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

# I t i s c lear t hat a s ystem is a symptotically

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: or t his reason, we shall merely mention here some o f these insights without discussion o r proof. T he interested reader should read Sec. 2.7 for more explanation. T he s ystem's behavior is s trongly influenced by t he c haracteristic roots (or modes) of t he system. T he s ystem responds strongly t o i nput signals similar to its characteristic modes, a nd responds poorly to i nputs very different from its characteristic modes. In fact, when t he i nput is a c haracteristic mode of t he s ystem, the response goes t o infinity provided t hat t he m ode is a nondecaying signal. This is t he resonance phenomenon. T he w idth o f an impulse response h[k] i ndicates the response time (time required t o r espond fully t o a n i nput) o f t he s ystem. I t is the t ime c onstant o f t he system. t Discrete-time pulses are generally dispersed when f h'flu[k] = f + n] + an-18[k + n - 1] + ... + a18[k + 1] + ao8[kJ) = bn8[k + n] + bn_18[k + n - 1] + ... + b18[k + 1] + bo8[k] k ~ 0 I f we s et k = 0 in this equation a nd recognize t hat 8[0] = 1 a nd 8[m] = 0 when m i ' 0, all b ut t he l ast terms vanish o n b oth sides, yielding Aoao = bo a nd bo A o=ao A Special Case: ao = 0 I n t his case Ao = £IL becomes indeterminate. T he p rocedure has t o b e modified ao tThis conclusion follows from the fact that (see Sec. B.7-4) k =-oo A o(8[k slightly. W hen ao = 0, Q [E] c an be expressed as E Q[E]' a nd Eq. (9.28) can be expressed as l'Yilk = 1 _l/'Yi! k =O E Q[E]h[k] = P [E]8[k] Therefore, if /'Yi! &lt; 1 for i = 1 ,2, . .. , n , ~ ~ k =-oo Ih[kll:S ~_1_ ~l-/'Yil &lt; 00 i =l Although we have assumed all the roots to be distinct in this derivation, it is also valid for repeated rO&lt;lts proVIded that they lie inside the unit circle. j:The converse is not true. See footnote on p. 152. tThis part of the discussion applies to systems with impulse response h[k] that is a mostly positive (or mostly negative) pulse. I f we recognize t hat i is a delay operator, this equation can be rearranged as j:Note that Q[E] (Yn[k]) = 0 for all k. But if we restrict the mode terms to be causal, the equation is valid only for k is, Q[E] (Yn[k]u[k]) = 0 k 2: 0 ~ 0; that 6 10 9 T ime-Domain Analysis o f D iscrete-Time Systems Q[EJh[kJ = P[EJ (~8[kJ) = P[EJ8[k - h[kJ = Ao8[kJ + A18[k - IJ c onditions exist: (i) a t l east one r oot is o utside t he u nit circle, (ii) t here a re r epeated r oots o n t he u nit circle. IJ I n t his case t he i nput vanishes n ot for k 2: 1, b ut for k 2: 2. T herefore, t he response c onsists n ot only of t he z ero-input t erm a nd a n i mpulse Ao8[kJ ( at k = 0), b ut also of an impulse A 18[k - IJ ( at k = 1). Therefore + Yn[kJu[kJ W e c an determine t he u nknowns Ao, A I, a nd t he n coefficients in Yn[kJ from t he n + 2 n umber o f initial values h [0], h [1], . .. , h [n + 1], d etermined as usual from t he i terative s olution of t he e quation Q[EJh[kJ = P[EJ8[kJ. Similarly, if ao = a l = 0, we need t o use t he form h[kJ = Ao8[kJ + A 18[k - IJ + A 2 8[k - 2J + Yn[kJu[kJ...
View Full Document

Ask a homework question - tutors are online