next up previous
Next: About this document ... Up: Random Processes II Previous: GAUSSIAN PROCESSES

BANDPASS PROCESSES

Must work about these for multiplying many signals using same channel (like radio)
Now need to worry about decomposition
Any X(t) can be decomposed into its sin portion and its cosine portion. What we want to do is represent X(t) as a modulated signal

\begin{displaymath}X(t)=X_c(t)\cos 2\pi f_0 t- X_s(t) \sin 2\pi f_0 t\end{displaymath}


\begin{displaymath}\hat{X}(t)=X_c(t)\sin 2\pi f_0 t+ X_s(t)\cos 2\pi f_0 t\end{displaymath}

The Xc(t) and Xs(t) are "baseband" processes
Hilbert tells us

\begin{displaymath}\hat{X}^(t)=Hil(X(t))\end{displaymath}


\begin{displaymath}X_c(t)=X(t)\cos 2\pi f_0 t+ \hat{X}(t)\sin 2\pi f_0 t\end{displaymath}


\begin{displaymath}X_s(t)=X(t)\cos 2\pi f_0 t- \hat{X}(t)\sin 2\pi f_0 t\end{displaymath}

FALLOUT
Thm.
X(t) is a stationary BP process (BW=2W)
$\Rightarrow X_s(t)$ and Xc(t) are jointly stationary lowpass (baseband) processes each having the power of X(t)


\begin{displaymath}R_{X_c}(\tau)=R_{X_s}(\tau)=R_X(\tau)\cos 2\pi f_0 \tau+ \hat{R}_X(\tau)\sin 2\pi f_0 \tau\end{displaymath}


\begin{displaymath}R_{X_cX_s}(\tau)=R_X(\tau)\sin 2\pi f_0 \tau - \hat{R}_X(\tau)\cos 2\pi f_0 \tau\end{displaymath}


\begin{displaymath}\hat{R}_X(\tau)=Hil(R_X(\tau))\end{displaymath}

BP

\begin{displaymath}S_{X_c}(f)=S_{X_s}(f)=
\left\{\begin{array}{ll}
\displaysty...
...
\displaystyle {0} & {\qquad \mbox{o.w.}}
\end{array}\right.\end{displaymath}

Since SX(f) limited to |f-f0|<W then SXc/s(f) limited to |f|<W


\begin{displaymath}S_{X_c X_s}(f)=\left\{\begin{array}{ll}
\displaystyle {j(S_...
...f_0} \\
\displaystyle {0} & {\qquad o.w.}
\end{array}\right.\end{displaymath}

same idea (notice SXs Xc(0)= always)

Important $\Rightarrow$ if X(t) is such that

\begin{displaymath}S_X(f+f_0)=S_X(f-f_0) \qquad \qquad (\vert f\vert<f_0)\end{displaymath}

then cross corr is zero!
i.e. $R_{X_c X_s}(\tau)\equiv 0$ everywhere!
So, say X(t) is also zero mean $\rightarrow X_c(t)$ and Xs(t) also zero mean (easy to see)
Therefore, Xc(t) and Xs(t) uncorrelated
Say X(t) is also gaussian $\rightarrow X_c$ and Xs jointly gaussian
Therefore, Xc and Xs INDEPT

FINALLY, PX=RX(0)=RXc(0)=RXs(0)
Powers in Xc(t) nd Xs(t) same as in X(t)

END

next up previous
Next: About this document ... Up: Random Processes II Previous: GAUSSIAN PROCESSES

1999-02-06