Reply by Jon Harris June 6, 20052005-06-06
Sorry to take so long to respond--missed your post the first time around.

"Clay S. Turner" <Physics@Bellsouth.net> wrote in message
news:mt7ne.53530$J25.25850@bignews6.bellsouth.net...
> > "Jon Harris" <jon_harrisTIGER@hotmail.com> wrote in message > news:119n17fsq109691@corp.supernews.com... > > > > I'm not sure if I agree with you here. I believe the feedback can happen > > at the > > frequency indicated by the round trip delay, _plus_ all harmonics (integer > > multiples) of that frequency. So even if you add delay to move the > > fundamental > > feedback frequency to say 20 Hz, it can still oscillate at 40, 60, 80, > > 100... > > Hz. Hence adding the delay usually makes the problem even worse, because > > the > > system has more frequencies at which to oscillate (not to mention the echo > > problem you mentioned). > > Jon, > > Certainly a bad amp (non-linear) can produce harmonics, but one hopes the > round trip gain for the harmonics is less than one. But in extreme cases, > yes a harmonic could be the cause of the feedback.
I think we have a slight disconnect here. I'm not talking about harmonics of the signal generated due to system non-linearities. I was trying to say that in an audio system, you are prone to oscillations at any integer-multiple of the frequency indicated by the round-trip delay. "Integer multiple of the fundamental frequency" is the idea I was trying to get across, and I used harmonic as a short version of that, but I'm not talking about harmonics generated due to distortion. To clarify, let's take a simple example of a microphone-loudspeaker system that is perfectly flat and linear and exhibits a pure 10ms round-trip delay (from loudspeaker to microphone back to loudspeaker). Ten milliseconds corresponds to 100Hz, so a 100Hz sinewave played from the loudspeaker and heard by the microphone will be perfectly in phase re-inforce itself perfectly. If the system gain is near unity at 100Hz, you will start to hear ringing and if increased further, eventually oscillation and run-away feedback. However, the same thing could happen at 200 Hz. This waveform will have gone through 2 complete cycles and be back perfectly in phase after the 10ms round-trip delay. Same for 300, 400, 500 ... Hz. In fact, when you hear feedback in an acoustic system, it is typically at a harmonic of the fundamental "round-trip" frequency. With typical microphone-speaker distances and other delays, the fundamental frequencies are generally so low (e.g. 20-40 Hz) that they aren't prone to oscillations due to limited low-frequency response of the speakers and microphones (not to mention intentional gain reduction at super-low frequncies). Feedback usually occurs in the voice range (300-3kHz), since systems tend to have the highest gain in this area, which is usually many times the round-trip delay.
> In fact certain crystal > oscillators can be made to oscillate using a harmonic rather than the > fundamental. Old style radio transmitters use frequency multiplication. > But a higher than normal gain must be used to offset the lower harmonic > production. A classic example is a HeNe laser. The highest gain for the > HeNe system is in the infrared. But we are all used to seeing visible HeNe > lasers. Basically the feedback mirrors aren't reflective at infrared. The > gain of the laser comes from the length of the lasing medium itself. The > minimum length for visible HeNe is about 10cm. But a HeNe laser can be made > to function in the green (instead of the usual red) by using special > mirrors, but the laser's cavity must be lengthend to provide enough gain to > lase. > > If a long enough delay is added to force the fundamental oscillation mode to > a frequency below what the system can reproduce, a much higher gain will be > needed to make the system oscillate using a harmonic. (Harmonics are almost > always weaker than what is causing them) The point I was getting at, is in a > system that is starting to cause feedback oscillation, then a delay can > solve the issue. It will certainly not solve all possible cases. Plus the > echo will be problematic.
And that is where I again disagree. If what you said was true, you would never hear feedback at large venues (football stadiums, etc.) because the microphone-loudspeaker distances usually correspond to frequencies in the <20Hz range. The audio system will happily oscillate at any of its harmonics where the gain is at least unity, and so adding delay will not help with the problem (other than making the build-up to run-away happen more slowly). In fact, the more delay you add to your system the more unstable it is because more harmonics are within the audio range and they are spaced closer together in frequency. -- Jon Harris SPAM blocked e-mail address in use. Replace the ANIMAL with 7 to reply.
Reply by Clay S. Turner May 31, 20052005-05-31
"Jon Harris" <jon_harrisTIGER@hotmail.com> wrote in message 
news:119n17fsq109691@corp.supernews.com...
> > I'm not sure if I agree with you here. I believe the feedback can happen > at the > frequency indicated by the round trip delay, _plus_ all harmonics (integer > multiples) of that frequency. So even if you add delay to move the > fundamental > feedback frequency to say 20 Hz, it can still oscillate at 40, 60, 80, > 100... > Hz. Hence adding the delay usually makes the problem even worse, because > the > system has more frequencies at which to oscillate (not to mention the echo > problem you mentioned).
Jon, Certainly a bad amp (non-linear) can produce harmonics, but one hopes the round trip gain for the harmonics is less than one. But in extreme cases, yes a harmonic could be the cause of the feedback. In fact certain crystal oscillators can be made to oscillate using a harmonic rather than the fundamental. Old style radio transmitters use frequency multiplication. But a higher than normal gain must be used to offset the lower harmonic production. A classic example is a HeNe laser. The highest gain for the HeNe system is in the infrared. But we are all used to seeing visible HeNe lasers. Basically the feedback mirrors aren't reflective at infrared. The gain of the laser comes from the length of the lasing medium itself. The minimum length for visible HeNe is about 10cm. But a HeNe laser can be made to function in the green (instead of the usual red) by using special mirrors, but the laser's cavity must be lengthend to provide enough gain to lase. If a long enough delay is added to force the fundamental oscillation mode to a frequency below what the system can reproduce, a much higher gain will be needed to make the system oscillate using a harmonic. (Harmonics are almost always weaker than what is causing them) The point I was getting at, is in a system that is starting to cause feedback oscillation, then a delay can solve the issue. It will certainly not solve all possible cases. Plus the echo will be problematic. Clay
Reply by Jon Harris May 30, 20052005-05-30
"Clay S. Turner" <Physics@Bellsouth.net> wrote in message
news:lskme.42562$CR5.31279@bignews1.bellsouth.net...
> "Ronny Nilsson" <rln-dsp@NOSPAM.arbetsmyra.dyndns.org> wrote in message > news:2287652.XaPTsQ50KC@gabby... > > > > Hi > > I'm having a question about echo cancellation and feedback reduction. > > > > > > --------- ----- > > | Speaker | --------air------> | Mic | > > --------- ----- > > ^ | > > | ------- ----- | > > +- | Delay | <--wire-- | Amp | <--+ > > ------- ----- > > > > > > The above setup represents an audio path, where the microphone is placed > > to > > near the speaker. Feedback would typically arise with the characteristic > > high pitch tone as a result. > > > > Now, a friend of mine tells me the feedback-tone can be somewhat reduced > > by > > tuning the delay in figure above. He tells me that if the signal delay > > from > > mic to speaker is long enough feedback won't occur. Personally I don't > > believe him. It can't be that simple, or can it?? Perhaps someone can > > clarify? > > > > Best regards > > /Ronny Nilsson > > > > Hello Ronny, > > Audio feedback is a case of turning your amp,speaker,mike combo into an > oscillator. The round trip delay is a multiple of the period of oscillation > and the round trip gain needs to be >= one. (there can be a half period > delay if there is an inversion somewhere). But if you make the round trip > delay long enough so that the feedback frequency is below the cutoff of the > amp,speaker, mike combo, there will not be any oscillation. However you may > have a very annoying echo. So this trick isn't always quite useful.
Clay, I'm not sure if I agree with you here. I believe the feedback can happen at the frequency indicated by the round trip delay, _plus_ all harmonics (integer multiples) of that frequency. So even if you add delay to move the fundamental feedback frequency to say 20 Hz, it can still oscillate at 40, 60, 80, 100... Hz. Hence adding the delay usually makes the problem even worse, because the system has more frequencies at which to oscillate (not to mention the echo problem you mentioned). However, one thing that adding delay can help with is that it causes the feedback to build up more slowly. This can be of some advantage if a human is operating the sound equipment, because it gives them more time to turn down the gain before the feedback reaches deafening levels. Another thing that is sometimes done is to use an all-pass filter to add a 180 degree phase shift to the feedback frequency. This might help a little if you have one main problematic frequency, but often it just moves the problem to another frequency. It's kind of like trying to push on a balloon--it keeps popping up somewhere else! Yet another solution that was tried was to introduce a slight pitch or frequency shift between input and output. This was usually only acceptable with speech (not music) and even then, if the system was close to feedback, you heard annoying "chirps" as the signal fed back at constantly changing frequencies. The most common commercial solutions today use automatic notch filters to reduce system gain at problematic frequencies. We are starting to see some solutions based on echo canceling hit the market, now that DSP processing is cheap enough to make this work with full bandwidth audio. -Jon
Reply by Jerry Avins May 29, 20052005-05-29
Ronny Nilsson wrote:

   ...

> Ok. Thanks. But is it ever useful? I haven't heard of anyone else using > this technique and I suppose it's for a reason, or? Does anyone know if > there is a pratical real-world use for this method?
It can sometimes be helpful, it the side effects are tolerable. (The delay from distant loudspeakers to the orator's ears is troublesome enough. Adding even more delay could tonguetie anyone without practice.) Delay is a poor tool for controlling feedback. It doesn't work very well and it has a high cost. A rock can be used to drive nails, and serves when no better tool is available. "It works" is a poor recommendation. Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Reply by John E. Hadstate May 29, 20052005-05-29
"Ronny Nilsson" <rln-dsp@NOSPAM.arbetsmyra.dyndns.org> wrote 
in message news:29407490.eatagXUJpj@gabby...
>>> I'm having a question about echo cancellation and >>> feedback reduction. >>> --------- ----- >>> | Speaker | --------air------> | Mic | >>> --------- ----- >>> ^ | >>> | ------- ----- | >>> +- | Delay | <--wire-- | Amp | <--+ >>> ------- ----- >>> >>> Now, a friend of mine tells me the feedback-tone can be >>> somewhat reduced by >>> tuning the delay in figure above. He tells me that if >>> the >>> signal delay from >>> mic to speaker is long enough feedback won't occur. >>> Personally I don't believe him. It can't be that simple, >>> or can it?? Perhaps someone can clarify? >>> /Ronny Nilsson > > >> Hello Ronny, >> half period delay if there is an inversion somewhere). >> But >> if you make the round trip delay long enough so that the >> feedback >> frequency is below the cutoff of the amp,speaker, mike >> combo, >> there will not be any oscillation. However you may have a >> very >> annoying echo. So this trick isn't always quite useful. >> Clay > > > Ok. Thanks. But is it ever useful? I haven't heard of > anyone else using > this technique and I suppose it's for a reason, or? Does > anyone know if > there is a pratical real-world use for this method? >
No amount of delay anywhere in an analog feedback loop is a good thing. Introducing delay in a system with a feedback loop causes the transfer function to "explode" with an infinite number of poles. This dramatically reduces the loop gain allowed before the system becomes unstable (oscillates) and consequently makes it almost impossible to get any benefit from feedback. What's worse, in the scenario you described, the delay itself is likely to vary both in duration and gain as the entertainer with the microphone moves with respect to the speakers. Perhaps the best approach is to suppress the feedback using a highly directional microphone and hope the person holding the mic keeps the speakers in a null.
Reply by Jim Adamthwaite May 29, 20052005-05-29
A single delay (with careful gain adjustment) would prevent oscillation at a
particular frequency, but if the system loop gain is high enough, it will
oscillate at wide range of other possible frequencies (usually fractionally
harmonically related to the original).

What is needed is a self-delay-learning, self-gain-adjusting system.  It
should be possible to "morph" an echo canceller to do the job.  This will
allow compensation for multiple echoes and will also satisfy the usability
requirement that there be no significant delay between microphone &
loudspeaker, lest the singer/orator develop a stammer.

The only difficulty is the time length of samples needed to cover the
longest likely echo delay time, which could be considerable in a large hall
and the audio system is to operate at high sound levels.  It takes a fair
bit of crunching power to process (for example) 0.2 seconds of samples if
the sampling rate is hi-fi.

Jim Adamthwaite.

If you can't be good, at least do it good - anon.


Reply by Ronny Nilsson May 29, 20052005-05-29
>> I'm having a question about echo cancellation and feedback reduction. >> --------- ----- >> | Speaker | --------air------> | Mic | >> --------- ----- >> ^ | >> | ------- ----- | >> +- | Delay | <--wire-- | Amp | <--+ >> ------- ----- >> >> Now, a friend of mine tells me the feedback-tone can be >> somewhat reduced by >> tuning the delay in figure above. He tells me that if the >> signal delay from >> mic to speaker is long enough feedback won't occur. >> Personally I don't believe him. It can't be that simple, >> or can it?? Perhaps someone can clarify? >> /Ronny Nilsson
> Hello Ronny, > half period delay if there is an inversion somewhere). But > if you make the round trip delay long enough so that the feedback > frequency is below the cutoff of the amp,speaker, mike combo, > there will not be any oscillation. However you may have a very > annoying echo. So this trick isn't always quite useful. > Clay
Ok. Thanks. But is it ever useful? I haven't heard of anyone else using this technique and I suppose it's for a reason, or? Does anyone know if there is a pratical real-world use for this method? Regards /Ronny
Reply by Clay S. Turner May 29, 20052005-05-29
"Ronny Nilsson" <rln-dsp@NOSPAM.arbetsmyra.dyndns.org> wrote in message 
news:2287652.XaPTsQ50KC@gabby...
> > Hi > I'm having a question about echo cancellation and feedback reduction. > > > --------- ----- > | Speaker | --------air------> | Mic | > --------- ----- > ^ | > | ------- ----- | > +- | Delay | <--wire-- | Amp | <--+ > ------- ----- > > > The above setup represents an audio path, where the microphone is placed > to > near the speaker. Feedback would typically arise with the characteristic > high pitch tone as a result. > > Now, a friend of mine tells me the feedback-tone can be somewhat reduced > by > tuning the delay in figure above. He tells me that if the signal delay > from > mic to speaker is long enough feedback won't occur. Personally I don't > believe him. It can't be that simple, or can it?? Perhaps someone can > clarify? > > Best regards > /Ronny Nilsson >
Hello Ronny, Audio feedback is a case of turning your amp,speaker,mike combo into an oscillator. The round trip delay is a multiple of the period of oscillation and the round trip gain needs to be >= one. (there can be a half period delay if there is an inversion somewhere). But if you make the round trip delay long enough so that the feedback frequency is below the cutoff of the amp,speaker, mike combo, there will not be any oscillation. However you may have a very annoying echo. So this trick isn't always quite useful. Clay
Reply by Ronny Nilsson May 29, 20052005-05-29
Hi
I'm having a question about echo cancellation and feedback reduction.


   ---------                           -----
  | Speaker |  --------air------>     | Mic |
   ---------                           -----
       ^                                 |
       |   -------             -----     |
       +- | Delay | <--wire-- | Amp | <--+
           -------             -----


The above setup represents an audio path, where the microphone is placed to 
near the speaker. Feedback would typically arise with the characteristic 
high pitch tone as a result.

Now, a friend of mine tells me the feedback-tone can be somewhat reduced by 
tuning the delay in figure above. He tells me that if the signal delay from 
mic to speaker is long enough feedback won't occur. Personally I don't 
believe him. It can't be that simple, or can it??  Perhaps someone can 
clarify?

Best regards
/Ronny Nilsson