Tom Rauch wrote:
> I just took a moment to measure group delay in a 250Hz wide crystal
filter.
> The swept pattern looks like a 45-degree tilted Z. Group delay error delta
> totaled 18 mS within the -6dB attenuation points of the filter.
Yes, but what does the frequency amplitude response of the filter look like?
> There is one point in the passband, near the flat top of the passband,
where
> a sudden spike in delay time occurs. This spike results in a 10mS time
delta
> in a frequency span of only a few Hz!
An artifact in a few Hz bandwidth is likely to be unnoticeable unless the
signal or noise happens to concentrate a lot of energy in those few Hz.
> When a signal is near "rough" noise, such changes are absolutely
> devastating. Sharp noise pulses smear out to cover weak signals, and the
> rise and fall of the signal are distorted and blended with the noise, so I
> never used this filter. Now I know why!!
Was the AGC on? If so, try turning it off. I know that in some situations,
it's not practical to run without AGC, but the AGC is an example of a
nonlinear system that can convert phase distortion into amplitude distortion
through gain modulation. In my IC-781, the AGC time constant is
continuously adjustable. With rough noise I can often improve copy of weak
signals by careful adjustment of the AGC.
73, John W1FV
|