Feedback: Why It Happens, How to Find It, How to Kill It

The Physics of Acoustic Feedback

Feedback is an oscillation that occurs in a closed-loop system when two conditions are simultaneously met at any single frequency:

1. Loop gain ≥ 0 dB: The sound energy arriving at the microphone from the speaker, after amplification, equals or exceeds the energy that was originally captured. In Nyquist terms, |G(f) × H(f)| ≥ 1.
2. Phase shift = N × 360°: The total phase rotation around the loop (mic → preamp → EQ → amplifier → speaker → room → mic) is a multiple of 360° at that frequency. This means the returning signal adds constructively to the original.

The frequency that satisfies both conditions first — the one with the highest loop gain — is the one that feeds back. In most rooms, this is between 1 kHz and 4 kHz where room resonances, speaker sensitivity, and microphone sensitivity all peak.

Gain Before Feedback (GBF)

GBF is the maximum system gain before the first feedback frequency reaches unity loop gain. It is the single most important number for any reinforcement system where microphones and speakers share the same space.

GBF depends on:

• Distance ratio: doubling the mic-to-source distance while halving the mic-to-speaker distance costs 12 dB of GBF
• Microphone pattern: a cardioid mic gains ~6 dB GBF over omni when the speaker is behind it
• Speaker directivity: a narrow-pattern speaker aimed away from the mic gains 3-6 dB over a wide-pattern speaker
• Room absorption: more absorption = fewer reflections reaching the mic = higher GBF. Reducing RT60 from 2.0s to 1.0s gains approximately 3 dB GBF.
• Number of open mics: each doubling of open mics costs 3 dB GBF (the NOM rule)

Identifying Feedback Frequencies

The Ring-Out Method

The traditional method: slowly increase system gain until feedback begins. The RTA shows a narrow peak rising above the surrounding spectrum. Note the frequency, apply a notch filter, and increase gain until the next frequency rings. Repeat until GBF reaches the target.

Automated Detection with SonaVyx

The problem detector automates this process. The feedback detection algorithm:

1. Computes the magnitude spectrum every 50 ms
2. Identifies peaks with Q > 10 (bandwidth < 1/10 octave)
3. Tracks peak prominence (height above surrounding spectrum) across consecutive frames
4. Flags frequencies where prominence exceeds 15 dB AND is rising over 3+ consecutive frames
5. Reports severity: "rising" (detected), "ringing" (sustained), "howling" (dominant)

Detection typically completes within 400 ms of feedback onset — fast enough to catch ring modes before they become audible howl.

Applying Notch Filters

The surgical fix for feedback is a narrow parametric EQ cut (notch filter) at the exact feedback frequency:

• Q factor: 8-16 (narrow enough to affect only the feedback frequency, wide enough to catch the peak)
• Gain: -6 to -12 dB (start with -6 dB; increase if feedback persists)
• Center frequency: exactly on the detected peak (within ±2% accuracy)

Each notch filter suppresses one feedback frequency without significantly affecting the overall tonal quality. The feedback elimination workflow guides you through the ring-out process with real-time spectrum display and interactive notch filter placement.

How Many Filters?

Typically 6-10 notch filters are needed to achieve 6-9 dB of additional GBF beyond the system's natural limit. Beyond 10-12 filters, the cumulative effect on overall sound quality becomes noticeable — at that point, physical solutions (speaker/mic repositioning, acoustic treatment, pattern selection) are more effective than more filters.

Physical Prevention Strategies

Notch filters are reactive — they fix symptoms. These physical strategies address root causes:

• Microphone placement: Keep mics as close to the source as practical. A lapel mic at 15 cm has 20 dB more GBF than a podium mic at 45 cm.
• Speaker placement: Main speakers should be forward of all microphones. Every meter of separation between the nearest mic and the nearest speaker adds ~2 dB of GBF.
• Directivity: Use cardioid or hypercardioid mics with the null pointed at the nearest speaker. Use speakers with controlled directivity to minimize energy aimed at mic positions.
• Room treatment: Absorption panels at first reflection points reduce the reverberant energy reaching the mic.
• In-ear monitors: Replacing wedge monitors with IEMs removes stage speakers from the feedback loop entirely.

Feedback in Different Venue Types

Feedback characteristics vary by venue. Houses of worship with hard parallel walls and long RT60 tend to feed back at lower frequencies (500-1500 Hz). Conference rooms with glass and hard ceilings feed back at higher frequencies (2-5 kHz). The AI diagnostic identifies venue-specific feedback risk patterns from your measurement data.