The conditions for feedback follow the
Barkhausen stability criterion, namely that, with sufficiently high gain, a stable
oscillation can (and usually will) occur in a feedback loop whose frequency is such that the phase delay is an
integer multiple of 360
degrees and the
gain at that frequency is equal to 1. If the small-signal gain is greater than 1 for some frequency, then the system will start to oscillate at that frequency because noise at that frequency will be amplified. Sound will be produced without anyone actually playing. The sound level will increase until the output starts
clipping, reducing the loop gain to exactly unity. This is the principle upon which
electronic oscillators are based; in that case, although the feedback loop is purely electronic, the principle is the same. If the gain is large but slightly less than 1, then
ringing will be introduced, but only when at least some input sound is already being sent through the system. Early academic work on acoustical feedback was done by Dr.
C. Paul Boner. Boner was responsible for establishing basic theories of acoustic feedback, room-ring modes, and room-sound system equalizing techniques. Boner reasoned that when feedback happened, it did so at one precise frequency. He also reasoned that it could be stopped by inserting a very narrow
notch filter at that frequency in the loudspeaker's
signal chain. He worked with
Gifford White, founder of
White Instruments to hand craft notch filters for specific feedback frequencies in specific rooms.
Distance To maximize
gain before feedback, the amount of sound energy that is fed back to the microphones must be reduced as much as is practical. As sound pressure falls off with 1/r with respect to the distance r in free space, or up to a distance known as reverberation distance in closed spaces (and the energy density with 1/r2), it is important to keep the microphones at a large enough distance from the speaker systems. As well, microphones should not be positioned in front of speakers, and individuals using mics should be asked to avoid pointing the microphone at speaker enclosures.
Directivity Additionally, the loudspeakers and microphones should have non-uniform
directivity and should stay out of the maximum
sensitivity of each other, ideally in a direction of cancellation.
Public address speakers often achieve directivity in the mid and treble region (and good
efficiency) via horn systems. Sometimes the woofers have a
cardioid characteristic. Professional setups circumvent feedback by placing the main speakers away from the band or artist, and then having several smaller speakers known as
monitors pointing back at each band member, but in the opposite direction to that in which the microphones are pointing taking advantage of microphones with a
cardioid pickup pattern which are common in sound reinforcement applications. This configuration reduces the opportunities for feedback and allows independent control of the sound pressure levels for the audience and the performers.
Frequency response Almost always, the natural frequency response of a
sound reinforcement systems is not ideally flat as this leads to acoustical feedback at the frequency with the highest loop gain, which may be a
resonance with much higher than the average gain over all frequencies. It is therefore helpful to apply some form of
equalization to reduce the gain at this frequency. Feedback can be reduced manually by
ringing out a sound system prior to a performance. The sound engineer can increase the level of a microphone until feedback occurs. The engineer can then attenuate the relevant frequency on an equalizer, preventing feedback at that frequency but allowing sufficient volume at other frequencies. Many professional sound engineers can identify feedback frequencies by ear but others use a real-time analyzer to identify the ringing frequency. To avoid feedback, an automatic
feedback suppressor can be used. Some of these work by shifting the frequency slightly, with this upshift resulting in a
chirp-sound instead of a howling sound of unaddressed feedback. Other devices use sharp notch filters to filter out offending frequencies. Adaptive algorithms are often used to automatically tune these notch filters. ==Deliberate uses==