Capacity-based design and where to actually put the access points

Coverage design asks "can the client hear the AP?". Capacity design asks "will the client get its work done?". Modern networks fail on the second question long before the first.

Sizing from the workload

Work through this in order:

1. Count concurrent devices, not people. A modern knowledge worker brings 2.5 to 3 associated devices. A school pupil brings 1 to 2. A conference attendee brings 2 to 3.
2. Assign an application bandwidth budget. Web and email 1–2 Mbps; standard video conferencing 2–4 Mbps; HD video 5–8 Mbps; a 4K stream 25 Mbps; a large software deployment as much as you give it.
3. Sum the aggregate demand for the busiest area at the busiest time. Peak matters, average does not.
4. Divide by realistic per-AP throughput. Not the data sheet. A well-designed 80 MHz 802.11ax cell in a real office delivers 300 to 500 Mbps aggregate, not the 4.8 Gbps on the box. Airtime overhead, management traffic, retries and mixed client capability eat the rest.
5. Cap clients per radio. 25 to 30 active clients per radio in an office; 15 to 20 where video conferencing is constant; 50-plus only in lecture-theatre designs with careful RF isolation.

The number you get from this is usually higher than a coverage design and lower than a vendor's enthusiastic estimate.

Placement rules that matter

Mount in the open. An AP above a suspended ceiling tile loses 3 to 6 dB immediately, more through mineral fibre with a foil backing. Above a plenum with ductwork, cable trays and pipework it can be far worse. Mount below the ceiling wherever aesthetics allow.

Never mount an AP on an outside wall pointing in. Half your radiated energy goes into the car park and the other half illuminates one end of the floor plate. Interior placement gives you a usable cell shape.

Understand the antenna pattern. A standard omnidirectional indoor AP radiates like a flattened doughnut below the ceiling. Directly beneath the AP is a relative null. In a high-ceiling space — warehouses, atria, sports halls — use directional or patch antennas aimed down, not omnis mounted at 12 metres.

Corridors are waveguides. Signal propagates far along a corridor and poorly through the rooms either side. Placing APs in corridors gives you excellent corridor coverage and disappointing office coverage. Place them in the rooms.

Density beats power. More APs at lower power always beats fewer APs at high power for capacity, roaming and battery life. The constraint is cabling cost, not RF.

Special environments

• Open-plan offices: uniform grid, roughly 12 to 18 m spacing indoors, all radios low power.
• Lecture theatres and auditoria: front-wall directional antennas or under-seat APs. An omni in the ceiling of a raked auditorium serves the back rows badly.
• Warehouses: racking is the network. Survey with the racking full, because a full rack of metal stock and a full rack of paper behave very differently. Aisle-aligned directional antennas beat a ceiling grid.
• Healthcare: coverage continuity for roaming telemetry matters far more than throughput. Design for -65 dBm continuous with generous overlap, and be conservative with DFS.
• Hotels and student accommodation: in-room or corridor-with-wall-plate designs. Density is driven by wall attenuation and per-room isolation, not by floor area.

Validate, always

A design is not finished when the drawings are approved. It is finished when someone has walked the floor with a survey tool after installation, on the real channel plan, with the furniture in place and — where possible — with people in the room. Bodies absorb 2.4 and 5 GHz energy noticeably; an empty-room survey of a full conference space is optimistic by several dB.