Steel-Racking Aisles: Good RSSI, Terrible Throughput
A regional retailer is fitting out a 14,000 m² distribution center with 12 m ceilings. APs are mounted at 10 m on top of steel racking and use 5 GHz with 40 MHz channels at 20 dBm. In the wide pick aisles throughput is acceptable, but handheld barcode scanners stall and drop sessions inside the narrow aisles between racking, even where the survey shows −67 dBm. Scanner logs show retry rates above 60% and links falling to 2 Mbps. The site RF engineer proposes adding 20 more APs at the same height. Capital budget allows either antenna and mounting changes or new APs, not both. The measured noise floor is −92 dBm at 2.4 GHz and −95 dBm at 5 GHz. Racking is 10 m high and 1.2 m wide with 2.5 m aisles. Forklifts move at 8 km/h. Dropped scans cost about 40 labor hours per week, and peak season starts in six weeks.
1. The scanner sees −67 dBm but retries above 60%. What RF causes are plausible, and what measurements at the scanner would separate them?
Read the worked answer
Start with the signal-to-noise ratio. A −67 dBm signal over a −95 dBm noise floor gives roughly 28 dB SNR, which is enough for high-order modulation, so raw signal strength is not the problem. That points to causes that corrupt or contend for the channel rather than weaken it:
- Multipath and nulling. Steel racking acts as a waveguide and reflector; the scanner can sit in a null a few centimeters from a strong signal, and delayed reflections cause inter-symbol interference at wide channel widths.
- Co-channel interference. APs at 10 m spill over racking tops into the same aisle, raising channel utilization without raising RSSI much.
- Hidden nodes. Forklifts and racking block the path between clients that can both hear the AP.
Measurements to separate them: per-channel utilization and retry counters from the infrastructure; a protocol capture at the scanner showing MCS/PHY rate, retries and frame durations; a spectrum analysis taken at scanner height inside the aisle; and a walk test holding the scanner at 1.5 m rather than at the AP's survey height. If the noise floor is clean and retries track MCS collapse with strong RSSI, multipath is dominant. If utilization is high with several APs audible, co-channel contention is dominant.
2. Recommend a fix, and explain why adding 20 APs at the same height is likely to make things worse.
Read the worked answer
The design error is mounting height, not AP count. At 10 m, the AP illuminates racking tops and the aisle becomes a shadowed canyon; adding APs at the same height adds co-channel energy and more contention inside the same metal structure, while doing nothing about nulls or the hidden-node problem.
Recommended changes, in priority order:
- Move radiators into the aisles. Use directional antennas (patch or 90° sector) at 4–6 m, aimed along the aisle, alternating aisles left and right so adjacent aisles use different channels. Alternatively drop APs to 5–6 m on aisle ends with downtilt.
- Narrow the channel. Move from 40 MHz to 20 MHz. This increases the number of non-overlapping 5 GHz channels, shortens the delay spread relative to the symbol time, and reduces ISI; the client population (scanners) does not need 40 MHz.
- Lower basic rates and disable lowest rates so clients roam before the link degrades to 2 Mbps, and enable band steering only if scanners support 5 GHz well.
- Cap 2.4 GHz power to reduce co-channel bleed from forklift-mounted and ceiling APs.
Re-survey at scanner height with the new antennas before committing to extra APs; the aisle-level survey will show whether any coverage holes remain.
3. Operations asks whether moving scanners to 2.4 GHz would be safer because it penetrates racking better. How do you answer?
Read the worked answer
Penetration is not the constraint here: the scanner already receives −67 dBm at 5 GHz. Moving to 2.4 GHz trades a non-problem for real ones. The 2.4 GHz band offers only three non-overlapping channels (1, 6, 11), and this facility already has a −92 dBm noise floor from other systems, so channel reuse across a 14,000 m² warehouse with dozens of APs would produce heavy co-channel contention and airtime waste. 2.4 GHz also has longer delay spread and greater susceptibility to interference from microwave and Bluetooth equipment, and lower peak data rates.
The defensible position: keep scanners on 5 GHz at 20 MHz, fix mounting and antenna pattern, and use 2.4 GHz only as a fallback radio on the same APs for legacy devices. If the scanners themselves handle roaming poorly, tune client roaming thresholds and validate with a pilot aisle before changing bands.