Features · Conference AV
Every seat in the room, scored before the install.
The same building model that computes your Wi-Fi and camera coverage already holds the conference room — its walls, its table, its chairs. The A/V engine takes each seat and scores what that person will see and hear, across four dimensions, before the purchase order exists.
Scored at every seat
- Display
- AVIXA DISCAS 4H/6H/8H, off-axis, brightness
- Speech
- STI per IEC 60268-16, Sabine RT60
- Microphone
- polar-pattern pickup and SNR
- Camera
- framing and face resolution to the far end
Four dimensions, one composite per seat, an A–F grade per room.
Display · AVIXA DISCAS
Can the far seat read the content?
Display checks start where AVIXA DISCAS starts: the ratio of viewing distance to image height, classified against what the room is for. Then the engine keeps going — off-axis position, panel behavior, brightness against the room's light, and how high the display hangs.
- Every seat classified into its DISCAS zone — 4H for analytical work like spreadsheets and CAD, 6H for slides and video calls, 8H as the passive-viewing limit — from its real distance and the actual screen size.
- Off-axis degradation from panel-type viewing angles — an IPS panel and a budget LED wall do not fail the corner seat at the same angle, and the engine knows which one is on the wall.
- Brightness judged against the room's ambient light, referenced to SMPTE screen-luminance guidance — enough nits to overpower reflections, or the seat is flagged.
- Mounting height checked as a vertical viewing angle — a display hung too high fails the front row even when the size is right.
- Text readability computed as character height in arc-minutes at the seat's actual distance.
The DISCAS zones
- 4H
- analytical — spreadsheets, CAD, dense text
- 6H
- basic decision-making — slides, video calls
- 8H
- passive viewing — the readability limit
- Beyond 8H
- flagged — not readable at that distance
Distance-to-height ratio computed per seat, against the room's content tier.
Speech · IEC 60268-16 STI
Will they understand the far end?
Loud is not the same as intelligible. The engine computes the Speech Transmission Index at each seat from the actual speaker layout and the room's acoustics — the same STI measure the pro-audio world specifies, not a coverage circle.
- Speaker level at the seat from the published on-axis SPL — inverse-square falloff over distance, cone attenuation off axis, and wall transmission loss when a partition sits in the path.
- Conference speakers modeled as what they are — self-powered video bars and ceiling speakers — not as 70-volt paging arrays.
- Reverberation per octave band via the Sabine formula, from room volume and acoustic environment.
- Speech referenced to ISO 9921 — a normal voice is 65 dBA at one meter — so "can they hear the presenter" is a computed number, not an assumption.
From datasheet to STI
- 01 Published SPL the speaker's on-axis level
- 02 Distance inverse-square falloff to the seat
- 03 Off-axis attenuation from the coverage cone
- 04 Walls transmission loss through partitions
- 05 Reverberation Sabine RT60, per octave band
- 06 STI intelligibility per IEC 60268-16
Scored at every seat — not averaged across the room.
Microphone · polar patterns
Does the mic hear the far chair?
A pickup radius on a spec sheet flattens everything that matters: pattern, angle, and noise. The engine models the microphone's actual polar pattern and asks, seat by seat, whether a normal voice arrives above the room's noise.
- Standard directivity models for omnidirectional, cardioid, supercardioid, and beamforming microphones — pattern math, not a circle on the floor plan.
- Beamforming bars treated as steered beams with finite width from the datasheet — a seat outside the beam is a seat the system strains to hear.
- Talker level at the capsule from distance and angle, referenced to ISO 9921 normal speech.
- Signal-to-noise graded on thresholds anchored in ITU-T P.800 speech quality — from transparent above 25 dB to unintelligible below 6 dB.
Pattern → pickup
- Omnidirectional
- equal gain in every direction
- Cardioid
- front-favoring, rear rejected
- Supercardioid
- tighter front lobe, slight rear pickup
- Beamforming
- steered beam, width from the datasheet
SNR at every seat, graded against the ambient noise floor.
Camera · framing
Is every face resolved for the far end?
The camera check reuses the same projection math as the security-camera POV engine — different discipline, same geometry. For each seat: is the face inside the field of view, and does it land on enough pixels to matter on the remote side?
- Field of view from the datasheet, with varifocal lenses interpolated between their published wide and tele endpoints.
- Every seat tested against the actual FOV cone — in frame or out, and where in the frame it lands.
- Face resolution in pixels from distance, field of view, and sensor resolution — anchored to a six-inch anthropometric face width.
- Graded for conferencing — 120 pixels where expressions read clearly, down to the 40-pixel floor where a face stops being useful remotely.
Face resolution at the seat
- 120 px and up
- excellent — expressions visible
- 80–120 px
- good — clear identification
- 40–80 px
- acceptable — recognizable
- Under 40 px
- flagged — not useful to the far end
Tiers adapted for conferencing from the security engine's DORI approach.
Per-seat scoring
Four scores per seat. One grade per room.
Seats are not a synthetic grid — they come from the conference table in your model. The engine resolves each chair position around the table, runs all four analyses at that exact point, and adds a grid of standing positions for the rest of the room.
- Chair positions resolved from the tables in the room — move the table, and the seats and their scores move with it.
- A weighted composite per seat — display 30%, audio 30%, microphone 25%, camera 15% — combined with a harmonic mean, so one failing dimension drags the seat the way it would drag the meeting.
- Seat scores roll up to an A–F grade for the room, with the worst-performing seat identified.
- Validation gates on every device: missing datasheet data is reported explicitly — never estimated, never silently defaulted — and the composite only exists when all four dimensions have real data.
The composite
- 01 Display 30% of the seat score
- 02 Audio 30% of the seat score
- 03 Microphone 25% of the seat score
- 04 Camera 15% of the seat score
Weighted harmonic mean — a failing dimension cannot be averaged away. Rooms grade A–F.
Remediation
Not just a grade — the fix, with the physics attached
The output is not a red seat map and good luck. Failing seats generate deterministic, physics-backed recommendations — prioritized, deduplicated, and escalated to critical when half the room is affected.
- Every recommendation cites the check that triggered it — display sizing per AVIXA DISCAS, intelligibility per IEC 60268-16, pickup from the polar-pattern model, framing from the camera geometry.
- Issues are tied to the specific seats that fail, so a one-chair problem reads differently from a back-half-of-the-room problem.
- Fix suggestions are grounded in your product library — concrete equipment and placement changes, not "add more speakers."
What trips a flag
- Display
- beyond 8× image height — not readable
- Speech
- STI under 0.45 — poor intelligibility
- Level
- under 55 dBA — masked by typical ambient
- Microphone
- SNR under 6 dB — unintelligible
- Camera
- face under 40 px — unusable remotely
Deterministic thresholds — the same room produces the same result, every run.
One model
The room was already in the model
No re-drawing the space in an acoustics tool, no export to a display calculator. The walls that attenuate Wi-Fi are the walls that block speech; the table that seats people defines the seats to score. Design the room once — every discipline reads the same geometry.
Coverage physics
Wi-Fi RF, camera optics, and the wider engine family — separate engines, same building model.
Learn more
Survey & design studio
Where the room, its table, and its devices are placed — on real floor plans, in 2D and full 3D.
Learn more
The six-tool stack
One model instead of six subscriptions — where the A/V seat fits.
Learn more
Frequently asked questions
Where do the seat positions come from?
From the model itself. The engine finds the conference tables in the room, resolves the chair positions around each table, and scores those exact points — plus a grid of standing positions for the rest of the space. Move the table, and the seats and their scores follow.
What happens when a device is missing spec data?
It is reported, not padded. Every device passes a validation gate; missing datasheet fields are called out explicitly and the affected dimension is excluded rather than estimated. The composite seat score only exists when all four dimensions have real data.
Is this the same engine that runs Wi-Fi and camera coverage?
No — and that is deliberate. Each discipline runs its own deterministic engine against the same shared building model. The conference camera check shares projection math with the security-camera engine, but display, speech, and microphone pickup are their own physics.
How is the room grade computed?
Each seat gets a weighted composite — display 30%, audio 30%, microphone 25%, camera 15% — combined with a harmonic mean, so a failing dimension cannot be averaged away. Seat scores roll up to an A–F grade for the room, with the worst-performing seat called out.
Who writes these numbers — SOC?
Never. The deterministic engines own every figure; SOC researches products, suggests fixes, and explains results. That boundary is architectural, not editorial.
Grade your next conference room before the hardware order
Bring a floor plan and the room you have in mind — table, displays, bar. The seats are scored in front of you. Talk to us about what your team needs.