| A good sensor obeys the following rules: | | | | If the output signal slowly changes |
| 1. the sensor should be sensitive to the | | | | independent of the measured property, |
| measured property | | | | this is defined as drift. |
| 2. the sensor should be insensitive to | | | | Long term drift usually indicates a slow |
| any other property | | | | degradation of sensor properties over a |
| 3. the sensor should not influence the | | | | long period of time. |
| measured property | | | | Noise is a random deviation of the |
| In the ideal situation, the output | | | | signal that varies in time. |
| signal of a sensor is exactly | | | | Hysteresis is an error caused by the |
| proportional to the value of the | | | | fact that the sensor not instantly |
| measured property. The gain is then | | | | follows the change of the property being |
| defined as the ratio between output | | | | measured, and therefore involves the |
| signal and measured property. For | | | | history of the measured property. |
| example, if a sensor measures | | | | If the sensor has a digital output, the |
| temperature and has a voltage output, | | | | signal is discrete and is essentially an |
| the gain is a constant with the unit [V | | | | approximation of the measured property. |
| K]. | | | | The approximation error is also called |
| If the sensor is not ideal, several | | | | digitization error. |
| types of deviations can be observed: | | | | If the signal is monitored digitally, |
| The gain may in practice differ from the | | | | limitation of the sampling frequency |
| value specified. This is called a gain | | | | also causes a dynamic error. |
| error. | | | | The sensor may to some extent be |
| Since the range of the output signal is | | | | sensitive for other properties than the |
| always limited, the output signal will | | | | property being measured. For example, |
| eventually clip when the measured | | | | most sensors are influenced by the |
| property exceeds the limits. The full | | | | temperature of their environment. |
| scale range defines the outmost values | | | | All these deviations can be classified |
| of the measured property where the | | | | as systematic errors or random errors. |
| sensor errors are within the specified | | | | Systematic errors can sometimes be |
| range. | | | | compensated for by means of some kind of |
| If the output signal is not zero when | | | | calibration strategy. Noise is a random |
| the measured property is zero, the | | | | error that can be reduced by signal |
| sensor has an offset or bias. This is | | | | processing, such as filtering, usually |
| defined as the output of the sensor at | | | | at the expense of the dynamic behaviour |
| zero input. | | | | of the sensor. |
| If the gain is not constant, this is | | | | Resolution |
| called nonlinearity. Usually this is | | | | The resolution of a sensor is the |
| defined by the amount the output differs | | | | smallest change it can detect in the |
| from ideal behaviour over the full range | | | | quantity that it is measuring. Often in |
| of the sensor, often noted as a | | | | a digital display, the least significant |
| percentage of the full range. | | | | digit will fluctuate, indicating that |
| If the deviation is caused by a rapid | | | | changes of that magnitude are only just |
| change of the measured property over | | | | resolved. The resolution is related to |
| time, there is a dynamic error. Often, | | | | the precision with which the measurement |
| this behaviour is described with a bode | | | | is made. For example, a scanning probe |
| plot showing gain error and phase shift | | | | (a fine tip near a surface collects an |
| as function of the frequency of a | | | | electron tunnelling current) can resolve |
| periodic input signal. | | | | atoms and molecules. |