A belt weigher wrong at low flow, but accurate when the plant is flat out, almost always has a zero problem, not a span problem. Zero is a fixed offset: whatever the scale reads on an empty belt is added to every reading afterwards. At high flow it vanishes into a big number; at low flow it can be most of what the totaliser counts. A span error does the opposite, sitting at the same percentage at every rate. So the shape of the disagreement names the fault before anyone touches the scale.
This is the deep dive on low flow accuracy. See also zero, span and repeatability explained and common belt weigher faults.
How do I tell a zero error from a span error without touching the scale?
Compare the percentage error at high rate against the percentage error at low rate, over several runs rather than one. That comparison is the diagnosis. A belt weigher (also called a weightometer, belt scale or conveyor belt scale) carries both errors at once, but they leave different fingerprints:
- The same percentage at every rate points at span, or at the belt speed signal, which scales every tonne alike. See belt weigher versus weighbridge.
- Small at high rate, growing as the rate drops, points at a zero offset, and is behind almost every “fine when we are flat out” call we take.
- Scattering run to run at every rate is instability, not bias. See erratic totals.
One low flow comparison is mostly noise: a light run moves so few tonnes that ordinary variation swamps the error.
How does a small zero offset turn into a large percentage error at low flow?
By plain arithmetic. The figures below are invented round numbers to show the shape of it, not site measurements.
Take a conveyor with a design rate of 1,000 t/h, and a weigher with a zero offset of 5 t/h: it reads 5 t/h with nothing on the belt. At 800 t/h it reports 805, an error of about 0.6% that nobody would find. At 200 t/h it reports 205, an error of 2.5%. At 50 t/h it reports 55, an error of 10%. Nothing changed but the material underneath.
Set a 1% span error beside that: 8 t/h high at 800, 2 t/h high at 200, half a tonne an hour at 50, which is 1% every time. Span is a percentage of what is there. Zero is a fixed lump added whether anything is there or not.
A zero sitting low does the reverse: the light belt under reads, and can fall below the integrator’s cutoff so nothing is counted.
What actually puts that zero offset there in the first place?
Physical change in the weigh area, nearly every time. The scale cannot tell a clean belt from one carrying an even coating, so anything that adds dead load or shifts the geometry becomes a standing offset:
- Build-up and carryback, the most common by a wide margin: wet fines packing onto the carriage and weigh idlers, plus material carried back past worn scrapers.
- A frame that cannot float. It must deflect freely onto the load cells, and corroded tie rods, a wedge of material against the structure, or a seized idler all stop that.
- Moisture. Water on the belt is real mass and reads as tonnes; water in a junction box gives an offset that comes and goes with the weather.
- Auto zero tracking a dirty belt, quietly learning a coated belt during empty periods and baking it in.
- A zero taken on a belt that was not clean or settled: a short zero, water or return side material present, or a new belt not yet bedded in.
For a zero that keeps moving, see zero drift.
Why does intermittent feeding, or a short run, make the error worse?
Because a conveyor fed in bursts lives in the region where the offset dominates, and picks up two penalties on top.
It lives at low flow. Near empty is the normal condition there, so the ballooning percentage above is what you get all day.
Ramp up and ramp down. Every run starts with the belt filling and ends with it clearing, both crawling through the lightest, worst part of the range. On a continuous conveyor that happens twice a week; on one fed in bursts, dozens of times a day.
The low flow cutoff throws real tonnage away. Most integrators have a setting, called zero drop out, low flow cutoff or a minimum flow threshold, that stops the totaliser counting below a set rate. It earns its place: without it, a small offset would clock up phantom tonnes overnight on an empty belt. But it cannot tell a genuine light load from an offset, so it discards real material at each end of every run. A few tonnes times dozens of runs a day is no rounding error by month’s end.
Can a calibration fix a conveyor that is running far below its design rate?
No, and we would rather say so plainly than dodge it. A belt weigher is built around a design loading: load cell capacity, weigh span and the calibration reference are all chosen for it, and it works well over a usable range beneath that. Run it at a small fraction of that and you are outside what any belt weigher does well: the load the cells are asked to resolve is a sliver of what they are built to carry.
The honest options are to treat the low rate figure as an indication rather than a measurement; to move the measurement to a conveyor that actually runs loaded; to feed the conveyor harder for shorter periods instead of trickling all day; or to re-rate the installation, a hardware conversation rather than a calibration. See choosing a belt weigher.
How should a material test or belt cut be run to represent the flow I care about?
At the flow you actually care about, not the flow that is convenient, because a test only proves accuracy at the loading it was run at. The common mistake is a full rate check, quoted afterwards for a scale that lives at a quarter of it.
Anything on or near the belt waits until the conveyor is isolated. A belt cut means people removing material from a stopped belt, so it needs your site’s lockout and tagout procedure, a proven zero energy state, trained and authorised people, and the site’s permits first. If that is not available, stop and call us on 1300 101 666.
With that in place:
- Test at both ends of the range, same day and same scale: either result alone is half the story.
- Take more cuts at low loading, because thin material lies unevenly and a scattered result wants re-running.
- Run enough tonnes that the difference beats the noise.
- Check the zero over a full belt revolution, which needs no material and measures the offset directly.
- Write the rate beside every result, or it compares to nothing later.
The method is in the belt cut test.
What do we actually do about a weigher that is only wrong at low flow?
We fix the offset mechanically first, prove it at the rate that matters, then say straight out if the installation is the limit. Adjusting numbers over an uncorrected fault only hides it, so the sequence is the usual one: clean back to bare, free off the frame, tie rods and idlers, align, then zero over a full belt revolution rather than a short one.
Two things are extra when low flow is the complaint. We check what the auto zero has been tracking, because a scale that re-tares onto a coated belt keeps recreating the fault. And where the belt can be run light on the day, we can verify at low rate as well as high, so the record covers the condition you actually run in. Ask for both when you book, so the visit is planned around them. If the conveyor runs too far below its design loading, we put that in writing rather than recalibrate every few months.
Frequently asked questions
Why does my belt weigher over read when the belt is nearly empty?
Because a zero offset is added to a small real number, so it becomes a large percentage of it. Build-up, carryback or water on the belt are the usual sources, and the fix is cleaning and a fresh zero, not a span adjustment.
Can I just recalibrate at low flow to fix it?
Not really. It tilts the calibration to suit one condition and puts the high rate readings out instead. The offset is physical, so it needs removing rather than dialling out.
What is a low flow cutoff, and should I turn it off?
It stops the totaliser counting below a set rate, so a small zero offset cannot invent tonnes on an empty belt. Turning it off is rarely the answer, since you then count the offset too, but check the threshold suits how your conveyor is fed.
Is there a minimum flow rate a belt weigher can measure?
Every installation has a practical lower limit, set by its load cells, weigh span and how large the fixed errors are next to the load. There is no universal figure, but a small fraction of the design rate is beyond any belt weigher.
My weigher agrees with the weighbridge on full trucks, so it must be fine?
It is fine at that loading, which is all a full rate comparison can tell you. If the conveyor also runs light, check it at low rate before treating that result as proof.