06.06.2022 Views

B. P. Lathi, Zhi Ding - Modern Digital and Analog Communication Systems-Oxford University Press (2009)

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

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

530 PERFORMANCE ANALYSIS OF DIGITAL COMMUNICATION SYSTEMS

of luck. On the other hand, Gram-Schmidt orthogonalization is a systematic approach to

extract the basis signals from the known signal set. The details of this approach are given in

Appendix C.

10.5 VECTOR DECOMPOSITION OF WHITE NOISE

RANDOM PROCESSES

In digital communications, the message signal is always one of the M possible wavefonns. It

is therefore not difficult to represent all M waveforms via a set of CON basis functions. The

real challenge, in fact, lies in the vector decomposition of the random noise n(t) at the receiver.

A deterministic signal can be represented by one vector, a point in a signal space. Is it possible

to represent a random process as a vector of random variables? If the answer is positive, then

the detection problem can be significantly simplified.

Consider a complete orthonormal (CON) set of basis functions { <f!k (t)} for a signal space

defined over [0, T 0 ]. Then any deterministic signal s(t) in this signal space will satisfy the

following condition:

(10.56a)

This implies that for t E [0, T 0 ], we have the equality*

s(t) = I>k<f!k (t)

k

However, for random processes defined over [0, T 0 ], this statement is generally not true.

Certain modifications are necessary.

10.5. 1 Determining Basis Functions for a Random Process

First of all, a general random process x(t) cannot strictly satisfy Eq. (10.56a). Instead, a proper

convergence requirement is in the mean square sense, that is,

E {

f lx(t) -

x,c, (t{ dt } = O

(10.56b)

This equality can be denoted as

x(t) s . I >k<f!k (t)

k

(10.56c)

If x(t) and y(t) are equal in the mean square sense, then physically the difference between

these two random processes have zero energy. As far as we are concerned in communications,

signals ( or signal differences) with zero energy have no physical effect and can be viewed as 0.

* Strictly speaking. this equality is true not for the entire interval [O, T0]. The set of points for which equality does

not hold is a measure zero set.

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

Saved successfully!

Ooh no, something went wrong!