Load Cell Zero Drift on a 24/7 Filling Line
On a 24/7 filling line, a 500 kg-rated strain-gauge load cell weighs 25 L drums. The cell is rated 2 mV/V, excited at 10 V, and connected by 30 m of unshielded four-core cable that runs in the same tray as a 7.5 kW pump VFD. The signal goes to a 24-bit ADC with a 2.5 V reference and a programmable gain of 128; the effective input-referred noise is about 0.6 µV RMS. Over an 8-hour shift the zero indication drifts upward by 0.9 kg while the ambient temperature rises 12 °C. The cell datasheet gives a temperature effect on zero of 0.02 % FS/°C, creep of 0.03 % FS/30 min, and non-linearity of 0.03 % FS. The operator re-zeroes every two hours, but quality records show filling weight drifting 0.4 kg low by the end of the shift. The line cannot stop for a full re-calibration.
1. Calculate the full-scale output, the sensitivity in µV/kg, the ADC LSB and the noise-limited resolution. Which effect actually limits usable accuracy?
Read the worked answer
Full-scale output = 2 mV/V × 10 V = 20 mV at 500 kg. Sensitivity = 20 mV / 500 kg = 40 µV/kg.
The ADC LSB is 2 × 2.5 V / (128 × 2^24) = 5 / 2,147,483,648 = 2.33 nV. Referred to weight, one LSB is 2.33 nV / 40 µV/kg = 5.8 × 10^-5 kg, which is negligible.
The 0.6 µV RMS input noise corresponds to 0.6 µV / 40 µV/kg = 0.015 kg RMS. Even with a few times that for peak-to-peak noise, the electrical resolution is far better than 0.1 kg.
Therefore the ADC and shunt are not the limiting effects. The observed 0.9 kg shift must come from the cell, the mounting, the excitation supply, thermal effects or interference. The temperature coefficient alone predicts 0.02 % × 500 kg × 12 °C = 1.2 kg, which is the same order as the observed drift. Creep and non-linearity are each about 0.15 kg over their specified intervals, so they are secondary.
2. Quantify the temperature and creep contributions over the shift, explain how re-zeroing can still leave a 0.4 kg underfill, and decide whether this is a zero or span problem.
Read the worked answer
Temperature effect on zero: 0.02 % FS/°C = 0.0002 × 500 = 0.1 kg/°C. Over 12 °C this is 1.2 kg, consistent with the measured 0.9 kg upward drift.
Creep: 0.03 % FS/30 min = 0.15 kg per 30 min. Creep is load- and time-dependent and partly recoverable, so it should not be treated as a simple cumulative 2.4 kg over 8 h, but it is large enough to matter if the cell sits loaded between fills.
Non-linearity: 0.03 % FS = 0.15 kg.
Re-zeroing every two hours hides the zero offset at the moment of zeroing, but it does not correct span or non-linearity. If the zero indication drifts upward between re-zeroes, the controller believes the drum already contains more product than it does, so it stops filling early. A gradual 0.9 kg zero drift with two-hour re-zeroing can easily produce the observed 0.4 kg low fill by the end of the shift.
This is primarily a zero-drift problem, with a possible span component. A loaded calibration check at mid-scale and full-scale is needed to separate them; an empty zero check only proves the zero.
3. Recommend corrective actions in mechanical, electrical and software areas, and state how you would verify them without a full shutdown.
Read the worked answer
Mechanical: check that the cell is mounted flat, with the correct torque on the mounting bolts and no pipe or frame binding. A binding mount converts thermal expansion into a false load. Verify that the load is applied through the cell axis and that the drum guide does not touch the platform.
Electrical: replace the unshielded cable with shielded twisted-pair, route it away from the VFD tray, earth the shield at one point only, and use a clean linear or low-noise excitation supply. Add common-mode chokes or ferrites at the enclosure entry and check the VFD output cable shielding and earths. Measure the excitation voltage at the cell, not at the indicator.
Software: implement temperature compensation using the cell's zero coefficient or a dummy gauge, shorten the re-zero interval during warm-up, and add a mid-scale daily check with a traceable test weight. Do not simply increase digital filtering, which would slow the fill and hide the drift.
Verification without a full stop: run an empty zero log for 30 min with the pump running, then a loaded 30 min creep test at a fixed weight, then compare the indicator with a calibrated test weight at 25 %, 50 % and 100 % of range. Repeat after a warm-up cycle and check that the 0.4 kg end-of-shift error is reduced.