Maker Pro
Maker Pro

Phase shift, audio filter

A

amdx

Jan 1, 1970
0
I ran two LC 3khz audio filters in LTspice.
#1 has a sharp cutoff down 44db at 4khz and continues down
to 92db at 4.9khz. But it shows a 420* phase shift at 3khz.

#2 filter is down 34db at 4khz and reaches 46db at 4.4khz
(that's all the attenuation it has).
The phase shift is 250* at 3khz.

How does the extra phase shift of filter #1 show itself in listening?

Evaluating for radio filter.

Thanks, Mike
 
B

Bob Eld

Jan 1, 1970
0
amdx said:
I ran two LC 3khz audio filters in LTspice.
#1 has a sharp cutoff down 44db at 4khz and continues down
to 92db at 4.9khz. But it shows a 420* phase shift at 3khz.

#2 filter is down 34db at 4khz and reaches 46db at 4.4khz
(that's all the attenuation it has).
The phase shift is 250* at 3khz.

How does the extra phase shift of filter #1 show itself in listening?

Evaluating for radio filter.

Thanks, Mike

What are these filters for? They look like elliptic filters given the steep
cut off or notch. Such filters are not good for audio except in special
circumstances. Why do you want to cut off at 3KHz? Is this to eliminate
noise in a poor audio channel in which case audio quality is not good anyway
and such filters would make little difference. Or, is this a low pass filter
for a crossover or other network where audio quality is important? If audio
quality is important, these are not the filters to use. These filters are
multi-pole high "Q" and will cause "gritty" unnatural sound in high quality
applications. The grittyness is caused by "ringing" that these filters
exhibit when hit with an impulse. Do an impulse response and see that there
is a lot of overshoot and undershoot or ringing. Such a response is a
disaster for audio unless you are making bells. Bells are supposed to ring
but not filters. Secondly, the high and varying phase shift will make a
nightmare response for a crossover network where the filter works in
conjuction with another, high pass unit. The sum of the two outputs in the
crossover region will be very uneven or bumpy. As a general rule, never use
a higher "Q" or more poles in an audio filter than absolutely necessary to
do the job.
 
J

Jon

Jan 1, 1970
0
I ran two LC 3khz audio filters in LTspice.
#1 has a sharp cutoff down 44db at 4khz and continues down
to 92db at 4.9khz. But it shows a 420* phase shift at 3khz.

#2 filter is down 34db at 4khz and reaches 46db at 4.4khz
(that's all the attenuation it has).
The phase shift is 250* at 3khz.

How does the extra phase shift of filter #1 show itself in listening?

Evaluating for radio filter.

Thanks, Mike
Mike,
I need more information. What kind of fiters are they (Butterworth,
Elliptic, etc)? How many poles? What the bandpass, bandstop
attenuation specs? corner frequency?
 
A

amdx

Jan 1, 1970
0
Bob Eld said:
What are these filters for? They look like elliptic filters given the
steep
cut off or notch. Such filters are not good for audio except in special
circumstances. Why do you want to cut off at 3KHz? Is this to eliminate
noise in a poor audio channel in which case audio quality is not good
anyway
and such filters would make little difference. Or, is this a low pass
filter
for a crossover or other network where audio quality is important? If
audio
quality is important, these are not the filters to use. These filters are
multi-pole high "Q" and will cause "gritty" unnatural sound in high
quality
applications. The grittyness is caused by "ringing" that these filters
exhibit when hit with an impulse. Do an impulse response and see that
there
is a lot of overshoot and undershoot or ringing. Such a response is a
disaster for audio unless you are making bells. Bells are supposed to ring
but not filters. Secondly, the high and varying phase shift will make a
nightmare response for a crossover network where the filter works in
conjuction with another, high pass unit. The sum of the two outputs in the
crossover region will be very uneven or bumpy. As a general rule, never
use
a higher "Q" or more poles in an audio filter than absolutely necessary to
do the job.
This site gives the explaination of use.
http://www.kongsfjord.no/dl/Audio/ELPAF.pdf
The filter came from this page.
http://www.kongsfjord.no/dl/Audio/AELPAF SAI.pdf

Thanks, Mike
 
B

Bob Eld

Jan 1, 1970
0
amdx said:
This site gives the explaination of use.
http://www.kongsfjord.no/dl/Audio/ELPAF.pdf
The filter came from this page.
http://www.kongsfjord.no/dl/Audio/AELPAF SAI.pdf

Thanks, Mike
As I thought, multi-pole elliptic filters of high "Q." They are the "special
circumstance" I mentioned above. Fading SW, AM is not high quality audio
under any definition. These filters are used to "clean up" an already very
poor audio signal and are appropriate for that purpose. Whether they work
very well or not is another matter. They ain't HI-FI and should NEVER be
used in your stereo for the reasons I mentioned! These filters are designed
to be used with speakers at 8ohms but they have some inconvenient, large
components. It is possible to do these filters at low level using op-amps as
active components with R's and C's, no inductors. That may be more
convenient, simpler and a lot cheaper.
 
T

Tom Bruhns

Jan 1, 1970
0
....
As I thought, multi-pole elliptic filters of high "Q." They are the "special
circumstance" I mentioned above. Fading SW, AM is not high quality audio
under any definition. These filters are used to "clean up" an already very
poor audio signal and are appropriate for that purpose. Whether they work
very well or not is another matter. They ain't HI-FI and should NEVER be
used in your stereo for the reasons I mentioned! These filters are designed
to be used with speakers at 8ohms but they have some inconvenient, large
components. It is possible to do these filters at low level using op-amps as
active components with R's and C's, no inductors. That may be more
convenient, simpler and a lot cheaper.

Ah, yes, I think the OP's posting name, "AM DX," gives a clue. ;-)

As Bob says, implement them with active filters. In fact, if you do,
you can re-tune them relatively easily and evaluate which works best
for you without having to rewind big coils.

Even better, if you can, implement your filter as an FIR filter in a
processor. FIR filters can easily have linear phase, at least, though
they likely will have some "ringing" to get the sharp cutoff you
probably want. An additional advantage is that it's then easy to
change the filter cutoff frequency to suit your needs.

If the stations you listen to are channelized, that is with uniform
frequency spacing, it's probably worthwhile to have a filter with a
deep null at the channel spacing frequency. For example, a filter
with a 10kHz notch is worthwhile for listening to US medium-wave AM
broadcast signals, because it kills the "whistle" from adjacent-
channel carriers.

Cheers,
Tom
 
Top