It’s well known that all measurement devices are affected by temperature – from the oscillators used in mobile phones to the need to measure tyre pressures ‘cold’. In fact, all sensors are sensitive to more than one thing, and an understanding of the interplay of factors is essential in order to get beyond a first-order sensor response to a robust, reliable measurement – removing or reducing the effects of the things we don’t want to measure.
Suppose that we’re trying to measure the pressure inside a barrel – the pressure gauge is significantly affected by temperature, making its output unreliable in a real-world application. You could compensate for temperature with another sensor – but a simple temperature compensation is complicated by the fact that the thermometer you might add to the system is similarly affected by pressure (a bit). In this case, the two measurements need to compensate each other using a coordinate transform. If one of the sensors is non-linear it’s more complex, but there are several standard methods.
By considering the sensors in this way – as measuring along axes – it’s possible to both choose them well and make them work harder for you. This also gives a clear view of what will be achievable, and what accuracy might be achievable.