The output of many modern bridge type sensors contains errors that must be eliminated through calibration. One major source of error is variation of sensor span and offset from device to device and their respective drifts with temperature. These drifts can be non-linear. Another source of error is the nonlinearity of the sensor output with applied stimulus. Signal conditioning electronics can also introduce its own errors. The complexity and duration of the calibration process that removes these errors is a major cost factor in modern sensor production. Multiple measurements have to be made while applying stimulus to the sensor in temperature chambers. This paper discusses a practical model of sensor behavior vs. input stimulus and temperature based on general pre-characterization data.
This paper was originally presented at the SENSOR 2005 conference (www.sensorfair.com) in Nürnberg, Germany and is reprinted by Texas Instruments with permission.