Parallel connection between two full-bridge load cells - Unstable readings

ericgibbs

Joined Jan 29, 2010
13,597
hi kaindub,
His sketch works just fine, I have it here, running on the bench.

His on bench hardware does not work when using my version of the sketch which works OK on my setup, it is also used by others.

Read post #14.

E
 

RPLaJeunesse

Joined Jul 29, 2018
183
Load cells are very lossy transducers, the output is in the low millivolts. Any stray electromagnetic noise could be readily picked up by those 30cm connections. Excitation pairs (+ & -) should be twisted together, Signal pairs should be twisted together and, for good practice, twist the pairs together. The goal is to minimize and loops in the wiring. Using shielded cable can also help. Cat-5 shielded cable is a great possibility, albeit with more wires than needed. Another noise mitigator is to add a filter capacitor across the signal lines at the HX711, for an example of how to do this look at https://cdn.sparkfun.com/assets/f/5/5/b/c/SparkFun_HX711_Load_Cell.pdf
 

ericgibbs

Joined Jan 29, 2010
13,597
hi RP,
If you are referring to my test setup, you can see that Signal and excitation wires are already twisted pairs, also there is a filter cap on the HX711 signal input lines.

The results I have posted shows that the setup works perfectly OK.

The TS has not posted any photo's of his set up, only a simple block layout.

BTW: Load Cells are not lossy, they have low signal output of a few 10's of millivolts.

E
 

Beau Schwabe

Joined Nov 7, 2019
100
We are using an HX711 with our own combiner circuit design for a 6 wire load cell (7 with shield) and 4 load cells on a 24/7 propane kiosk. The cable distance from the load cell to the combiner was at least 5 feet. The load cells are rated for 2500kg. Load cells are notorious for noise and must allow for an adequate settle time. One filter we have used on the "RAW" data that has really worked well is to create a "Peak detector" in software for both HIGH and LOW peaks. A conventional Peak detector only looks at the high peaks and has some auto decay to zero. We implemented an additional peak detector that does the same thing for low peaks with a "climb" towards zero. Combining the two peaks by summing them and then dividing by two produced surprisingly ideal (stable) results for our setup. Instead of using an Arduino, we used a PIC16F15323, but the principal is the same.
 

RPLaJeunesse

Joined Jul 29, 2018
183
Eric I was not referring to your setup, I was just pointing out some general principles for the benefit of the original poster. I find many non-engineers using Arduino devices see sensors that produce 4V swings with a 5V excitation, these often work fine with sloppy wiring. A load cell produces, as you said, millivolts of signal, and sloppy wiring causes issues. Good on you for having achieved a properly working setup, now let's get the OP there.

FWIW I consider any device with an output that is below 1% of input to be quite lossy. And yes, I do think of the excitation voltage as an input, because it will affect the output.
 

ericgibbs

Joined Jan 29, 2010
13,597
hi RP,:)

I think this is the basis of the TS's problem

clip:
Part of the problem is that they gave me these two load cells for this project, but they do not have the specifications anymore. I realise that that makes solving the issue a lot harder. I have ordered two new ones, but I don’t have them yet.

He was originally suspecting his Sketch, so I patched up a HX711 with UNO and a 5kG LC.
His Sketch ran OK, I suspect his LC's and/or wiring, asked for a photo shot, still waiting.

E
 
Top