Digital Signal Processing Reference
In-Depth Information
of the transmitted symbol stream s ( n ), of the form
s ( n )= s ( n )+ e ( n ) ,
(2 . 20)
where e ( n ) is the reconstruction error. Recall here that s ( n ) , and hence
s ( n ) ,
are complex numbers. It is usually reasonable to assume that the error e ( n )is
complex and has the form
e ( n )= e r ( n )+ je i ( n ) ,
where e r ( n )and e i ( n ) are independent zero-mean Gaussian random variables
with identical variance 0 . 5 σ e . In this case the total variance of the complex error
e ( n )is
0 . 5 σ e +0 . 5 σ e = σ e .
The joint pdf of the variables [ e r ( n ) ,e i ( n )] is given by
f E ( e r ,e i )= e −e r e
e −e i e
πσ e
= e ( e r + e i ) e
πσ e
πσ e ×
.
(2 . 21)
Figure 2.13 demonstrates this for σ e =0 . 01. This is a special case of a so-called
circularly symmetric complex Gaussian random variable. 4 Next, the complex
symbol s ( n ) is of the form
s r ( n )+ js i ( n ) ,
where s r ( n )and s i ( n ) are PAM symbols. Since
|
| 2 = s r ( n )+ s i ( n ) ,
it follows that the average energy of the constellation is the sum of the average
energies of the real and imaginary parts. Thus, for a b -bit QAM constellation
with average energy E ave , the real and imaginary parts are ( b/ 2)-bit PAM con-
stellations with average energy E ave / 2 , and each of these PAM constellations
sees an error source with variance σ e / 2. For the real-part PAM the probability
of error can be obtained from Eq. (2.17) by replacing b, E ave ,and σ e
s ( n )
with half
their values:
3 E ave
(2 b
.
2 −b/ 2 )
P e,re =2(1
Q
(2 . 22)
1) σ e
Since the factor of one-half cancels out in the ratio E ave e , this is nothing but
the error probability for a ( b/ 2)-bit PAM constellation with energy E ave and
noise variance σ e . Similarly for the imaginary part
3 E ave
(2 b
.
2 −b/ 2 )
P e,im =2(1
Q
(2 . 23)
1) σ e
The QAM symbol is detected correctly if the real part and imaginary part are
both detected correctly. The probability for this is
(1 −P e,re ) 2 .
4 A detailed discussion of circularly symmetric complex random variables can be found in
Sec. 6.6.
Search WWH ::




Custom Search