Skip to content
Sign inGet Started

Fleet Fuel Numbers Do Not Add Up? When CAN Bus Is Enough and When You Need a Sensor

By Saeed Baatiah · Published:

Short answer

The CAN bus reports what the engine control unit computes about fuel, which is accurate enough for efficiency trends but not for guarding inventory. A dedicated tank sensor measures the tank directly, independent of the engine, which is what a fuel-theft or short-delivery question actually requires.

Key takeaways

  • CAN bus reports an engine-computed value, not a direct tank measurement, it answers efficiency questions.
  • A peer-reviewed NREL study found CAN fuel data suitable for relative trends, not accurate consumption rates.
  • A dedicated tank sensor is the only instrument that sees the tank independent of engine state.
  • CAN fuel level resolution is coarse by protocol, often a single 8-bit percentage with no litres figure.
  • Fit both where the question is both: CAN for the engine, a sensor for the inventory.

01Two different questions, two different instruments

The choice between reading fuel from the CAN bus and fitting a dedicated tank sensor is not a budget decision first, it is a question of what you are actually trying to know. "How efficiently is this engine running?" is answered well by data the vehicle already computes and transmits. "How much fuel is physically in this tank right now, and did any of it leave without an engine event to explain it?" is a different question, and the engine has no part in answering it. Fleets that treat this as one decision usually end up with a CAN-only deployment that cannot support the fuel-theft investigation they eventually need, or a sensor-only one that never gives them an efficiency baseline. Deciding what you need before deciding what you buy avoids both.

02What the CAN bus actually reports

SAE J1939 is the standard governing heavy-vehicle CAN networks, running at 250 kbit/s under J1939/11 or 500 kbit/s under the newer J1939/14, both designed for immunity to electrical interference on a truck. Fuel data travels as specific parameters inside this protocol: SPN 183 is fuel rate, SPN 96 is fuel level as a percentage, SPN 250 is total fuel used, and SPN 182 is trip fuel, the exact SPN-to-name mapping documented in commercial J1939 reference manuals used across the telematics industry. None of these figures come from a sensor in the tank. They come from the engine control unit's own model of combustion, built from injector timing, airflow and other internal parameters the ECU already tracks for engine management, not for fuel accounting.

03How a dedicated sensor measures instead

A dedicated fuel-level sensor is fitted inside or against the tank and measures the tank directly, it has no relationship to the engine at all. The Eurosens Dominator RS, for example, publishes a maximum measurement error of ±1%, operates from −40 to +85 °C, and reports over RS-232, RS-485 or K-Line, independently confirmed on both the Wialon sensor catalogue and the manufacturer's own site. The SOJI LIGO Pro publishes a measuring error of ±0.5% under IP69K protection with galvanic isolation to 2,500 V. Because these instruments read the tank rather than compute a model, they keep reporting a fuel level whether the engine is running, idling, or switched off entirely, which is precisely the case a CAN-only system cannot see.

04The resolution problem with CAN-reported fuel level

Beyond the source of the number, its precision is coarser than most buyers expect. Under the OBD-II standard, SAE J1979, fuel tank level input is PID 0x2F: a single byte scaled from 0 to 100% using the formula 100/255 × A, which means the smallest step it can represent is about 0.392 percentage points, and it never reports litres, only a percentage. J1939's own SPN 96 for fuel level uses a similar percentage encoding at 0.4% resolution per bit. The companion rate figure, OBD-II PID 0x5E for engine fuel rate, is coarser in a different way: it is a two-byte value covering a 0 to 3,212.75 L/h range, built for engine diagnostics rather than fleet-level fuel accounting, and its granularity was never designed around distinguishing a normal trip from a small, gradual loss. A percentage of a tank whose exact capacity may not be known precisely to the reporting system is not the same measurement as a litre figure from a calibrated sensor, and the gap between the two grows in absolute terms on a large tank.

05What CAN cannot see

This is the limit that decides the choice for a fuel-security question, and the owner's view puts it plainly: the CAN bus tells you what the engine consumed by its own internal model, but it has no awareness of a hose siphoning fuel from a stationary vehicle. There is no mechanism by which an engine-side computation can register fuel leaving through the filler neck or a valve while the engine is off, the ECU is not involved in that event at all, so no SPN reflects it. A tank sensor, by contrast, sees the level fall regardless of engine state, because it is watching the tank rather than inferring from combustion. Where the threat is physical loss of inventory rather than engine inefficiency, only the sensor is watching the right thing.

06What a sensor cannot see

The limit runs the other way too. A tank sensor tells you what is in the tank; it says nothing about whether the engine is burning that fuel efficiently. It cannot separate a legitimate high-consumption trip, a fully loaded truck on a steep grade, from wasteful idling or a mechanical fault raising fuel use, because it has no visibility into engine load, RPM or injector behaviour at all. Consumption-per-kilometre trends, comparisons between drivers on identical routes, and diagnosing why a specific vehicle is burning more than its sister units all need engine-side data. A sensor also cannot explain a change over time: if a truck starts consuming noticeably more litres per 100 km than last quarter, the sensor reports the higher draw faithfully but says nothing about whether the cause is a clogged air filter, a failing injector, or simply a heavier load, that diagnosis needs the engine's own reported parameters. A fleet that only fits sensors and never reads CAN has excellent inventory protection and no way to answer why its fuel bill moved.

07Consumption versus inventory: the distinction that decides it

A peer-reviewed study published in the SAE International Journal of Fuels and Lubricants, testing CAN-broadcast fuel data against precise laboratory instrumentation across eight vehicles and two engines, reached a conclusion worth quoting directly rather than paraphrasing away: discrepancies ran in both directions with no consistent vehicular trend, and the data were judged suitable for relative fuel consumption trends but not for accurate or precise consumption rates, though drive-cycle differences were captured within about 4%. A separate, earlier peer-reviewed validation of J1939 CAN fuel rate against reference laboratory instrumentation, published in Computers and Electronics in Agriculture, reached a compatible conclusion working from agricultural equipment rather than road trucks, the CAN signal tracked real consumption well enough for comparative work without matching laboratory-grade absolute accuracy. That is the academic version of the owner's distinction. CAN is a trend instrument; a sensor is a measurement instrument. Use CAN to ask whether Tuesday's route burned more than Monday's. Use a sensor when the number itself, in litres, at a point in time, is what a decision depends on.

08Vehicle coverage: why CAN reading is not universal

A contactless CAN reader like the Teltonika LV-CAN200 covers a defined parameter set, mileage, fuel level, fuel consumption, RPM and coolant temperature, for light vehicles, vans, trucks and buses, while the ALL-CAN300 adds roughly a hundred further parameters and extends to special, agricultural and construction transport. But which parameters are actually available on a given truck depends on that vehicle's own FMS gateway configuration, not on the adapter. Teltonika's own support channel states this directly: on some truck models only fuel percentage is available and litres is not listed and may be impossible to obtain. A fleet with mixed makes and model years should expect mixed CAN fuel coverage, and should verify what a specific vehicle actually exposes before assuming CAN alone will answer the fuel question fleet-wide.

09Calibration: the sensor's real cost

The comparison is not free on the sensor side either. A dedicated fuel sensor needs to be calibrated to the specific geometry of the tank it is fitted to, because a tank's cross-section is rarely uniform and a fixed height-to-volume table has to be built for it before the sensor's readings mean anything in litres. This is fitting labour and time that a CAN connection, which reads a value the vehicle already computes, does not require. A wired sensor also needs a serial interface: an FMC650 supports up to five digital tank sensors on its RS-485 line, or a single sensor on RS-232, so a multi-tank vehicle needs the higher-capacity interface planned in from the start. The trade-off is symmetric: CAN reading is fast to deploy and reads a coarse, engine-side estimate; a sensor takes longer to commission and reads the tank directly at high precision. Neither cost disappears, the question is only which one buys you the data your actual question needs.

10Temperature affects both measurements

One complication applies whichever instrument you choose: diesel changes volume with temperature, which is why the petroleum industry corrects measured volumes to a standard reference temperature under ASTM D1250, also published as API MPMS Chapter 11.1, using base temperatures of 15 °C and 20 °C. A tank sensor reading a warm tank at midday and a cool tank at night will show a real physical difference that has nothing to do with theft or engine efficiency. CAN-reported consumption figures are less directly exposed to this because they describe fuel burned rather than fuel volume in a tank, but any comparison against a delivered-litres figure on a hot Gulf afternoon should account for the same correction before treating a discrepancy as a finding.

11When to fit both, and what each one then owns

For a fleet running fuel-price-sensitive routes with genuine theft exposure, long-haul diesel trucks, tankers, or vehicles parked overnight at unattended sites, fitting both instruments and giving each one a distinct job is the position that holds up under scrutiny. CAN answers "is this vehicle's engine running efficiently compared to its own history and its sister units," using the trend-grade data the NREL study validated for that purpose. The tank sensor answers "is the fuel that should be in this tank actually there," using a direct, calibrated, litre-accurate reading independent of engine state. Trying to make one instrument answer both questions is where most fuel-monitoring programmes go wrong, because it asks a trend tool to make an inventory claim, or an inventory tool to explain an efficiency trend.

12How to specify this in a tender

Writing the requirement in terms of the two questions rather than a generic "fuel monitoring" line item protects you from receiving a CAN-only bid when you actually need inventory protection, or an expensive sensor fleet when your real problem is efficiency benchmarking. State explicitly which vehicles need litre-accurate, engine-independent tank measurement, usually because they carry theft or short-delivery risk, and which only need engine-side consumption trends for benchmarking and maintenance triage. Ask the vendor to name the exact CAN parameters (SPN or PID) they will read on your specific vehicle makes, since coverage is vehicle-dependent, and ask for the sensor's published accuracy figure with its source rather than a marketing percentage. A specification that names the instrument by the question it answers is a specification a vendor cannot quietly substitute.

13How ENTEK.AI positions the two inputs

ENTEK.AI does not present CAN-derived fuel data and tank-sensor data as interchangeable, because the evidence says they are not. Where a deployment brings in CAN fuel figures, the platform treats them as engine-side trend data, matching the boundary the NREL study drew. Where a deployment adds a tank sensor, the Eurosens Dominator RS or SOJI LIGO Pro used elsewhere in our fuel-monitoring work, that reading carries the weight for inventory and theft questions, independent of whatever the engine is doing. This is not a hedge; it is the same distinction the owner draws when specifying a system: CAN calculates operational consumption, and the sensor protects the physical stock, and conflating them is how a fleet ends up trusting a number for a question it was never designed to answer.

CAN bus vs a dedicated tank sensor, by question
QuestionCAN busDedicated tank sensor
Is the engine running efficiently vs its own history?Yes, validated for relative trendsNo visibility into engine load or injectors
Is there fuel physically missing from the tank right now?No, cannot see a stationary-vehicle lossYes, reads the tank directly
What is the fuel level while the engine is off?Not reported (engine-derived value)Reported continuously
How precise is the figure?~0.4% resolution, percentage only on many vehicles±0.5–1% published accuracy, litre-calibrated
Deployment effortFast, reads an existing vehicle valueSlower, needs tank-specific calibration
CoverageDepends on the vehicle's own FMS configurationUniversal, independent of vehicle electronics

Figures and sources

Questions

Can CAN bus fuel data replace a fuel sensor for theft detection?

No. CAN reports an engine-computed value with no visibility into a stationary vehicle losing fuel through a hose or valve. A peer-reviewed NREL study found CAN fuel data suitable for relative trends, not accurate rates, the wrong precision for a theft claim even where the engine event existed.

Why does CAN fuel level only show a percentage on some trucks?

Because the standard PIDs and SPNs for fuel level are defined as a percentage, not litres, and which parameters a truck actually exposes depends on that vehicle's own FMS gateway configuration. Teltonika's own support documentation confirms some models never expose a litres figure at all.

Do we need both a CAN connection and a tank sensor?

For vehicles with real theft or short-delivery exposure, long-haul trucks, tankers, overnight-parked vehicles, yes. Give CAN the efficiency-trend job and the sensor the inventory job; asking one instrument to do both is where fuel-monitoring programmes go wrong.

Does temperature affect which measurement to trust?

It affects both. Diesel volume changes with temperature, which is why the industry corrects to a standard reference temperature under ASTM D1250. A tank reading taken at midday and one at night will differ for real physical reasons that have nothing to do with theft or efficiency.

Sources

Related insights