Legacy Hospital Wireless Refresh: What to Measure Before You Touch the Design
A 420-bed regional hospital runs a Cisco 5508-250 controller pair with 240 Aironet 3502i APs deployed five years ago. Clinical voice badges and legacy medication scanners use 2.4 GHz; laptops and carts use 5 GHz. Nursing reports badge dropouts in the emergency department and eMAR cart stalls in the west wing. Radiology PACS images intermittently stall on 5 GHz. The current configuration uses 2.4 GHz channels 1, 6, and 11 at 20 MHz with all data rates enabled, 5 GHz at 40 MHz on eight channels, RRM enabled, no band steering, and local DHCP. CleanAir is not licensed. The CIO wants a redesign proposal in three weeks but will not approve a full outage or a site survey during peak hours. You have read-only access to the controllers, Prime Infrastructure, switch CLI, and DHCP logs.
1. What information would you collect first, and how would you prioritize it?
Read the worked answer
Start with data that separates coverage, capacity, and interference, because the three symptoms do not point to one cause. From the controllers, pull per-client SNR, retry and discard counters, data rates used, roam history, AP neighbor lists, channel utilization, and RRM channel/power changes for at least seven days. From Prime Infrastructure, generate reports for the ED and west wing rather than the whole campus. From the switches, check port errors, PoE draw, and duplex; from DHCP, check scope utilization. Prioritize badge dropouts first: badges are 2.4 GHz, low-power, and roam, so capture client details for a few badge MACs and compare serving AP RSSI at known drop locations. PACS stalls are different: check 5 GHz retransmits, channel width, and AP uplink utilization, because 40 MHz on eight channels can create co-channel interference. Because CleanAir is unlicensed, plan a spectrum analyzer pass in the ED after hours. Do not propose AP counts or controller replacement until you can show whether the ED problem is coverage, interference, or roaming behavior.
2. How would you validate the root cause without disrupting clinical operations?
Read the worked answer
Use passive and targeted active measurements, not a full redesign. Walk the ED with a survey laptop in monitor mode at shift change to capture RSSI, SNR, and retry behavior without associating or changing APs. If the controller supports monitor-mode APs, convert one or two non-critical APs after hours, but document the change and revert. Use a spectrum analyzer in the ED to identify non-Wi-Fi interference, microwave leakage, or rogue APs. For badges, enable controller client debugging for a small set of MAC addresses and correlate disconnect reasons with roam events and DHCP leases; do not debug all clients. For PACS, run a wired-to-wireless throughput test from a cart in radiology after hours and compare with the controller counters. Check switch port errors and PoE on the serving APs. Validate in one wing before touching the ED. The goal is to prove which symptom is caused by 2.4 GHz congestion, which by 5 GHz channel width, and which by roaming or backhaul, because each has a different fix.
3. Given the data, what design conclusions would you draw?
Read the worked answer
If the data confirms the suspected pattern, treat 2.4 GHz as a controlled legacy band: set 20 MHz, channels 1/6/11, disable 1, 2, 5.5, and 11 Mbps, make 12 and 24 Mbps mandatory, and enable band steering for laptops so they leave 2.4 GHz. Add APs or adjust power in the ED only where SNR and retry data show a hole; do not blanket-increase power. For badges, enable fast roaming where supported and verify the badge vendor's roaming requirements; check that AP coverage overlaps enough at entrances and elevator lobbies. For 5 GHz, reduce 40 MHz to 20 MHz in high-density areas, keep 40 MHz only in low-density offices, and check AP uplink utilization for PACS. License CleanAir and use it to find interference. The conclusion should be evidence-based: if retries are high with good SNR, fix interference or channel width; if SNR is low, fix placement; if roams fail, fix roaming and coverage overlap. Do not replace the 5508 pair until client count, throughput, and AP capacity data justify it.