Municipal Water Metering: Choosing the Access Technology
Metro Water operates 12,000 residential meters across a dense city. About 60% sit below grade in concrete basements, where the legacy 2G meters measure -120 to -130 dBm and miss 8% of monthly readings. No meter has mains power; the utility wants 8-10 years from a 3.6 V lithium pack. Each meter must send three readings a day plus a daily tamper and valve status, and a burst alarm must reach the platform within 5 minutes. The utility owns no licensed spectrum and has four field engineers. A vendor proposes a private LoRaWAN network with 40 rooftop gateways; the national NB-IoT network covers the city at -95 dBm outdoors and the operator offers a managed SIM tariff. A third option is a Zigbee mesh that would route meter traffic through residents' always-powered smart plugs.
1. Choose the access technology for the 12,000 meters and justify it on coverage, battery and operations.
Read the worked answer
NB-IoT is the defensible choice.
- Coverage: NB-IoT has a maximum coupling loss of about 164 dB versus 144 dB for GSM, a 20 dB improvement. Meters currently at -120 to -130 dBm are exactly the case coverage enhancement was designed for, using repetitions and higher CE levels.
- Battery: PSM lets the modem sleep with microamp-level current between the three daily reports. At roughly 100 bytes per report, the radio is awake for seconds per day, which supports the 8-10 year pack.
- Operations: licensed spectrum means no interference liability, and the operator owns cell planning, backhaul and power. Four engineers cannot realistically maintain 40 rooftop LoRaWAN gateways, their backhaul and their power.
- Data volume: about 3 x 100 bytes per day per meter is trivial; NB-IoT's low throughput is not a constraint here.
LoRaWAN is technically close on link budget (about 157 dB) and would work in a private deployment, but duty-cycle limits, gateway capex and the maintenance burden decide against it at this scale. Zigbee is unsuitable: 2.4 GHz penetrates basements poorly and the mesh would depend on third-party plugs the utility does not control.
2. In one district 40% of meters fail to attach, and measured RSRP at the meter is -128 dBm. What do you do, in order?
Read the worked answer
Separate RF failure from provisioning failure before buying hardware.
- Classify: pull attach and PDN failure codes from a sample. SIM not activated, wrong APN, or a dead battery looks like no coverage but is not.
- RF fixes at the meter: confirm the antenna is not inside a metal pit lid or against a concrete wall; fit an external antenna routed to the basement riser or window; re-seat the meter if the installation blocks radiation.
- Radio parameters: ask the operator to raise coverage enhancement, that is, higher repetition counts and CE level 2 or 3, and to check whether the serving cell is oversubscribed on PRACH. Stagger retries across the fleet so 12,000 devices do not collide at the same minute.
- If the district is a genuine RF hole, a single LoRaWAN gateway or a small NB-IoT repeater site for that area is cheaper than truck rolls for 4,000 meters.
Set an acceptance threshold, for example -125 dBm with attach success above 98%, and verify with a 50-meter pilot before rolling out.
3. Which radio settings would you request from the operator to protect battery life while still meeting the 5-minute burst alarm?
Read the worked answer
First, note that the burst alarm is uplink-initiated and does not need paging, so the meter can stay deeply asleep. The 5-minute requirement is a data-path and processing requirement, not a reachability requirement.
Ask for:
- PSM with a long periodic TAU and a short active timer, so the modem returns to sleep seconds after each report.
- If the platform must be able to query a meter on demand, eDRX with a cycle around 40.96 s rather than PSM-only, because a device in PSM is unreachable until its next TAU.
- A single-tone or narrow uplink configuration if link budget is marginal; it trades throughput, which is irrelevant here, for coverage.
- No unnecessary downlink polling from the platform, since every page wakes the radio.
Then validate with a power model: average current must stay in the tens of microamps for a 3.6 V pack to last 8-10 years. A meter waking every 5 minutes to listen would miss that target by orders of magnitude.