Boardroom Echo Cancellation: AEC Reference and Mic Gain Note

Far-end participants hear echo in a boardroom when the DSP's acoustic echo canceller (AEC) is fed the wrong reference signal, when microphone gain is set so high that the room loudspeakers overwhelm the canceller, or when non-linear processing is used to hide a poor acoustic design. Fix the reference routing first, then gain staging, then the AEC, NLP and AGC settings, in that order. Most echo complaints are solved in the first two steps.

This note is written by Singapore-based AV and IPTV integrator Prestige Solutions for AV consultants, facility managers and system integrators who specify or troubleshoot conference audio systems in hotel boardrooms, function rooms and corporate meeting spaces. Settings below are planning starting points as of 2026; each DSP platform labels and scales its controls differently, so confirm ranges against the manufacturer's documentation before commissioning.

What does the AEC actually need from the system?

The AEC needs one clean copy of exactly what the room loudspeakers play from the far end, and nothing else. It compares that reference with each microphone signal and subtracts the part that matches. If the reference contains local programme audio, misses a far-end source, or is taken after a different processing chain than the loudspeakers, the canceller models the wrong signal and echo leaks through.

The last two rows cause the most difficult faults. A single canceller after an automixer has to re-converge every time the mixer switches microphones, so echo appears in short bursts when people take turns speaking. External delay in the loudspeaker path pushes the echo outside the canceller's modelled window.

How should microphone gain be staged?

Set microphone gain from the talker, not from the loudspeakers. A normal speaking voice at the intended seating distance should reach the DSP input at a healthy level with clear headroom, and the far-end voice at the microphone should stay well below that local voice level. When the loudspeaker signal at the microphone approaches the local talker level, no AEC setting will fully remove it.

1. Preamp gain: with a talker speaking normally at the furthest seat the microphone must cover, set preamp gain so peaks sit roughly 15–20 dB below the input clip point. That leaves room for raised voices and laughter.
2. Loudspeaker level: set far-end playback for comfortable listening at the furthest seat, typically around 65–70 dBA for speech, measured with a sound level meter rather than by ear.
3. Echo return check: play far-end pink noise at that level and read the microphone input meter. The loudspeaker signal at each microphone should be clearly lower than a local talker at the same microphone. If it is not, move the loudspeaker or microphone before changing DSP settings.
4. Automixer: set the automixer after gain staging, so its gating decisions are based on balanced inputs.

Which AEC, NLP and AGC settings are sensible starting points?

Start conservative and change one control at a time while a real far-end participant listens. The table gives the starting values we use when a platform exposes these controls; names vary between DSP manufacturers.

Heavy NLP is a symptom, not a setting. If residual echo only disappears at a high NLP level, the reference routing or gain structure is still wrong. A far-end participant will then notice clipped sentence starts, which is often reported as "bad internet" instead of an audio fault.

How many microphones does a boardroom table need?

Coverage should be proven per seat, not per square metre. Ceiling array and beamforming microphones cover a zone, while table boundary microphones cover a few seats each; the deciding factor is the distance from mouth to microphone and how many microphones are open at once.

Every extra open microphone adds room noise and reverberation to the send. The automixer reduces that, but fewer microphones closer to talkers always gives the AEC an easier job. In rooms with high reverberation, acoustic treatment on parallel hard walls usually improves the far-end result more than adding microphones.

Which faults appear after handover, and what is the first check?

Echo heard by the far end is caused in the local room, even though local users hear nothing wrong. Remote participants usually report it, so log the date, call platform and room configuration with each complaint.

The USB row is common in rooms that added bring-your-own-meeting connectivity after the original design. Our reference on USB routing for BYOM meeting rooms covers how the USB audio device should be presented to the laptop. For the loudspeaker side of the gain chain, see the note on amplifier sizing, headroom and limiters.

Acceptance tests that prove the echo control works

Test with a real remote participant on headphones, on every call path the room supports. Headphones matter because the remote listener must hear only what the room sends.

1. Double talk: local and far-end talkers speak at the same time for 30 seconds. Both sides must stay intelligible, without words cut off.
2. Convergence: start a fresh call and note how long residual echo is heard. It should settle within the first few seconds of far-end speech.
3. Furthest seat: confirm a normal voice from every seat reaches the far end at a similar level.
4. Every call path: repeat on the room codec, USB/BYOM and telephone bridge, if fitted.
5. Room states: repeat in each partition or combine mode.
6. Record: save the final DSP file with the date and the call platforms tested, so later changes can be compared.

FAQ

Does the conferencing software's own echo cancellation replace the DSP AEC?

No. Laptop and cloud platforms cancel echo using the signals they can see, but in a room with a separate DSP and loudspeakers the room processor should cancel its own echo. Running both is normal; the platform's processing then has little left to remove.

Can acoustic panels reduce the need for DSP processing?

Yes. Lower reverberation shortens the echo tail the canceller must model and raises the direct voice level at each microphone. Absorptive treatment on at least one of each pair of parallel hard walls is a common starting point for glass-walled boardrooms.

Should hotel function rooms use the same settings as a corporate boardroom?

Not without re-tuning. Function rooms change layout, occupancy and partition state between events, so they need saved DSP presets per configuration and a test for each preset rather than one set of values.

Visit Prestige Solutions for the wider AV and IPTV service range.

Recommended next step

Contact Prestige Solutions to review your site drawings and current operating pattern. Call +65 8010 2337 or +65 9825 8218, message us on WhatsApp (+65 8010 2337), or email sales@prestigesolutions.com.sg. You can also browse the full product range before the site walk.