Why Do Wi-Fi Clients Refuse to Roam? 802.11k, 802.11v and 802.11r Explained

Who decides when a device changes access point?

The client decides. An access point can inform, suggest and, in some firmware, refuse to hold a very weak association, but the handover is triggered by the phone, laptop or handheld itself. That one fact explains most roaming complaints: the levers you control are where the second AP sits and how strong each cell is, not a checkbox in the controller.

The link-layer handover starts when the client driver reaches its own threshold, commonly somewhere between -70dBm and -75dBm and never identical across handset vendors. At that point, if no clearly stronger AP is in view, the device stays where it is — which is the whole reason a phone can sit under a new AP while still associated to the old one.

What does each of 802.11k, 802.11v and 802.11r actually do?

They cover three different segments of one problem, not three interchangeable switches. 802.11k shortens the search for a target, 802.11v lets the network suggest a target, and 802.11r shortens the re-authentication after the move. None of them forces a client to leave, and none of them substitutes for placement and power design.

How much cell overlap does roaming actually need?

Published design guidance lands inside one narrow band: overlap neighbouring cells by roughly 10 to 15 percent of their coverage area for data, 15 to 20 percent where people talk while walking, with each cell bounded at about -67dBm. Below that, the client finds no clearly better candidate at the transition point; far above it, both cells look equally strong and the client stops choosing.

Overlap is produced by power and spacing, not by adding units. Running every AP at maximum makes coverage bigger and cells mushier, which makes the transition point worse rather than better; a fixed medium transmit power that closes each cell at -67dBm is what gives a client something to choose. At the transition point, the target typically needs to be around 10dBm stronger than the current AP before the client commits.

Why does one handset stay put when others roam fine?

Because the definition of too weak belongs to the handset vendor, not to the access point. On one floor, iOS, Android, a scanner and a conference tablet can sit 5 to 8dBm apart in their thresholds, and the conservative one holds on until -80dBm. The problem in that case is not coverage; it is the absence of a clearly better next step.

• Fix overlap at the transition point before considering eviction: kicking a client with no usable candidate turns slow into disconnected
• Drop transmit power from maximum to a fixed medium level so each cell closes at the -67dBm boundary
• Where firmware offers minimum-RSSI eviction, start between -75dBm and -78dBm and move one step at a time
• Confirm the chosen firmware actually enables 802.11k and 802.11v, and that the whole floor runs one SSID and one security profile

What should an access point datasheet say about roaming?

Ask four things per model: which of 802.11k, 802.11v and 802.11r the firmware implements; whether FT needs a separate controller or AC to work; whether minimum-RSSI eviction and band steering exist; and whether a firmware upgrade can take those features away again. Those four answers decide whether seamless handover is achievable on a given floor plan, and the rate column says nothing about it.

Our product pages publish what can be checked: radios, antennas, ports and power, such as the 2.5G uplink on CF-E391AX, the dual-band 3000Mbps of WA-690 with two 2.4G plus two 5G internal antennas, and the in-wall 1800Mbps format of WA-580. Firmware-level k/v/r support is not on the public page, so ask for the datasheet per model and we return it together with the matching firmware release.

How do you prove roaming works before handing the floor over?

Walk the route a person walks rather than a grid: desk to meeting room to stairwell to lift lobby, with a continuous voice call or a one-second ping running the whole way. Record where the association changes, how long the gap is and which AP took over. When one handset roams cleanly and another stalls, the difference is client support rather than coverage, and adding APs at that point only muddies the result.

• Test with the models actually used on site, not a single engineering laptop
• Run 2.4GHz and 5GHz separately: the two transition points rarely coincide
• Walk during a voice call; the length of the gap says more about the design than the number of signal bars
• Check the AP association logs to see whether clients roamed or were evicted and reconnected

How does this land on your floor plan?

Send the floor plan, the ceiling height, the wall construction and what the people walking the floor actually carry. We answer with a model per zone and the power setting that puts the transition point where the overlap should be, rather than a number of units per room. Corridors and plant rooms can run entry tiers while meeting rooms and desks run high-bandwidth tiers, provided the whole floor keeps one SSID and one security profile.

The place a mixed-tier floor breaks is exactly at roaming: if the entry-tier units do not implement 802.11r, a client crossing between tiers falls back to full re-authentication. Checking the firmware capability of every tier before mixing them is far cheaper than chasing dropped syllables in a meeting room afterwards.