This preview shows page 1. Sign up to view the full content.
Unformatted text preview: der nonrecursive filter transfer function is given by ( ao = a l = a 2 =
... = a nl = 0).
(12.27a)
(12.27b)
T he inverse ztransform o f t his equation yields (a) h[k] = bn 6[k] + bn  1 6[k  1] + ... + h6[k  n + 1] + bo6[k  n] (12.28) Observe t hat h[k] = 0 for k > n .
Because nonrecursive filters are a special case o f recursive filters, we e xpect t he
p erformance of recursive filters t o b e superior. This expectation is t rue in t he sense
t hat a given amplitude response can be achieved by a recursive filter o f a n o rder
smaller t han t hat r equired for t he corresponding nonrecursive filter. However, nonrecursive filters have t he a dvantage of having linear phase characteristics. Recursive
filters can realize linear phase only approximately. F [zl (b) F ig. 1 2.7 Digital filter realization: 1 2.4 (a) recursive filter ( b) nonrecursive filter. Nonrecursive Filters
I f t he recursive coefficients a o, a l, a nd a 2 a re zero, H[z] in Eq. (12.23) reduces to (12.25a)
(12.25b)
T he difference equation corresponding t o t his system now reduces t o y[k] = b3f[k] + b2/[k  1] + b d[k  2] + bof[k  3] (12.26) N ote t hat y[k] is now computed from t he present a nd t he t hree p ast values of t he
i nput f[k]. Such filters are called n onrecursive f ilters. F igure 12.7b shows a
canonical realization of H[z], which is identical t o t he r ealization in Fig. 12.7a, with
all t he feedback connections eliminated. I f we a pply a n i nput t [k] = 6[k] t o this Filter Design Criteria A digital filter processes discretetime signals t o yield a discretetime o utput.
D igital filters can also process analog signals by converting them into discretetime
signals. I f t he i nput is a c ontinuoustime signal f (t), i t is c onverted into a discretetime signal f[k] = f (kT) by a C jD (continuoustime t o discretetime) converter.
T he signal f [k] is now processed by a "digital" (meaning discretetime) system with
transfer function H[z]. T he o utput y[k] of H[z] is t hen converted into a n "analog"
(meaning continuoustime) signal y(t). T he s ystem in Fig. 12.8a, therefore, acts as
a continuoustime (or " analog") s ystem. O ur objective is t o d etermine t he "digital"
(discretetime) processor H [z] t hat will make t he s ystem in Fig. 12.8a equivalent t o
a desired "analog" (continuoustime) system w ith t ransfer function H a (s ), shown
in Fig. 12.8b.
We may strive t o m ake t he t wo systems behave similarly in t he t imedomain or
in t he frequencydomain. Accordingly, we have two different design procedures. Let
us now determine t he equivalence criterion of t he two systems in t he t imedomain
a nd in t he frequencydomain. 1 2.41 The TimeDomain Equivalence Criterion By timedomain equivalence we m ean t hat for t he same input f (t), t he o utput
y(t) of t he s ystem in Fig. 12.8a is equal t o t he o utput y(t) of t he s ystem in Fig. 732 12 Frequency Response a nd Digital Filters 1 12.4 Filter Design C riteria 73...
View Full
Document
 Spring '13
 Bayliss
 Signal Processing, The Land

Click to edit the document details