LoadCell with ADS1232, not linear, not exactly same value

Other Hardware Sensors
1

Hi,

I try to make as lot of people a scale with cheap load cell (40-50Kg) from AE and an ADS1232 (gain 128, 3.3v) for reading. The construction is like that. Its a “half scale” (there is only half Kg on the load cells part), the other side is at the height of that part :

Screenshot_20251016_093713
Screenshot_20251016_0937131573×269 5.51 KB

First, i discovered that there is a way for Load Cell (no sign on mine), so one gave value going up, and the second going down, which make useless result.

I connect the 2 load cell in parallel as described in different solution.

So, i get value increasing when i put more load on it, which make me feel like it is basically working but the problem is the value is not accurate. Each time, i make average of value given by the ADS1232 :

0Kg    : -140200
8.8Kg  :  -23400 (13.27)
18.8Kg :  104200 (13)
28.8Kg :  233700 (12.98)
38.8Kg :  382100 (13.46)
48.8Kg :  518900 (13.50)
38.8Kg :  381100 (13.43)
28.8Kg :  242000 (13.27)
18.8Kg :  103800 (12.97) 
8.8Kg  :  -24100 (13.19)
0Kg    : -140300
58.8Kg :  585200 (12.33)
68.8Kg :  692400 (12.10)

So, at beginning, the ratio is more less 13.20 but if i put too much weight, it going down to 12, thats first problem. The second is the value is decreasing other time.

At 68.8Kg, it start at 692400, but few minutes after, it is 686000 and continu going down. So, it makes difference around 1-2Kg if a apply a ratio as 12.5, but if value allways change, there is no accuracy.

So, my questions are :

I know it is cheap load cell (< 5€) and industrial one cost hundred euros. I don’t need really good precision in fact but is there a way to get more precise result ? I can scale it to the maximum load and make some conversion after for example, applying a scale of 13 for result beetween 0Kg and 60Kg then a scale of 12 for result beetween 60Kg and more, but would like to know if there is other way. Until for a load > 60Kg, the value decrease other time, it’s hard to make something.

Is my construction not as good as it could be, for example using a unique 50Kg cell (load is maximum 110Kg) or using 4 cells ?

I don’t really get why the value is decreasing so much other time at > 60Kg, i can understand that the load is undercontraint and need to stabilize, but it’s should not take so much time.

Thanks in advance,

2

Ok, so … :smiley:

The problem is repartition of weigth. I don’t have good way to test it, i have a some wooden board (not rigid) to make equality beetween my construction and the other part. The weight was in the center of that wooden board. If i put the weight all around, it get good number (804400 - 13.75), it means there is some kind of deformation under my construction.

3

Is it two, three, four or six wire?

Show a schematic of how you have everything connected?

3.3v

You should use 5V

Are the header pins soldered on the ADS1232?

4

Its 4 wire, forgot to mention it. I can”t use 5V in my final project, i know it should be better

The pins are not soldered on the ADS1232, but its not dupont wire. I made a PCB with JST connector for the loadcell

5

Why not?
What Arduino are you using?

The pins are not soldered on the ADS1232

Whether you use pins or wires, they must be soldered, that is probably your main problem.

6

Hi, @froggycorp

Can you please post a copy of your circuit, a picture of a hand drawn circuit in jpg, png?
Hand drawn and photographed is perfectly acceptable.
Please include ALL hardware, power supplies, component names and pin labels.

Can you post some images of your project?
So we can see your component layout.

Thanks.. Tom.... :smiley: :+1: :coffee: :australia:

7

Should be wired like this, with all wires soldered or with header pins soldered to the ADS board

image
image743×442 118 KB

8

Have you calibrated the load cell?

9

you can add a non linear interpolation, assuming that the points measured are stable and do not drift. Have a look at my MultiMap code, then you can translate it to Kg or pounds right away.

GitHub - RobTillaart/MultiMap: Arduino library for fast non-linear mapping or interpolation of values

1 Like
10

So, i change the loadcell to one loadcell (now in the middle), and get better value, not changing and more linear converstion. But same at that previous 2 load configuration, sometimes there is a deviation. I think its not a big deal when you get a value every 1h because the deviation will end at a moment and i can see it because i get a value every 2s for testing.

@jim-p yep, it should be better to solder it, but its a pcb “homemade” for testing, so i will do it on the next one. I don’t think it makes a big différence about deviation, because i think it is more physical problem than wire problem. Actually i made gain to + and temperature compensation to GND, so i can’t test without another option. It’s a 1 face PCB, it’s hard to do a lot with.

About the 5v, its a good and big question. As my experience, there is more “module” working at 3.3v than at 5v. I’m using a RFM69 which run on 3.3v, so at the end, its better to stay on 3.3v. I’m using attiny85 because … i got some and not using them. I will try with PIC which are less expensive.

@TomGeorge

20251024_100151
20251024_1001511920×864 146 KB

20251024_100017
20251024_1000171800×4000 724 KB

20251024_100010
20251024_1000101800×4000 819 KB

@sonofcy yep, now its not that bad

@robtillaart will get a look, thanks

11

Forgot to mention that there is “huge” deviation depending on temperature, especially when the sun is on it. I took a look at some overpriced bee scale and i can see the same deviation.

So, maybe its because the bees go outside, but i don’t think around 4Kg of bee go outside at same time

12

In my experience you will achieve better accuracy, stability and precision at 5V but you are the boss.

Forgot to mention that there is “huge” deviation depending on temperature.

Well most loadcells are temperature sensitive, that's why you need to calibrate.