01.05.2017 Views

563489578934

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

440<br />

Random Processes and Spectral Analysis Chap. 6<br />

The output PSD is<br />

y (f) = ƒ H(f) ƒ 2 x (f)<br />

(6–83)<br />

where H( f ) = [h(t)].<br />

Equation (6–83) shows that the power transfer function of the network is<br />

G h (f) = P y(f)<br />

P x (f)<br />

= ƒ H(f) ƒ<br />

2<br />

(6–84)<br />

as cited in Eq. (2–143).<br />

But<br />

Proof. From Eq. (6–80),<br />

R y (t) ! y(t)y(t + t)<br />

q<br />

q<br />

= c h(l 1 )x(t - l 1 )dl 1 d c h(l 2 )x(t + t - l 2 ) dl 2 d<br />

L L<br />

- q<br />

=<br />

L<br />

q<br />

q<br />

-q L-q<br />

-q<br />

h(l 1 )h(l 2 )x(t - l 1 )x(t + t - l 2 ) dl 1 dl 2<br />

(6–85)<br />

x(t - l 1 )x(t + t - l 2 ) = R x (t + t - l 2 - t + l 1 ) = R x (t - l 2 + l 1 )<br />

so Eq. (6–85) is equivalent to Eq. (6–82a). Furthermore, Eq. (6–82a) may be written in terms<br />

of convolution operations as<br />

q<br />

q<br />

R y (t) = h(l 1 ) e h(l 2 )R x [(t + l 1 ) - l 2 ] dl 2 f dl 1<br />

L L<br />

-q<br />

q<br />

= h(l 1 ){h(t + l 1 ) * R x (t + l 1 )} dl 1<br />

L<br />

-q<br />

q<br />

= h(l 1 ){h[-((-t) - l 1 )] * R x [-((-t) - l 1 )]} dl 1<br />

L<br />

-q<br />

-q<br />

= h(-t) * h[-(-t)] * R x [-(-t)]<br />

which is equivalent to the convolution notation of Eq. (6–82b).<br />

The PSD of the output is obtained by taking the Fourier transform of Eq. (6–82b).<br />

We obtain<br />

or<br />

[R y (t)] = [h(-t)][h(t) * R x (t)]<br />

y (f) = H * (f)H(f)P x (f)<br />

where h(t) is assumed to be real. This equation is equivalent to Eq. (6–83).

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

Saved successfully!

Ooh no, something went wrong!