Is it the load cell, the cable, or the junction box?

For what a load cell is and how one is tested, see belt weigher load cells; for unstable readings, erratic totals; for the wider list, common belt weigher faults.

Why is the junction box the first thing we open, not the load cell?

Because on a wet Australian site that is where the faults are. A steel box of fine low voltage terminals on a conveyor stringer gets rained on, hosed down at every clean and left out through a whole wet season. Water gets in past a perished lid seal, a dropped hinge, an unused gland or a conduit sloping downhill into it, then bridges the terminals, corrodes them and drops the insulation, shifting the summed signal with no cell at fault.

Work on or near a conveyor, or inside an electrical enclosure, needs isolation under your site’s procedures and trained, authorised people. If that’s not you, stop and call us on 1300 101 666.

Drying and resealing gets the scale running that day, but unless the path the water took is closed off the box fills again next wet season, so perished seals or green terminals want a new box.

What do cable faults look like, and why do they imitate a dead cell?

Because the cable is part of the measurement, and likelier to be damaged than the cell. Chafe where it crosses a sharp edge or a tray section has gone missing, crush damage from something dropped, rodent damage on plant that runs warm. Worst is a small nick in the sheath: water wicks along inside the cable and surfaces at the terminals or in the cell, well away from the visible damage.

Then the fault that reads as a steady error rather than noise. On a four wire connection the cable's own resistance sits inside the measurement, and it climbs as the cable warms, so a long run reads differently on a cold morning than a hot afternoon. It looks like a scale tracking the weather, and recalibrating never settles it: the fix is the wiring, and the sense connection that cures it is in our load cells guide.

What does a genuinely failed load cell look like, and what causes it?

It shows itself in ways a wiring fault does not. The clearest is a zero balance well away from the cell's datasheet figure: the metal has been pushed past its elastic range and no longer springs back. That is shock overload: a lump down the chute, a seized idler dumping its share onto the frame, or somebody jacking the frame during roller work without unloading the cells. Next is a cell that will not repeat under the same known weight, then physical evidence, a distorted body or corrosion inside.

One failure does not look like a fault at all. A cell can take a step change in its millivolt output and then behave impeccably: the scale stays linear, calibration runs repeat, and the tonnes sit offset by a fixed proportion. Only the raw signal read against the asset's history catches it. Lightning is the same family: a strike can kill a cell outright, or leave it alive and reading badly.

How does a multi-cell frame hide one bad cell, and how do we separate them?

By summing them. On a two or four cell frame the outputs are combined in the junction box before the integrator sees anything, so the display shows one number. A lazy cell is partly covered by its neighbours taking more of the load, and the scale still calibrates and still repeats well enough to pass. It goes wrong the moment loading shifts off centre, and the error surfaces months later in a reconciliation.

Separating them is the most useful test here: apply known check weights over each cell in turn and read the outputs individually. Cells on one frame should track together under equal load, so one reading differently is damaged or on a degraded cable. It also splits the fault families: every cell reading the same wrong thing points at a common cause in the junction box, the excitation or the input, while one standing out points at that leg. On scheduled visits we inspect every load cell for mechanical balance and health, not only when a fault is reported.

Can earthing be the fault, and why will recalibrating never fix it?

Yes, and it is the fault recalibrating cannot touch, because nothing is out of calibration. If the structure is earthed at one point and the integrator cabinet at another at a slightly different potential, current flows between them through the signal wiring and lands as noise on a measurement working in millivolts. A shield earthed at both ends does the same, making a loop of the screen meant to protect the signal. The tell is a reading that moves when something else on the plant runs, not with load or weather.

Calibrating over it just bakes the interference in. Surge protection belongs in the same conversation: see protecting belt weigher electronics.

What can the integrator's own diagnostics tell me, and what can they not?

They narrow it down, but cannot name the part. What they give you is real: whether the excitation reaching the cells is in range, the raw millivolt value the input sees and whether it sits still, and often the load cell utilisation, how much of the cells' rated capacity the frame uses. Too little and the signal rides in a small corner of the range; too much and one surge deforms it. A load cell alarm means the input saw a value outside its window: note it, clear it, and watch whether it returns.

They cannot see past the end of the chain. The integrator reads a summed signal that has come through two cable runs and a junction box, so it cannot tell a corroded terminal from a cracked strain gauge, cannot pick the lazy cell in a summed frame, and will report a textbook signal from a cell held off its load by a binding frame.

Can a load cell be electrically perfect and still read the tonnes wrong?

Very often, and it is the outcome of many load cell calls. A cell measures only the load that reaches it, so anything stopping the weigh frame floating sends part of that load into the structure, and the scale reads low, drifts, or stops repeating. Tie rods stiffened by corrosion after years in mine water are the usual offender, along with a bind between frame and stringer, dirt packed in the gap, a seized weigh idler, and loose mounting hardware.

Every one is a mechanical fault wearing a load cell mask, and every one tests as a healthy cell, which is why we confirm the frame floats before pulling wiring apart. The geometry side is in zero drift.

What order do we work through it, and what can site check safely?

The order runs cheapest and most likely first.

1. Read the symptom shape against the asset's history: a steady offset is bias, a drift following the weather points at moisture or a long four wire run, and scatter points at a connection or earthing.
2. Look before measuring. Open the junction box and inspect for water, corroded terminals, loose screws, seals and glands, then walk the cable run.
3. Prove the supply: confirm the excitation the cells actually receive is in range.
4. Split the frame with check weights on each cell in turn.
5. Test the cell only then, using the checks in the load cells guide.
6. Prove the input by moving the connection to a spare port, calibrate last, then re-verify.

Site can help without going near any of that: after heavy rain, look at the junction boxes from a safe position for staining or dribble marks, and check the lids are dogged down. Note whether the error tracks the weather, the time of day, or a particular machine running.

Leaving it is the expensive option. A slow drift is not a breakdown: nothing alarms, nothing stops, and quietly wrong tonnage goes out against a figure nobody questions until a calibration or weighbridge comparison puts a size on it. Call us when a box has been dried out and the fault came straight back, when the reading tracks the weather, when a scale has stopped repeating, or after a storm or a lump through the plant.

Frequently asked questions