Designing a Wi-Fi network means choosing where the access points go and how many, then proving that the result does what it was meant to. This topic gives you a sizing method and a validation routine.
1. DefinitionDesign, survey and validate
- Predictive design: a software model of the building places access points and predicts coverage. It is only as good as its wall and material settings.
- Site survey: measurements in the real building. A passive survey listens (signal, noise, channels); an active survey connects and tests throughput and roaming.
- Validation: proving after installation that the network meets the agreed targets, with the evidence saved.
- Capacity planning: making sure the access points can carry the traffic, not just reach every corner.
2. PurposeWhy this is a method, not a feeling
A network that has coverage everywhere can still be useless at 9 a.m. when a hundred people join. Capacity and coverage are two different limits, and fixing one does not fix the other. Writing the targets down first, and measuring against them at the end, is what separates engineering from guessing.
3. MechanismHow many access points?
Enter the clients, their demand, the link rate and the floor area. The capacity figure follows from the airtime lesson: only about half of a link rate is usable, and keeping access points well under full load keeps delay low.
Try this. (1) Leave the area alone and raise the number of clients until capacity becomes the limit. (2) Then increase the area and watch coverage take over. (3) Lower the share of the link rate you really get to 40% (a busy, mixed network) and see the effect.
4. VisualThe validation routine
A checklist is the simplest tool in the book. Work through the five phases and tick off each step; your progress is saved in this browser only.
5. MathematicsThe sizing rules
- Capacity: access points = total demand / (link rate × usable share × target load). With 120 clients at 6 Mbit/s each (720 Mbit/s), a 600 Mbit/s link, 55% usable and a 50% load target: 720 / (600 × 0.55 × 0.5) = 4.4, so 5 access points.
- Coverage: access points = floor area / the area one access point serves well. For 3000 m2 and 400 m2 each that is 7.5, so 8. You need the larger of the two numbers, here 8.
- Channel reuse: access points that can hear each other on the same channel share its airtime, so extra access points add capacity only if they use different channels. In 2.4 GHz there are three; 5 and 6 GHz offer many more.
- Link balance: the client’s transmit power is lower than the access point’s. If an access point is much louder than clients, the downlink looks fine while the uplink is poor, so match the design to the weaker side.
- Per-client rate is roughly the usable throughput of the access point divided by the number of active clients.
6. Real-world behaviourStandard tests and RF exposure
TR-398 from the Broadband Forum defines repeatable tests of Wi-Fi equipment in a lab, so that results can be compared between products and over time. Its test cases include:
| Group | Test cases |
|---|---|
| Radio performance | Receiver sensitivity |
| Baseline performance | Maximum connection; maximum throughput; airtime fairness |
| Coverage | Range versus rate; spatial consistency |
| Multiple stations | Multiple stations performance; association and disassociation stability; downlink MU-MIMO |
| Stability and robustness | Long-term stability; coexistence with other access points |
From TR-398 Issue 1 Corrigendum 1 (October 2019); later issues add more.
- Acceptance criteria belong in the contract. Agree the minimum signal, SNR, throughput, roaming time and packet loss and where they apply (for example, at the edge of every working area), and test against them.
- Test when busy. A survey at 7 a.m. in an empty building says little about a full one.
- Document everything: final positions, channels, power levels, firmware and the survey data, so that the next person can reproduce the network.
- RF exposure. Wi-Fi power is low, and installations must comply with the exposure guidelines the regulator applies (ICNIRP guidelines are the international reference). Keep the documentation to prove compliance when asked.
Practical validationProve it in a capture and on a spectrum analyser
Reading about a band is one thing; seeing it in your own evidence is the skill. These checks work with any monitor-mode capture:
- Write down the targets before the survey and mark each as pass or fail afterwards.
- Run a passive survey on all bands, an active survey with a real device and a throughput test with
iperf3 -c server -t 30 -P 4at the worst point of coverage. - Test roaming along every walking route with a client that is capturing, and record the reassociation time.
- Repeat the throughput test at the busiest time of the week and compare.
Check your understandingQuick quiz
ReferencesSources and further reading
- Broadband Forum: TR-398 Wi-Fi In-Premises Performance Testing
- ICNIRP: Guidelines for limiting exposure to electromagnetic fields (100 kHz to 300 GHz), 2020
- IEEE Std 802.11-2020 (IEEE Xplore)
- iPerf3 documentation
The sizing method is an engineering approximation, not a standard: the usable share of the link rate and the target load are inputs you should choose from your own measurements. The TR-398 test case list was checked against the Issue 1 Corrigendum 1 document; later issues differ. RF exposure limits are set by national regulators; the ICNIRP guidelines are the international reference they commonly use.