L18: FM Threshold Effect & Pre /de Emphasis: N S T M P D N S

L18:
FM Threshold Effect & Pre‐/De‐emphasis
Last lecture
S 3D 2 PM S
 FM output SNR is    2  
 N  o max t | m(t ) |  N  b
Can we make this arbitrarily large by increasing D
(& thus the required transmission BW?)
 2 N0
(2f ) 2 , if | f | W
 LPF’ed output noise psd is S Z ( f )   Ac .
 0, if | f | W
I.e. noise affects higher freq. message components worse.
How can we mitigate this?
L18 1
Threshold Effect
Output FM noise actually depends in a complicated &

highly nonlinear way on both message and channel noise.
(just like AM)
If channel noise power N 0 B  0.5Ac2 , then output noise is

additive & uncorrelated with message
 Output SNR is well-defined & also decreases linearly
with carrier power.
But as carrier power drops, nonlinearities start to
dominate. Beyond some threshold, output signal quality
deteriorates faster than linear.
L18 2
Clicks
Received FM signal is R(t)  Re e j 2f t  Ace j (t )   Vn (t)e j 

c n (t ) 

If N 0 B  0.5Ac2 then the noise envelope Vn (t)  Ac
with prob.  1, so phase of signal is  (t) , as desired.
But if not, then with significant prob., Vn (t)  Ac . Small

variations in-phase & quadrature channel noise then lead
occasionally to large step-like phase changes of  2
The demodulator differentiates these, yielding impulse-like

effects that are heard as crackling or clicking noises.
L18 3
Threshold in terms of SNR
In terms of baseband SNR, we need

S Pr
Pr  0.5 A  N 0 B  N 0 2W (1  D)    
2
 2(1  D)
 N b N 0W
c
In fact, to avoid mutilation, it suffices to have
S S
   20 (1  D ) :  
 b
N  N b ,th
S 60 PM S
   2
D (1  D) :  
2
 N o max t | m(t ) |  N o ,th
L18 4
L18 5
Bandwidth‐ or Threshold‐Limited?
E.g., given 4kHz sinusoidal message, baseband SNR=20dB

& a BP channel with BW= 48kHz.
If all available BW’s used, then D  5.

S
 N   20(1  5)  20.8dB - below threshold!
 b,th
The maximum permissible deviation ratio’s given by
20(D  1)  10(S/N) b /10  100  D  4  B  40kHz .
Hence, this system’s threshold-limited.
L18 6
Choosing FM Parameters
To summarise, when designing FM systems, we need to find

deviation ratio D & received power Pr such that
2
 
S 3 P D Pr S
    
M
2
 N o max t | m(t ) | N 0W  N o , desired
Pr
while ensuring that threshold is met, 20(D  1) 
N 0W
and that available BW isn’t exceeded, 2W (D  1)  Bavail
(If channel has power attenuation  , then transmission power
Pt  Pr )
L18 7
Pre-emphasis & De-emphasis
 Quadratic FM output noise psd High frequencies in

message won’t sound as clean as low freq’s.
 In some cases, “message” is a multiplexed combination of
independent messages  Channels located further away
from carrier freq. experience worse performance.
Soln: pass demodulated signal through a (non-ideal) LP filter

H d ( f ) which attenuates the noise at high freqs.
However, this also attenuates high message freq.s.

 Also put message through highpass filter 1/ H d ( f ) before
modulation.
L18 8
First Order Pre‐/De‐emphasis
Simplest design is a 1st order RC de-emphasis filter,
1 1
Hd ( f )   ,
1  j2fRC 1  jf / f 0
H e ( f )  1/ H d ( f )  1  jf / f 0
where f 0 = 3dB BW of filter.
The message portion of the output is unaffected, since the pre- &
de-emphasis filters are reciprocals of each other & therefore
cancel out.
L18 9
Output Noise PSD with Pre‐/De‐emphasis
The pre-emphasis filter doesn’t affect the channel noise, only the
message.
 The new output noise psd is given by
N0 2 N0 f2
S Z , pd ( f ) | H d ( f ) | (2f ) 2  4
2 2
Ac Ac2 1  ( f / f 0 ) 2
If we choose f 0  W , then at high message freq.s the noise psd

doesn’t keep increasing but levels off.
L18 10
Output SNR Improvement with Pre-De-Emphasis
2 2
W N0 f 2 N0 W f
PZpd ( f )  W 4 A2 1  ( f / f ) 2 df  8 A2 0 1  ( f / f ) 2 df
2
c 0 c 0
N 0 f 03  W W  
 8
2
2 
  arctan  
Ac  f 0  f0  
As the message remains the same, ratio of new SNR to old is

ratio of old noise power to new,
S  S  1 (W / f 0 )3 
N     
 o, pd  N o 3  W / f 0  arctan(W / f 0 ) 
L18 11

L18: FM Threshold Effect & Pre /de Emphasis: N S T M P D N S

Uploaded by

Copyright:

Available Formats

L18: FM Threshold Effect & Pre /de Emphasis: N S T M P D N S

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

L18: FM Threshold Effect & Pre /de Emphasis: N S T M P D N S

Uploaded by

Copyright:

Available Formats

L18:

FM Threshold Effect & Pre‐/De‐emphasis

Output FM noise actually depends in a complicated &

If channel noise power N 0 B  0.5Ac2 , then output noise is

Received FM signal is R(t)  Re e j 2f t  Ace j (t )   Vn (t)e j 

But if not, then with significant prob., Vn (t)  Ac . Small

The demodulator differentiates these, yielding impulse-like

In terms of baseband SNR, we need

In fact, to avoid mutilation, it suffices to have

 N o max t | m(t ) |  N o ,th

E.g., given 4kHz sinusoidal message, baseband SNR=20dB

If all available BW’s used, then D  5.

The maximum permissible deviation ratio’s given by

20(D  1)  10(S/N) b /10  100  D  4  B  40kHz .

Hence, this system’s threshold-limited.

To summarise, when designing FM systems, we need to find

 Quadratic FM output noise psd High frequencies in

Soln: pass demodulated signal through a (non-ideal) LP filter

However, this also attenuates high message freq.s.

Simplest design is a 1st order RC de-emphasis filter,

where f 0 = 3dB BW of filter.

 The new output noise psd is given by

If we choose f 0  W , then at high message freq.s the noise psd

As the message remains the same, ratio of new SNR to old is

You might also like