The signal from a load cell is small. At 2 mV/V and 5 V excitation, full scale is 10 mV, and a resolution of 1 in 10,000 is 1 µV. Anything that couples a few tens of microvolts into the cable becomes visible on the display.
Where the noise comes from
Three sources dominate: mains frequency coupling through an unshielded or badly shielded cable, earth potential differences between the cell frame and the indicator, and switching noise from nearby drives or contactors.
| Source | Typical amplitude | Symptom |
| Mains coupling, unshielded cable | 20 to 100 µV | Reading oscillates at 50 Hz |
| Earth potential difference | 10 to 500 µV DC | Zero offset that changes with machine state |
| VFD switching noise | 50 to 200 µV | Random jumps at drive frequency |
| Thermocouple effect at a bad joint | 5 to 40 µV per K | Slow drift with temperature |
The shielding rule
Connect the shield at one end only, normally at the indicator. Connecting both ends creates a loop in which circulating current induces a voltage in the signal pair, which is the opposite of what a shield is for.
Run the signal cable separately from power cables, cross at right angles where they must cross, and keep the pair twisted right up to the terminal.
The diagnostic that takes five minutes
Disconnect the cell and short the indicator input at the terminals. If the display is quiet, the indicator and its earth are fine and the problem is in the cable or the cell. If it still wanders, the problem is upstream of the cable.
Then reconnect and disconnect the shield at the indicator end. A reading that improves with the shield floating points to an earth loop; a reading that gets worse points to genuine coupling that the shield was suppressing.
Standards referenced: OIML R76 for non-automatic weighing instruments, IEC 61000-4-3 and -4-6 for radiated and conducted immunity.
The excitation supply is part of the problem
A load cell is a bridge, and a bridge is only as stable as its excitation. A 5 V supply that moves by 1 mV changes the output by 0.02 % of full scale, which is the same order as the temperature specification.
Use a stable, low-noise supply, and measure it if the readings drift. Many weighing problems blamed on the cell are actually excitation problems.
Cable length and the six-wire connection
With a four-wire connection, the resistance of the excitation leads is in series with the bridge and reduces the effective excitation. A 20 m run of 0.5 mm² cable adds about 1.4 Ω per conductor, which on a 350 Ω bridge is a 0.4 % loss.
A six-wire connection adds sense leads that let the indicator regulate the excitation at the cell, removing the cable loss from the measurement. For any run over 10 m, six-wire is worth the extra pair.
Checking the earth path once
Measure the AC voltage between the cell body and the indicator chassis with the machine running. Anything above 0.5 V indicates a potential difference that will drive current through the shield. Bond them at one point and re-measure.
The five-minute layout review
Walk the cable route with the machine powered. A signal cable running parallel to a motor cable for 2 m picks up more noise than one that crosses it at right angles, and the difference is often 30 to 50 µV.
Keep 200 mm of separation where parallel runs cannot be avoided, and use a twisted pair for the signal. The twist cancels magnetic coupling; the shield handles electric coupling. Both are needed.
