01.05.2017 Views

563489578934

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

496<br />

Performance of Communication Systems Corrupted by Noise Chap. 7<br />

also be a Gaussian process. For baseband signaling, the processing circuits would consist of<br />

linear filters with some gain. For bandpass signaling, as we demonstrated in Chapter 4, a superheterodyne<br />

circuit (consisting of a mixer, IF stage, and product detector) is a linear circuit.<br />

However, if automatic gain control (AGC) or limiters are used, the receiver will be nonlinear,<br />

and the results of this section will not be applicable. In addition, if a nonlinear detector such as<br />

an envelope detector is used, the output noise will not be Gaussian. For the case of a linearprocessing<br />

receiver circuit with a binary signal plus noise at the input, the sampled output is<br />

r 0 = s 0 + n 0<br />

(7–10)<br />

Here the shortened notation r 0 (t 0 ) = r 0 is used. n 0 (t 0 ) = n 0 is a zero-mean Gaussian random<br />

variable, and s 0 (t 0 ) = s 0 is a constant that depends on the signal being sent. That is,<br />

s 0 = e s 01, for a binary 1 sent<br />

(7–11)<br />

s 02 , for a binary 0 sent<br />

where s 01 and s 02 are known constants for a given type of receiver with known input signaling<br />

waveshapes s 1 (t) and s 2 (t). Since the output noise sample n 0 is a zero-mean Gaussian random<br />

variable, the total output sample r 0 is a Gaussian random variable with a mean value of either s 01<br />

or s 02 , depending on whether a binary 1 or a binary 0 was sent. This is illustrated in Fig. 7–2,<br />

where the mean value of r 0 is m r01 = s 01 when a binary 1 is sent and the mean value of r 0 is<br />

m r02 = s 02 when a binary 0 is sent. Thus, the two conditional PDFs are<br />

1<br />

f(r 0 |s 1 ) = e -(r 0-s 01 ) 2 >(2s 2 0 )<br />

(7–12)<br />

12p s 0<br />

and<br />

1<br />

f(r 0 |s 2 ) = e -(r 0-s 02 ) 2 >(2s 2 0 )<br />

(7–13)<br />

12p s 0<br />

s 2 0 = n 2 0 = n 2 0 (t 0 ) = n 2 0 (t) is the average power of the output noise from the receiver processing<br />

circuit where the output noise process is wide-sense stationary.<br />

Using equally likely source statistics and substituting Eqs. (7–12) and (7–13) into<br />

Eq. (7–8), we find that the BER becomes<br />

P e = 1 2 L<br />

V T<br />

-q<br />

1<br />

12p s 0<br />

e -(r 0-s 01 ) 2 >(2s 0 2 ) dr 0 + 1 2 L<br />

q<br />

12p s 0<br />

e -(r 0-s 02 ) 2 >(2s 0 2 ) dr 0<br />

V T<br />

1<br />

(7–14)<br />

This can be reduced to the Q(z) functions defined in Sec. B–7 (Appendix B) and tabulated in<br />

Sec. A–10 (Appendix A). Let l =-(r 0 - s 01 )/s 0 in the first integral and l = (r 0 - s 02 )/s 0 in<br />

the second integral; then<br />

P e = 1 2 L<br />

q<br />

1<br />

-(V T -s 01 )>s 0<br />

12p e-l2 >2 dl + 1 2 L<br />

q<br />

(V T -s 02 )>s 0<br />

1<br />

12p e-l2 >2 dl<br />

or<br />

P e = 1 2 Q a -V T + s 01<br />

b + 1 s 0 2 Q a V T - s 02<br />

b<br />

s 0<br />

(7–15)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!