Load cells, how to and general questions

[darko31]

Hi everyone, I'm looking into making a automatic system for measuring weight. Since the max weight should go to 100kg, and the container for the material is relatively big, the plan is to use 4 load cells, amplifying their signal with op-amps and measuring each signal (still don't know whether I'll be using PLCs or MCUs, that will depend on the budget and desired reliability). Container and the platform are to be maximally balanced.

After measuring readings from every load cell, average of the values should be the real weight. Now, I presume there are couple of traps for young players, so is there anyone with some experience to point me in the right direction, and to confirm that this can work?

[f5r5e5d]

cross axis sensitivity can be a problem for accuracy with kinematically overconstrained load cells - but they can be made to work together if you have to

single tension S-beam cell holding up the container would be cheapest and most accurate if you can manage

metal foil gages may need few 1000x amplification, doing that with high DC accuracy isn't a job for a generic op amps

there are plenty of Delta-Sigma ADC chips for weigh cells now

[woodchips]

Ok, what does maximally balanced mean? Surely you should sum, not average, the weights from the four, or whatever, load cells?

What does cross axis sensitivity mean? I think I know what kinematically overconstrained means.

There are other solutions, a spring and an LVDT? Optical sensor? You don't say what accuracy you are looking for, or what overload and shock loading it has to cope with. Been many years since I have been involved with load sensors but then one part in 2000 was the best accuracy obtainable, is this better now? Note, accuracy, not resolution. And of course if the sensor had to cope with a 100% overload then that is half the accuracy gone straight away.

[PA0PBZ]

Is this a one-off project? If yes, get yourself a bath scale, most have 4 load cells (one in each leg) and of course the electronics so $20 could get you almost everything you need.

[darko31]

@f5r5e5d

Care to elaborate on the first sentence further? S-beams are available, it could be good solution. Regarding the ADCs, I could run into some problems with interfacing the ADCs and PLCs, a instrumental amp circuit should do the trick.

@woodchips

Maximally balanced as in container would be symmetrical and reverse cone shaped and material in it would be pretty much evenly spread. Brainfart moment, yeah, definitely sum of the load cells.
For the accuracy, I haven't yet spoke in much detail with the man who requested this system, but +-0.5kg or even 1kg should be fine.

@PA0PBZ

It's a potentially not an one off, but a prototype. A local owner of a workshop for farming machines and accessories is looking to make a automatic system for measuring and bagging of grainy materials like cattle food, grain, wheat and so on. I'm still a student, and this would be a nice opportunity for getting some hands on experience, and I get to play with some relatively expensive gear. The owner of the workshop is fully aware of this, and I'm grateful, so I'm looking to do a good job here, and not mess something up.

Regarding the overload shock, there should not be any, the bags will typically be 50kg, and the material would be transported via screw conveyor which won't overload the container.


 

[darko31]

And I've chosen the load cells because they are readily available, relatively cheap and mechanics for them is really simple. LVDTs would require a custom spring, same thing goes for opto sensors. Load cells and optos are more or less similar prices, so load cells do seem like optimal solution.

[edavid]

Just use an HX711 ADC, it's designed to connect directly to the load cell.

[Paul Moir]

I had a similar project a couple years ago, and here's some things I learnt from it:
1) Load cells are not scales, and a display is not a glorified voltmeter.  The scale does *significant* processing of the data coming from the load cell to compensate for noise and thermal drift.  Boxcar averaging, zero and stable load tracking are big things.  I would hate to have to try to implement this on a PLC.  The software is going to be likely more work than the hardware.  (Note:  your  accuracy requirements may save you here, but do have a look at the thermal drift of your load cells. )

Helpful reference:  http://www.analog.com/library/analogDialogue/archives/39-12/weigh_scale.html

2) I used three load cells so they would be somewhat evenly loaded.  For the same stability think of a squared triangle for the footprint, or if reduced stability is OK you're triangle can be somewhat inside the square.  This was important for 3)

3) For multiple load cells, "load cell summing boxes" are typically used.  These are usually just some decent trimmers in a box, so really they average the cells rather than sum them.  If I were to do the project again, I would simply use the A/D converter's input mux or multiple A/D converters to read each load cell, then sum in software.