> In terms of group delay, there is probably a tradeoff
> between shape factor and delay. Whereas the 60dB
> points in a filter don't contribute significantly to noise
> bandwidth, they have a great deal of bearing on group
> delay.
Group delay changes should be as small as possible within low attenuation
area of the filter because that is the only area that contributes to
significant desired signal or significant noise, while delay changes
at -15dB or -20dB points or higher attenuation would be for all practical
purposes meaningless.
I'd think it quite possible to build a steep skirt filter with low group
errors in the nose of the passband.
73 Tom
|