In this guide: What feedback is · Why one frequency first · The five factors · What to fix first · Ringing out · Wedges and IEMs · Channel discipline · Suppressors · Problem cases · Speech in halls · Common mistakes · FAQ

What feedback actually is

Every reinforced microphone sits in a loop: voice into microphone, amplified, out of a loudspeaker, back through the air, into the same microphone. That path is unavoidable; the question is how much signal survives it.

While the loop loses more than it gains, the system is stable. When the gain round the loop reaches unity, which is 0 dB, at a frequency where the total phase shift is a whole number of 360 degrees, it sustains itself: the Barkhausen condition for oscillation. The system stops reinforcing your voice and generates a tone of its own.

Every real fix therefore either reduces the gain in the loop or reduces the acoustic coupling from loudspeaker back to microphone. Nothing else works, and how far you can push a channel is your gain before feedback.

Why it always starts at one frequency

Feedback arrives as one note, not a broadband roar, because the loop response is nothing like flat: microphone and loudspeaker response, room modes, reflections and comb filtering combine into peaks and dips several dB deep.

Raise the fader and the highest peak reaches unity first. Dr C. Paul Boner established this in the 1960s: feedback happened at one precise frequency, and a narrow notch there restored stability. Push further and the second peak joins, then the third, which is why an over-driven system howls on several notes at once, and why the same rig rings on a different note in every venue.

The five things that set your gain before feedback

Five variables matter, in rough order of what they buy you. The last is the one most people reach for first.

1. Microphone to source distance

The biggest single win, and it is free. Sound obeys the inverse square law, so halving the source to microphone distance raises the direct signal by 6 dB while everything else at the microphone stays put. Shure call moving the microphone closer the single easiest way to reduce feedback.

A singer working 100 mm off the grille rather than 200 mm has gained 6 dB. A lectern gooseneck brought from 300 mm to 75 mm has gained 12 dB, more than any equalisation can hand you.

2. Microphone to loudspeaker distance and geometry

The mirror image: each time the loudspeaker to microphone distance doubles, the system can go 6 dB louder. Geometry matters more than raw distance, and one rule gets broken more than any other. The loudspeakers must be in front of the microphones. If any part of a main loudspeaker’s coverage falls behind the plane of the open microphones, most of your headroom is gone before the first channel is unmuted. Aim matters too: a box pointed into a hard rear wall returns energy to the whole stage. Our small venue PA guide covers getting the mains right.

3. Polar pattern and where the microphone points

A directional microphone buys gain before feedback by rejecting sound from the directions the loudspeakers are in, which only works if they are in the rejection zones. Aiming a monitor into a null can be worth as much as 20 dB of rejection.

A well-aimed cardioid beats a badly aimed hypercardioid. Note too that a cardioid held parallel to the floor cannot get a wedge into its null, so either the microphone angles upwards or the wedge moves. For pattern choice by source, see our guide to choosing live microphones.

4. Room acoustics

In a reverberant room much of the level arriving at the microphone is the reverberant field rather than direct sound, and it is roughly the same everywhere, so beyond the critical distance moving the microphone away buys almost nothing. That is why a PA which behaves impeccably in a carpeted function room becomes unmanageable in a stone church, and why drapes, carpet and a full audience all help.

5. System equalisation

Last, not first. Flattening the peaks does raise the point at which the first one reaches unity, but the amount on offer is modest: Shure put the realistic gain from ringing out a whole system at 3 dB to 9 dB, against 6 dB for halving one microphone’s working distance. Placement first, always.

The order to work through on a real gig

1. Mute anything not needed. Free, instant, often the whole fix.
2. Fix the geometry. Loudspeakers in front of the microphone plane, monitors aimed into nulls, microphones off hard walls.
3. Get every microphone closer to its source. This is where the dB are.
4. Sort out gain structure. Preamp gain set for a well-placed microphone, faders near unity, nothing at its limit compensating for distance.
5. High-pass everything that is not a kick or a bass. Speech microphones lose nothing above about 80 Hz.
6. Undo tone boosting, then equalise, as a deliberate ring-out rather than panic cuts mid-speech. Broad boosts are borrowed gain, repaid at the first ring.
7. Leave headroom, 3 dB to 6 dB back from wherever it started to ring.
8. Re-check with the room full. Bodies absorb, so the operator turns it up.

Ringing out a system properly

Provoke the first feedback frequencies at a controlled level and notch them, so the loop response is flatter and the system goes louder before the first peak hits unity.

1. Start flat. Zero every filter, including whatever was left in last month.
2. Set the scene honestly. Microphone where it will actually be used, with a person standing at it, because a body in front of a lectern microphone changes the response.
3. Raise the level slowly until it just rings. The first hint of a sustained tone, not a full howl.
4. Identify the frequency. An analyser makes this trivial; otherwise cut a narrow band heavily and sweep until the ring disappears.
5. Notch narrow and shallow. Narrow enough to be inaudible, and no deeper than the peak that caused it plus a small margin: a couple of dB, not tens.
6. Bring the level up and repeat. Three or four notches is a normal, healthy result. Stop while it still sounds good, then pull back 3 dB to 6 dB for working headroom.
7. Repeat for every monitor mix. Each wedge is its own loop with its own peaks.

Why heavy-handed equalisation ruins the tone

A notch cannot tell a feedback tone from a note the singer is holding, so every dB you remove comes out of the programme too. Ten deep cuts give a thin, phasey voice that still feeds back at the eleventh frequency, and needing that many means the problem is geometry.

Wedges, in-ear monitors and the loop you can delete

A floor wedge is by definition a loudspeaker aimed towards a microphone from close range. It is the tightest loop on the stage, and on most gigs it is where feedback starts, not the mains. Monitor level is part of your gain before feedback whether you like it or not. If you are keeping wedges, use as few as the band will accept, aim each into the null of the microphone in front of it, and ring out each mix on its own.

In-ear monitors do something no equaliser can: they take the monitor loudspeaker out of the room entirely, so the monitor path stops contributing to the loop at all. The stage gets much quieter and vocal microphones only hear the voice plus a little front-of-house spill. The trade-offs are a mix per performer, ambience microphones if the band wants to feel the room, and level discipline. Many bands run a hybrid: in-ears for the vocalists, a wedge for the drummer. Our in-ear monitor systems guide covers wired against wireless and how many mixes you need.

Channel discipline, gates and automatic mixers

Here is the principle that catches out almost every multi-microphone speech system: every open microphone adds to the total gain in the loop. Each time the number of open microphones doubles, one to two, two to four, four to eight, potential acoustic gain drops by 3 dB. One lectern microphone to eight costs 9 dB before anyone touches a fader.

So the cheapest headroom in the building is muting: mute buttons between items, or mute groups, DCAs and scene recall on a larger desk, a strong argument for a digital mixing system in a multi-use room. Gates help too, but a threshold set too high or an attack too slow clips the consonants that carry intelligibility.

For multi-microphone speech, an automatic microphone mixer does the job properly: it gates channels faster than a person can and applies number-of-open-microphones attenuation, nominally 3 dB per doubling, so total gain stays constant however many are live. Refinements include a threshold that opens for speech but not steady room noise, a bus limit allowing one microphone per talker, and a last-microphone lock-on. Many install system processors include automatic mixing alongside equalisation and delay. It is not a feedback suppressor, though, and will not save a lectern microphone under a loudspeaker.

Feedback suppressors: what they can and cannot do

Automatic feedback suppressors, sometimes sold as feedback eliminators, work one of two ways. Adaptive notching listens for a building ring and drops a narrow filter on it automatically: Boner’s method under software control. Frequency shifting moves the signal up or down by a few hertz so the return no longer lines up in phase, breaking the phase condition instead of the gain condition.

• Automatic notching typically provides only a couple of dB of extra gain before feedback, and none at all where the loop response is already flat, because there are no peaks left to remove.
• Frequency shifting depends heavily on the room. Published measurements show a dry lecture hall benefiting by slightly less than 2 dB, and a reverberant space with a reverberation time over one second gaining nearly 6 dB. Shifts stay under about 12 Hz to keep the pitch change inaudible.

A few dB is genuinely useful, and there are rooms where a suppressor earns its rack space: installed multi-use spaces with a volunteer or nobody on the desk, roving handhelds that get carried in front of the loudspeakers, and halls where the microphone position changes weekly and nobody will ring the system out again. What they are not is a substitute for placement, and they are a mixed blessing on music, because an adaptive notcher can mistake a sustained organ pedal for feedback. Fix the filters found during a ring-out and leave two or three dynamic.

The cases that cause most of the trouble

A lectern gooseneck under a ceiling speaker

The worst geometry in most buildings, and common, because lectern and ceiling speaker were positioned by different trades. A speaker directly above a lectern fires down into the microphone with no useful null. Fixes in order: move the lectern out of its coverage; zone those speakers to drop while the microphone is live; get the head closer to the mouth; fit a tighter capsule. Our guide to conference and gooseneck microphones covers capsule patterns, and goosenecks come in a range of lengths.

A headset microphone on a loud stage

A headset starts strong, with the capsule a few centimetres from the mouth and never moving, but it picks up the whole stage at a fixed level and it walks. The wearer will eventually stand in front of a main loudspeaker, where no pattern saves you. Make that area off limits, and mute when they are not speaking.

Choir microphones

Choir reinforcement fights the physics twice: the microphones must be back far enough to cover a group, giving a weak direct signal, and several are open at once, so the 3 dB penalty per doubling applies. Hang them close to the front row, use the fewest that cover evenly, keep them behind the plane of the mains, and accept that a choir will never sit as loud as a soloist.

Hard, reflective halls

The surfaces nearest the microphone matter as much as the room. A microphone 150 mm above a hard lectern top receives a delayed reflection that combs its response, creating exactly the peaks feedback exploits. Move microphones off hard boundaries, or lay something absorbent on the lectern top.

A singer cupping the microphone head

Measurable, not folklore. Microphones in the cardioid family get their directionality from rear entry ports that let sound reach the back of the diaphragm with a controlled delay. Wrap a hand round the grille and you block those ports: the pattern collapses towards omnidirectional and the rear rejection disappears. A cardioid picks up roughly a third as much ambient sound as an omni, so cupping throws that away. No processing fixes it.

Why this matters for speech in churches and halls

Feedback control and speech intelligibility are the same problem in a different hat. Both come down to the ratio of direct to reverberant sound, so every step that raises gain before feedback also raises the direct proportion, and the room hears words rather than wash.

Critical distance ties them together: the distance from a loudspeaker at which direct sound and the reverberant field are level. Inside it speech is clear; beyond it consonants smear into the reverberation. It lengthens with loudspeaker directivity and room absorption, and shortens as reverberation time rises.

• More loudspeakers, each quieter and closer to people, so most of the congregation sits inside the critical distance.
• Directional loudspeakers aimed at the seating, not at stone walls or a vaulted ceiling.
• Close, directional microphones, so you amplify speech, not the room’s reverberation of it.
• Absorption: drapes, banners and rear carpet lengthen critical distance.

Common mistakes

• Turning the microphone up instead of moving it closer. The first raises the whole loop, the second only the signal you want.
• Reaching for the equaliser before fixing placement, then cutting notches too wide. An octave-wide slider dropped 12 dB takes a chunk of the voice with it.
• Leaving no headroom, so the first louder speaker squeals.
• Loudspeakers behind the microphone line, and channels left open all night.
• Cupping the grille, and ignoring the room. Absorption and tighter coverage change the answer; a bigger amplifier does not.

Popular picks from the shop

Frequently asked questions

How do I stop feedback quickly in the middle of a show?

Pull the offending fader down, then mute anything not in use. Ask the performer to work closer and stop cupping the grille. If it is a monitor, drop that mix, not the mains.

Why does my microphone feed back?

Because your loudspeaker output is arriving back at it strongly enough that the loop gain has reached unity. Almost always one of five things: too far from the source, too close to a loudspeaker, pointed the wrong way, a room that is too reverberant, or too many microphones open.

What is gain before feedback, and how do I get more of it?

How far you can turn a microphone up before the system rings. The biggest lever is source distance: halve it and you gain 6 dB. Then increase the loudspeaker to microphone distance, aim into nulls, cut the open microphone count, and only then equalise.

Do feedback eliminators actually work?

Modestly. Automatic notching typically buys a couple of dB, and none at all if the loop response is already flat. Frequency shifting measured slightly under 2 dB in a dry lecture hall and nearly 6 dB in a reverberant space. A safety net, not a fix.

How do I stop feedback in a church or village hall?

Get microphones much closer to the talkers, keep loudspeaker coverage off the microphone positions, zone the speakers nearest live microphones so they can drop, mute everything not in use, and use more loudspeakers at lower level rather than one loud pair.

What frequency does feedback usually happen at?

Wherever the loop response peaks, which differs in every room. A hoot or howl is usually 250 Hz to 500 Hz, a singing tone around 1 kHz, a whistle above 2 kHz, and very rarely below 80 Hz or above 8 kHz.

Will switching to in-ear monitors get rid of feedback?

It removes the monitor loop entirely, which on most gigs is where feedback starts: with no wedge radiating at the vocal microphones, monitor level stops contributing. The front-of-house loop remains, so placement still matters.