Tom Rauch wrote:
> I'm sure we agree the upper and lower sidebands are what allow us to have a
> change in signal level. After all, the CW signal is actually amplitude
> modulation of a carrier.
The presence of upper and lower sidebands does become a factor when passed
through a filter if the signal is not properly centered in the filter and/or if
the filter for some reason has an asymmetric response around band center. In
this case the upper and lower sidebands get distorted differently and when
mixed down to baseband the result is a distorted frequency response. That is a
separate issue from merely having group delay present.
>
> I don't see how I could possibly hear passband amplitude ripples of a few dB
> as "ringing" or any other deleterious effect, and expect a time delay error
> in the sidebands three to five times the rise and fall time of the signal to
> be inaudible. Especially when that group error would extend the
> time-duration of any sharp noise pulse by the group delay error.
There is no question that phase or group delay errors will cause waveform
distortion that can sometimes look "gross". When "machines", i.e non-humans,
do the data detection, as is the case in digital data systems, then the
distortion can be a serious issue. However, when the human auditory system is
involved, then we need to be careful about assuming what is audibly significant
and what is not. I would point out that there is a large body of evidence in
the field of psychoacoustics that demonstrates our hearing systems do not
necessarily respond the way one might expect intuitively based on waveform
shapes, particularly when phase distortions are involved.
>
> Many people claim broad filters improve copy when weak signals are in rough
> noise. This is because the reduction in group delay errors and the increase
> in bandwidth allow for less "blending" of signals and noise. What I find is
> a filter with less group delay error will allow me to use narrower
> selectivity under conditions of weak signals in rough noise.
The filter response data (http://www.freqdev.com/guide/amp_phase.html) I
referenced in my previous post shows that "significant" group delay (at least a
few msec and quite likely a lot more) must necessarily be present with the
bandwidths and cutoff characteristics of the types of crystal filters we are
discussing. Theory predicts this has to be so and the numbers you measured for
group delay are entirely consistent with the theory. This does not necessarily
make the filters "bad" because that is exactly the way they are supposed to
behave. Now there are more complex filter designs that can compensate for
delay errors but I really doubt that these are incorporated into any of the
inexpensive filters used in amateur equipment. If someone knows otherwise,
please chime in.
>
> I guess we have to agree to disagree on this one. I'm firmly convinced
> significant group delay errors can make a difference. You are not.
>
I have pointed out that group delay errors can make a difference when coupled
with other nonlinear effects. But that's OK...there are plenty of other areas
we can agree upon.
73, John W1FV
|