Feedback

Why feedback happens

Every stage in the loop contributes a gain (or loss) at every frequency: the microphone's sensitivity and polar pattern, the preamp and channel gain on the mixer, the amplifier's gain, the loudspeaker's output level, and finally the acoustic transfer of sound through the room back into the microphone. Multiply all of these together at a given frequency and you get the loop gain for that frequency.

If, at some frequency, the loop gain reaches 1 (0 dB) at the moment the signal arrives back "in phase" with itself, that frequency becomes self-sustaining — energy at that frequency keeps regenerating every time it goes around the loop. If the loop gain exceeds 1, the level at that frequency grows with every pass. This is the same stability condition (sometimes called the Barkhausen criterion) used to analyze any feedback loop, electronic oscillators included.

What pushes a frequency over 0 dB

• Gain too high for the physical separation between the microphone and the loudspeakers — the closer and louder the speaker is relative to the mic, the less headroom before the loop reaches unity.
• Microphone aimed at, or placed near, a loudspeaker or monitor wedge — directly increasing the acoustic transfer back into the mic.
• Wide or mismatched polar patterns — an omnidirectional mic, or a directional mic facing the wrong way, picks up more of the room (and the speakers) than a well-aimed cardioid or hypercardioid mic would.
• Room acoustics — hard, reflective surfaces and resonant room modes effectively add extra gain at specific frequencies, on top of whatever the electronics provide.
• Coincidental peaks stacking up — a microphone's response peak, a loudspeaker's response peak, and a room resonance landing near the same frequency add together, making that frequency the first to cross 0 dB.
• EQ boosts that happen to land on one of these peaks, adding gain exactly where the system has the least headroom.
• Multiple open microphones — every additional live mic adds more paths back into the loop, raising the effective gain even if no single channel's fader has changed.

Stable, ringing, or runaway — the loop gain decides

The behaviour of any feedback loop at a given frequency falls into three regimes, based purely on whether the loop gain at that frequency is below, at, or above unity (0 dB):

• Loop gain < 0 dB (gain < 1): any energy at that frequency loses a little each time around the loop, so it dies away. The system is stable.
• Loop gain = 0 dB (gain = 1): energy neither grows nor shrinks — it sustains itself indefinitely as a continuous tone. This is the classic feedback "ring."
• Loop gain > 0 dB (gain > 1): energy grows on every pass. In practice it grows exponentially until something else stops it — usually the amplifier or loudspeaker running out of headroom and clipping (see Section 04).

What's physically happening in the electronics

Sections 01–03 described feedback in terms of loop gain and frequency response — useful for understanding why and where feedback occurs. This section goes one level deeper: what's physically happening inside the preamp, the power amplifier, and the loudspeaker itself once that loop gain crosses 0 dB.

The preamp stage: clean amplification until it isn't

A typical microphone preamp or channel input is built around an op-amp (or transistor) gain stage. In the common inverting-amplifier configuration, the gain is set by the ratio of two resistors — a feedback resistor (R_f) and an input resistor (R_in) — so gain ≈ −R_f/R_in.

The op-amp's output can only swing within roughly its supply rail voltages (minus a volt or two lost to the output transistors' own saturation — sometimes called "headroom"). If the input signal, multiplied by the stage's gain, would call for an output beyond that range, the output simply can't go any further. The waveform's peaks get cut off flat — this is clipping, and it's the first place in the chain that a feedback signal runs out of room.

The power amplifier: where the energy is

Most power amplifier output stages use complementary pairs of transistors (a "push-pull" or class-AB arrangement) — one device handles the positive half of the waveform, the other handles the negative half, and together they reproduce the full signal across the loudspeaker.

Whatever isn't delivered to the speaker as useful power is dissipated inside these transistors as heat, removed by a heatsink (and sometimes a fan). Heatsinks are sized around the average power of normal program material — music and speech have large peaks but spend most of their time well below those peaks, so the average heat load is much lower than the momentary peak load.

Inside the loudspeaker: the voice coil heats up

Electrically, a loudspeaker's voice coil behaves mostly like a resistor (its DC resistance, R_DC) with some added inductance, sitting as the load on the amplifier's output. Current flowing through R_DC dissipates power as heat (P = I²×R_DC) — this heat is what makes the cone move, but it's also a byproduct that has to go somewhere.

As the coil heats, the resistance of its copper winding rises — roughly 0.4% per °C. At a constant drive voltage, this slightly reduces the current and therefore the output level, an effect called power compression. It's self-limiting in the sense that it slightly reduces the power being dissipated, but it's also a direct sign that the coil is heating toward its limit. If a sustained tone keeps pushing past the temperature rating of the coil's wire insulation or the adhesives bonding the windings together — depending on construction, commonly somewhere in the range of roughly 150–300°C — the coil can deform, short turns together internally, or detach from its former entirely. That's a "blown" driver.

Clipping doesn't just get louder — it changes shape

A clean sine wave contains energy at one frequency only. When that wave is clipped — its peaks flattened because it exceeds a supply rail or the amplifier's rated output — the result is no longer a pure sine. Mathematically, a flat-topped (square-ish) wave is equivalent to a sine at the original (fundamental) frequency plus a series of additional odd harmonics (3×, 5×, 7×... the fundamental frequency) at progressively smaller amplitudes.

This has two consequences. First, the added harmonic content increases the signal's total RMS power for a given peak voltage — a fully clipped square wave carries roughly twice the power (about +3 dB) of an undistorted sine wave with the same peak voltage, directly increasing the heat loads described above for both the amplifier and the speaker. Second, if the feedback frequency itself is already in the upper-midrange — a common range for feedback to ring at — these added harmonics land at even higher frequencies, squarely in the range tweeters and compression drivers are both most active in and least equipped to dissipate continuously.

What feedback does to hearing

The cochlea — the spiral, fluid-filled organ of hearing in the inner ear — contains rows of hair cells with tiny hair-like projections (stereocilia) that bend in response to sound-induced vibration and convert that motion into nerve signals. Loud enough, or long enough, vibration can bend, fatigue, or permanently destroy these stereocilia and the hair cells they belong to. In mammals, including humans, cochlear hair cells do not regenerate once lost — the damage is permanent.

Two different injury mechanisms

• Cumulative noise-induced hearing loss (NIHL): repeated or prolonged exposure to moderately loud sound (roughly 85 dB and above) gradually fatigues and kills hair cells, particularly those tuned to higher frequencies first. This produces a slow, often unnoticed, high-frequency hearing loss and is the standard concern behind workplace noise regulations.
• Acoustic trauma: a single, very loud, sudden event — exactly what an uncontrolled feedback spike is — can damage hair cells (or, at extreme levels, rupture the eardrum) in an instant, regardless of normal daily exposure limits. Feedback is dangerous specifically because it can jump from a normal performance level to a piercing high-SPL tone in well under a second, with no warning.

Symptoms after exposure

• Temporary threshold shift (TTS): hearing feels muffled and there may be ringing for minutes to hours after exposure, then it recovers. TTS is a sign the ear has been pushed past a safe level, even though it "got better."
• Permanent threshold shift (PTS): hearing loss that does not recover — the actual damage that NIHL and acoustic trauma cause.
• Tinnitus: a persistent ringing, hissing, or buzzing sound with no external source, often the first noticeable sign of hair-cell damage.
• Hyperacusis: an increased sensitivity to, and discomfort from, everyday sounds — sometimes a consequence of the ear's altered response after acoustic trauma.

A special case: in-ear monitors

Performers wearing in-ear monitors (IEMs) have a transducer sitting at or inside the ear canal. If feedback occurs in a monitor mix feeding an IEM, the resulting spike is delivered at very close range with essentially no distance for the sound to lose energy before reaching the eardrum — making IEM feedback one of the highest-risk scenarios for acute hearing damage in live sound.

What feedback does to gear

Feedback is dangerous to equipment largely because of what kind of signal it is, not just how loud it gets. Normal music and speech have a crest factor — a large gap between average (RMS) level and peak level — made up of transients, pauses, and varying dynamics. A feedback tone, by contrast, is close to a continuous sine wave at a single frequency: its average power is nearly equal to its peak power. Components rated assuming "musical" signals can be pushed well past their continuous-power rating by a sustained feedback tone at the same peak level. Section 04 covers the circuit-level detail behind each point below.

Mixing console / mixer

A mixer operates on line- or mic-level signals, so feedback rarely causes direct physical failure inside the console itself. What it does cause is clipping and distortion in the input preamp or channel circuitry once the signal exceeds the console's headroom — heard as harsh, gritty distortion on top of the howl. Repeated, severe, sustained clipping can add unnecessary thermal stress to output stages over the long term, but the console's main role in a feedback event is usually as the point where the problem is created (via gain) and where it must be controlled (via faders and EQ), rather than as the component most likely to be damaged.

Active loudspeakers

Loudspeaker drivers fail in two distinct ways, and feedback can cause either:

• Thermal failure (overheating the voice coil): a sustained, near-full-power feedback tone delivers continuous I²R heating with none of the "rest" musical transients provide, overheating the coil's insulation or adhesive bonds far faster than music at the same peak level would. High-frequency drivers (tweeters, compression drivers / horns) are especially vulnerable: feedback frequently rings at high frequencies (often 1–8 kHz), and HF drivers have small voice coils with limited surface area to dissipate heat and correspondingly low continuous-power ratings.
• Mechanical failure (over-excursion): less common but possible with low-frequency feedback ("rumble" or "motorboating"), where the cone or diaphragm is driven beyond its designed travel limit, tearing the suspension (spider or surround) or causing the voice coil former to strike the magnet structure.

Power amplifiers

When an amplifier is driven beyond its rated output by a feedback signal, its output stage clips (Diagram 3), reshaping the waveform and increasing its RMS power for the same peak voltage (Diagram 6). Combined with the continuous nature of a feedback tone (Diagram 4), this can push the output transistors well past their intended thermal envelope, triggering thermal-protection shutdown, blowing fuses, or — in amplifiers without adequate protection circuitry — causing outright component failure. The clipped, harmonic-rich signal is also itself more likely to damage the loudspeaker it's feeding, compounding the problem.

Other instruments and signal sources

• Electric guitar pickups: the controlled feedback guitarists deliberately use (sustained notes ringing from the amp back through the strings and pickup) doesn't meaningfully stress the pickup itself — pickups are passive magnetic coils with no moving parts to overdrive. The real risk in an uncontrolled situation is, again, to the amplifier and speaker reproducing that sustained signal.
• Acoustic instruments with piezo or internal mic pickups: feedback can excite a resonance in the instrument's body at certain frequencies, but actual structural damage to the instrument from this is extremely rare and would require sustained, extreme sound pressure levels well beyond typical feedback events.
• Wireless and in-ear systems: as noted in Section 05, the equipment risk here is secondary — the primary risk is to the performer's hearing, because IEM transducers sit so close to the ear.

How to tell if damage has already happened

Hearing

• Ringing, hissing, or buzzing (tinnitus) that persists for hours after exposure, or doesn't go away at all.
• Sounds feeling muffled or "underwater," or needing the TV/conversation volume turned up more than usual — a possible temporary threshold shift if it resolves, or a permanent shift if it doesn't.
• Difficulty understanding speech in noisy environments, often an early sign of high-frequency hearing loss.
• The only way to confirm and quantify a hearing change is a professional audiogram from an audiologist, which measures the quietest sound you can hear at each frequency and compares it to a baseline.

Loudspeakers

• Distortion, buzzing, rattling, or a "scratchy" quality on signals that used to sound clean.
• A driver that produces no sound at all, or is noticeably quieter than its matching driver in a pair.
• A scraping or rubbing sound, often meaning the voice coil is no longer centred in the magnetic gap and is physically rubbing.
• A burning smell, or visible discoloration/melting around the dust cap or surround.
• Multimeter test: with the speaker disconnected from any amplifier, measure resistance across its terminals. A reading of infinite resistance (open circuit) means the voice coil winding has failed. A working coil typically reads somewhat below the driver's nominal impedance — for example, roughly 5–7 ohms of DC resistance on an "8 ohm" driver is normal; a reading far outside that range, or no reading at all, indicates a problem.

Amplifiers

• No output at all, or one channel dead while the other works.
• Distortion or buzzing even at low volume, when the signal feeding it is clean.
• A protection or fault indicator LED lit up, or the unit refusing to come out of standby/protect mode.
• Heatsinks or chassis noticeably hotter than during normal operation, or a burning smell.
• A blown internal fuse (only inspect or test this with the unit unplugged, and only if you're comfortable opening the chassis).

Mixing console

• A channel's clip indicator lighting persistently, even at normal gain settings.
• Crackling, static, or intermittent dropouts on a specific channel.
• Phantom power no longer reaching a condenser microphone that previously worked on that channel.

Preventing feedback — and the damage it causes

Gain structure ("gain staging")

Set the gain at each stage of the chain — microphone preamp, channel fader, master, amplifier input — so that no stage is pushed unnecessarily high only to be turned back down later. A well-staged system has more headroom available before any frequency's loop gain reaches 0 dB, giving you a larger margin before feedback starts.

Microphone choice and placement

Directional microphones (cardioid, supercardioid, hypercardioid) reject sound from certain directions far more than an omnidirectional mic does. Positioning a directional mic so that loudspeakers and monitor wedges sit in its rejection zone — rather than its main pickup lobe — directly lowers the acoustic-path gain in the loop.

Room acoustics

Absorptive panels on hard, reflective surfaces (especially walls and ceiling areas facing the stage) reduce the reflected energy that adds extra "gain" at specific frequencies. Bass traps address low-frequency room modes, which can otherwise contribute to low-frequency feedback or "boom."

"Ringing out" the system and notch filters

Before a performance, sound engineers often deliberately bring the system gain up until feedback just begins, identify the ringing frequency on a graphic or parametric EQ, and apply a narrow reduction — a notch — at that frequency, then repeat at a slightly higher gain to find the next-weakest frequency. A notch is simply a sharp, narrow cut at one frequency: it brings the loop gain at that specific frequency back below 0 dB while leaving the rest of the spectrum essentially untouched, so the system gains usable headroom without sounding noticeably different.

Automatic feedback suppressors

Dedicated feedback-suppression processors continuously monitor the signal for the rapid, narrow-band buildup characteristic of feedback and automatically apply a notch like the one in Diagram 10 — often faster and more precisely than a human operator could react.

Limiters, compressors, and filters

• Limiters on amplifier inputs or loudspeaker processing cap the maximum signal level, protecting drivers and amplifier output stages even if a feedback spike occurs.
• High-pass filters on vocal and instrument channels remove unnecessary low-frequency energy (handling noise, stage rumble) that can otherwise contribute to low-frequency feedback without adding anything useful to the sound.
• In-ear monitors instead of floor wedges remove an entire acoustic path from the loop — there's no open speaker on stage for a vocal mic to "hear."

Reducing feedback once it starts

When feedback breaks out during a live performance, speed matters more than precision. A practical order of operations:

1. Pull down the fader on the channel that's ringing — this is the fastest way to drop the loop gain for that path below unity.
2. Mute any microphones not currently in use — every open mic is another path back into the loop.
3. Move the microphone or performer away from the loudspeaker or monitor that's feeding the loop, or reorient a directional mic so the speaker falls into its rejection zone.
4. Notch the ringing frequency on a parametric or graphic EQ if one is available and you can identify the pitch quickly.
5. Reduce the monitor or main system level generally if the problem persists — this lowers the loop gain across all frequencies at once, at the cost of overall volume.

Addressing the cause (gain, placement, EQ) rather than just "riding it out" matters because, as covered in Sections 04–06, every second a feedback tone continues is a second of potential thermal stress on amplifiers and loudspeakers, and a second of potential acoustic trauma for anyone within range — especially performers wearing in-ear monitors.

Summary by component