Correcting load cell drift in software

Hi all,

I've searched and found several discussions here and on other sites about correcting load cell drift, but they seem to center on hardware i.e. the reference, or tempco compensation in hardware, or AC excitation of the cell vs DC.

I have load cells on the way from ebay to play with, and some software ready to go - my application is a long term (months) static load (container of liquid - beer actually) to be weighed, with the remaining quantity of liquid to be read using the load cell. A change in liquid (tap opened) will give a sudden drop in weight which is many times greater than the drift i have read would be typical, so should be easily detectable.

My question is - am i thinking this is simpler than it really is, or is it not talked about much simply because is it so simple?

What's wrong with watching the load cell reading, and if it changes by less than a threshold ignore it, but when it changes by more than the threshold we take the weight change as valid?

I have an idea commercial cheap kitchen scales do something similar based on observations - if you very slowly pour more water onto them they often won't read any change - i can get 2 or 3X the weight onto my scales without a change in reading if i do it very very slowly.


Have you considered a pseudo calibration mode? Set it up so when you press a button it zero’s out. What you would do is when the load cells are empty like when changing kegs press the calibration button. I would assume you will use a lot more kegs then the drift time.
Good Luck & Have Fun!

I think its simple, after each large discrete change the software should remember the weight, and over longer periods recalibrate to this value regularly if no large change.

It won't handle evaporative loss of course.

It's worth having this as a switchable mode - the cells might be fine as they are, drift may not be a problem, or it may happen initially and then tail off considerably.