Conference Room Audio System Design Guide for System Integrators

Quick answerConference room audio is a signal chain, not a product. Design it in four decisions, in this order: how the room picks up voices, how those voices are processed, how the far end is reproduced in the room, and how everything is cabled and gain-staged. Room size and system architecture drive all four — an integrated device handles a huddle room and can serve a mid-size room, while separate microphones, an external amplifier or multiple audio zones call for a dedicated DSP. Get the pickup and the processing right and the room will outperform a far more expensive one with a better camera.

Audio is where conference rooms are judged. Participants tolerate a mediocre picture; they abandon a room where the far end can’t hear them, or where every call starts with “you’re breaking up.” It is also the part of the system most often under-specified — the display and camera get a page each in the tender, and audio gets one line.

This guide is written for the system integrators and project companies who have to size microphone coverage, decide whether a room needs a DSP, and defend both choices in a bid document — not for an office manager comparing speakerphones. It walks through how to design conference room audio that works, room size by room size, in the order the decisions actually have to be made.

Start with the room, not the product list

Before choosing any device, establish four facts about the space: the seat count and table length, the ceiling height, the acoustic condition (hard glass and plasterboard, or carpet and soft furnishings), and the background noise from HVAC. These decide how far a microphone has to reach, how much speaker coverage the room needs, and how much processing it takes to keep the far end intelligible.

A useful reality check: a room with glass on two walls, a hard ceiling and a noisy air conditioner will defeat a good microphone array, while a modest system in a treated room sounds excellent. If the room is acoustically hostile and the budget can’t change that, compensate with closer microphone placement — not with more gain.

Audio design by room size

This table is the starting point for a bill of materials, and it maps onto our video conference room deployment levels. Treat the seat counts as guidance, not law — table length, ceiling height and whether the room uses integrated or separate components move the boundaries.

Decision 1 — pickup: match the microphone to the table, not the room label

Microphone choice is a coverage calculation, not a preference. The distance from the farthest talker to the nearest microphone element is what determines whether the far end hears a person or a room. As that distance grows, the direct sound falls away faster than the reverberant sound, so the voice arrives thinner and washier — which no amount of gain fixes, because gain raises the room noise with it.

• All-in-one speakerphones and integrated video bars. A speakerphone combines the microphone array, loudspeaker and echo cancellation in one unit. An integrated video bar goes further, adding the camera, auto-framing and the room compute or platform client. Both keep the acoustic relationship between microphone and speaker fixed and pre-tuned by the manufacturer, which is why they are reliable in huddle and small rooms — and why extended-array video bars now serve many mid-size rooms that would once have required separate components.
• Boundary and gooseneck microphones remain a mainstay in boardrooms, council chambers and courtrooms. They sit close to the talker, behave predictably, cost less per position and are simple to service — which matters in rooms that must work every day for years. On long tables, several units covering zones outperform one unit straining to cover everything.
• Beamforming arrays are increasingly common in medium and large rooms. They steer pickup toward the active talker and reject off-axis noise, which keeps the table clear and adapts as people move. They are a strong option rather than an automatic answer: compare them against fixed microphones on cost, serviceability and how much the room actually changes.
• Table or ceiling placement. Table microphones sit closer to talkers, which generally means better signal-to-noise and more tolerance of a reverberant room. Ceiling arrays keep the table clear and survive furniture being moved. Which performs better depends on ceiling height, room acoustics and the array technology — modern ceiling arrays perform very well in a quiet, well-proportioned room, and poorly under a noisy diffuser in a high, hard-surfaced space.
• Wireless microphones are for a person, not a room: a presenter who walks, a trainer, a chairperson. Add them when someone leaves the table, not as a substitute for table coverage. Check the UHF allocation in the country of installation before committing a model to the bid — parts of the band are restricted in some markets.

Whatever the type, specify pickup in terms an evaluator can verify — the coverage pattern and effective range against the actual table length — rather than a model name. That is also what keeps the spec defensible; the same principle applies across the whole system, as covered in our guide to writing conference room tender specifications.

Decision 2 — processing: when you need a DSP, and what it does

The trigger for a dedicated DSP is architecture, not seat count. An integrated speakerphone or video bar contains its own echo cancellation, beamforming, gain control and noise suppression, and modern extended-array units cover rooms well beyond the huddle size they are often associated with. A dedicated DSP becomes the preferred architecture once the room needs separate microphones, external amplification, multiple audio zones, or routing between several sources — because at that point the acoustic relationship between microphones and loudspeakers is no longer fixed by one manufacturer, and someone has to model it.

When there is a DSP, it is doing several jobs at once:

• Acoustic echo cancellation (AEC). The far end’s voice comes out of the room speakers, is picked up again by the room microphones, and would be sent back — so they hear themselves, delayed. AEC models that path and subtracts it, and is what allows natural full-duplex conversation where both ends can talk at once. This is the single most important function in the chain. In large or reverberant rooms, check that the canceller’s tail length is long enough for the room’s reflection time; a tail set for a small office will leave residual echo in a hall.
• Automatic mixing / gating. With several open microphones, every unused mic adds room noise. An automatic mixer attenuates inactive channels so the far end hears the talker, not the sum of the room.
• Automatic gain control. Quiet and loud participants arrive at the far end at a usable, similar level. AGC is not a cure on its own — it works alongside automixing, a limiter and a correctly set gain structure. Used to paper over bad gain staging, it simply pumps the room noise up and down.
• Noise suppression, EQ and filtering. HVAC rumble and fan noise get filtered; EQ is used to improve intelligibility, not to make voices sound pleasant. Note the limit: EQ cannot compensate for poor microphone placement or excessive reverberation. If the room sounds wrong, move the microphones or treat the room before reaching for a filter.
• Matrix routing. Which sources go to which destinations — microphones to the far end, far-end audio and content audio to the room speakers, and program audio to a recorder if there is one.
• Feedback suppression and delay. Needed where local voice reinforcement exists (see below), and in long rooms where distributed speakers require time alignment.

One detail that decides whether AEC works at all: the canceller needs a clean reference of what is being played into the room — the far-end signal only. If the reference feed accidentally includes the local microphones, the algorithm is modelling the wrong signal and echo returns under load. Getting this routing right, often described as a mix-minus arrangement, is a commissioning step, not a box to tick on a datasheet.

Decision 3 — reproduction: and whether the room needs reinforcement at all

This is the point most often misunderstood. In a typical small or medium conference room, the loudspeakers exist to reproduce the far end and content audio — not to amplify the people sitting in the room. Participants around a table hear each other directly. Adding local voice reinforcement to a room that does not need it creates an open-microphone-plus-loudspeaker loop, which costs you gain before feedback and complicates the entire design for no benefit.

Local reinforcement becomes genuinely necessary in large rooms, tiered or training spaces, divisible halls, and any room where a presenter addresses an audience rather than a table. Those rooms need careful gain structure, feedback suppression, and speaker placement that keeps microphones out of the direct coverage — a different design exercise from a boardroom, worked through in our lecture hall sound system guide.

For the loudspeakers themselves, there are three common topologies:

• Powered (active) speakers — amplifier built in, fed a line-level signal. Simple and cost-effective for small and medium rooms. Position them near the display, commonly as a left/right pair, so the far end’s voice appears to come from the screen — which is what people instinctively expect and what keeps voice and image coherent.
• Amplifier with passive wall speakers — the standard approach once you need several speakers or higher output. One amplifier, distributed boxes, consistent level across the room, and speakers positioned near the display or distributed along the room as coverage requires.
• Amplifier with ceiling speakers — used in larger rooms and open plans for even coverage, spaced according to ceiling height and the speakers’ coverage angle so levels don’t swing as people move around the table.

Where a system uses many loudspeakers — training centres, hotel function rooms, government conference halls with twenty or more units — a constant-voltage (70 V / 100 V) distributed line is the normal choice, because it lets you parallel a large number of speakers and set each one’s share of the power with a simple tap setting, instead of calculating impedance across the whole run. Easier long cable runs come along with it, but the multi-speaker distribution is the reason to use it. Rooms with a handful of speakers on short runs are usually low-impedance.

Aim for even, intelligible coverage rather than maximum volume. If the level at the far end of the table is noticeably lower than at the front, add distribution — not power.

Decision 4 — signal levels, cabling and connection

A large share of “the audio doesn’t work” callbacks are level and cabling problems, not equipment problems.

• Know which level you are on. Microphone level, line level and speaker level are three different things. Most commercial displays and interactive panels provide a line-level output intended for powered speakers or an amplifier — check the model, since panels vary between line out, headphone-level out, S/PDIF and HDMI ARC/eARC. A line-level output will not usefully drive a passive speaker.
• Balanced connections for anything long. Balanced runs on XLR or Euroblock reject induced noise; unbalanced consumer connectors over long distances in a ceiling void are how rooms acquire hum.
• Separate audio from mains. Keep audio cable runs away from power cabling, and cross at right angles where they must meet.
• Decide analogue or networked early. Analogue is simple and predictable for a single room. Networked audio (Dante, AES67 and similar) makes sense across multiple rooms, divisible spaces or long distances — but it is a network project as much as an audio one. It needs proper clock synchronisation, managed switches configured for QoS and multicast handling, and someone who owns that network after handover. Confirm who that is before specifying it.
• Keep an eye on latency. Processing, network transport and the display’s own picture pipeline all add delay. Total end-to-end latency has to stay low enough that audio and image remain in sync — a room that is technically correct but visibly out of lip-sync will be rejected at acceptance.
• Watch USB distance if the compute is remote. Where microphones or a speakerphone connect by USB, standard passive cable lengths are short; longer runs need active USB cables, USB-over-CAT extenders or fibre.

Commissioning: the checks that decide whether the room is accepted

Design and installation get you to a room that powers on. These checks get you to a room the client signs off:

• Confirm the AEC reference is the far-end feed only, then test by talking over the far end at realistic volume — full-duplex behaviour is what you are checking.
• Walk the room and check level consistency from the nearest to the farthest seat, for both far-end voice and content audio.
• Test from the farthest chair, speaking normally — not leaning toward the microphone. That is how the room will actually be used.
• Set gain structure through the whole chain rather than compensating at one point; check nothing is clipping at the DSP input while the room volume sits low.
• Confirm how users control volume and mute day to day — the platform controller, a wall keypad or a touch panel — and that mute status is visible in the room. Rooms fail on this more often than on frequency response.
• Confirm who supplied and who owns the equipment rack, the structured cabling and the network switch. On an export project these are almost always the installing contractor’s scope, and a room stalls at commissioning when nobody has ordered them.
• Run a real call on the client’s actual platform, with the HVAC running, before handover.
• Document the settings and hand them over. A room that drifts out of tune with no documentation becomes your support burden.

Frequently asked questions

Can the display’s built-in speakers be used instead of a speaker system?

In a small huddle room with a few people close to the screen, sometimes. Beyond that, built-in speakers generally lack the output and coverage needed for intelligible speech across a room — they are designed for content playback at close range. For anything above a small room, specify a proper reproduction path: powered speakers near the display, or an amplifier with distributed speakers.

When does a room need a dedicated DSP?

When the architecture calls for it, not at a fixed seat count. An integrated speakerphone or video bar carries its own echo cancellation and beamforming, and extended-array units serve many mid-size rooms without any external processor. A dedicated DSP becomes the preferred architecture once you need separate microphones, external amplification, multiple audio zones, or routing between several sources — including any room with local voice reinforcement.

Ceiling microphones or table microphones?

It depends on ceiling height, room acoustics and the array technology, not on a general rule. Table microphones sit closer to talkers, which is more forgiving in a reverberant or noisy room and usually cheaper per position. Ceiling arrays keep the table clear, survive furniture being moved and look cleaner in executive spaces; in a quiet, well-proportioned room a good ceiling array performs excellently. Assess the actual room rather than defaulting to either.

How much amplifier power does a conference room need?

Far less than a hall, and over-powering is the more common error. Work from the speakers: total the continuous (RMS) rating of every speaker on the amplifier and size the amplifier at roughly twice that figure so it never clips on speech peaks. A typical meeting room needs speech at a comfortable conversational level, not concert output — an amplifier specified far beyond that ends up running at a fraction of its range, which makes the gain structure harder to set and buys nothing. Check the rating at the impedance actually being driven, and for a 70 V or 100 V distributed line, total the tap settings and add about 20% margin.

Does a conference room audio system need a PoE switch?

Often, yes — more often than in a classroom. Ceiling microphone arrays, networked audio endpoints and IP cameras are commonly PoE-powered, so the switch stops being a network detail and becomes part of the audio design. Confirm the PoE class and total power budget against the devices specified, not just the port count. The switch itself normally sits in the network cabinet under the installing contractor’s scope rather than shipping with the equipment, and if the design uses Dante or AES67 it must be a managed switch configured for QoS and multicast handling.

Which audio clauses keep weak bids out of a conference room tender?

Three do most of the work. Require microphone coverage to be stated against the actual table length and seating positions rather than by model name. Require acoustic echo cancellation with full-duplex operation, and state that the AEC reference must be the far-end feed only. And require the bidder to supply a signal-flow diagram and documented gain structure at handover — a supplier who cannot produce one has not designed the room, only quoted the parts.

Who supplies the rack, cabling and mains distribution?

The installing contractor, on almost every export project. We ship the equipment, the system design, the signal-flow diagram and the specification wording; the equipment rack, structured cabling, conduit, mains distribution, ceiling penetrations for microphones or speakers, and the commissioning labour are local scope. Writing that split into the bid protects your margin and prevents a scope argument at handover.

Why does the far end hear an echo of their own voice?

Almost always an echo cancellation problem: no AEC in the chain, an AEC reference feed that includes the local microphones, room volume set so high that the microphones are driven beyond what the canceller can model, or a reverberant room whose reflection time exceeds the canceller’s tail length. Check the reference routing first, then reduce room level and re-test before changing hardware.