1,200 Seats, 12 APs, 80 MHz Channels: Designing for Airtime, Not Peak Rate
An 1,200-seat auditorium has 12 ceiling APs, all 5 GHz-capable, and will host a three-day conference with roughly 3,000 client devices. The design was copied from an office template: ARM enabled, 80 MHz channel width on 5 GHz, 40 MHz on 2.4 GHz, and both a staff SSID and a guest SSID broadcast on every AP. During the last event, median downlink per client was 1.2 Mbps, 5 GHz channel utilisation sat between 80 and 90 percent on most APs, and transmit retries exceeded 35 percent. Post-event analysis showed only six usable 80 MHz channels in the band, and that 60 percent of clients were on 2.4 GHz at -72 dBm or worse from the AP they were associated with.
1. Which configuration choices are self-defeating at this density, and what exactly would you change in the channel plan? Show the arithmetic behind your choice.
Read the worked answer
At 80 MHz the 5 GHz band yields about six non-overlapping channels, fewer if you avoid DFS. Twelve APs on six channels is a 2:1 reuse factor: every AP shares its channel with a neighbour, all twelve radios contend for the same airtime, and each 80 MHz transmission occupies four times the spectrum of a 20 MHz one, so the interference footprint grows as well. This is the core error.
Rescale:
- 20 MHz everywhere. The US 5 GHz plan yields roughly 24-25 usable 20 MHz channels (36-48, 52-64, 100-144, 149-165), so 12 APs each get their own channel with no in-room reuse.
- If some 40 MHz must remain, cap it to a couple of low-density corner APs. Never 80 MHz in a single room.
- 2.4 GHz at 20 MHz only, on 1/6/11, and only for the SSID that truly needs it.
- Set these widths in the ARM profile rather than leaving them to auto-negotiation.
Caveat that must be stated: channel width alone will not save this room. 3,000 devices over 12 APs is roughly 250 clients per radio; at 20 MHz a realistic per-AP aggregate is 100-150 Mbps of airtime, which is well under 1.5 Mbps per client. This room needs 20-30 APs, not 12.
2. Sixty percent of clients sit on 2.4 GHz at -72 dBm or worse. What steering and SSID-level changes do you make, and why does each one work?
Read the worked answer
This is a cell-size plus steering problem, not a band-capability problem.
- Lower 2.4 GHz transmit power (cap ARM max TX power, for example 8-11 dBm). High 2.4 power creates enormous cells that reach clients the AP cannot hear well, which is exactly how you get -72 dBm associations.
- Disable the 1, 2, 5.5 and 11 Mbps rates on 2.4 GHz. Removing the slowest rates shrinks the practical coverage edge, because a -72 dBm client no longer has a 1 Mbps option that looks usable to it.
- Enable band steering and ClientMatch so dual-band clients are pushed to 5 GHz and sticky clients are moved off overloaded radios.
- Cut back to two SSIDs. Every SSID costs beacon airtime, and management frames are a fixed tax on 2.4 GHz.
- Enable airtime fairness so a single legacy client cannot monopolise a cell.
- If no staff device needs 2.4 GHz, consider making it guest-only.
Why it works: the client's choice of band and AP is driven by what it can hear and by the rates the AP advertises. You are changing both, so the client's own decision-making produces the outcome you want without any client-side configuration.
3. Which metrics would you capture before and after, and what acceptance thresholds would you set given the physical constraints of the venue?
Read the worked answer
Capture per AP and per band, at peak, before and after:
- Channel utilisation.
- Transmit retries as a percentage of frames.
- Client count and airtime consumed per client.
- Percentage of clients on 5 GHz, plus their RSSI distribution.
- DHCP, authentication and roam failure counts.
Acceptance thresholds I would set: channel utilisation below 60 percent at peak, retries below 15 percent, at least 80 percent of associated clients on 5 GHz, and 90 percent of clients better than -67 dBm to their serving AP.
On throughput, be honest about physics: with 250 clients per radio, a fair target after the channel plan is a median of 3-5 Mbps, not 20. Promising more sets up a failed event. If those numbers do not move after the width and steering changes, the binding constraint is AP count, and the conversation shifts to budget, cabling and mounting points rather than to more radio tuning.