Wi-Fi Roaming Without Dropped Connections: A Configuration Field Guide

The Conference-Room Disconnection Nightmare

Last Wednesday afternoon, a partner at a law firm in Tuanjiehu walked out of a third-floor meeting room with his MacBook, joining a conference call on the move. When he reached the corner of the corridor, the Wi-Fi signal dropped for a full six seconds — the person on the other end assumed he had hung up. I have seen this scene no fewer than twenty times. The problem is usually not insufficient AP coverage, but poorly tuned roaming decision logic: the client wavers between two APs and ultimately latches onto the one with the weaker signal. At the heart of enterprise Wi-Fi roaming optimization is building a pre-agreed understanding between clients and APs, rather than scrambling for a fix once the signal collapses. The 802.11k, 802.11v, and 802.11r protocols are the rulebook for that understanding.

802.11r: Turning Roaming from Improvisation into Advance Notice

With traditional WPA2-Enterprise roaming, a client switching APs requires a full four-way handshake, typically adding 200-500ms of latency. For voice or video, that is long enough for the other party to hear a “Hello?” and then fall silent. 802.11r (Fast BSS Transition) solves this problem. At the client’s first association, it establishes an FT (Fast Transition) key hierarchy that lets the new AP derive the decryption key, so the switch can complete without a full authentication exchange. In tests I ran at a technology park in Zhongguancun, roaming latency fell to under 50ms, and WeChat voice calls and video conferences showed virtually no stutter. When deploying wireless networks for enterprise clients, Siwenlide generally recommends enabling 802.11r on APs in core areas and ensuring that FT capability negotiation between the AP controller and clients works properly. When configuring it, keep the following in mind:

1. Enable FT mode on the AP side (FT-Over-The-Air or FT-Over-DS); most vendors default to the latter;
2. On the client side, Windows 10/11, macOS, and iOS all support 802.11r, but drivers for some industry-specific devices may be incomplete;
3. The RADIUS server needs FT-R1KH and FT-R0KH policies configured, or key derivation will fail.

BSS Transition: The Art of the AP Politely Asking Clients to Leave

802.11v (BSS Transition Management) gives the AP the upper hand. It can send a “suggestion” to the client — “The AP next door has a better signal, would you consider switching?” This mechanism is especially useful in high-rise office areas, where an employee walking down a corridor clings to an old AP at -72dBm while a new AP at -60dBm sits just three meters away. BSS Transition can be implemented in two ways: a BSTM Request pushed proactively by the network, and a BSR Response requested by the client. The former requires the AP to analyze the load and signal distribution of nearby APs in advance; the latter relies on the client’s own judgment. In practice, I lean toward enabling a combined “load balancing + signal steering” strategy on the AP side: when an AP has more than 40 associated clients and neighboring APs have spare capacity, BSS Transition is triggered proactively. A project at a logistics warehouse in Shunyi stands out in my memory. Forklift drivers carrying PDAs between the racks had been plagued by frequent disconnections for half a year. The cause was that the APs were placed too densely (about 15 meters apart), so clients bounced back and forth among three APs. After we adjusted the BSS Transition sensitivity, the number of switches fell from 200+ per hour to under 30, and packet loss essentially disappeared.

RSSI Thresholds and Power Balancing: Giving Roaming a Soft Landing

RSSI thresholds are the gatekeepers of a roaming policy. A common approach is first-come-first-served: roaming is triggered only once the client’s signal falls below the threshold. But this tends to produce a cliff effect — the client begins trying to roam at -65dBm, yet association response delays mean the signal has already dropped to -70dBm by the time the switch completes, and the voice call has already dropped. A more reliable approach is to set dual thresholds: -65dBm as the roaming trigger threshold and -70dBm as the forced roaming threshold. If a client’s signal falls below -70dBm without roaming, the AP can initiate a BSS Transition to force it off and make it re-associate. Power balancing is often overlooked. Most APs ship with full transmit power (20dBm or 23dBm) as the default, so densely deployed areas easily form “signal overlap zones.” The client sees a cluster of APs at -50dBm and has no reason to switch to any of them. I usually keep the transmit power difference between neighboring APs within 3dB, so clients roam naturally through a balanced transition zone instead of being locked to one “signal king.”

Final Thoughts

Roaming optimization is not a one-and-done configuration task; it is scenario-specific tuning that requires repeated verification. Having worked on many projects in Shunyi, Yizhuang, and Wangjing, my biggest takeaway is that client diversity is the largest variable. Two identical iPhone 15 units will roam differently on different iOS versions; two identical Windows laptops will show uneven 802.11r support depending on whether the wireless card driver comes from Intel or MediaTek. During on-site acceptance testing, walking the site with several different devices is more convincing than any parameter table. If your company faces similar issues, Siwenlide offers one-stop service from wireless site surveys to device tuning — feel free to get in touch.