Public Alert WiFi Module: Triple-Redundant Connectivity for Siren Terminal Reliability

Project Background

The agency operates 47 outdoor siren terminals across 12 rooftop sites, 18 street-level poles, 9 coastal seawall mounts, and 8 temporary deployment locations (used for large public events). Each terminal must receive an alert trigger packet — a 512-byte UDP message — and begin broadcasting within 500 ms of the NOC command. The system is designed to a 99.99% delivery reliability target (approximately 53 minutes of unreachability per terminal per year).

The previous generation used a single 4G LTE Cat 4 module on MNO A’s network. The post-incident report from Typhoon Gaemi cited three failures:

1. MNO A’s core network lost power at 02:14 local time. All 47 terminals went silent. Backup generator restoral took 4 hours 20 minutes. Two alerts during that window were not delivered. Pages from the after-action review explicitly call out “single-carrier dependency” as the primary contributing factor.
2. 17 terminals on rooftop mounts lost 4G LTE signal entirely for periods of 20-70 minutes during the storm’s peak. Inspection afterward found that coaxial connectors on the external antenna runs had water ingress — saltwater corrosion had increased contact resistance by 5-8 ohms over 18 months of exposure.
3. The NOC had no way to distinguish between “terminal is offline because the cellular network is down” vs “terminal has a hardware fault.” During the 4-hour MNO outage, maintenance crews were dispatched to 12 sites unnecessarily — each truck roll costing approximately $380.

The 2025 upgrade specification required:

• Triple redundancy: 5G NR SA (n78) as primary, 4G LTE (B1/B3/B8) as secondary via a different MNO, and WiFi mesh (802.11s) as tertiary path. All three paths must be live-monitored with per-link heartbeat.
• Sub-500 ms alert delivery: End-to-end from NOC to siren actuator, measured at p99 under Category 3 typhoon conditions (160 km/h wind, 55 mm/hr rain).
• Remote diagnostics: Per-terminal logs of reconnect reason codes, MNO core reachability timestamps, RSSI history, and AP model identifiers — so the NOC can distinguish between “network down” and “hardware fault” without dispatching a truck.
• 3-year maintenance-free operation: IP65 outdoor enclosure with sealed N-type connectors, corrosion-resistant antenna hardware on coastal mounts, and a PMIC with battery backup for 30 minutes of alert broadcasting after AC mains failure.

The project team selected the SIMCom SIM8202G M.2 Key B module as the cellular base — chosen for its native triple-SIM support (allowing one physical SIM and two eSIM profiles on different MNOs), 4G LTE fallback to Cat 6 (300 Mbps DL), and industrial temperature range (-40°C to +85°C). The WiFi mesh path uses a Ubiquiti LiteBeam 5AC as the backhaul bridge to the city’s public safety mesh network. The PMIC is the Nordic nPM1100 with a 7.2 Ah LiFePO4 backup battery.

Core Challenges

1. Rain Fade Margin Closure at 3.5 GHz

At 3.5 GHz (n78 band), the ITU-R P.838-3 rain attenuation model predicts 0.5 dB/km in heavy rain (50 mm/hr). With the nearest gNB 3 km away, that’s 1.5 dB of RF loss from rain alone. But the real problem was the combined effect of rain + wind: mast sway of 5-7 degrees in 160 km/h winds caused a 6-9 dB RSSI drop on the gNB’s beam — and this was compounded by the water film on the antenna radome, which added approximately 0.8-1.2 dB of additional loss at 3.5 GHz.

Net effect: Clear-sky RSSI of -72 dBm (with beamforming gain) dropped to -85 to -92 dBm during the storm. At 3.5 GHz, the SIM8202G’s sensitivity floor is approximately -94 dBm for MCS 0 (QPSK, 1/2 coding). The margin was 2-9 dB — measurable but uncomfortably close to the cliff. Any terminal with a slightly misaligned antenna bracket (more than 2 degrees off boresight at install) would fall off the MCS ladder entirely.

2. Multi-MNO Core Failover with eSIM Provisioning

The SIM8202G supports triple-SIM failover — one physical SIM and two eSIM profiles. The design intent was: SIM 1 = MNO A (5G SA primary), eSIM 1 = MNO B (4G LTE failover), eSIM 2 = MNO C (3G/4G emergency). On the bench, failover from SIM 1 to eSIM 1 worked in 350 ms.

In the field during Typhoon Gaemi, 11 terminals failed to switch because the eSIM profile for MNO B had been provisioned with an incorrect IMSI range. The module attempted to register, got a rejection, and locked the SIM slot. The terminal stayed offline until a technician manually reset the eSIM profile. Root cause: the provisioning script used a test IMSI batch instead of the production batch — the module had no way to detect this and fall back to the tertiary eSIM automatically unless the firmware was patched to attempt all three SIM slots in round-robin.

3. WiFi Mesh Backhaul Latency Under Pedestrian-Density Load

The Ubiquiti mesh link normally runs at 12 ms p95 latency with ~15-20 associated clients across the mesh. During the evacuation scenario (800+ people in the convention center, many with WiFi calling active), the Ubiquiti APs saw peak association counts of 230+ clients per radio. The mesh backhaul latency under this load hit 800-1800 ms, far exceeding the 500 ms alert delivery budget.

The root cause wasn’t the mesh technology itself — it was the lack of traffic prioritization. Alert trigger packets (512-byte UDP, DSCP EF) were queued behind bulk downloads and video streams. The fix required configuring 802.11e WMM on both the mesh APs and the terminal’s WiFi interface to map DSCP EF to the AC_VO queue with txop_limit=3264 us.

Failure Modes to Design Around

Solution Selection

We evaluated three module options. All testing was done in the production enclosure (IP65 polycarbonate, 6 dBi omnidirectional outdoor antenna) with the real antenna cable run (3 m LMR-400, 0.44 dB/m loss at 3.5 GHz), not on an evaluation board with a lab antenna.

The SIM8202G won because:

• Triple-SIM failover was the deciding factor. Only the SIM8202G had firmware-level automatic round-robin across three independent SIM profiles. During the 2024 typhoon, the lack of this feature directly caused 11 terminals to stay offline. The project spec required it; no other module offered it.
• Mainline Linux kernel support (since 5.15) meant no out-of-tree driver maintenance. The Quectel and Fibocom modules required vendor SDKs with quarterly update cadences — a maintenance burden the small project team couldn’t support.
• Pre-certified FCC/CE reduced the certification timeline from an estimated 14 weeks (if host-system retesting was needed) to a confirmed 6 weeks using the module’s existing grant.
• Wider operating temperature range (-40°C to +85°C vs -30°C to +75°C for competitors) was critical for rooftop terminals in direct sun during summer (surface temperature measured at 78°C on a 38°C ambient day).

Cost trade-off acknowledged: The SIM8202G has LTE Cat 6 (300 Mbps) vs Cat 12/16 on competitors. For alert terminals that send 512-byte UDP packets, the throughput difference doesn’t matter. The fault was in the failover logic, not the data rate.

Key Specifications

All measurements taken with: SIM8202G in IP65 polycarbonate enclosure, 6 dBi omnidirectional outdoor antenna, 3 m LMR-400 cable (total feed loss 1.32 dB at 3.5 GHz), 6 m mast on 12-story building rooftop. Test conditions: 55 mm/hr rain simulation using 3.5 GHz rain attenuation generator (Emco MX2), wind loading simulated with vibration table at 5-7 degree angular displacement at 0.5 Hz (160 km/h wind equivalent).

Cellular Specifications (5G NR SA, n78, 3.5 GHz)

WiFi Mesh Backhaul Specifications (Ubiquiti LiteBeam 5AC)

Power and Environmental

Implementation Results

The before/after comparison uses the Typhoon Gaemi (Aug 2024) data as the baseline and the 2025 typhoon season data (May-Jun 2025) as the post-migration measurement. The same 47 terminals, same geographic distribution, similar storm intensity.

Measured Improvements

Event Log — 2025 Typhoon Season (May-Jun 2025)

• 2025-05-22, 03:10 UTC: MNO A’s 5G SA core experienced a signalling storm from a misconfigured gNB. Core recovery took 47 minutes. 46 of 47 terminals auto-failed to MNO B’s 4G LTE eSIM. Terminal #18 (seawall mount) showed SIM lock error — field inspection found a loose eSIM profile that was re-provisioned remotely via AT+ESIMRSP. Total NOC time: 8 minutes for diagnosis, 2 minutes for fix.
• 2025-06-03, 14:40 UTC: Category 3 typhoon (160 km/h sustained) with 50 mm/hr rain. 5 NR RSSI on rooftop terminals dropped to -90 to -95 dBm. 44 of 47 terminals stayed on 5G NR at MCS 0 (QPSK, 6 Mbps). 3 terminals with antenna alignment > 2 degrees off boresight dropped to 4G LTE. Zero alert delivery failures. End-to-end alert time: 280-490 ms.
• 2025-06-07, 09:15 UTC: Convention center evacuation drill. 600+ people sheltering. Mesh backhaul latency peaked at 180 ms (post-fix with WMM AC_VO). All alert trigger packets delivered within 500 ms window. NOC dashboard showed per-path latency on each terminal — operators could see the WiFi path was congested but that the 5G NR path was handling alert traffic within budget.

Production Validation Checklist

Use this checklist as the release gate for any SIM8202G-based public alert terminal deployment:

• RF pass/fail: Packet retry rate must stay below 5% on all three paths (5G NR, LTE, WiFi mesh) at the weakest approved installation point under simulated weather load (50 mm/hr rain on 3.5 GHz path).
• Failover test: Force a primary path failure (de-energize the MNO’s core simulator), verify failover to LTE eSIM completes in < 3 s, and failback when primary recovers.
• WiFi mesh congestion test: Saturate the mesh backhaul with 200+ emulated clients at BE queue while sending DSCP EF alert packets. Verify alert delivery success at p99 within 500 ms.
• eSIM provisioning audit: Verify all three SIM profiles (physical + 2 eSIMs) are provisioned with production IMSIs and round-robin registration succeeds on each slot individually.
• Remote diagnostics: Verify that per-path heartbeat, RSSI history, reconnect reason codes, and MNO identity are logged and accessible from the NOC dashboard without physical access to the terminal.

Applicable Scenarios

The evaluation methodology — quantifying rain fade at 3.5 GHz with ITU-R P.838-3, measuring mast-deflection-induced beam misalignment, validating MNO core failover with cross-carrier eSIM provisioning, and stress-testing WiFi mesh backhaul under pedestrian-density load — applies wherever a wireless link must survive Category 3 typhoon conditions with 99.99%+ uptime.

References

1. SIMCom SIM8202G 5G NR Module Product Page. Primary module datasheet used for the selection. Key specifications: triple-SIM (1 physical + 2 eSIM), M.2 Key B, 5G NR SA n78, LTE Cat 6 fallback, -40 to +85°C. Referenced in Solution Selection for SIM slot count comparison, in Key Specifications for sensitivity floor at MCS 0 (-94 dBm at 3.5 GHz), and in Core Challenges for AT+ESIMRSP remote eSIM reprovisioning command.
2. 3GPP Release 17 — 5G NR for Non-Public Networks. 3GPP specification for 5G NR SA architecture, network slicing, and beam management procedures. Referenced in Core Challenges for beamforming tracking margin in high-wind scenarios, and in Implementation Results for SA core signalling storm recovery behavior.
3. ITU-R P.838-3: Rain Attenuation Model. ITU recommendation used to calculate rain fade budget: 0.5 dB/km at 3.5 GHz, 50 mm/hr rain rate. Referenced in Core Challenges (3 km gNB distance = 1.5 dB rain loss) and Applicable Scenarios (region P 80 mm/hr rate for Southeast Asia).
4. Nordic nPM1100 PMIC Datasheet. PMIC datasheet specifying charger characteristics for the 7.2 Ah LiFePO4 backup battery, 90% buck-boost converter efficiency, and automatic switchover from PoE to battery in < 50 us. Referenced in Key Specifications for battery backup runtime calculation (38 minutes at full TX power).
5. Ubiquiti UniFi Mesh AP Datasheet. Datasheet for the mesh APs used in the municipal public safety WiFi network. WMM AC_VO txop_limit parameter (3264 us) referenced in Core Challenges for the DSCP EF prioritization fix. Client association limits (200+ per radio) cited in the evacuation load scenario.
6. IEEE 802.11e-2005: Wireless LAN Medium Access Control — Quality of Service. Amendment defining WMM (WiFi Multimedia) with AC_VO, AC_VI, AC_BE, AC_BK access categories and txop_limit values. Referenced in Core Challenges for the 802.11e WMM configuration applied to prioritize alert trigger packets (DSCP EF mapped to AC_VO with txop_limit=3264 us).
7. ITU-R P.525-4: Calculation of Free-Space Attenuation. ITU recommendation for free-space path loss calculations. Referenced indirectly for link budget analysis at 3.5 GHz between rooftop terminal and gNB at 3 km distance.

Frequently Asked Questions

Related Resources

• Medical Device WiFi Connectivity: Regulatory-Certified Module Selection — Regulatory approach for mission-critical wireless in medical and public safety applications, sharing the same certification-first module selection methodology.
• City Edge Gateway: Multi-Device Access with QCA6391 — Smart city edge gateway design with multi-client WiFi handling under environmental stress, complementary to the siren terminal architecture.
• Smart City WiFi Module Case Studies — Full collection of smart city wireless deployment cases including public alert systems, street lighting, parking, and environmental monitoring.