Page 247 - Analog and Digital Filter Design
P. 247

244 Analog and Digital Filter Design




                        some  means  of  correcting  for  the  nonlinear  phase  shift  of  Buttenvorth,
                        Chebyshev, and Cauer filters is desirable.

                        ”Group delay” is the term used to describe the time delay versus frequency rela-
                        tionship of the transmitted signal. It is defined as the rate of phase change with
                        frequency. The term “group delay” is very descriptive, in that it is the delay seen
                        by a group of frequencies that are being transmitted through a filter. A constant
                        group delay implies that all frequencies experience the same delay. A frequency-
                        dependent group delay implies that  some frequencies are delayed more than
                        others.


                        Bessel filters have a constant group delay, because the phase change of  signals
                        passing through them is proportional to the frequency. Other filter types, such
                        as the Butterworth, have a group delay that  is frequency dependent, and the
                        rate  of  phase  change  generally increases  as  the  filter’s cutoff  frequency  is
                        approached. The amount of  group delay variation with frequency depends on
                        the filter type, and generally increases for filters that have a rapid increase in
                        attenuation outside their passband (a steep skirt response). Group delay varia-
                        tions can be minimized by  the use of  phase-equalizing all-pass filters. All-pass
                        filters can be designed to have a group delay that is virtually complementary to
                        a lowpass filter, so the two filters connected in series produce an almost con-
                        stant group delay.


                        Detailed Analysis
                        Impulsive signals contain many harmonics, and Fourier analysis can be used to
                        show that summing all the odd harmonics can produce a square wave. Consider
                        a square wave of amplitude “A”; each harmonic will have an amplitude of 4A/n
                        multiplied by the inverse of  the harmonic number. The sum of harmonics, up
                        to the fifth order, is thus: square wave = 4A/n x fundamental + 4A/(3n) x  third
                        harmonic + 4A/(5n) x fifth harmonic.

                        Some  distortion  is  inevitable  if  the  signal passes  through  a  lowpass filter,
                        because restricting the bandwidth will reduce the amplitude of the higher har-
                        monics. Generally the distortion caused by restricting the bandwidth is pulse-
                        edge rounding and some amplitude ripple in the pulse. This is illustrated in
                        Figure 9.1.
   242   243   244   245   246   247   248   249   250   251   252