/ˈfiːdbæk/
Feedback is the phenomenon where an audio signal's output is routed back into its own input, creating a loop that can produce delay trails, resonant sustain, or uncontrolled oscillation depending on the loop gain.
Every producer has winced at the squeal of runaway feedback — but the greatest electronic textures in recorded history were built by people who learned to walk that edge on purpose.
Feedback, in its most fundamental definition, is the routing of a system's output signal back to its input. In audio, this creates a loop: a signal enters a processing chain, exits, and some proportion of that exited signal is reintroduced at the input. The sonic result depends entirely on the gain of that loop. When loop gain is less than unity — meaning less than 100% of the signal is being fed back — the recirculating copies decay over time, producing the familiar trails of a delay or reverb. When loop gain reaches or exceeds unity, the signal sustains indefinitely or grows, ultimately resulting in the pure-tone oscillation known as howlround or, in extreme cases, the destructive screech familiar from PA systems and live stages.
In a production context, feedback is not a single effect but a principle underlying an enormous family of tools. Every delay plugin with a Feedback or Regeneration knob is a controlled feedback loop. Every analog tape echo unit — the Roland Space Echo, the Maestro Echoplex — used physical tape loops to achieve the same recirculation. Spring and plate reverbs develop their characteristic shimmer through mechanical feedback resonances. Comb filters, flangers, and chorus units all exploit feedback paths to colour their frequency responses. Even the resonance parameter on a synthesizer filter is a feedback coefficient: at maximum resonance, the filter self-oscillates because its output is fed back into the input at unity gain.
The critical variable in any feedback system is loop gain, often expressed as a percentage (0–100%) or a dB figure relative to unity (0 dB). Below unity, each recirculating pass attenuates the signal; the number of audible repetitions and their decay rate are direct functions of this attenuation per pass combined with the loop delay time. At unity the signal repeats forever without growing or shrinking. Above unity, gain increases exponentially with each pass, producing the runaway oscillation that can damage speakers and hearing. Practical delay units deliberately limit their feedback range to just below unity — typically around 95–98% maximum — as a safety measure, though some boutique designs and Max/MSP patches allow over-unity feedback for experimental purposes.
Frequency-selective feedback is equally important to understand. A plain feedback loop recirculates all frequencies equally, but inserting an equaliser, filter, or even a simple tone control inside the loop changes which frequencies accumulate fastest. High-frequency shelving cuts inside a delay loop create warmer, darker repeats that simulate tape saturation and head absorption. A bandpass filter inside a feedback path produces a resonant, pitched whistling effect used extensively in dub production and in the creation of Karplus-Strong string synthesis. Producers who understand this principle can sculpt the spectral character of their feedback-based effects far beyond what a single Feedback knob allows.
Spatially, feedback interacts with room acoustics in ways that matter in both live sound and studio recording. A microphone placed in front of a speaker in the same acoustic space forms a natural feedback loop through the air itself — the physical version of the electrical routing. Studio engineers exploit controlled versions of this by placing mics near resonant objects, guitar amps in live rooms, or signal chains that feed a speaker and re-record it. Jimi Hendrix's sustained notes and Sonic Youth's noise sculptures are physical-acoustic feedback events that happened to be captured on tape. The principle is identical whether the loop exists in circuitry, in software, or in the air between a speaker cone and a condenser capsule.
At its core, a feedback system can be modelled as a signal summing node feeding into a delay line, whose output is then multiplied by a gain coefficient and added back to the summing node. In a simple digital delay, the summing node is the plugin's input buffer; the delay line is a circular buffer of length N samples (where N = delay time × sample rate); the gain coefficient is the Feedback parameter, typically a value between 0.0 and 1.0. On every processing cycle, the dry input signal and the delayed-and-attenuated feedback signal are summed, written into the delay buffer, and also sent to the plugin output. This loop means that a single input transient produces an infinite geometric series of echoes, with amplitude scaling by the feedback coefficient on each pass. At a feedback value of 0.75 and a 500 ms delay, echo 1 is at –2.5 dB, echo 2 at –5.0 dB, echo 3 at –7.5 dB, and so on, producing a smooth exponential tail.
Stability is governed by the Nyquist-stability criterion: a feedback loop is stable only when the loop gain magnitude is strictly less than 1.0 at every frequency simultaneously. Inserting gain-boosting elements inside the loop — saturation, upward expansion, or resonant EQ boosts — can push specific frequencies above the stability boundary even while the aggregate broadband gain appears safe. This is how a guitarist with a cranked amp can sustain a single pitch indefinitely: the pickup, amp, and speaker system form a loop where one frequency (determined by the guitar string's natural resonance and the room's standing waves) achieves exactly unity loop gain while others decay. The frequency that stabilises depends on phase relationships throughout the loop — which is why moving even slightly closer to or farther from the amp shifts the feedback pitch.
In the digital domain, feedback loops inside a single processing block must be handled carefully to avoid zero-sample delay loops, which are mathematically unsolvable and crash most DAWs. Every practical feedback implementation introduces at least one sample of latency in the feedback path, creating a minimum delay of 1/sample-rate seconds (roughly 20 microseconds at 48 kHz). For musical delays this is irrelevant, but for comb-filtering effects — where the delay time is measured in samples rather than milliseconds — even one or two samples of forced latency changes the comb filter's notch frequencies. Hardware analog units avoid this entirely because their feedback signals travel at the speed of electricity through physical components, with no quantization of delay time.
Frequency-selective accumulation inside feedback paths deserves special attention. If a first-order low-pass filter with a –3 dB point at 4 kHz is inserted in the feedback path, then high-frequency components attenuate more per pass than low-frequency components. After 10 recirculations at 0.85 loop gain, a 10 kHz component might be 40 dB below the fundamental while the low-mids are only 7 dB down. The result is a warm, rounded tail that darkens with each repeat — the defining tonal signature of tape echo units. Conversely, a high-frequency boost inside the loop causes treble content to accumulate faster, producing the bright, edgy feedback character of certain bucket-brigade device (BBD) chorus and flanger circuits.
Understanding feedback mathematically allows producers to predict rather than discover by accident. The time constant τ of a feedback tail — the time to fall by 1/e, approximately 8.7 dB — equals –delay_time / ln(feedback_coefficient). At a 375 ms delay and a feedback coefficient of 0.707 (–3 dB), τ ≈ 1.08 seconds, meaning the tail is effectively inaudible after roughly 3τ ≈ 3.2 seconds. This formula lets a producer calculate exact feedback settings to match a reverb tail length, align decay to a bar boundary, or set up a self-oscillating loop that will take exactly eight bars to blow up — a compositional tool, not an accident.
Diagram — Feedback: Signal flow diagram showing a feedback delay loop: input enters a summing node, passes through a delay buffer and optional low-pass filter, then a gain stage (g), exits as output, and loops back to the summing node via a dashed feedback path.
Every feedback — hardware or plugin — operates on the same core parameters. Know these and you can work with any implementation.
Expressed as 0–100% or a dB offset from unity. Values below 95% produce decaying repeats; 95–99% produces very long tails; 100% produces infinite sustain. Most DAW delay plugins hard-limit this at 99% to prevent runaway oscillation, but modular and hardware units may allow values beyond unity for intentional self-oscillation.
Sets the spacing of repeats in milliseconds, samples, or note divisions relative to project BPM. Short delay times (1–30 ms) in a feedback loop produce comb filtering and flanging. Medium times (80–500 ms) produce classic slapback and rhythmic echoes. Long times (500 ms–2 s) produce ambient tails. BPM-synced settings should be chosen to align feedback reflections with the rhythmic grid.
A low-pass or high-pass filter placed inside the feedback loop alters the tonal character of each successive repeat. A –3 dB point at 3–6 kHz in the LP position mimics the high-frequency absorption of analog tape heads, producing warm, darkening tails. A high-pass at 100–200 Hz prevents low-frequency build-up that can cause muddy resonance in long feedback chains with bass-heavy material.
Many delay plugins include a saturation stage inside the feedback path (Valhalla Delay's 'Mode' options, Soundtoys EchoBoy's 'Drive'). Driving this stage increases harmonic density with each recirculation, simulating the non-linearity of worn tape heads or overdriven preamps. At moderate settings (a few dB of saturation) this adds warmth; at heavy settings it creates a self-reinforcing distortion bloom that can become the central texture of a sound design element.
Low-frequency oscillation applied to the delay time within the feedback path creates chorus, vibrato, or tape-wow artifacts on each recirculation. Depth values of 0.1–0.5 ms at rates of 0.3–1.5 Hz are typical for subtle vintage-tape emulation. Higher depths (1–5 ms) at faster rates produce more pronounced chorusing. Because modulation compounds over multiple feedback passes, even a small depth setting can produce wide pitch fluctuations after many recirculations.
In stereo feedback configurations, each repeat can alternate between left and right channels (ping-pong mode) or maintain independent feedback amounts per channel with optional cross-feedback. Cross-feedback coefficients control how much of the left delay output feeds into the right input and vice versa, creating complex stereo interweaving. At matched cross-feedback levels, a mono input produces a wide stereo field; at unmatched levels, complex rhythmic stereo patterns emerge.
Session-ready starting points. Values are starting points for a 120 BPM session; scale delay times proportionally at other tempos and test feedback levels against mix density before committing.
| Parameter | General | Drums | Vocals | Bass / Keys | Bus / Master |
|---|---|---|---|---|---|
| Feedback % | 30–65% | 15–40% | 25–55% | 20–50% | 0–30% |
| Delay Time | 1/8 or 3/16 note | 1/16 or 1/8 note | 1/8 dotted | 1/4 note | 1/4–1/2 note |
| LP Filter Cutoff | 4–8 kHz | 8–12 kHz | 5–8 kHz | 3–6 kHz | 6–10 kHz |
| HP Filter Cutoff | 80–150 Hz | 200–350 Hz | 150–300 Hz | 60–120 Hz | 100–200 Hz |
| Modulation Depth | 0.1–0.3 ms | Off or minimal | 0.2–0.5 ms | 0.1–0.4 ms | Off |
| Drive / Saturation | Light (2–4 dB) | Off or 1–2 dB | 1–3 dB | 2–6 dB | Off |
| Wet Level (parallel) | 15–30% | 8–20% | 10–25% | 12–25% | 5–12% |
Values are starting points for a 120 BPM session; scale delay times proportionally at other tempos and test feedback levels against mix density before committing.
The first systematic observation of audio feedback in electrical systems dates to the early 1920s, when telephone engineers noticed that amplified telephone circuits would occasionally break into sustained oscillation if the handset receiver coupled acoustically to the mouthpiece. Lee de Forest's triode vacuum tube amplifier, patented in 1907, created the first practical electronic feedback loops, and by 1927 Harold Black at Bell Labs had formalised the principle of negative feedback in amplifier design — a concept that would underpin every studio preamp and power amplifier manufactured since. Black's key insight was that deliberately feeding a fraction of the output back to the input out of phase (negative feedback) dramatically reduced distortion and improved frequency-response linearity, a principle still central to every studio-grade operational amplifier.
Positive feedback — feeding output back in phase to reinforce the input — became the foundation of early reverb and echo technology. Les Paul, working in his garage studio in the late 1940s, experimented with multiple tape machines connected in series, routing the output of one back into the record head of another to create layered echo effects. The Echoplex, designed by Don Sommer and introduced commercially in 1959, placed a moveable playback head over a tape loop to allow variable delay times, with a feedback control that routed playback audio back to the record head. By the early 1960s, the Roland RE-201 Space Echo had refined this concept with spring reverb integration and multiple play heads, becoming the defining echo tool of reggae, dub, and early electronic music. Lee 'Scratch' Perry and King Tubby both used the Space Echo's feedback control as a live performance instrument, riding the regeneration knob to create swelling, chaotic noise events that defined the dub aesthetic.
Acoustic feedback as a compositional tool was pioneered in rock music by Link Wray, whose 1958 recording 'Rumble' featured the first commercially released guitar sustain achieved through speaker-to-pickup feedback. Jimi Hendrix elevated the technique to an art form between 1966 and 1970, using his Marshall stacks' extreme volume levels at concerts and in Olympic Studios, London, to produce feedback notes that he could hold, bend, and fade at will — heard definitively on the studio recording of 'Purple Haze' (1967, produced by Chas Chandler) and the live Monterey Pop Festival performance. The Who's Pete Townshend and Sonic Youth's Thurston Moore and Lee Ranaldo developed entirely different feedback vocabularies, the former using it for punctuation and power, the latter treating sustained feedback as a compositional drone texture throughout their mid-1980s output on SST Records.
Digital feedback loops arrived with the first DSP-based delay units in the early 1980s. The Lexicon 224 (1978) and AMS DMX 15-80S (1978) both implemented digital delay with regeneration controls, bringing the feedback loop into the studio rack with precise repeatability unavailable on tape or spring units. Software implementations followed with the first DAW plug-in delay units in the mid-1990s; Digidesign's early Pro Tools delay plug-ins (circa 1994) included feedback parameters modelled directly on hardware nomenclature. By the 2000s, developers including Valhalla DSP, Soundtoys, and Sugar Bytes had built entire plug-in architectures around sophisticated multi-tap feedback networks, filter-per-tap feedback systems, and self-oscillation modes that could be played as instruments in their own right.
On guitars and electric instruments, feedback is used both as a recording technique and as a performance tool. For studio recording, an electric guitar signal is sent to an amplifier in an isolation booth or live room while the guitarist stands in a position that allows the pickup to couple acoustically to the speaker. The angle, distance, and pickup-selector position all determine which note sustains. Engineers typically set room microphone levels and amp volume to allow one specific frequency to sustain cleanly before blending with close-miced signals in the mix. Hendrix's engineer Eddie Kramer described regularly checking the studio's natural feedback nodes before recording sessions to map which amp positions would yield which feedback pitches.
In electronic music and sound design, feedback is deployed as a synthesis engine. Routing a delay plugin's output back into itself with the feedback at or near unity, combined with a slowly sweeping bandpass filter inside the loop, creates an evolving resonant drone that requires no sustained input after the initial trigger. This is the basis of tape loop composition pioneered by Terry Riley and Steve Reich in the 1960s and carried into modern production by artists working in ambient, noise, and experimental club music. Physically Informed Synthesis techniques such as Karplus-Strong plucked string simulation explicitly use a filtered feedback delay loop as their entire synthesis architecture — a short burst of noise fed into a feedback loop with a delay time corresponding to the desired pitch and a loop filter simulating string damping.
For vocals and melodic elements in contemporary pop and hip-hop, feedback within delay effects is typically kept conservative: 20–40% feedback with an LP filter cutting above 5–6 kHz creates a single audible repeat that enriches without cluttering. The classic 'Slapback plus one' technique used on Elvis Presley's Sun Records sessions (1954–1955, recorded and mixed by Sam Phillips) used a short feedback delay at extremely low regeneration to double the vocal in a way the ear perceives as width rather than discrete echo. Modern producers achieve similar results using 100–120 ms delays with 15–25% feedback on vocal aux sends, often automating the feedback upward during held notes and drops for dramatic effect.
On synthesisers and samplers, filter resonance is the most ubiquitous application of feedback thinking in a producer's daily workflow. Every synthesiser — hardware or software — implements resonance as a positive feedback coefficient within the filter circuit. At moderate resonance values (50–70% on most synths), the filter emphasises its cutoff frequency, adding character to sweeps. At maximum resonance, the filter self-oscillates at the cutoff frequency, becoming a pure sine-wave oscillator entirely independent of any input signal. This self-oscillation technique is central to analogue techno and acid house production: the Roland TB-303's resonance control was routinely maxed during live performances to produce screaming pitched blips that became the defining sound of acid house.
One email a week. The techniques behind the terms — curated by working producers, not algorithms.
Abstract knowledge becomes practical when you can hear it in music you know. These tracks demonstrate feedback used intentionally, at specific moments, for specific purposes.
The opening guitar figure bleeds into the verse with a sustained feedback pitch hovering around E4, achieved at Olympic Studios, London, with a Marshall Super Lead at performance volume. Listen on headphones to the way the feedback pitch shifts slightly between takes — evidence that Hendrix was managing room standing waves and his distance from the amp rather than relying on an effect unit. The feedback sustain is allowed to decay naturally before each chord hit rather than being truncated, a deliberate dynamic choice that gives the track its aggressive breathing quality.
Recorded at Black Ark Studio in Kingston, Perry rode the Roland RE-201 Space Echo's Intensity (feedback) control as a live performance instrument throughout the mixing session, allowing it to oscillate into full self-resonance before pulling it back. The result is the swelling, pitch-shifting white noise bursts that punctuate the mix every four to eight bars. This was not a mistake but a conscious compositional choice — Perry treated feedback as rhythm. Note how the feedback events are timed to the snare hits, creating a call-and-response between drums and electronic oscillation.
The melodic lead line's heavy pitch instability and progressive tonal darkening across each eight-bar phrase is characteristic of high-feedback delay with a strong LP filter in the loop — each iteration loses high-frequency content as if the signal is ageing in real time. The Autechre-school production aesthetic here embraces feedback's tendency to colour material over time rather than simply repeat it cleanly. By bar 32, the lead melody has lost most of its attack transient information; only the sustained harmonic content survives into the tail, creating a natural compression effect from the filter's frequency-selective attenuation.
Nigel Godrich and the Radiohead engineering team used acoustic feedback between room microphones and the string ensemble's monitors during the orchestral recording at Abbey Road Studio 3 to create the slowly building dissonant overtone wash that peaks at the two-minute-forty mark. This is not a plug-in effect but a physical feedback event captured to tape, distinguished from electronic feedback by its irregular, acoustic quality — the frequencies that sustain shift continuously as the room modes interact with the live performance. The technique is cited in Godrich's mixing notes as an intentional decision to blur the line between orchestration and ambience.
A master class in feedback as texture rather than effect, 'Hubble' builds its entire sonic environment from feedback loops running at near-unity gain through heavily filtered signal chains. Cunningham is known for running audio through degraded hardware chains where the feedback path includes genuine electronic noise and component nonlinearity. The result is a constantly evolving, self-modifying signal that sounds alive rather than processed. Notice how the apparent key of the piece shifts over the track's duration — a consequence of feedback loop phase drift rather than any programmed pitch change.
The physical routing of the playback head's output back to the record head via a pot labelled Intensity or Regeneration. Each pass subjects the signal to tape compression, head-gap frequency roll-off, and mechanical flutter, producing repeats that darken, saturate, and pitch-waver progressively. At high intensity settings the loop self-oscillates into a sustained wash that preserves the tonal character of the most recent input transient.
A precise numerical coefficient applied to each sample in a circular delay buffer before it is summed back to the input. Unlike tape, digital feedback is perfectly consistent per pass with no tonal colouration unless a filter or saturation stage is explicitly inserted in the loop. The TC 2290's feedback path included a controllable high-frequency damping stage that was one of the first to offer producers the ability to emulate tape-style tonal decay in a digital unit.
The physical phenomenon where sound from a speaker is captured by a microphone, amplified, and returned to the speaker, forming an acoustic loop. The frequency at which this system oscillates is determined by room resonances, microphone and speaker positioning, and system gain. Used deliberately in studio recording to create sustained guitar tones, ambient noise textures, and the drone backdrops of experimental and noise music.
The resonance control on an analogue or modelled filter is a feedback coefficient: as resonance approaches maximum, the filter's internal feedback path reaches unity gain at the cutoff frequency, producing a sustained sine-wave output independent of any input signal. The pitch of this oscillation tracks the cutoff frequency exactly, making the filter a playable oscillator. The TB-303's Cutoff and Resonance interaction when both are maximised produces the screaming, pitch-sliding acid sound at the centre of acid house and techno.
A very short feedback delay (1–25 ms) with high regeneration creates a comb filter whose notches are spaced at intervals of 1/delay_time Hz. Modulating the delay time with an LFO sweeps these notches through the spectrum, producing the characteristic jet-plane flanger sound. The feedback amount controls notch depth: low feedback creates shallow colouration; high feedback creates the aggressive metallic resonance associated with studio flanger on drum busses.
In Reaper, Max/MSP, and modular-style environments like VCV Rack or Ableton with third-party routing utilities, it is possible to create audio routing feedback loops at the DAW level — sending a track's output back to its own input or to an earlier point in the signal chain. These macro-level loops can incorporate entire mixing chains, effects sequences, and parallel processing paths within the feedback circuit, creating feedback systems of arbitrary complexity not available in any single hardware unit.
These MPW articles put feedback into practice — specific techniques, real tools, and applied workflows.