Differential group delay does occur in one variety of digital filter - if I
remember correectly, the FIR (Finite Impulse Response) design has differential
group delay, while the IIR (Infinite Impulse Response) can be designed without
differential group delay.
Ringing is not harmonics - it's the output at a constant frequency which dies
away, usually exponentially. Like hitting a high Q tuned circuit with a pulse,
and watching the oscillation gradually die away - a technique used in echo boxes
for radar, and indeed, in early radars for producing synchronised timing (range)
calibration.
Now, paraphrasing Zverev,
An ideal filter response has a rectangular amplitude response and linear phase
inside and outside the passband. This is not ideal in the time domain because of
ringing and overshoot. Phase correction will not improve the time domain
response of an ideal filter because the phase is already linear.
Things get a bit heavy from hereon: he starts using the Heaviside operational
method. However, after several pages of Laplace transforms and integrating
everything within sight, he comes down to the conclusions that 'constant group
delay is a necessary requirement for good pulse transmission. Consider the
increasing delay characteristics associated with Chebyshev filters. The high
frequencies which are required to give a fast pulse rise time are delayed more
than the low frequencies, and thus give an amplitude peak at a later time. Thus
rise time can be expected to increase as the high frequencies become delayed
more and more with respect to the low frequencies.
There's whole chapter on the subject, and it's really too much either for this
forum, or for e-mail discussion generally. I recommend getting the book.....
73
Peter G3RZP
--
FAQ on WWW: http://www.contesting.com/FAQ/amps
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
|