Digital Uncertainty
Error introduced into a temperature measurement when an analog signal is converted into a discrete digital value by an electronic circuit. Temperature quantization noise represents the difference between the actual thermal state and the nearest available digital step. The size of this error is determined by the resolution of the analog-to-digital converter and the range of the sensor.
This noise is inherent to all digital sensing systems and cannot be removed by simple filtering once the conversion has occurred.
Step Resolution
Maximum possible error for any single measurement is defined by the step size of the converter. If a sensor covers a range of one hundred degrees with a ten-bit converter, each step represents approximately zero-point-one degrees. The temperature quantization noise will fluctuate within this range, creating a floor for the precision of the system.
Increasing the bit depth of the converter reduces the noise power but may increase the power consumption and the conversion time. Designers must choose a resolution that meets the requirements of the application without adding unnecessary complexity.
Signal Processing
Averaging can be used to improve the effective resolution of the measurement beyond the hardware limit. By taking multiple samples and calculating their mean, the random component of the quantization error is reduced. This technique relies on the presence of a small amount of thermal or electrical noise to dither the signal across the quantization boundaries.
If the signal is perfectly static, the averaging will not improve the result. Temperature quantization noise becomes a limiting factor in control loops where small changes in heat must be detected to maintain stability.
Metrological Floor
Verification of the noise level is performed by applying a stable temperature to the sensor and observing the fluctuations in the digital output. The variance of the signal under these conditions provides a measure of the quantization effects and any additional electronic interference. Calibration does not eliminate this noise but ensures that the digital steps correspond to the correct physical values.
A temperature quantization noise specification is usually provided as a root-mean-square value in the device datasheet. Engineers use this figure to determine the minimum detectable change in the thermal environment.