Distribution Center Cell-Edge Collapse After Racking Change
A parcel company runs a distribution center 200 m by 110 m with 11 m ceilings and double-stacked steel racking. Fifty-two Wi-Fi 6 APs were installed five years ago against a 5 GHz design target of -67 dBm, validated when the racking was half its current height. RRM/DCA runs with 20/40/80 MHz enabled. Pickers with 1x1 handheld scanners now report 10-20 second stalls at aisle ends; the controller shows those clients at -74 to -76 dBm, 40-50% retries, and pinned at 24 Mbps. Six APs near a newly installed conveyor variable-frequency drive report noise floors of -82 to -78 dBm on UNII-3 channels. Each scanner needs under 200 kbps, but a missed scan costs a shift. Management wants to know whether to add APs, widen channels, or raise power.
1. Quantify the 5 GHz channel budget available to 52 APs in a US regulatory domain at 20, 40, and 80 MHz, and state the channel width and reuse plan you would run.
Read the worked answer
US 5 GHz yields roughly 25 non-overlapping 20 MHz channels: UNII-1 (36-48) = 4, UNII-2A (52-64) = 4, UNII-2C (100-144) = 12, UNII-3/4 (149-165) = 5. Sixteen of those are DFS.
- 40 MHz bonds adjacent pairs: about 12 channels, of which only 8 are non-DFS.
- 80 MHz bonds four: about 5-6 channels, only 2 non-DFS, and each bond consumes DFS spectrum you may have to vacate on radar detection.
With 52 APs the reuse factor is ~2 at 20 MHz, ~4.3 at 40 MHz, ~10 at 80 MHz. In an open warehouse the racking does not contain RF, so co-channel reuse is already the limiting factor; widening channels multiplies it. Run 20 MHz only, enable DFS (this alone roughly doubles the pool), pin RRM to 5 GHz-only DCA with a stable channel list, and cap power rather than raising it. Widening channels or raising power both increase the interference radius and make the -74 dBm stalls worse, not better.
2. A scanner associates at -67 dBm on an AP near the conveyor drive. What SNR does it have, what modulation can it sustain, and in what order do you remediate?
Read the worked answer
SNR = -67 - (-78) = 11 dB. That is marginal: roughly MCS0-MCS2 for a single stream at 20 MHz, and in a real channel with fading and retries the effective rate collapses well below that. Compare against a quiet channel at -95 dBm noise, where the same -67 dBm would give 28 dB and support high MCS. The lesson: always reason on SNR, never RSSI alone.
Remediation order, cheapest and most causal first:
- Fix the emitter: VFD shielding, line filters, motor cable routing away from AP cabling, grounding, and physical separation between the AP/antenna and the drive.
- Re-measure the noise floor with a spectrum sweep to confirm the improvement, then check whether the affected APs can be moved off UNII-3 into DFS spectrum that is clean.
- Keep channels at 20 MHz; wider channels integrate more noise and lower SNR further.
- Only then add APs or directional antennas to restore SNR at the aisle ends. Adding APs before fixing a -78 dBm noise floor just buys more expensive interference.
3. Why do the stalls happen specifically at aisle ends, and which configuration changes address the root cause rather than the symptom?
Read the worked answer
Aisle ends are the cell boundary. At -74 to -76 dBm the client is running low MCS with heavy retries, and because its own roam threshold is often around -75 to -80 dBm it holds the far AP instead of moving to a closer one. Raising power makes this worse by extending the far AP's reach.
Changes that attack the root cause:
- Re-survey: double-stacked steel racking changed attenuation and multipath, so the original -67 dBm validation is void. Re-run a predictive model against the current racking and validate with an on-site survey at scanner height.
- Rise the cell edge: disable 6/9/12 Mbps on 5 GHz so the AP's usable boundary tightens and clients are forced to roam earlier. Verify scanners do not need those rates.
- Set a per-WLAN minimum RSSI around -70 dBm so the infrastructure politely evicts the client, and enable 11k neighbor reports plus 11v BSS transition so the scanner is told where to go instead of scanning blindly.
- Add APs or re-aim antennas along the aisles so coverage follows the aisle geometry rather than forming lobes across racking.