A new approach is put forward for obtaining formulas of narrowband lowpass FIR digital filters(NLFDFs) with linear phase based on sinc sum function. Five NLFDF formulas are deduced as examples with stopband attenuation from 47dB to 90db and corresponding relations between filter order and transition width are given. Filter orders designed using four formulas are at least 8.2℅ smaller than the ones designed using Kaiser window with same stopband attenuation and same transition width and using the remaining formula even 27℅ smaller. For the design of filters calculation using Kaiser window is incomparable with the calculation of these formulas because the later is not more difficult than calculation using fixed window. In addition, by the same steps as the examples much more NLFDF formulas can be obtained with different stopband attenuation and good performances.
IntroductionAbout the NLFDF design with linear phase there are some methods available. We can use fixed window method[1], which is the simplest one, but the order of filters is big. Using Kaiser window and Chebyshev window[1,2] method, we can design filters with relative good performances, but the calculation is some large. About the frequency sampling method, which is suitable for narrowband filter design indeed, but the transition width is some large [1]. As to the optimal method, filter order is almost the smallest in comparison by other types of filters because of equal passband ripple and equal stopband attenuation, but the calculation is very complicated. The three methods are common methods for FIR filter design [1]. In addition we have other methods for selection. Narrowband frequency sampling FIR filters using Zwindow[3] need significantly smaller number of terms than the standard window technique with sharp cut-off amplitude characteristic, it is only suitable for the stopband attenuation in the range of 18-42 dB. The frequency masking technique(FRM) is mainly for the design of narrowband filter [4,5].The basic idea behind the FRM technique is to compose a sharp FIR filter using several short subfilters. There are two sectors in an FRM approach. The first sector forms the sharp transition band and arbitrary bandwidth by a pair of complementary interpolated bandedge shaping filters, whereas the second sector removes undesired periodic frequency components from the bandedge shaping filters by