Far-field voice: how many microphones does a desk robot need?

The most common voice complaint on desk robots is not sensitivity — it is directionality and echo. Understanding what microphone count buys you helps decide where the money should go.

What two microphones give you

A two-mic array with beamforming handles a quiet to normal room at two to three metres, and provides enough spatial cues to suppress a ceiling fan or keyboard clatter. For a desk companion that mostly answers from arm’s length, this is usually sufficient, and it is what mainstream voice platforms design their reference around.

What the third and fourth microphones buy

Extra mics widen the aperture: better suppression of diffuse noise, more robust performance when the talker is off-axis, and stronger far-field pickup in larger rooms. But on many platforms going past the native array means an external audio codec on a different bus — new components, new clocking, new firmware paths, and a structural change to fit the openings.

Echo is the real villain

If the robot plays its own prompts through speakers, its mics hear them. Acoustic echo cancellation is mandatory, and its quality depends on the platform’s DSP resources as much as on mic count. A clean AEC on two mics beats a muddy AEC on four.

Mechanical placement outranks electronics

Mic openings need air paths, isolation from vibration sources (motors, speakers), and protection from wind and dust without foam that muffles highs. A well-placed two-mic array on a stable mount outperforms a poorly-placed four-mic array every time.

Our practical default

Start with the platform’s native array and invest in placement and AEC tuning. Move to a larger array only for products whose use case is genuinely room-scale — and budget the structural and firmware changes that choice implies.