Wireless Site Design and Roaming

RF propagation and attenuation

Radio signal strength doesn't fall off linearly with distance — it follows an inverse-square relationship, so doubling the distance from an AP reduces received power to roughly a quarter, not half. Real environments make this worse through attenuation: drywall barely affects a signal, but concrete, brick, and especially metal (filing cabinets, elevator shafts, HVAC ductwork) absorb or reflect it heavily, and water — including the water in a crowded room full of people — specifically absorbs 2.4 GHz energy, the same effect a microwave oven exploits deliberately. Signals also bounce off hard surfaces, creating multipath, where delayed reflected copies of the same signal interfere with the direct path and degrade the effective link. A site's actual coverage is the product of all of this, not just a tidy circle radiating from each AP.

Site surveys: predictive, active, and passive

A predictive survey uses software to model coverage from a floor plan and building-material data before any hardware is installed — fast and cheap, but only as accurate as its assumptions about walls and interference. A passive survey involves walking the physical site with a spectrum or signal analyzer before APs go live, to find existing interference sources and confirm the predictive model against reality. An active survey — sometimes called "AP-on-a-stick" — involves walking the site with a temporary AP actually transmitting and a client actively connected and measuring real throughput and signal quality, the closest a survey gets to how the finished network will actually perform. Most serious deployments use predictive design first, then validate (and adjust) with an active survey after installation.

AP placement and cell design

Adjacent APs on the same SSID need deliberate channel reuse — assigning non-overlapping channels so neighboring cells don't create co-channel interference, where two APs' transmissions compete for the same airtime even though they belong to the same network. A long-standing guideline for voice-grade networks calls for roughly 15-20% cell overlap and a minimum signal strength around -67 dBm at every point a client might use voice or video, specifically so a client always has a strong next AP available before its current one becomes unusable. Modern dense deployments increasingly favor lower transmit power with more, smaller cells rather than fewer APs blasting at maximum power — smaller cells mean less co-channel interference and more total capacity, even though it sounds counterintuitive to deliberately shrink an AP's range.

The sticky client problem

The AP placement above only creates an opportunity to roam smoothly — it doesn't guarantee the client actually takes it. A client decides entirely on its own when to roam, based on its own (often conservative) read of its current signal, and most client radios are tuned to avoid the overhead and brief interruption of scanning and roaming for as long as the current link still basically works. The result is the sticky client problem: a device can stay attached to a weakening, distant AP long after a much stronger one is available nearby, because nothing is forcing it to check. 802.11k and 802.11v, covered in the previous lesson, exist specifically to counteract this by giving the network a way to inform — or outright suggest — a better move.

Interactive the sticky client and the roaming trigger

Watch a client cling to a weakening AP well past the point a stronger one was already available, then replay the same walk with 802.11k and 802.11v enabled — same physical overlap, a dramatically earlier and smoother roam.

Capacity vs. coverage design

Older wireless design asked one question: can every point in the building reach an AP at all? Modern dense environments — open offices, stadiums, lecture halls — have to ask a second, often more important one: how many clients does each AP actually need to serve well, not just reach? A single powerful AP covering a huge area looks efficient on a coverage map, but if hundreds of clients associate to it, every one of them gets a shrinking slice of its airtime — "full bars" doesn't mean fast. Capacity-based design deliberately adds more APs at lower power specifically to divide the client load across more radios, even in an area that one AP could technically have covered.

Inspect site design evidence

A Wi-Fi scanning tool that logs RSSI, channel, and BSSID over time while walking a site is the practical equivalent of an active survey, and the same captured data can retroactively reveal exactly where a sticky client stayed attached to a distant AP for too long.

Guided site design practice

Work through why a predictive survey alone is never sufficient for a mission-critical deployment. A predictive survey modeled a warehouse's coverage using standard attenuation values for drywall and concrete, and the resulting AP placement looked solid on paper. After installation, an active survey revealed several dead zones near tall metal shelving racks that the predictive model's generic material library didn't account for — metal attenuates and reflects RF far more aggressively than the model assumed. The fix wasn't to question the predictive methodology itself, but to treat its output as a starting hypothesis that the active survey is specifically there to validate (or correct) against the physical reality.

Troubleshoot coverage and roaming

1. When a user reports "weak Wi-Fi" in a specific spot, check for metal, water, or dense building materials near that exact location before assuming an AP placement or power problem — the physical environment is very often the actual cause.
2. Distinguish a coverage problem (RSSI is genuinely low everywhere in a given area, meaning more or better-placed APs are needed) from a capacity problem (RSSI looks fine, but throughput craters when many clients are active — meaning the area needs more APs at lower power, not stronger signal).
3. For a reported "sticky client," check whether 802.11k and 802.11v are actually enabled on the network and supported by the specific client — a client without that support is relying entirely on its own, often conservative, roaming logic.
4. When two APs on the same SSID both show strong signal at one point, check their configured channels — if they're the same or adjacent, this is co-channel interference, not a redundant coverage benefit, and it actively hurts performance for every client in that overlap zone.
5. If users report problems specifically during peak occupancy (all-hands meetings, large classes) but not at other times, suspect a capacity rather than a coverage issue — this is exactly the symptom capacity-based design with more, lower-power APs is meant to prevent.

Knowledge check and summary

Pro deep dive