Running power and signal cable to a remote tank is not an instrumentation problem — it is a civil works problem. A buried instrument run costs $80–150 per meter once you count excavation, conduit, cable tray, junction boxes, terminations, permitting and site restoration, and road crossings push individual sections to $250–400 per meter. At 500 meters from the nearest power drop, that is $40,000–75,000 to connect a $1,200 level transmitter: 30 to 60 times the price of the instrument itself. Solar-powered wireless monitoring deletes that line item — but it does not delete cost, it moves it. What follows is a model you can defend in a capital review, and the two engineering constraints that decide whether it holds.
The Buried Run Is the Project
Instrument cable is cheap. Getting it into the ground is not. The five line items below account for almost all of a buried level installation, and only the last one is usually missing from the first estimate:
| Cost element | Typical unit cost | What drives it |
|---|---|---|
| Trench, sand bed, backfill — open soil | $30–60 / m | 0.6–0.8 m cover depth, spoil removal |
| Trench through road or hardstanding | $150–300 / m | Saw-cutting, traffic control, reinstatement |
| Conduit, tray, shielded twisted pair | $25–45 / m | Instrument cable, not power cable |
| Junction box, termination, grounding | $150–400 / point | One per field splice or branch |
| Design, permits, inspection, restoration | 15–25% of subtotal | The line that gets forgotten, then added back |
Put the numbers on one site and the scale becomes obvious. Twelve tanks on a 3 km site, average buried run 250 meters, three of them needing a 40-meter road crossing: 2,880 m of soil route at a blended $120, 120 m of road crossing at $300, twelve junction points at $250, and 20% for design and restoration — about $460,000 before a single instrument is bought. The same site on LoRaWAN: twelve transmitters at $1,300, one gateway at $1,800 plus $1,500 to mount and commission it, twelve brackets at $250 — under $22,000 installed. That is not a percentage difference. It is a factor of twenty.
Where Wireless Wins — and Where It Does Not
The crossover is not a distance, it is a topology. If ten tanks sit inside a 100-meter radius served by one power drop, you run a single trunk and branch to each tank. The marginal cost of the tenth tank is a branch, a junction box and a few meters of cable — $800–1,500 — and the wired installation has a simpler operating profile: no batteries, no gateway, no data plan. Wire them.
Reverse the geometry and the arithmetic flips. Tanks scattered over 2–5 km, each with its own 200–500 meter run, mean every additional tank carries its own five-figure cable bill. That is the point where a wireless node's fixed cost — transmitter, bracket, and a share of one gateway — stops being a premium and becomes the cheap option. On a per-tank, ten-year basis the gap is roughly an order of magnitude:
| 10-year cost, per tank | Wired, 250 m run | Battery / solar wireless |
|---|---|---|
| Instrument | $1,200 | $1,300 |
| Cable, conduit, trench, junction box, design and permits | $36,000 | — |
| Controller channel and cabinet | $400 | — |
| Gateway or concentrator, shared | — | $275 |
| Installation labour | $600 | $250 |
| Data plan | — | $0 LoRaWAN / $360 cellular |
| Battery replacement | — | $60 |
| Calibration and inspection | $800 | $800 |
| Total | ≈ $39,000 | ≈ $2,700 |
LoRaWAN or Cellular: the Choice Sets the Power Budget
LoRaWAN operates in the unlicensed ISM band — 868 MHz in Europe, 902–928 MHz in North America — and reaches 2–15 km depending on terrain and antenna height. One gateway typically serves 50–200 nodes inside a 5 km radius on flat ground, and there is no per-device subscription, so that gateway cost amortises away on any site with more than a handful of tanks.
Cellular, whether LTE-M or NB-IoT, removes the gateway entirely: every node talks to the nearest base station. The trade is a per-device data plan at roughly $1–5 per month, and a far higher transmit peak — an NB-IoT uplink can pull 200 mA for several seconds, against tens of milliamps for LoRaWAN. That single difference, not the monthly fee, is what decides the power architecture.
The Battery Arithmetic Nobody Publishes
Battery life is usually quoted as a number of years, which hides the assumption behind it. Take a transmitter with an internal replaceable 3.6 V lithium cell rated for 20,000 transmissions — the figure the Ellenex PLS2-L datasheet publishes, alongside “more than 10 years for most applications”. Both numbers are true, and both depend on one variable: how often you transmit.
The datasheet default sampling period is four hours, so the arithmetic closes neatly: six uplinks a day into 20,000 transmissions gives about nine years. Change only the reporting cadence and the same cell tells a different story — at fifteen-minute intervals the budget is roughly 208 days, and at one-minute intervals roughly 14 days. Nothing about the sensor changed. This is why “ten-year battery life” and “solar powered” are not competing claims: they describe the same device at two duty cycles.
That framing also settles the specification argument. If the tank needs a level reading four times a day for inventory, a battery alone is the correct and cheapest answer, and the cell is replaced once a decade. If operations wants quarter-hourly data for leak detection or delivery scheduling, the battery becomes a consumable and a panel becomes mandatory.
Sizing the panel is then almost trivial, because the energy is tiny: a fifteen-minute LoRaWAN node draws single-digit milliwatt-hours per day, and a 1–2 W panel delivers 0.3–0.8 Wh even on the worst overcast December day — two orders of magnitude of margin. The constraint is not generation, it is the cell. Capacity falls 20–40% at −20°C, so the battery must still carry the load through the coldest week when the panel contributes almost nothing. A 5 W panel is not needed for a LoRaWAN level node. It is needed when the node is cellular, or when the sensor is radar.
Which Sensor — and What It Costs in Power
The sensing principle moves the power budget more than the radio does. A submersible hydrostatic transmitter such as the Ellenex PLS2-L measures the pressure head at the bottom of the tank through a vented cable that compensates for barometric change, with ±0.2% of span accuracy and ±0.01% of span resolution. Its IP66 enclosure and IP68 sensor head suit water, diesel and chemical tanks, and it draws almost nothing because there is no transmitter to fire and no echo to resolve.
Non-contact ultrasonic is the other low-power option, but foam, vapour layers and condensation on the transducer face all erode the return echo. Radar, either FMCW or guided wave, handles foam and vapour far better and is the right answer for agitated or steaming tanks — at roughly twice the measurement energy and a materially higher peak current, which is usually what puts a panel on the pole. Browse the full level sensors category if the medium is the open question rather than the distance.
One caution belongs in every specification: a battery-powered level transmitter is not automatically rated for a hazardous area. Check the certification against the actual area classification instead of inferring it from the enclosure rating.
The Decision Rule
Put the civil hour and the capital hour side by side and the answer usually falls out on its own. Tanks inside 50 meters of power, or clustered where one trunk serves many, should be wired — the marginal cost is trivial and the operating model is simpler. Tanks beyond roughly 200 meters, or dispersed across kilometres, should be wireless, and the only remaining question is the reporting interval. Four-hour reporting means a battery and a nine-year replacement cycle. Fifteen-minute reporting means a 1–2 W panel and a rechargeable cell. One-minute reporting, or a radar sensor, means a 5 W panel and a hard look at the coldest week of the year.
The cable you do not bury is the only instrument cost that never needs calibrating, never corrodes at a junction box, and never appears on a shutdown schedule. Count the trench before you count the radios — everything else in this decision is downstream of it.
