This preview shows page 1. Sign up to view the full content.
Unformatted text preview: s n + 0 0. Yet t he c urious
fact, as seen f rom F ig. 3.11, is t hat even for large n , t he t runcated series exhibits
a n o scillatory b ehavior a nd a n o vershoot approaching a value of a bout 9% in t he
v icinity of t he d iscontinuity a t t he first p eak o f oscillation. Regardless of t he value of
n , t he o vershoot r emains a t a bout 9%. T his s trange b ehavior a ppears t o c ontradict
t he m athematical r esult derived in Sec. 3.32 t hat t he e rror energy + 0 as n + 0 0.
I n f act, t his a pparent c ontradiction puzzled m any p eople a t t he t urn of t he c entury.
Josiah W illard G ibbs gave a m athematical e xplanation of this behavior (now called
called t he G ibbs p henomenon). We can reconcile t he two confiicting notions by
observing from F ig. 3.11 t hat t he frequency of oscillation of t he s ynthesized signal
is n , so t he w idth o f t he spike w ith 9% o vershoot is a pproximately 1 /2n. As we
i ncrease n , t he n umber of t erms i n t he series, t he frequency o f o scillation increases
a nd t he spike w idth 1 /2n diminishes. As n + 0 0, t he e rror energy  + 0 b ecause t he
e rror consists m ostly o f t he spikes, whose widths  > O. T herefore, as n + 0 0, t he
c orresponding F ourier series differs from f (t) b y a bout 9% a t t he i mmediate left
a nd r ight of t he p oints of discontinuity, a nd y et t he e rror energy + O.
W hen we u se o nly t he first n t erms in t he F ourier series t o s ynthesize a signal,
we a re a bruptly t erminating t he series, giving a u nit weight t o t he first n h armonics
a nd zero weight t o all t he r emaining harmonics beyond n . T his a brupt t ermination
o f t he series c auses t he G ibbs phenomenon in synthesis of discontinuous functions.
More discussion o n t he G ibbs phenomenon, its ramifications, a nd c ure a ppear i n
section 4.9. E xercise E 3.8
B y inspection of signals in Figs. 3.7b, 3 .l0a, a nd 3.lOb, determine t he a symptotic r ate of
decay o f t heir a mplitude s pectra.
Answer: l in, l /n 2 , a nd l in, respectively. 'V A Historical Note on the Gibbs Phenomenon N ormally speaking, troublesome functions w ith s trange b ehavior are invented
by mathematicians, although we rarely see such oddities in practice. I n t he case
of t he G ibbs phenomenon, however, t he t ables were t urned. A r ather p uzzling
behavior was observed in such a m undane o bject as a mechanical wave synthesizer,
a nd t hen wellknown m athematicians of t he d ay were dispatched o n t he s cent of it
t o discover its hideout.
A lbert Michelson (of MichelsonMorley fame) was an intense. practical m an
who developed ingenious physical i nstruments o f e xtraordinary precision, mostly in
t he field of optics. I n 1898 he developed a n i nstrument ( the h armonic analyzer)
which could c ompute t he first 80 coefficients of t he F ourier series of a signal f (t)
specified by any graphical description. T he h armonic analvzer could also be used
as a harmonic synthesizer, which could plot a function f(t)" g enerated by s umming
t he first 80 h armonics (Fourier components) of a rbit...
View
Full
Document
 Spring '13
 Bayliss
 Signal Processing, The Land

Click to edit the document details