The Dock Doors, the Conveyor, and the Rebooting AP
A 120 m by 80 m distribution warehouse with 8 m steel racking runs 40 handheld Wi-Fi scanners on 2.4 GHz and 14 access points set to channels 1, 6, and 11 at full transmit power. Between 06:00 and 08:00, when the north dock doors open and three 480 V conveyors run, scanners near the dock lose association and re-roam every 30 to 90 seconds. A wired IP camera on the same access switch also drops frames in the same window. A contractor recently ran the dock AP's Cat6 alongside the conveyor feed for the last 25 m of a 92 m horizontal run and terminated it on a punch-down block rather than a jack. The switch port reports the AP drawing 11.4 W; the AP spec sheet lists 802.3at and 12.95 W typical with both radios at full power.
1. The wired camera and the wireless scanners fail in the same window. Explain what that tells you and set out your diagnostic order.
Read the worked answer
The camera is the most valuable clue in the building. A wired device on the same switch cannot be affected by RF, roaming, or client density. If it drops frames in the same window as the scanners, the common element is the switch, its uplink, the power, or the copper - not the Wi-Fi design. Chasing channel plans first would waste the morning.
Order of work:
- Switch counters for both the AP port and the camera port: FCS/CRC/align errors, runts, drops, duplex mismatch. Errors on both ports point upstream to the uplink or the switch itself; errors on the AP port alone point at that run.
- PoE negotiation on the AP port: which class was negotiated, and what is the switch's total PoE budget across 14 APs plus cameras? See the next task.
- The cable. 92 m horizontal plus patch cords at both ends exceeds the 90 m permanent-link and 100 m channel budget. The last 25 m run parallel to a 480 V VFD conveyor is an EMI antenna, and a punch-down block instead of a jack adds an impedance discontinuity that shows up as CRC errors under load.
- Only then the RF layer. 14 APs at maximum power across three non-overlapping channels is roughly five APs per channel - real co-channel interference and sticky clients, and a genuine problem, but not one that turns a camera's frames to noise.
Correlation: the conveyors and dock doors start at 06:00. If the error counters climb only during that window, you have a physical-layer cause tied to equipment operation, and the fix is in the cable, not the controller.
2. The AP draws 11.4 W. What does that number mean, and what do you check next?
Read the worked answer
Know the ceilings. 802.3af delivers 15.4 W at the PSE and guarantees 12.95 W at the powered device. 802.3at delivers 30 W at the PSE and 25.5 W at the device. The AP's spec sheet says 12.95 W typical with both radios at full power - that is exactly the 802.3af device ceiling, meaning the AP was designed to sit at the top of the older standard's budget with zero margin.
11.4 W delivered is therefore a warning, not a comfort. It strongly suggests the port negotiated 802.3af. Add the I-squared-R loss of a 92 m run and a questionable punch-down termination, and the AP may have under a watt of headroom. When both radios peak - cold morning, doors open, scanners reconnecting in bulk - the AP browns out and reboots. To a user, a reboot is indistinguishable from "loses association and re-roams every 30 to 90 seconds."
Check next:
- The negotiated PoE class on the port, not just the wattage reading.
- Total PoE budget consumed by all 14 APs plus cameras against the switch's rated budget - you may be near the supply limit, which would affect every AP, not just the dock.
- Whether a midspan injector is in the path and what class it provides.
- The AP's event log for unexpected reboots and boot timestamps correlated with the 06:00 window. A reboot record converts this from a hypothesis to a fact.
Fix: force 802.3at, or power the AP locally, and re-test before touching the radio configuration.
3. Write the remediation plan and specify how you prove it worked.
Read the worked answer
Physical layer (do this once, do it properly): replace the 92 m copper run with fibre. One change removes the length violation, the parallel-to-VFD EMI exposure, and the PoE voltage-drop problem simultaneously, and fibre is immune to the noise source entirely. If fibre is impossible, use shielded Cat6a routed away from the conveyor, terminate both ends on proper jacks, and certify the link - never accept a punch-down block in a production run.
Power: move the AP to an 802.3at port or a local power supply so both radios have real margin.
RF layer: enable 5 GHz for the handhelds if the model supports it, drop 2.4 GHz radios to 20 MHz channels and reduce transmit power so APs stop shouting over one another, and tune roaming thresholds. Steel racking at 8 m creates deep nulls, so survey at the dock and in the aisles - not at a desk in the office. Add APs to fill genuine coverage holes; do not add power.
Verification, all three of these:
- 72 hours of switch counter monitoring showing zero new CRC/align errors on the AP and camera ports.
- Continuous association logs from three scanners through the 06:00-08:00 window showing a stable association and no reboots.
- A post-change walk test through the racking with a survey tool, comparing signal and roaming behaviour against the pre-change baseline.
"It seems better" is not a closure criterion, and this fault returns every time the weather or the shift pattern changes if you accept it.