load cell temperature drift



  • Well I know that this is not directly related to the pycom device, but the code and load cell are connected to the device so i am hoping to gain from the wealth of knowledge and experience on this forum.

    Until a few weeks ago, I knew nothing at all about load cells and have never used one, now I know enough to be able to read one and understand the many factors that effect drift.

    The one I am trying to crack right now is temperature drift, and rather than re-inventing the wheel, I wondered if any one else had been through this exercise and can save me some time in understanding how to cope with it.

    Obviously, the characteristics of the cell material will change with temperature, but will this be a linear offset that can easily be fixed with a factor variable or is it as I suspect much more than that ?

    Any insight into this would be most gratefully received !



  • @kjm As far as I understood, they were mostly interested in changes during the day, and could therefore cope with long term drifts; they had to, because the hive itself my change it's weight over time, not only to the changing amount of bees & honey, but also due to humidity etc. Still, an error of <1% is doable.
    For any long time measurement you will need way more precise devices. But at ~1 USD, you get what you pay for.



  • @robert-hh Mate I take my hat off to the beehive people if they've managed to use that chip in a weighing application! Presumably they can't lift the hive off the scale to get a null reading between weighings, so most impressive. Apologies to crankshaft for the thread drift.



  • @kjm Quite a while ago I made a few test series with the HX711. What I found:

    • the output is noisy. Take averages, and/or discard the lower 8-10 bits. The claim of 24 bits resolution is simply nonsense.
    • with temp tests up to the range of 50 °C variation I had an average value drift of about 0.2% over the full span, after filtering off the noise using the median over 100 single samples. I used a DS18B20 placed next to the HX711 for the temp reading.
    • Each test cycle ran between 3 hours and 24 hours. I did not make longer test series for along term drift. But there is a group of beehive people who use the hx711 for their purpose. They have experience with long term usage.
    • the popular "green board" is bad. It does not follow the hx711 vendor's recommendation on circuit design. Especially, the input impedance of the analog inputs doe not match, for the sake of saving the costs of 1 (!) smd resistor. That increases the temp drift. Analog and digital GND are not connected, and it is a single layer PCB, mimicking a double layer PCB.


  • @robert-hh Agreed but I've seen more than a few HX711s in depth sensing applications where it's not easy to take null readings between pressure readings. Random number generators!



  • @kjm said in load cell temperature drift:

    The only way you can make any use of them is to take a null reading with nothing on the load cell then quickly grab a reading with the load before the damn thing drifts.

    That is the typical use for e.g. a bathroom scale. You switch it on and let it settle to 0, get on it and quickly get off when you see the number.



  • I've found the HX711 type chips hopeless. The only way you can make any use of them is to take a null reading with nothing on the load cell then quickly grab a reading with the load before the damn thing drifts.



  • Hi hope you are doing well, possibly this website on load cell calibration could assist you.



  • @pchirouze - Thanks for the help, what I think I have discovered is that rate of change has far more effect than the absolute temperature change.

    What I mean is if you change the temperature rapidly, (approx > 0.1 deg / min) then the readings spike up and down severely, but they eventually settle again and the temperature drift is less significant than I originally thought.



  • Hi,
    I don't know which load you have. But a load cell you have drift on offset and a drift on span, but span drift is lower than zero drift.
    An industrial load cell is internally compensated,but you must have high performance electronic amplification and numeric conversion for good result.
    But if your goal is to obtain a result about 1% of precision it's little easy. The stability and quality of the voltage applied to load cell is very important, because the output signal is proportional to this voltage.
    I hope this explanations help you


Log in to reply
 

Pycom on Twitter