The simple answer is that a calibration reference load cell should have an uncertainty that is four times better than the load cell it is going to be used in calibrating. The more complete answer is, however, that it is not strictly the load cell that has the uncertainty, but the results obtained by that load cell. In order to get results from a calibration reference load cell, the load cell needs a display, signal conditioning, a calibration stand or means of loading and a calibration procedure.

If you look at the contributing uncertainties in the results obtained from a calibration load cell, the largest contributors to error in descending order are likely to be:

• Test Fixture(s)

• Calibration rig or means of loading

• Reference load cell display and signal conditioning

• Reference calibration load cell calibration

• Reference load cell creep

• Reference load cell hysteresis

• Reference load cell linearity

If a low uncertainty calibration result is desired, then it makes sense to reduce individual contributors to the overall uncertainty budget. If the test fixture, calibration rig, display and signal conditioning are poor, it would seem pointless to spend time and money employing a low (good) uncertainty load cell. If the test fixture, calibration rig, display and signal conditioning are good, then it would be worthwhile spending time and money to get a low uncertainty calibration carried out on a low-creep, low-hysteresis, high-linearity load cell.