rando@skipper.dfrf.nasa.gov (Randy Brumbaugh) (06/13/90)
An idea for a device which would be very useful, but for some reason doesn't seen to be available: In sound reinforcement work, especially indoors, feedback is a problem. Certain frequencies tend to be "hotter" than others in a given room. Usually equalizationis used to cut these frequencies (actually wavelengths) before a show, during set-up. But because of the horrible effect of feedback on an audience and change in room acoustics with an audience, it is normal to leave large margins in the system gain to be sure feedback doesn't happen. Wouldn't it be possible to use DSP to eliminate feedback in real time before it is noticed? The processor would "look" at the spectral contents of the program and watch for any "runaway" frequency bands, where the energy is increasing exponentially (??). It could also check for ringing. It might test the room before a show to find frequencies to watch more closely. If it was really slick, it would distinguish between input channels, realizing that feedback would only come through a mic, while a synth might make a sound which "looks" like feedback. Once it found feedback starting, it could reduce the gain at the exact frequency by just enough to stop it. Thus, the sound reinforcement system would continuosly adapt to changing conditions, and much more gain could be squeezed out of a system because feedback margins could be eliminated. So . . . Does such a thing exist? If not why not? It seems 1) possible 2) useful 3) saleable Am I missing something? Randy Brumabugh rando@skipper.dfrf.nasa.gov
chuck@umiami.miami.edu (06/13/90)
In article <539@skipper.dfrf.nasa.gov>, rando@skipper.dfrf.nasa.gov (Randy Brumbaugh) writes: > > ...Wouldn't it be possible to use DSP to eliminate feedback > in real time before it is noticed? ... > > So . . . Does such a thing exist? If not why not? > It seems 1) possible 2) useful 3) saleable > Am I missing something? I've done live sound work for a few years now. I am certain that many folx such as yourself have thought of such a monster, but I have some thoeries as to why it's not out yet. However, it is a *great* idea! This may not be much of a factor for large sound companies, but something like this seems expensive! The software (and hardware) for this would seem to take quite a bit of time developing, and start out costing loads of dough, so that the sound quality would not be compromised, and so that it would be easy enough for any sound jockey to use. Although, I'm sure costs could be kept within range of most sound-company budgets with some time. Also, you'd either have to decide on a specific computer to run it on, or develop your own CPU, operating system and user interface to run with the unit. And, if you decided on a computer, should it be a Mac, IBM, etc? How many sound companies have road-worthy computers? Sound gear on the road gets *real* banged up during load-ins and -outs. I don't see an efficient way of interfacing with a console to control individual input gains. You'd have to build a patch bay of some sort, or a computer interfacing scheme to make the whole thing work, and the most efficient way for the computer (the interface) is the most "costly" (not just money), and the inefficient (patch-pay method) is old technology. Am I making sense? I'm sure there are a few other things that hold back this "technological wonder". I certainly see how a profitable sound company could afford some machine like this, cause the best way to go would be to buy a console with the computer built-in, but this is a great expense. The ideal way to go (so it would be available to everybody) would be to make it interface with present boards, and that is a bit more difficult. I am open to more discussion on this topic, cause I would certainly like to see something like this developed! > > Randy Brumabugh > rando@skipper.dfrf.nasa.gov I hope this fuels a little discussion on the subject for ya, Randy. -- =========================================================================== Chuck Urwiler University of Miami Music Engineering Voice & Keys chuck@umiami | This space for rent... e1mvqi9z@umiamivm | chuck@umiami.miami.edu | chuck@miavax.ir.miami.edu | =========================================================================== Disclaimer: I only work and learn at this University. I don't speak for it!
cdc@uafhcx.uucp (C. D. Covington) (06/13/90)
In article <539@skipper.dfrf.nasa.gov>, rando@skipper.dfrf.nasa.gov (Randy Brumbaugh) writes: > Wouldn't it be possible to use DSP to eliminate feedback > in real time before it is noticed? The processor would > "look" at the spectral contents of the program and watch > for any "runaway" frequency bands, where the energy is > increasing exponentially (??). It could also check for This is a great idea that lots of people have had over the years but we are all waiting on the cost to get down. In order to do a good job, lots of bits are needed to keep the quantization noise down. I even went to the trouble of measuring the impulse response of my church's sanctuary. Very impressive. I then took a *long* FFT of the results. More impressive. Troublesome resonances tend to be *very* narrowband due to the distances involved. Of course if anyone turns their head - all bets are off. The system needs to be very fast in producing good parametric estimates of the feedback path characteristics. What I have thought of is an array of possible suboptimal solutions. I did spend a great deal of time at the library studying architectural acoustics. I had no idea people took it all that seriously. You ought to look up a few references. They call impulse response an echogram and measure it in European concert halls using a 9mm blank pistol! C. David Covington (WA5TGF) cdc@uafhcx.uark.edu (501) 575-6583 Asst Prof, Elec Eng Univ of Arkansas Fayetteville, AR 72701
bhanafee@ADS.COM (Brian Hanafee) (06/13/90)
In article <4762@uafhp.uark.edu> cdc@uafhcx.uucp (C. D. Covington) writes: > [Much stuff deleted] They call impulse response an echogram and >measure it in European concert halls using a 9mm blank pistol! > When I took the course in acoustics, I was told that popping a large balloon works much better than a pistol, because the frequency distribution is beter. Gunshots are somewhat deficient in the low end of frequency. I also want to point out that there are really two implementations of this device (feedback assassin) that we could be discussing: the first is a big, expensive version that sound companies would want to buy, with complete tweaking of all the parameters. The cheaper alternative might be a "black box" which does nothing more than look for frequency spikes and attenuate them. This could be very useful for cases when sound quality is less important than not having feedback. Brian Hanafee
bill@bilver.UUCP (Bill Vermillion) (06/13/90)
In article <6418.267529f7@umiami.miami.edu-> chuck@umiami.miami.edu writes: ->In article <539@skipper.dfrf.nasa.gov>, rando@skipper.dfrf.nasa.gov ->(Randy Brumbaugh) writes: ->> ->> ...Wouldn't it be possible to use DSP to eliminate feedback ->> in real time before it is noticed? ... ->> ->I've done live sound work for a few years now. I am certain that many folx ->such as yourself have thought of such a monster, but I have some thoeries ->as to why it's not out yet. However, it is a *great* idea! -> ->Also, you'd either have to decide on a specific computer to run it on, or ->develop your own CPU, operating system and user interface to run with the ->unit. And, if you decided on a computer, should it be a Mac, IBM, etc? ->How many sound companies have road-worthy computers? Since it (the automatic FB eliminating DSP) would be a dedicated piece of equipment it would make sense to make it a single piece of gear. It doesnt' need to be an add in board to a computer, nor should it. However, as wonderful as this device sounds, I'll bet you it will (when it happens) get overused, and improperly used. I can radio stations who are now using 4 band compressors to get a louder signal than those with single band units, now going to the "infinite band" unit. Then what music sounds like will only be a memory for many people. :-(. bill -- Bill Vermillion - UUCP: uunet!tarpit!bilver!bill : bill@bilver.UUCP