Why do three sources give three numbers?
The card counts litres at the pump. The tank sensor measures the level in the tank. The CAN counter adds up what the engine control unit injected. Fuel can be lost between each pair, and each instrument has its own error.
- Card minus sensor fills: fuel bought that never showed up in the tank.
- Sensor use minus CAN: fuel that left the tank but was not burned by the engine.
- Card, corrected for the level change, minus CAN: the overall balance of the period.
How big is the error of each source?
Each source gets a one-sigma error. A sensor error counts once per level reading, and a fill needs two readings. Independent errors add as the root of the sum of squares, as in the JCGM guide to measurement uncertainty.
Worked example
A 400 L tank went from 220 L to 180 L. The card shows 1,850 L, the sensor saw 9 fills totalling 1,800 L, and CAN reports 1,812 L.
| Comparison | Difference | Band | Result |
|---|---|---|---|
| Card vs sensor fills | +50 L | ± 38.7 L | Check |
| Sensor use vs CAN | +28 L | ± 114.5 L | Within the band |
| Balance: card vs CAN | +78 L | ± 110.9 L | Within the band |
The CAN band is wide because 3% of 1,812 L is 54 L. The card and the sensor fills are the tightest pair, so the 50 L gap shows there first: ask where those litres went before you blame the sensor.
How to read the result
- Within the band: the difference can be explained by the error of the sources.
- Check: up to twice the band. Look at the fills one by one.
- Investigate: more than twice the band. The sources alone do not explain it.
- A result that is always on the same side points to calibration, not to chance.
Assumptions and limits
- Errors are treated as independent and random. A wrong tank table or a slow counter gives a bias that this model does not show.
- The error values are typical inputs, not Tracom specifications. Use your sensor datasheet and your calibration act.
- Fuel temperature changes the volume. For hot tanks, correct the litres first.
- A difference is a reason to look, not proof of theft.
- The calculations run in your browser. Your inputs reach our server only if you ask for the emailed report, and are then kept for up to 45 days.
Sources
- OIML R 117-1: dynamic measuring systems for liquids other than water (fuel dispensers)
- JCGM 100 (GUM): combining independent uncertainties
- J1939 explained: engine data on the truck CAN bus
Sources and date checked 3 November 2026: OIML R 117-1 (class 0.5, 0.5 %) and JCGM 100 (root sum of squares for independent inputs; about 1.96 sigma for 95 % of a normal distribution). The J1939 guide is background reading.