Estimating Ambient Temperature using Internal Sensors and Thermal Modelling in Mobile Phones

Abstract: This thesis studies the possibility of estimating the ambient temperature around a mobile phone using its internal temperature sensors. The thesis aims to develop and evaluate ambient temperature models that use internal temperature sensors within the device and, from a manufacturer’s perspective, investigate if the models for ambient temperature can be enhanced by tailoring them to the internals of specific devices. The chosen approach was to use thermal circuits and, based on them, derive state-space models. The parameters of these models were then estimated using collected data. A linear polynomial approach was also evaluated but gave worse estimations than the state-space approaches. The best-performing model, one of the state-space models, has an estimation accuracy of within ±1–3°C and an average error of below 1.5°C when evaluated on a broad collection of testing scenarios. Once parameter estimations have been performed, all the models have a low processor resource utilisation, which is ideal for an on-device implementation. The thermal models should theoretically be generalisable and could be used on other mobile phone models with similar internal layouts with only the need for new parameter estimations. The thermal models’ principles could theoretically also be able to be used on almost all embedded devices with similar internal temperature sensors.