22 Oct
2016
22 Oct
'16
4:30 p.m.
Hi,
I'm looking for a solution for adaptive feedback cancellation. The type
of feedback that occurs between a microphone and speaker, not echo
cancellation. Preferrably in the form of a plugin (LADSPA, LV2,
DSSI, ...) or even as a standalone JACK app. This must be a common
problem, but I haven't found a solution.
Thanks,
J.
22 Oct
22 Oct
5:50 p.m.
On Sat, Oct 22, 2016 at 04:30:00PM +0200, Johannes Kroll wrote:
I'm looking for a solution for adaptive feedback
cancellation. The type
of feedback that occurs between a microphone and speaker, not echo
cancellation. Preferrably in the form of a plugin (LADSPA, LV2,
DSSI, ...) or even as a standalone JACK app. This must be a common
problem, but I haven't found a solution.
Because there isn't any that really works.
The basic reason is that the combined response of speaker, room
and mic in a typical PA situation is very irregular with hundreds
if not thousands of very narrow peaks, and the frequency of those
peaks is very sensitive to small movements of almost everything.
Simple tricks like adaptive notch filters will maybe give you a dB
or so extra, but there it ends.
The only thing that can work sometimes is shifting all frequencies
up or down by a few Hz, smoothing out the peaks. Simple to implement,
usable for speech but usually not for music because it will appear
to be out of tune.
One step beyond that is quasi proportional frequency shifting,
but that's quite complicated, has latency issues, and may also
affect music in ways that are not acceptable.
So in practice: use the right tools (mic, speakers), take room
acoustics into consideration when deciding on placement, and be
careful with EQ and effects.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
23 Oct
23 Oct
1:13 a.m.
On Sat, 22 Oct 2016 15:50:22 +0000
Fons Adriaensen <fons(a)linuxaudio.org> wrote:
On Sat, Oct 22, 2016 at 04:30:00PM +0200, Johannes
Kroll wrote:
[...]
Because there isn't any that really works.
The basic reason is that the combined response of speaker, room
and mic in a typical PA situation is very irregular with hundreds
if not thousands of very narrow peaks, and the frequency of those
peaks is very sensitive to small movements of almost everything.
Simple tricks like adaptive notch filters will maybe give you a dB
or so extra, but there it ends.
The only thing that can work sometimes is shifting all frequencies
up or down by a few Hz, smoothing out the peaks. Simple to implement,
usable for speech but usually not for music because it will appear
to be out of tune.
That's a good hint, thanks. I could try the AM Pitch Shifter or Bode
Pitch Shifter LADSPA plugins. That could be a good enough solution for
me, if it doesn't change the sound too much.
22 Oct
22 Oct
7:49 p.m.
On Sat, 22 Oct 2016, Johannes Kroll wrote:
I'm looking for a solution for adaptive feedback
cancellation. The type
of feedback that occurs between a microphone and speaker, not echo
cancellation. Preferrably in the form of a plugin (LADSPA, LV2,
DSSI, ...) or even as a standalone JACK app. This must be a common
problem, but I haven't found a solution.
That would be because it requires a different solution every situation.
What happens when the "fat lady" sings? The plugin nulls out her vocal
thinking it is feedback. As happens, my acoustic instrument amp includes
an anti-feedback section. It works in two parts:
polarity reversal
notch filter
The polarity of the FOH/monitor speakers does matter in this case. Piezo
pickups are bad in this respect because the polarity may be different
depending on mounting position (inside vs outside of sound plate). Mics
_should_ be wired corectly and a FOH mixer/amp/speakers _should_ be too.
Those are things to check. I have seen FOH mixers where the main outs are
purposely 2- and 3+ for this reason. I don't know if this is standard or
not.
Aim for as low of stage monitor level as possible. Setting mains level
first may allow talent to use lower stage levels... but some are just
deaf! In ear would be better in such a case. Note that I am using my
(limited) experience with small venue work. A large venue is different.
Notch filters are the thing we think of first, but should probably be
last. As already noted, this will not give you much more level and it does
change the sound. The frequency to use changes with the hall, mic
placement, talent mic know how, etc, etc. (teach your talent if they must
hold the mic to hold it away from the capsule)
The other thought I have is that maybe you mean something to remove
feedback noise on an already recorded file... notch filter is the tool for
that... but it is really already too late by then.
--
Len Ovens
www.ovenwerks.net
8:02 p.m.
I don't know why you guys think this is impossible. The dbx unit, or
copies thereof, you see in small PA systems all the time works fine. Not
as good as an experienced engineer with a 1/3rd octave EQ, but good enough
to be very useful to performers running sound from the stage or for ad hoc
PA systems in small venues.
I don't know of an open source implementation, but the engineering problem
itself is not intractable.
On Sat, Oct 22, 2016 at 1:49 PM, Len Ovens <len(a)ovenwerks.net> wrote:
On Sat, 22 Oct 2016, Johannes Kroll wrote:
I'm looking for a solution for adaptive feedback cancellation. The type
of feedback that occurs between a microphone and
speaker, not echo
cancellation. Preferrably in the form of a plugin (LADSPA, LV2,
DSSI, ...) or even as a standalone JACK app. This must be a common
problem, but I haven't found a solution.
That would be because it requires a different solution every situation.
What happens when the "fat lady" sings? The plugin nulls out her vocal
thinking it is feedback. As happens, my acoustic instrument amp includes an
anti-feedback section. It works in two parts:
polarity reversal
notch filter
The polarity of the FOH/monitor speakers does matter in this case. Piezo
pickups are bad in this respect because the polarity may be different
depending on mounting position (inside vs outside of sound plate). Mics
_should_ be wired corectly and a FOH mixer/amp/speakers _should_ be too.
Those are things to check. I have seen FOH mixers where the main outs are
purposely 2- and 3+ for this reason. I don't know if this is standard or
not.
Aim for as low of stage monitor level as possible. Setting mains level
first may allow talent to use lower stage levels... but some are just deaf!
In ear would be better in such a case. Note that I am using my (limited)
experience with small venue work. A large venue is different.
Notch filters are the thing we think of first, but should probably be
last. As already noted, this will not give you much more level and it does
change the sound. The frequency to use changes with the hall, mic
placement, talent mic know how, etc, etc. (teach your talent if they must
hold the mic to hold it away from the capsule)
The other thought I have is that maybe you mean something to remove
feedback noise on an already recorded file... notch filter is the tool for
that... but it is really already too late by then.
--
Len Ovens
www.ovenwerks.net
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
8:19 p.m.
On Sat, Oct 22, 2016 at 10:49:11AM -0700, Len Ovens wrote:
The polarity of the FOH/monitor speakers does matter
in this case.
Maybe if mic and speaker are close. But consider a mic that is
5 meters from a FOH speaker. That's 14 ms at the speed of sound.
In each 70 Hz frequency interval there will be one frequency
which returns exactly in phase, no matter what polarity. And
that's considering only the direct path.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
9:28 p.m.
I have a cheap Behringer noise suppressor (an old model now out of
production). Its adaptive notch filters can be "freezed" once they are set
for the particular application, so they don't move around during the
performance. It saved my day several times with an amped acoustic guitar
that simply wouldn't behave nicely. Yes, of course it changes the sound,
but the extra headroom can be HUGE: not just a couple of dBs, but the
difference between having to keep the guitar almost completely unamplified
and hearing it clearly in the mix.
I can't see why something similar wouldn't be attainable with a plugin.
-- Emanuele Rusconi
23 Oct
23 Oct
1:16 a.m.
On Sat, 22 Oct 2016 21:27:35 +0200
Emanuele Rusconi <emarsk(a)gmail.com> wrote:
I have a cheap Behringer noise suppressor (an old
model now out of
production). Its adaptive notch filters can be "freezed" once they are set
for the particular application, so they don't move around during the
performance. It saved my day several times with an amped acoustic guitar
that simply wouldn't behave nicely. Yes, of course it changes the sound,
but the extra headroom can be HUGE: not just a couple of dBs, but the
difference between having to keep the guitar almost completely unamplified
and hearing it clearly in the mix.
I can't see why something similar wouldn't be attainable with a plugin.
That sounds like what I'm looking for. How do these units work? It
can't be all that complex if it's implemented all in (old, cheap)
hardware.
22 Oct
22 Oct
10:59 p.m.
On 10/22/2016 08:19 PM, Fons Adriaensen wrote:
On Sat, Oct 22, 2016 at 10:49:11AM -0700, Len Ovens
wrote:
And this is without any delay in signal processing. Any additional delay
in the signal processing will increase the density of in-phase
frequencies, thus minimizing the processing delay might increase stability.
- Giso
The polarity of the FOH/monitor speakers does
matter in this case.
Maybe if mic and speaker are close. But consider a mic that is
5 meters from a FOH speaker. That's 14 ms at the speed of sound.
In each 70 Hz frequency interval there will be one frequency
which returns exactly in phase, no matter what polarity. And
that's considering only the direct path.
Ciao,
23 Oct
23 Oct
1:21 a.m.
On Sat, 22 Oct 2016 22:59:36 +0200
Giso Grimm <gg3137(a)vegri.net> wrote:
On 10/22/2016 08:19 PM, Fons Adriaensen wrote:
[...]
[...]
[...]
And this is without any delay in signal processing. Any additional delay
in the signal processing will increase the density of in-phase
frequencies, thus minimizing the processing delay might increase stability.
I try to run with a Jack buffer size of 64 and 2 buffers at 48Khz,
creating a roundtrip latency of 5.3ms. I say I "try" because depending
on processing time spent in applications and their behaviour (some apps
don't seem to be written with RT-safety in mind...), I have to switch
to higher block size sometimes.
7:52 p.m.
On Sat, 22 Oct 2016 18:19:28 +0000, Fons Adriaensen wrote:
On Sat, Oct 22, 2016 at 10:49:11AM -0700, Len Ovens
wrote:
Reminds me of the useful Sengpiel homepage.
http://www.sengpielaudio.com/calculator-soundpath.htm
Perhaps he provided information regarding feedback, too. I didn't take
a look, but it's worth to search. Hopefully his son continues site.
The polarity of the FOH/monitor speakers does
matter in this
case.
Maybe if mic and speaker are close. But consider a mic that is
5 meters from a FOH speaker.
24 Oct
24 Oct
12:44 p.m.
The wavelength of a 10kHz sound is 3.4cm, which means that, assuming a
single point source (monitor or PA) and single point mic, you go from a
full positive feedback (perfectly in phase) to full negative feedback
(perfect counter-phase) at that particular frequency simply by changing the
distance between them by 1.7cm. Things get awfully more complicated with
multiple sources, non-puntiform sources/mics, reverberant room, mic
directional patterns, etc., and of course each frequency will dictate its
own phase relationship. And then there's sound processing introducing phase
shifting…
It's just impossible - in a real world situation - to calculate the exact
mic placement that reduces the feedback, it's a lot more practical to use
some good sense (closest to the instrument as possible, low-sensitivity
direction pointed as much as possible toward the closest monitor) and then
move around the mic until you find a good spot - if there is one at all.
(And of course it's misguided to think - in this context - that there is a
"right" and "wrong" wire phase for speakers and mics.)
Once you placed your mics with some sense, the solution to the feedback
problem must necessarily be reactive, not proactive: you hear the feedback,
you lower the gain - possibly just at that frequency with a notch filter;
where "you" could be a human or a piece of hardware or software.
-- Emanuele Rusconi
2750
days inactive
2752
days old
linux-audio-user@lists.linuxaudio.org
11 comments
7 participants
participants (7)
-
Emanuele Rusconi -
Fons Adriaensen -
Giso Grimm -
Johannes Kroll -
Len Ovens -
Ralf Mardorf -
Robert Edge