Load cell repeatability

[gaminn]

Hi,
I use these load cells https://www.alibaba.com/product-detail/GML670-half-bridge-flat-micro-load_60574221859.html for platform scale (4 load cells).

I have a problem with repeatability of my measurements. Once I put my prototype of the scale from room temperature (20 °C) to approx. 10 °C environment for 1 hour and then come back to room temperature and leave there the prototype for 1 hour, reading drifts by about 1500 points (which corresponds to 3 uV and to 75 grams in my case) - compared to what I previously measured at the same room temperature.

To sum it up, I get e.g.:
At 20 °C: 100 000 raw value from ADC
At 10 °C after 1 hour: 97 000
Back at 20 °C after another hour: 99 000

Load cells itself are not mechanically coupled with anything, they freely lie on wooden surface. They are not under stress. I don't manipulate with load cells itself during the experiment.

Load cells are connected to HX711 ADC. The repeatability problem is not related to HX711. When fixed resistor bridge network is connected to HX711 instead of load cells then measurements are perfectly repeatable with changing temperature.

I have no clue why this is happening and would be glad if someone can advise. Maybe it is characteristic for these cheap load cells but can someone explain why it happens?

[danadak]

Thermoelectric effects possibly caused by dissimilar metal electrical
connections and joints....

http://www.analog.com/media/en/technical-documentation/application-notes/294542582256114777959693992461771205AN280.pdf

Repeatability spec of cell is .05% of full scale, 1.6V = 800 uV.

1 % = 16 mV
.01% = 160 uV



Regards, Dana.

[gaminn]

Thanks Dana.

After changing temperature I leave the scale for 1 hour before I take a reading. The temperature of cold and hot junctions of parasitic thermocouples should then be the same.

[BrianHG]

That's 4x50kg load cell.
Do you have the data sheet for the load cells?
Did you follow the load cell wiring guide for the HX711?
How did you wire all 4 into the HX711?
Where the load cells mounted so the centers were open to air instead of sitting on a surface?

(Just wait until you add an initial load, like a platform support on the load cells, that 75 gram drift will most likely multiply.)

[jbb]

Some thoughts:
1) What's the drift as a % of full-scale?  You can't expect too much of low-cost components.
2) A full-bridge might be more resilient to whatever is causing it.
3) A 3% change in output for a 10 K temperature swing is a lot.  Are you putting the whole circuit or just the load cell in the low temperature environment?

[tecman]

The cell is a half bridge.  Did you "complete" the bridge with resistors ?  That will require a differential amp to interface to an A/D.  If you are just using one leg and measuring the resistance you will see a large change due to the tempco of the resistive element material.

paul

[Kleinstein]

Cheap load cells may not be long time stable. So for use as a scale one might have to do a periodic zero.
Especially plastics parts (insulation / sealing) can change with temperature and humidity. A 2 µV error is also possible due to thermal EMF, if not careful.

So one possible mechanism is that during the cold time the plastic incorporated some extra humidity. This would show up as a kind of dirft / relaxation after temperature changes. Also thermal mismatch (spring - glue - DMS itself) can cause internal stress and thus trigger relaxations.

If the spring itself is steel too, magnetism can have an influence too.

[ebclr]

You get what you paid for