Signal Processing and Linear Systems-B.P.Lathi copy

# Signal Processing and Linear Systems-B.P.Lathi copy

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: is f ascinating behavior, let us define F(w), a c ontinuous function of w, as (4.3) 1 Ll.w = 21r To a nd Eq. (4.5) becomes fro(t) = A glance a t E qs. (4.2c) a nd (4.3) shows t hat Dn = - F(nwo) To As To - + 00, Wo becomes infinitesimal (wo - -t 0). Hence, we shall replace wo b y a more a ppropriate n otation, Ll.w. I n t erms o f t his new n otation, E q. (4.2b) becomes (4.4) T his m eans t hat t he F ourier coefficients Dn a re ( liTo) t imes t he s amples of F(w) u niformly s paced a t i ntervals of wo, a s depicted in Fig. 4.2a.t Therefore, ( lITo)F(w) is t he envelope for t he coefficients Dn. We now let To - + 00 by doubling To r epeatedly. Doubling To halves t he f undamental frequency wo [Eq. (4.2b)], so t hat t here a re now twice a s m any c omponents (samples) in t he s pectrum. However, by doubling To, t he e nvelope ( liTo) F(w) is halved, as shown in Fig. 4.2b. I f we continue t his p rocess of doubling To r epeatedly, t he s pectrum progressively becomes denser while i ts m agnitude becomes smaller. Note, however, t hat t he r elative s hape of t he envelope r emains t he s ame [proportional t o F(w) in Eq. (4.3)]. In t he l imit as To ~ 00, wo ~ 0 a nd Dn ~ O. T his r esult means t he s pectrum is so dense t hat t he s pectral c omponents a re spaced a t zero (infinitesimal) intervals. A t t he s ame t ime, t he a mplitude o f each component is zero (infinitesimal). We have n othing o f e verything, y et we have something! T his p aradox s ounds like A lice in Wonderland, t For t he sake of simplicity, we assume D n , a nd therefore F(w), in Fig. 4.2, t o b e real. T he a rgument, however, is also valid for complex D n [or F(w)). nf;oo [F(n~:)Ll.w] e(jnt>.w)t (4.6a) Equation (4.6a) shows t hat fTo(t) c an b e e xpressed as a s um o f everlasting exponentials of frequencies 0, ±Ll.w, ±2Ll.w, ±3Ll.w,. '" ( the F ourier series). T he a mount of t he c omponent o f frequency nLl.w is [F(nLl.w)Ll.wl/21r. I n t he l imit as To ~ 0 0, Ll.w - -t 0 a nd f ro(t) - -t f (t). T herefore 00 f (t) = lim fro(t) = Jim 2.. " F(nLl.w)e(jnt>.w)tLl.w t>.w-o 21r n ~ =-oo To-oo (4.6b) T he s um o n t he r ight-hand side of Eq. (4.6b) c an b e viewed as t he a rea u nder t he function F(w )e jwt , as i llustrated i n Fig. 4.3. Therefore 1 00 f (t) = - 1 21r F(w)ejwtdw (4.7) - 00 T he i ntegral o n t he r ight h and side is called t he F ourier i ntegral. We have now succeeded in representing a n a periodic signal f (t) by a Fourier integral ( rather tIf nothing else, t he reader now has a n irrefutable p roof of t he proposition t hat 0% ownership of everything is b etter t han 100% o wnership of nothing. 238 4 C ontinuous-Time Signal Analysis: T he F ourier Transform 4.1 239 Aperiodic Signal R epresentation by Fourier Integral t han a F ourier s eries).t T his i ntegral is b asically a Fourier series (in t he limit) w ith f undamental frequency ~w - t 0, a s seen from Eq. (4.6). T he a mount o f t he e xponential ejnL1wt is F(n~w)~w/21':. T hus, t he function F (w) given by Eq...
View Full Document

## This note was uploaded on 04/14/2013 for the course ENG 350 taught by Professor Bayliss during the Spring '13 term at Northwestern.

Ask a homework question - tutors are online