One-Way Audio and a Ping-Pong Handset
An office runs VoWiFi on 200 handsets. Users report one-way audio and calls dropping when walking between the east and west wings. AP-A is on channel 36 at 40 MHz and AP-B on channel 40 at 40 MHz, both at 20 dBm. At the boundary, AP-A reads -64 dBm and AP-B reads -66 dBm. Handsets roam every 20 to 40 seconds while walking the corridor; 802.11k and 802.11v are enabled and 802.11r is disabled. A capture at one handset shows 120 ms jitter spikes during roams. The upstream switch is remarking the handsets' DSCP 46 to 0. WMM is enabled on both APs.
1. Separate the symptoms: which mechanism explains the one-way audio and which explains the dropped calls?
Read the worked answer
One-way audio points at QoS or path asymmetry, not RF. The upstream switch remarking DSCP 46 to 0 puts voice frames into best effort. Under contention, one direction gets queued and dropped while the other still flows, which is exactly the signature of one-way audio. RSSI and retries do not produce one-way audio - they produce clipping, robotic audio, or a clean drop in both directions.
Drops while walking point at the roam. A 20 to 40 second roam interval means the handset is ping-ponging, and with 802.11r disabled each transition is a full re-authentication and key exchange, typically a few hundred milliseconds - long enough to tear down the call. The 120 ms jitter spikes at the roam confirm the disruption is at the transition.
Order of work. Fix the DSCP remarking and the WMM user-priority mapping first, because it is a configuration change with no RF risk. Then fix the channel plan, then tune power so the roam decision is unambiguous, and only then consider 802.11r after testing the handset fleet.
2. Evaluate the channel plan on the two APs and correct it.
Read the worked answer
The plan is broken. Channels 36 and 40 are the two 20 MHz halves of the same 40 MHz channel. With both APs at 40 MHz, AP-A occupies 36-40 and AP-B occupies 36-40 as well. They are co-channel, fully overlapped, with a 2 dB difference at the boundary.
Consequences. Co-channel interference drives retries up and eats airtime right where voice needs it; and because the two BSSIDs appear nearly equal in signal, the handset has no clear roaming preference, which is why it oscillates. This single misconfiguration explains both the roof and the ping-pong.
Correction. Put AP-A on 36-40 and AP-B on 44-48, or move both to 20 MHz in this area given the density. Where 40 MHz is retained, confirm the secondary channel placement so the two wide channels do not merely touch but overlap. Re-check the neighbors on both sides - a corrected pair can still be wrong if the third AP in the row was planned around the old layout.
3. What is your first change, and how do you prove the fixes worked?
Read the worked answer
Sequence. DSCP trust and re-marking on the upstream switch first - cheap, reversible, addresses the symptom users complain about most. Then the channel plan. Then coverage and power tuning at the wing boundary to create a deliberate primary/secondary relationship instead of a 2 dB coin flip. 802.11r last, on the voice SSID only, after verifying the handset fleet supports it and does not break under protected management frames.
Proof. End-to-end DSCP capture at the handset and at the wired egress showing 46 preserved through the switch and mapped to the voice user priority in WMM. A walk test across the boundary recording roam times and packet loss - target transition under roughly 50 ms with fast transition, and loss low enough not to be audible. Jitter and retry statistics before and after, plus a call-quality score over several passes. Repeat the walk test at peak hours, because a channel plan that passes at 08:00 can fail at 14:00.