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

I t is desirable to have a frequency resolution of

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

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: n t he v alue j ust o btained by using 1% a mplitude criterion. T he second issue is to determine To. Because t he signal is n ot t imelimited, we have t o t runcate i t a t To such t hat f (To) « 1. A r easonable choice would be To = 4 because f (4) = e - 8 = 0.000335 « 1. T he r esult is No = T o/T = 254.6, which is n ot a power of 2. Hence, we choose To = 4, a nd T = 0.015625 = 1 /64, yielding No = 256, which is a power of 2. N ote t hat t here is a g reat deal of flexibility in determining T a nd To, d epending on t he a ccuracy desired a nd t he c omputational capacity available. We could j ust as well have chosen T = 0.03125, yielding No = 128, although this choice would have given a slightly higher aliasing error. Because t he s ignal has a jump d iscontinuity a t t = 0, t he first sample ( at t = 0) is 0.5, t he averages of t he values o n t he two sides of t he discontinuity. We compute Fr ( the D FT) from t he s amples of e - 2t u(t) according t o Eq. (5.18a). Note t hat Fr is t he r th s ample of F (w), a nd t hese samples are spaced a t:Fo = l /To = 0.25 Hz. (wo = t r/2 r ad/s.) Because Fr is No-periodic, Fr = F(r+256) so t hat F256 = Fo. Hence, we need to p lot Fr over t he range r = 0 t o 255 ( not 256). Moreover, because of this periodicity, F - r = F( -r+256), a nd t he values of Fr over t he r ange r = - 127 t o - 1 a re identical to o LF(oo) -0.5 -I -1tI2 10 20 I 30 40 00- I ~ ~ t- \ '-, Exact F ig. 5 .15 ...., j FFT values D iscrete Fourier transform of an exponential signal e - 2t u(t). o C omputer E xample C 5.1 Use D FT ( implemented by t he F FT, t he fast Fourier transform algorithm) t o c ompute t he Fourier transform of e - 2t u(t). T =O.015625jTO=4jNO=TO/Tj t =O:T:T* ( NO-I) j t=t ' j f =T*exp(-2*t)j 5 Sampling 346 f (1)=T*O.5 ; F =fft(f); [ Fp,Fm}=cart2pol(real(F) , imag(F»; k =O:NO-l; k =k'; w =2*pi*k/ TO; s ubplot(2l1.) , plott w (l: 1 28) , Fm(l :128» s ubplot(2l2) , plot ( w(1:128) , Fp(1:128» 5.2 Numerical Computation of the Fourier Transform: T he D FT 347 (a) t_ o • E xample 5 .7 Use O FT t o c ompute t he F ourier transform of 8 rect ( t). T his g ate f unction a nd its Fourier transform are illustrated in Fig. 5.16a a nd b. To determine t he va.lue of t he s ampling interval T , we must first decide o n t he essential b andwidth B . I n Fig. 5.16b, we see t hat F(w) decays r ather slowly with w. Hence, t he essential b andwidth B is r ather large. For instance, a t B = 15.5 Hz (97.39 r ad/s), F{w) = - 0.1643, which is a bout 2% of t he p eak a t F(O). Hence, t he essential b andwidth is well above 16 Hz if we use t he 1% criterion for computing t he essential bandwidth. However, we s hall d eliberately take B = 4 for two reasons: ( I) t o show t he effect of aliasing a nd (2) t he use of B > 4 will give enormous number of samples, which cannot b e conveniently displayed on t he b ook size page without losing sight of t he essentials. Thus, we s hall intentionally accept approximation in order t o clarify t he c oncepts...
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