A scale that will not hold zero is usually reported as "the load cell is drifting". That single word covers at least three different mechanisms, and treating them the same way leads to a lot of unnecessary sensor replacements. The quickest way to separate them is to watch what the reading does when the load is removed.
Three behaviours worth distinguishing
No-load zero drift is a change in the bridge output with nothing on the platform, typically driven by temperature, warm-up or moisture. Remove the load and the reading slowly walks.
Creep is a change in output while the load is held constant, usually measured over 30 minutes at rated load. The reading changes under load, then largely recovers when unloaded. It comes from the elastic element, the adhesive and the strain gauge backing, not from the electronics.
Zero return error is the offset left behind immediately after unloading following a full-load excursion. A cell with poor zero return looks like a zero drift problem but is really a mechanical hysteresis issue.
| Test | What a healthy cell does | What to suspect if it does not |
|---|---|---|
| 30 min no-load, stable temperature | Zero stays within a few µV/V | Moisture ingress, cable strain, excitation drift |
| 30 min at rated load | Output changes within the creep spec (often ~0.02–0.05 %FS) | Bond degradation, overload history |
| Immediate zero after unloading | Returns close to the starting zero | Mechanical hysteresis, mounting preload |
| Temperature step of 10 K | Zero shift within the compensated range | Compensation resistors, gauge mismatch |
Warm-up is not a myth
Bridge excitation puts a few hundred milliwatts into the gauges. A cell connected to an indicator needs time to reach thermal equilibrium, and a 30-minute warm-up is not unusual for high-resolution weighing. During that period even a good cell will show a monotonic zero shift. If a system is being judged on a cold start, fit the warm-up into the operating procedure instead of trying to compensate it out.
Six-wire systems sense the excitation voltage at the cell and cancel most of the cable resistance effect. Four-wire installations cannot, so a long cable in a sunlit tray can produce a zero shift that follows ambient temperature rather than load. When the drift correlates with the time of day, check the cable run and the excitation sensing before ordering a new cell.
Mounting causes more "creep" than the sensor
Before blaming the load cell, look at what is touching it. A mounting bolt torqued unevenly, a hard stop with 0.2 mm of clearance that closes under load, a load button rubbing against a bracket, or a cable tie pulling on the cable gland — all of these introduce a slow, load-dependent change that looks exactly like creep. Loosen and re-torque the mount to the specified value, confirm the load is applied axially through the designed load introduction point, and make sure no rigid conduit is bypassing the sensor.
- Check for debris under the load foot and for a warped mounting plate.
- Verify the excitation voltage at the cell terminals, not at the indicator.
- Insulation resistance between bridge and body should be high and stable; a drop points to moisture.
- Log zero before loading, output at 30 minutes, and zero immediately after unloading.
When the cell really is at fault
A history of overload is the usual cause. A single excursion well beyond rated capacity can permanently shift the zero and worsen both creep and hysteresis even though the cell still produces a plausible span. Shock loads from dropping material onto a hopper do the same thing over time. If the zero return error grows progressively across a series of load cycles, retire the cell rather than recalibrating the indicator to hide the offset — a zero that has been trimmed away will come back with the next temperature change.
Sales Team