## Possible sources of error

Two possible error groups are distinguished in the context of strain gauge-based force transducers:

- Load-independent error: Errors that are not determined by the applied force, but whose error signal is always the same, no matter what the measured value is. They are calculated as a percentage of the force transducer’s nominal (rated) force
- Errors relative to the actual value: Errors that are proportional to the force applied at the moment of observation, which can therefore be calculated as a percentage of the measured value

The temperature effect on the zero point (TCzero) is an example of a load-independent error. This measurement deviation has a specific absolute value that is independent of the measured force. Considering such an error relative to the output signal shows that the TCzero is always particularly pronounced when just a small portion of the nominal (rated) force is used. The absolute value is always the same, which, due to the small useful signal, increases the relative portion in this situation.

For example, a typical force transducer using conventional technology should have a nominal force of 100 kN; we assume a TCzero of 0.5 % per 10 Kelvin.

This means that a temperature change of 10K generates a measurement uncertainty contribution of 0.5 % of the rated force. This corresponds to an uncertainty of 0.5 kN. If we do not use the force sensor at 100 kN but at 20 kN, then the measurement uncertainty contribution will still be 500 N. However, due to the smaller force, the relative measurement uncertainty contribution will now be 2.5 %.

Let's take a look at the same situation when using a C10. The nominal force of 100 kN is also available with the C10 series. However, the TCzero is specified at only 0.075 % per 10 Kelvin.

Thus, the error portion is 75 N and, similar to the above example, this portion remains constant over the entire measuring range. When using the modern C10 at 20 kN, the error portion of the TCzero drops from 2.5 %, achieved with conventional technology, to only 0.375 %.

Besides the TCzero, the linearity error too is related to the full-scale value. Hence, the above considerations can also be transferred to these error portions.

Errors that are relative to the actual value are calculated relative to the currently applied signal. This includes, for instance, the temperature dependency of sensitivity (TCS), creep, or even the tolerance of a calibration that may have been performed.

When evaluating an error, the individual errors are added geometrically, i.e. a significant improvement in measurement accuracy can only be achieved if the largest individual errors are improved. In many cases, TCzero , linearity, and hysteresis are of vital importance. Since these errors – as elaborated above – are relative to the full-scale value, i.e. the output signal at full utilization of the nominal (rated) force, any improvement of these parameters will be particularly effective, also enabling the force transducer to be used in the so-called partial load range, i.e. utilizing only a partial range of the nominal (rated) force