13.07.2015 Views

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

Passive, active, and digital filters (3ed., CRC, 2009) - tiera.ru

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Two-Dimensional IIR Filters 23-25–1H – N× + H N–1+FIGURE 23.152D high-pass filter configuration.As can be seen in Figure 23.15, the complexity of the high-pass configuration tends to increase rapidlywith the number of rotations, <strong>and</strong> consequently, the number of rotations should be kept to a minimum.It should also be mentioned that the coefficients of the rotated <strong>filters</strong> must be properly adjusted, by usingEquation 23.65, to ensure that the zero-phase is approximated. However, the use of this transformationleads to another problem: the 2-D <strong>digital</strong> transfer function obtained has spurious zeros at the Nyquistpoints. These zeros are due to the fact that the transformation in Equation 23.65 does not have typepreservation in the neighborhoods of the Nyquist points but their presence does not appear to be ofserious concern.With the availability of circularly symmetric low-pass <strong>and</strong> high-pass <strong>filters</strong>, b<strong>and</strong>pass <strong>and</strong> b<strong>and</strong>stop<strong>filters</strong> with circularly symmetric amplitude responses can be readily obtained. A b<strong>and</strong>pass filter can beobtained by connecting a low-pass filter <strong>and</strong> a high-pass filter with overlapping passb<strong>and</strong>s in cascade,whereas a b<strong>and</strong>stop filter can be realized by connecting a low-pass filter <strong>and</strong> a high-pass filter withoverlapping passb<strong>and</strong>s in parallel.23.4.3 Design of Filters Satisfying Prescribed SpecificationsA similar approach to that described in Section 23.3.3 can be used for the design of circularly symmetric<strong>filters</strong> satisfying prescribed specifications. Assume that the maximum=minimum passb<strong>and</strong> <strong>and</strong> themaximum stopb<strong>and</strong> gain of the 2-D filter are (1 Dp) <strong>and</strong> Da, respectively, if K rotated filter sectionsare cascaded where half of the rotations are in the range of 1808–2708 <strong>and</strong> the other half are in the range2708–3608. Then, we can assign the passb<strong>and</strong> ripple d p <strong>and</strong> the stopb<strong>and</strong> loss d a to be [2]<strong>and</strong>d p ¼ DpK(23:70)d a ¼ Da 2=K (23:71)The lower (or upper) bound of the passb<strong>and</strong> gain would be achieved if all the rotated sections were tohave minimum (or maximum) passb<strong>and</strong> gains at the same frequency point. Although it is possible for allthe rotated sections to have minimum (or maximum) gains at the origin of the (v 1 , v 2 ) plane, the gainsare unlikely to be maximum (or minimum) together at some other frequency point <strong>and</strong>, in effect, thepreceding estimate for d p is low. A more realistic value for d p isd p ¼ 2DpK(23:72)If Dp <strong>and</strong> Da are prescribed, then the passb<strong>and</strong> ripple <strong>and</strong> minimum stopb<strong>and</strong> attenuation of the analogfilter can be obtained from Equations 23.72 <strong>and</strong> 23.71 as KA p ¼ 20 log(23:73)K 2Dp

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!