I have a set of load cells connected to a AD620. Everything is working good except when I change the temperature. The temperature can vary between -10 and 50. I have the AD620 connected to an MCP3208, the MCP is connected to Pi via SPI.
For instance, when the exterior temperature is around 20 degrees the calculated weight is correct. But when the exterior temperature is 5 degrees, the weight is 5 kg less than the expected.
How can I control this ? Ideas ?
Thanks.
(More explanations added)
the problem could be the temperature coefficient of span of the load sensor. Or even temp coefficient of zero of load sensor. One has to set up the aims of the design.
You say that 5kg less than expected. 5kg change in 50 kg full scale is 10%. The same in 5000kg system is just 0.1%. So one has to indicate the maximum weight that you propose to weigh. In this uncertain world everything changes. But everything does not change by 1% or 5%. When the change is adequately low it becomes negligible. So identify this percentage first.
Say you can only neglect 0.1%, then a change of 0.1% in temperature range of -10 to 50 demands that you use a load sensor which has a zero/span tempcon of less than 1000/60 or about 16.6 ppm. [Is this available at all??] It may be very costly but that is what you need. If you can neglect 1%, you can live with load sensors with tempco of span/zero of about 1666 ppm or 0.016%. So start with some aims. Looks like 80 kg is maximum. What is the resolution that you aim at? 100gms? Then choose the load sensor with a rating of 100kg. the tempco of spna and zero has to be less than 0.016%/degC, if an accuracy of 0.5% over -10 to 50 deg is desired. Choose an amplifier with drift with temperature is less than 1/3rd of that due to load sensor. Identify each contribution to drift and identify whether you are using a proper amplifier. Sometimes drift can arise even due to drift of component values in the amplifier.
