Creep is the slow change in output that continues after a load cell has been loaded and left alone. It is the single most common reason a strain gauge load cell fails type approval, and it is easy to test badly. OIML R60 treats creep as a separate error component with its own limit rather than folding it into linearity, because creep behaves differently: it grows with time, not with load.
What the test actually does
The cell is loaded to its rated capacity, held for 30 minutes, and read at fixed intervals; the load is then removed and the return to zero is measured after a further 30 minutes. The maximum deviation during the hold, expressed as a fraction of the rated output, must stay within 0.7 times the maximum permissible error for the accuracy class. For a C3 cell the MPE at that load is 0.7 of a verification interval, so the creep allowance is small enough that a poorly bonded gauge will fail immediately.
Where the error comes from
Three mechanisms dominate: adhesive creep in the gauge bond line, elastic after-effect in the spring element, and thermal drift from self-heating of the bridge. The first two are cured by the manufacturer through bonding process control and by pre-loading the element before final calibration. The third is a measurement problem, not a cell problem, and shows up as an apparent creep that disappears when the excitation voltage is reduced.
Testing without a deadweight machine
A calibrated proving ring or a stack of certified weights on a lever gives enough stability for a 30 minute hold if the room is thermally stable. Record ambient temperature alongside the reading; a 1 K change during the hold on an uncompensated cell can look exactly like creep. Repeat the test at 50% and 100% of capacity, because creep compensation is trimmed at one point and can be over-corrected at another.
Kevin Huang