presentation in rectangular
and polar coordinates
 Modeling detectors for linear and exponential modulation
 Analysis of post-detection SNR
– Synchronous detector
– PM-detector
– FM-detector

2 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Noise in carrier wave modulation
systems:basic definitions
 Objectives: Define bandpass noise and use it to analyze post
detection SNR of analog CW systems
 Assume signal is ergodic, e.g., all ensemble averages E[ ] equal
the corresponding time averages <>. Then, for instance
 v(t )  E[v(t )] average value
 v 2 (t )  E[v 2 (t )] average power
 v(t )v(t   )  E[v(t )v(t   )] autocorrelation
where the time average is defined by
1 T /2
 vi (t )  lim  T / 2 vi (t )dt
T 
T
or
1 T /2
 vi (t )   T / 2 vi (t )dt (for a known period)
T
3 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 Assignment

 v(t )  E[v(t )] 1 T /2
 vi (t )  lim  T / 2 vi (t )dt
T 
T
 v 2 (t )  E[v 2 (t )]
1 T /2
 v(t )v(t   )  E[v(t )v(t   )]  vi (t )   T / 2 vi (t )dt
T
4 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 The system model
 We consider normalized ergodic analog message whose
amplitude and power are normalized
x(t )  1, S x  x 2  x 2 (t ) , S x  1
Channel loss Post-detection filter
Modulated signal Pre-detection filter

Detector

Transmitted power
Received power Pre-detection noise (after HR)
S R  ST / L  xc2 Received signal (not altered by HR)

5 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Detection models
 Pre-detection signal v(t) is presented in quadrature-carrier form:
v(t )  Av (t ) cos[ ct   v (t )] vi (t )  Av (t ) cos v (t )

 vi (t )cos( ct )  vq (t ) cos( ct ) vq (t )  Av (t )sin v (t )
 Detection models:

vi (t ) Synchronous detector


 Av (t )  Av Envelope detector
y (t )  
v (t ) Phase detector
d (t ) / dt Frequency detector
v

(Remember that FM was defined by:

xC (t )  AC cos[ C t  2 f  t x( )d  ], t  t0 )
t

6 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Pre-detection noise in bandpass channel

 Signal and noise are statistically independent and therefore their

power can be added to form the total pre-detection power:
v 2  xc2  n 2  S R  N R
 The pre-detection (bandpass) noise power is filtered from the
channel noise:
N R   ( / 2) H R ( f ) df  20 ( / 2)df   BT
 2 BT

from to detector
channel

7 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Pre-detection SNR
 Pre-detection signal-to-noise ratio for bandpass channels is defined by

 Note that above BT isSthe

R
/N R
 S R /( Bbandwidth
transmission T
) passing channel noise
power to the detector


For comparison,Nwecan write
 the received signal-to-noise in terms of
 2

R
BT
( / 2) H
 R ( f ) df
baseband system (BW = W) SNR defined by

and therefore also

  S /(W )
R
 Note that always (limiting case is the SSB with BT = W)
S R / N R  S R /( BT )  ( S R / W )(W / BT )   W / BT

(We will see, however, that post detection SNR can be much larger than
!) BT  W  S R / N R  


8 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 Assignment

S R / N R  S R /( BT )  ( S R / W )(W / BT )   W / BT

9 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Bandpass noise: n(t )  A (t ) cos[ t   (t )] n c n

 We assume stationary, zero mean Gaussian noise process for

which
n  0, n 2   n2  N R
 Bandpass noise in terms of lowpass equivalent signals
n(t )  ni (t )cos( c t )  nq (t )sin( ct )
 The in-phase and quadrature components nq An
are independent and hence n

n
ni (t )nq (t )  0 i

 Their average is zero ni  nq  0 and their average power is the

same: 2 2 2
ni  nq  n  N R

10 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Bandpass noise has Rayleigh distributed
envelope and evenly distributed phase
 I-Q components of the bandpass noise can be presented in
envelope - phase format: Two independent
nq r.v.:s - sum of their
An  ni  nq , n  arctan
2 2 2
variances equals variance
ni of the envelope
 The PDF of envelope is Rayleigh
distributed defined by
An  An2 
p A ( An )  exp    u ( An )
n
NR  2NR  nq An
n
 Therefore mean and variance for the n i
bandpass noise are (integrate from above, how?)
An   N R / 2, An2  2 N R
 N R  An2 / 2  (ni2  nq2 ) / 2 n n n2  n 2 2 2
i q

11 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Assignment

12 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Post detection noise in synchronous detection
 Signal component of synchronous detector:

vDSB (t )  x (t ) exp( j ct )
received DSB signal
vDSB (t )  Ac x(t ) cos( ct )  jAc x(t )sin( ct )
Ac x(t )
vDSB (t ) cos( ct ) 
2 detected message
A x(t ) A x(t )
 c cos(2 c t )  j c sin(2 ct )
2 2
 Noise component of synchronous detector:

n(t ) cos( ct )  cos( ct )  ni (t )cos( ct )  nq (t )sin( ct ) 

 ni (t )  1  cos(2 ct )  / 2  nq (t )sin(2 ct ) / 2
 Detector extracts i-components and removes double frequency
components
yD (t )  Ac x(t )  ni (t ) cos( x) cos( x)  1/ 2  cos(2 x) / 2
cos( x)sin( x)  sin(2 x) / 2
13 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 Post-detection SNR for DSB

 Obtain signal and noise power after detection from:

yD (t )  Ac x(t )  ni (t )
where average noise and signal power are
N D  ni2 (t )  , S D  Ac2 x 2 (t )  Ac2 S x
Received average signal power is
x 2 (t ) cos 2 ( 0t )  Ac2 S x / 2  S x  2S R / Ac2
S R  Ac2 
   
Sx 1/ 2

and therefore SNR after DSB detector is

S D / N D  Ac2 S x / N D
Ac2 2 S R
2S R 2S R SR
 2      (DSB)
Ac N D N D  BT BT 2 W (DSB)
W

14 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Comparing SNR for DSB and AM

 It can be show, that for AM the post detection SNR is

Sx
SD / N D 
AM
1  Sx
 Comparison of this to the SNR of DSB can done by noting that
in practice
Sx  0.5 tone modulation  SD  0.5 /(1.5)   / 3

S x  0.1 speech signal  SD   /11
 Hence AM performs usually much worse than DSB
 It can be shown that for SSB performance is the same as for
DSB, e.g.
SR
S D / N D USB ,DSB  
W

15 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Exponential modulation and channel noise

 Both PM and FM have constant envelopes so the received

power is constant
xc (t )  Ac cos[ ct   (t )]
S R  xc2  Ac2 / 2
SR Ac2 Ac2
 Received SNR is   yielding for wideband FM
N R 2 N R 2 BT
SR SR SR 
  
N R  BT  2 DW 2 D
where for wideband modulation
BT  2(   2) f m  1, f m W
 2W
 2 DW  2 f  (D  f  / W )

16 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Detection of exponential modulation
assuming small noise power

v(t )  Ac cos[ ct   (t )]  An (t ) cos[ ct   n (t )]

              
Signal Noise
carrier+noise
An (t )sin[ n (t )]
 (t )  arctan
Ac  An (t ) cos[ n (t )]
v

small compared to Ac

An (t )sin[ n (t )]
 (t ) 
v
noise
Ac Detected noise component
17 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 Post-detection noise spectra for PM
An (t )sin  n (t )  S R  Ac2 / 2
v (t )   Note that after
Ac nq  An sin  n
detection signal
bandwidth is W and
nq (t )  2 thus a post detection
 n
 i  nq
2
 N R   BLP filter is required to remove
2S R out-of-band channel noise

 The channel noise is bandpass noise filtered at the transmission

bandwidth and therefore the respective post-detection noise power
spectral density GPM(f) and the total noise power ND are
2
 nq (t )   f 
GPM ( f )     
 2 S R   BT 
  f 
  
2 S R  BT 
N D  W GPM ( f )df  W / S R
W

BLP
18 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 Assignment

An (t )sin[ n (t )]
v (t )  arctan
Ac  An (t )cos[ n (t )]

An (t )sin  n (t )  S R  Ac2 / 2
v (t )  
Ac nq  An sin  n
nq (t )  2
 n
 i  nq
2
 N R   BLP
2S R

Explain how the results shown connects to the phasor diagram!

19 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Post-detection SNR for FM
 Recall the definition of FM-signal
xc (t )  AC cos  C t  v (t )  , v (t )  2 f   x( )d 
t

t0
 Frequency discriminator (detector) differentiates the
instantaneous phase to cancel out the inherent integration in
phase. Now  (t )  2 f  x(t )  dv (t ) / dt  Not effected by f
due to small angle appr.
 
dv (t ) 1  dS (t ) d N (t )  1  d S (t ) d  nq (t )  
x(t ) f         
2 dt 2   dt  dt  2  dt 2 S R dt 
 Signal Noise 
 Inspection in frequency domain (In order to find the respective
noise PSDs) after detector
X ( f )  j 2 f  v ( f )
and the noise PSD
2
is
GFM ( f )  Xˆ ( f )  ( 2 f ) 2  N ( f ) / (2 ) 2  f 2  N ( f )
2 2

  2 f (t )  d C (t ) / dt A (t )sin  n (t ) nq (t ) d n x(t )
v (t )  n   ( j 2 f ) n X ( f )
 2 [ fC  f  x(t )] Ac 2S R dt n

20 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Post-detection SNR in FM (cont.)
 Therefore, the post-detection noise PSD can be written as

GFM ( f )  f  N ( f ) with 2 2
and now the PSD for FM post detection noise is

Nq ( f )
N ( f ) 
and the respective total noise power is

2SR

f  f  2

G (f)
FM
 
2S  B  R T

N D  W GFM ( f ) df  W 3 /  3S R 
W

21 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Destination S/N for PM and FM

 For PM we have
S 2

S /N    S  , where  S  
 x 2 2 2

W / S
D D  x  x
R

 For FM we have  W
N D ( PM )  W df 
W

f 2 S x 2S S
SD / N D  R R

W 3 /  3S R  W
f 2
W 3

2 N D ( FM )   df 
f  S W 2S 3S
 3    S x R  3 D 2 S x R R

W   W
D 
Note that SD/ND can
 Under wideband condition D  1 and be increased just by
increasing deviation!
BT  2(   2) f m  1, f m W
 2 W 2
3 B 
 2 DW  D  BT /(2W ) S D / N D   T  S x
4W 
22 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen
 FM preemphases and deemphases filters
 FM related noise emphases can be suppressed by pre-
distortion and post detection filters (preemphases and
deemphases filters):

receiver filter Hde ( f )

transmitter filter H pe ( f )

Hde ( f )  1  j ( f / Bde )
1 R
S
1, f  B de
Q: What would happen
if the filters would be TB / ( jf ), f  B
de de
LPF
reversed? (TX filter in
 j ( f / Bde ), f u  f  Bde
receiver & vice versa) H pe ( f )   1  j ( f / Bde )    HPF
 1, f  f u

23 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 FM post-detection S/N with deemphases

 Deemphases filter (that is a lowpass filter connected after

detector) can suppress noise further. FM post-detection noise
PSD and total noise power without deemphases:
f  f  2
W 3
G (f)    N   G ( f )df  W
W
2S B 
FM D FM
R 3S T R

 With deemphases filter (for simplification assume W/Bde>>1):

 
 Bde3  W W 
N D  W GFM ( f ) H de ( f ) df    arctan  
W 2

 B 2
W / SR
S R  Bde Bde
de

where     
1  )
H de ( f ) 
2
W / B  arctan(W / B de de

1  ( f / Bde ) 2
2 2 W  Bde
f Sx SxSR  f 
2
 f 
SD / N D  2     S x  
 BdeW / S R W  Bde   Bde 

24 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Example
 FM radio
f   75 kHz,W  15 kHz, D  5, S x  1/ 2, Bde  2.1kHz
 Without deemphases
S D / N D  3D 2 S x
 (3  52  1 2 )  38 SR

 With deemphases W
S D / N D  ( f  / Bde ) 2 S x  640
 Therefore if DSB or SSB system could be exchanged to FM
system 640 fold transmission power savings could be achieved.
 Note, however that the required transmission bandwidth is now
about 220 kHz /15 kHz = 15 times larger! Also, a problem is the
FM threshold effect that causes the more abrupt post detection
SNR drop, the larger the expansion factor BT/W gets.

25 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Comparison of carrier wave
modulation systems

26 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen

 Note to earlier lecture: LP-filter is an
approximation of the ideal integrator
Zin-> <-Zout
Vin Vout

H ( f )Vin ( f )  Vout ( f )
H ( f )  I ( f ) Z out ( f ) /[ I ( f ) Z in ( f )]  Z out ( f ) / Z in ( f )
( j C ) 1 1 1
H( f )   H( f )  ,  1
R  ( j C ) 1 j RC  1 j RC
Ideal integrator is defined by
t  Vin ( f ) 1
Vout ( f )  …  vin ( )d     Vin (0) ( f )
   j 2 f 2
1 1
Vout ( f ) V   (f)
in ( t )  (t )
j 2 f 2
…  (t   d )   exp( j d )
27 Helsinki University of Technology,Communications Laboratory, Timo O. Korhonen