What temperature-related drift do you have to reckon with?

What are the causes and magnitudes of temperature drift? What measures are there?

A temperature-related drift cannot be completely ruled out for sensors with strain gauges.
This is due to the fact that the supply voltage for the Wheatstone bridge causes a current to flow through the strain gauges, thereby heating them up.
Especially after switching on and in the first 20...30 minutes you have to deal with a drift due to self-heating.
The same also applies to the measuring amplifier: The first amplification stage of the measuring amplifier is particularly affected.
In a first approximation, the drift decreases with increasing time according to an exponential function.
If you look at the sensor over a longer period of several hours, the drift can be alternately positive and negative. this is because thermal equilibrium is never fully established.
Especially with 6D sensors, the Z-axis shows stronger fluctuations, which is due to the fact that in the compensation matrix all 6 channels are added, while for Fx and Fy some factors are positive and some are negative.
To make matters worse, in the case of force/torque sensors, the z-axis has a measuring range that is two to four times higher, and the absolute drift is therefore also two to four times higher than the drift for the Fx or Fy axis.
A drift of a few Newtons in the measuring range of 2 kN is quite "normal".
However, other possible causes of drift can also be related to the test setup itself if the sensor is firmly clamped in from two sides: The sensor then measures the thermal expansion of the test arrangement, and a few micrometers of expansion can generate forces of N or even kN.
What remedies are there:
If the drift due to self-heating or temperature fluctuations is critical because the measurement lasts several hours, then the following measures can help:
  1. Give the sensor 1 hour to warm up
  2. in exceptional cases, the supply voltage can also be halved, which reduces the power loss to a quarter. This is definitely recommended for a miniature sensor such as KM10 or KM10z, because these become noticeably warm at 5V and even hot at 10V bridge supply voltage. With a larger K6D40 sensor, it doesn't bring much relief.
  3. No heat input into the sensor shortly before the measurement: do not touch, do not place any cold or warm weights on the sensor, do not open the window. Any asymmetry in the temperature distribution overrides the temperature compensation.
  4. As a rule, one should try to zero just before the measurement. Even with long-term measurements, one can consider whether one should relax for a few seconds in order to measure or compensate for the current zero.