Noise in Frequency

Modulation
Chapter-4
Contents
1. FM representation signals
2. Noise calculation in FM
3. Noise in FM phasor diagram
4. SNR Calculation of noisy FM
5. Figure of merit in noisy FM
6. SNR Improvement: Pre‐emphasis and de‐emphasis
7. Threshold effect in Noisy FM
8. Threshold in Frequency Modulation
9. Numerical problems and solutions
 Representation of FM Signal
Let the angle modulated signal is given by Then f i t   f c  1 d
 t 
2 dt
ut   Ac cos t 
If m(t) is a message signal ,then in FM system, we have
Where  t  is the phase of the signal ,
f i t   f c  k p mt  
1 d
 t     (2)
2 dt
and its instantaneous frequency is given by
Where kf =frequency deviation constant
f i t  
1 d
 t 
2 d Take integration of both the sides of eqn (2) then
Since u(t) is a band pass signal, it can be represented as k p m(t ), for PM ( Phase mod ulation

ut   Ac cos2f c t   t           (1)  t    t

2k f m d , for FM System


 

From eqn(2) it is found that the demodulation of an FM signal involves finding instantaneous frequency of the
modulated signal and then subtracting the carrier frequency from it. f max  k f max mt  
The maximum frequency deviation in an FM system is given by
when m(t )  a cos 2f m t or a cos 2fwt
 t
 t   2k f m d
Noise in FM


t
 2k f  cos 2f

m  d

kf
or  t   a sin 2f m t
fm
The signal is given by
kf  
let  or  f 
t
a  Modulation indices
fm st   Ac cos2f c t  k f m d  ,where k f is in time domain

 
 
then u t  s (t )  Ac cos 2f c t   f sin 2f m t 
  

 t 
For non-sinusoidal signal, s t   Ac cos2f c t  2k f  m d  , where k f is in frequency domain
 
 
 

k f max mt 
f  The noise is given by
w or f m
nt   n s t  cos2f c t   n c t  sin 2f c t 
Where w=bandwidth of the signal nt   r t  cos2f c t   t 
So, the noise also can be written as
nt   r t  cos2f c t  t 

Where r t  and  t  are random variables

So, we know
 t 
s t   Ac cos2f c t  2k f  m d 
 
  

After passing through the band pass filter

xt   st   nt 

 t 
x t   Ac cos2f c t  2k f  m d   r t  cos2f c t   t 
 
  

xt   Ac cos2f c t   t   r t  cos2f c t  t 

 Here Ac and r(t) are two vector , has two angles  t  and  t 
 In this the signal taken as reference i, e amplitude and phase both are taken as reference.
 Now, the angle of r(t) = reference angle minus(-) r(t) vector angle
=  t    t 

 Now, let us take the angle made by x(t) is  t  , but x(t0 is making angle with reference
 t 
 So, that the actual angle becomes  t    t 
 Where  t  is the reference angle.
 xt   Ac cos2f c t   t   r t  cos2f c t  t 
C

r(t) makes an angle ψ(t)

AB  r t  cos t    t  Ac

BC  r t  sin  t    t 
O

 r t  sin  t    t  
 t    t   tan 1  
 Ac  r t  cos t    t  
AB  r t  cos t    t 

BC  r t  sin  t    t 

 r t  sin  t    t  
 t    t   tan 1  
 Ac  r t  cos t    t 

As r t   Ac , we can write

r t  sin  t    t 
 t    t   tan 1
Ac

And  t  is the phase angle of x(t)

Now the discriminator will do differentiations and detect the envelope.
The differentiation of the phase angle  t  will gives us

vt  
1 d
 t 
2 dt

 t   1 d  r t  sin  t    t  
 
vt  
1 d
2 dt 2 dt  Ac 

t
 t    m d
1
 2k f
2


Taking differentiation

d t  d  
t
 m d 

1
  2k f
dt 2 dt  
 

So, now the differentiation and integration cancel each other, and we get
d t 
 k f mt dt
dt
 1 d  r t  sin  t    t 
Let n d t          (X )
2 dt  Ac 

Here  t  related to noise and  t  related to signal.( so, there is no issue with  t  .

The equation(x) shows that the noise part still have message signal information due to  t  .

The eqn(x) shows the noise variation is from 0 to 2 due to sin t    t  , we can write with
ot loss of generality then

n d t  
1 d
r t  sin  t 
2Ac dt

Quadrature components
n d t  
1 d
r t  sin  t  For ideal differentiator
2Ac dt
d
Quadrature components  2jf  j
dt

The PSD of nd t  is given by

n d t  
1 d
2Ac dt

nc t  or nQ t  
S nd f  Hf  2 S nQ  f . 1
2Ac
Quadrature components
 2f 2 S nQ  f . 1
2Ac

 f 
2
BT
S nd

 f    Ac
 SN
  f , f 

q
2

0, , Otherwise

If the noise we take, the nQ t  can be represented with in the desired band. The meaning of
n d t  is the noise nQ t  passed through the ideal differentiator circuit
 2
 f 
Now plot S nd  f     SN f 

 Ac 
q

When f=0, S nd  f   0 , when f varies , we get the characteristics as shown in figure of demodulation.
W 2
f
Noise power
PNoise  
W
A 2
c
N 0 df

N 0 2W 3
PNoise  2
Ac 3
And the signal part is given by k f mt 

Now, the signal power is given by

Now put the LPF, to pass noise which
has a band as –w to +w PSignal  k 2f PM

3k 2f PM AC2 3PM AC2

SNR FM   k 2f
S Nq  f   N 0 2 N 0W 3 2 N 0W 3
3PM PR
 f 
2
 k 2f
S nd  f      N 0        (Y )
 N 0W 3
 Ac 
Figure of merit
Ac2 3PM PR
Input signal power = S i or PR  SNR FM  SNR 0  k 2f
2 N 0W 3

PR  Re ceived signal power  Ac2 Figure of merit is given by FOM 

SNR 0
SNR i
Input noise power
Ac2
As we know S i or PR 
w 2
N0
Ni  
w
2
dw
3k 2f PM AC2 3PM AC2
SNR FM   k 2f
N0 2 N 0W 3 2 N 0W 3
Ni   2W  N 0W
2 3PM AC2 2 N 0W
FOM  k 2f 
2 N 0W 3 AC2
Ac2 2
Si 2  A 3k 2f PM
 SNR i  c
FOM 
Ni N 0W 2 N 0W W2
So the Figure of Merit is given by
When increases, which means it is Noise is higher at higher frequency
wider band of frequency modulation. In If noise is lower, the frequency also
WBFM we will get better noise lesser, which gives invalid reason for
cancellation. frequency
So, we have to lower the noise and
frequency should also at higher side.
So, we will go for pre-emphasis and
de-emphasis

At this position the signal

strength is lower.
PRE-EMPHASIS AND DEEMPHASIS
By employing pre-emphasis and de-emphasis 13db gain increase

H pe  f  H de  f   1
So

We have to know what should be the noise and signal after de-emphasis

S Nd  f 

The signal will be same as before, but the noise has to go through Rx and De-emphasis
After receiver ,the noise PSD is S Nd  f  , it will go through the de-emphasis filter and it is
chosen to be a LTI system
S No  f   H De  f  S Nd  f 
2
So S No  f   H De  f  2 S Nd  f 

S
 
Average noise power without pre & de  em filter P no
 N  OPD
Let I  Average noise power with pre & de  em filter
 
P nPD
S
 
 N O
S
 
 f 
2  N  OPD P no 2W 3
     N 0        (Y )  I W
S
We know nd f  A  S P nPD
 c   
 H De  f  df
2 2
3 f
 N O
Noise power with Pe and De W

W
N0 In other way we can write
A f H De  f  df
2
PnPD  2
2

c
W

Noise power without Pe and De

f 
W
N0 N0

3 w 1
Pno  f 2 df  w 
W
A2 c A2 c 3
N0 2 3
Pno  2 3
w
A c
Now the pre-emphasis filter is a high pass filter to keep higher frequency intact
and lower frequency component by a LPF i,e de-emphasis filter

S 
 
 N  OPD P no 2W 3 For FM with pre-emphasis and
 I W
We know  S  P nPD de-emphasis filtering, we have
 
 H De  f  df
2 2
3 f
 N O
For a low pass filter W

2W 3 2W 3
I  
H De  f  
1 W W 2

 H De  f  df 
f 2 2 f
3 f 3 df
1 j  f 
2
f0 W W
1   

 0
f
3
Where f 0 
1  w 
2RC  
 f 
 0 
H Pe  f  is the inverse of H De  f  . I 
 w 
1  w 
3  tan  

 0
f  0  
f
H Pe  f  
1 f
 1 j
H De  f  f0 If f0=2.1 kw, w=15khz( for musical instrument)
Then I=22 or I in dB=13dB
Threshold effect of noise in FM
If FM captured by the noise, which is bad in analysis, I,e called threshold effect of noise or click
noise of FM.

 Noise is uncorrelated

t
 t   k f  m d
xt   st   nt  

 t  And k f =instantaneous frequency

xt   Ac cos2f c t  2k f m d   r t  cos2f c t   t 
 
  

xt   Ac cos2f c t   t   r t  cos2f c t  t 

Noise part  r t  cos2f c t  t 

 If you see, when the amplitude Ac is larger
than length of r(t) in vector diagram, there
will not be any such more problem in noise r(t)
analysis as r(t) noise component is lesser.

 Now when you see , the length of r(t) is more

than Ac component length, the circle is bigger,
which indicate that the noise also bigger which
can be seen from the phasor diagram

 The noise is random, it may happen, it goes

to –ve side and have a (180 degree)
When r(t) is bigger, then angle  t  also varies as  t    t  to  t    t 
As Ac and  t  is taken as reference, when Ac is small,
obviously  t  becomes small or nearly zero.
When r(t) becomes –ve, due to random nature, there will be a shift of (pi) or 180 degree
shift is there.

When r(t)>>Ac, then the angle becomes  t    t 

If the noise in case of Ac>>r(t), then it is given below in figure.

Note :
 S   S 
  SNR 0  3 f PMn 
2
3PM PR  
We know SNR FM  SNR 0  k 2
f  N O  N b
N 0W 3
For non-sinusoidal signal, Using carson;s rule the band width is given by

f 
k f max mt    BC  2  1W
W BC
and  2  1
kf f W

W max mt    Let
BC

W
2
 S  3PM PR  f 
SNR FM     SNR 0   And   2  1
 N O N 0W  max mt    

 S  PR 
If we denote    Or   1
2
 N b N 0W
2
 S   f   S     
2
   SNR 0  3PM       1 
Then SNR FM
 N O 
 max mt     N  S   2   S 
     SNR0  3PM   
 
b

PM  N O

max m t  

 N b
The is the average to peak power ratio of the message signal or
maxmt  
2  

normalized message PMn

Threshold effect in Angle modulation
The threshold effect shows SNR as

S 
   20  1
 N b

For a sinusoidal message signal

PM 1

max mt  2 2

In such case
 S  3 2 S 
   SNR 0    
 N O 2  N b

 S 
   SNR0  60   1PMn
2
 N O
Q-1
Note
Q -1
Q-3