Load cell strangeness

[eTimesV]

I have a cheapo 150kg WH-C100 load cell (https://www.aliexpress.com/item/32891808125.html) which was working fine until I tried to measure the force produced by a small hydraulic ram system I made and it overloaded it.  It stopped reading any weight after that.  I took it apart and hooked up the load cell element/sensor itself to another load cell project I happened to have lying around with an HX711 chip and was able to see that both pulling and compressive forces were measured by the 150kg "broken" load cell.  So I am pretty sure the load cell sensor piece itself works.  The other strange thing is that when I connect the 150kg load cell back into the original device it can read compressive (negative) force values, but when I pull on the cell it reads zero no matter how hard I pull.  I tried changing the E- E+ S- S+ wires around to see if it would let it read pulling force but to no avail.  I visually inspected the PCB and can't see anything out of the ordinary.

FWIW, I believe the sensor glued to the 150kg load cell is something like one of these:
https://www.aliexpress.com/item/1005001564881864.html full bridge Wheatstone type
https://www.aliexpress.com/item/1005005417981134.html not full bridge?

Some pics: the device opened up, detail of the E-E+S-S+ connection for the four wires coming from the 150kg load cell, test of load cell component using my HX711 project, making a little breakout board for the 4 wires between the device and the load cell to make it easy to try different combinations of the connections.

Additionally, there is some kind of 'admin' mode where if you push both the buttons (power and units) it boots into a different state, maybe for calibration or something but I have no info on how to use that mode.

Is the device broken, or is there maybe something I can fix? I don't get how it broke...I'd understand if the aluminum load cell piece had cracked or stopped functioning due to the overload, but I see no reason why the PCB would be affected by an overload.  It's possible the overload is simply a coincidence and the load cell stopped functioning for some other reason entirely.

[wraper]

Most likely load cell has been damaged and now has very large zero offset.

[eTimesV]

Thanks for the fastest reply in the known universe.
  When I attach the aluminum load cell to my HX711 project, as pictured, it starts to register both positive and negative small forces as I push with my hands, 1kg or even less.  Wouldn't that indicate there is no such zero offset?

[N2IXK]

Disconnect the load cell from the board, and check the resistance across E+ and E-, and compare to the resistance across S+ and S-.

If all 4 strain gages are still good, the resistances should be very close on both pairs.  If one gage has failed, you will have 2 different readings.

[eTimesV]

ah, sorry, I forgot to mention I had measured the wires before too.  The resistance across blk-red and grn-wht were 1kohm (very precisely, like plus/minus 1 ohm).   red-wht, red-grn, blk-wht, blk-grn were all about 750ohms.

[eTimesV]

when you say the resistance across E- E+ and S- S+...you mean the wires from the load cell or?  Not the solder pads on the PCB labelled E-, E+, S-, S+ right?

[eTimesV]

Here are some pics of the display.  The load cell is clamped in a vise.  Compressing it causes a negative 'weight' to be measured or at least shown on the display.  By taring the device while under compressive load then relaxing the vise I can get positive readings.  "O-Ld" was an overload condition, accompanied by a loud beeping, when I clamped the vise too much.  Interestingly it went into overload (compressive) at far less than 150kg...more like 40kg or so on the display.
The other display images show the 'admin' mode when I push both buttons during power on and/or with power already on.  Not sure what their function is although one display which I didn't get a picture of since it flashed only briefly said "Cali"...calibrate I suppose.

[eTimesV]

two more pics

[magic]

You can test these things with a sensitive DMM - full scale output voltage is several mV given a few volts of excitation. Even 10μV resolution meter sees steps of 0.2% full scale or thereabouts.

No load offset voltage is usually specified as a very low percentage of full load output, this is easily verified.

My guess: you have permanently deformed it, resulting in change of offset voltage which the "bad" device somehow isn't able (or trying) to cope with.

[wraper]

Thanks for the fastest reply in the known universe.
  When I attach the aluminum load cell to my HX711 project, as pictured, it starts to register both positive and negative small forces as I push with my hands, 1kg or even less.  Wouldn't that indicate there is no such zero offset?

Zero offset means large voltage offset from zero when load sell is not loaded. It does not mean that load cell won't react to load change. Normally it's not worse than a few % of full scale. For example if full scale is 5mV, then output voltage in non loaded state should be not higher than 0.1-0.2mV

[eTimesV]

Zero offset means large voltage offset from zero when load sell is not loaded. It does not mean that load cell won't react to load change. Normally it's not worse than a few % of full scale. For example if full scale is 5mV, then output voltage in non loaded state should be not higher than 0.1-0.2mV

  So...if I hook the load cell back up to the device, turn it on, do not apply any force to the load cell, then measure the voltage between the black and red wires (the excitation voltage as far as I understand it) the voltage between the green and white wires should not be more than ` few % of that?
If so, I did it an measured:
2.394V blk-red
0.008V grn-wht no load
0.007V grn-wht compression (maybe 15-20kgf?)
0.009V grn-wht tension

Does that tell you anything?
 

[eTimesV]

You can test these things with a sensitive DMM - full scale output voltage is several mV given a few volts of excitation. Even 10μV resolution meter sees steps of 0.2% full scale or thereabouts.

No load offset voltage is usually specified as a very low percentage of full load output, this is easily verified.

My guess: you have permanently deformed it, resulting in change of offset voltage which the "bad" device somehow isn't able (or trying) to cope with.

Given your hypothesis, I figured at this point I might as well try squashing the aluminum load cell piece in a vise to see if that maybe counteracts any deformation incurred when I overloaded it with the hydraulic load.
IT DID SOMETHING!
I squashed it in a vise with some, but not a huge amount of force.  I attached it back to the device, turned it on with no load, then put it in the reversed (tensioning) clamp to make a tensile load.  I got up to 17kg tension before it overloaded and started beeping!  Very encouraging. 
  The aluminum piece has 3mm wide slots cut into both sides in the middle.  I went back to squash it even more in the vise, not sure how much force but at least 100-150kg if I had to guess, to the point where the slots reduced to ~2.41mm.  I did it a few times, trying to center the piece so each slot showed equal reduction in width (if it was at all off center one slot showed more width than the other).  I hesitated to apply more force than I did for fear of the aluminum plate with those slots cut in it, buckling up or down.  Maybe I'll try again, sandwiching the aluminum piece in something so it can't buckle.
  I hooked the reversed clamp back up and applied tension to the load cell again and got all the way up to 48kg.  I think the clamp maxed out at that point since the load cell didn't start beeping to indicate an overload state.
  I don't know if the reading is accurate anymore.  I will try to find some known 10-20-30-40-50kg weights to test it with.
Even if it doesn't really work anymore, I'm happy to have an idea of where the problem lay.  I am surprised such a tiny deformation of the aluminum, probably sub-mm, could cause such a problem.
Oh, also when I redid the grn-wht voltage check with no load it now reads 0.003V.  Maybe I should take that reading between vise-squashing sessions to see if I can get it to 0?  Would that indicate I had compensated all the deformation then?

[wraper]

Why you are measuring such small voltage in V range and not mV? And if zero load voltage is really 0.008V = 8mV, then load cell is certainly broken.

[eTimesV]

forgot to add this pic

[eTimesV]

Why you are measuring such small voltage in V range and not mV? And if zero load voltage is really 0.008V = 8mV, then load cell is certainly broken.

holy cow! duh.  wow, no idea why I didn't switch it.  Brain glitch.

[magic]

  So...if I hook the load cell back up to the device, turn it on, do not apply any force to the load cell, then measure the voltage between the black and red wires (the excitation voltage as far as I understand it) the voltage between the green and white wires should not be more than ` few % of that?
If so, I did it an measured:
2.394V blk-red
0.008V grn-wht no load
0.007V grn-wht compression (maybe 15-20kgf?)
0.009V grn-wht tension

Right about excitation voltage.
Offset is the differential output voltage (between green and white) under no load, it should ideally be zero or close to zero in the real world.
Full scale output is the output voltage under maximum rated load, minus offset voltage if offset is not negligible.

Offset should be a few % of full scale. In your case full scale appears to be near 10mV and offset is 8mV. So it's damaged.

[eTimesV]

Offset should be a few % of full scale. In your case full scale appears to be near 10mV and offset is 8mV. So it's damaged.

Thanks for the clarifications.
As noted above, after squashing the aluminum load cell part in a vise, I got the offset down to 3mV and could now read at least ~50kg of tensile force.  I will try squashing the aluminum piece harder once I figure out a way of preventing it from buckling out of it's major plane.  Not sure if the end result will totally work but I think it's worth trying.

[wraper]

  So...if I hook the load cell back up to the device, turn it on, do not apply any force to the load cell, then measure the voltage between the black and red wires (the excitation voltage as far as I understand it) the voltage between the green and white wires should not be more than ` few % of that?
If so, I did it an measured:
2.394V blk-red
0.008V grn-wht no load
0.007V grn-wht compression (maybe 15-20kgf?)
0.009V grn-wht tension

Right about excitation voltage.
Offset is the differential output voltage (between green and white) under no load, it should ideally be zero or close to zero in the real world.
Full scale output is the output voltage under maximum rated load, minus offset voltage if offset is not negligible.

Offset should be a few % of full scale. In your case full scale appears to be near 10mV and offset is 8mV. So it's damaged.

With 2.4V excitation voltage full scale is unlikely to be more than 5mV, and quite likely half of that as 1mV/V load cells are much more common than 2mV/V. And 3mV/V load cells are extremely rare.

[magic]

As noted above, after squashing the aluminum load cell part in a vise, I got the offset down to 3mV and could now read at least ~50kg of tensile force.  I will try squashing the aluminum piece harder once I figure out a way of preventing it from buckling out of it's major plane.  Not sure if the end result will totally work but I think it's worth trying.

The usual solution to broken load cells is replacement, so you have noting to lose

The device which fails to read this cell probably has very narrow range of voltage it can measure, and offset limits the range in one direction. So you are gaining range by reducing offset.

Accuracy of the "repaired" cell is another matter; you can't truly bend a deformed metal back into exactly its original shape. If you want to play with it you could also try electronic means: add large resistance in parallel with one or two opposite arms of the bridge. Thermal drift of the extra resistor may cause problems, though.

[N2IXK]

Yes, the wires from the load cell, when disconnected from the PCB.

[eTimesV]

Yes, the wires from the load cell, when disconnected from the PCB.

thanks for the reply.
I think at this point I've confirmed it's what "magic" suggested, that the sensor and/or aluminum was deformed.
I've partially deformed it back but squeezing in a vise but have yet to test the max tensile load it can now read.  I got up to about 48kg tensile load before I think my clamp maxed out, but anyway that was very encouraging.

[eTimesV]

As a followup, here are the final results of the repair, which by the way I'm super surprised actually seems to have worked.
I made a much more robust tensioning test jig to explore the upper load limit.
The meter can now read up to just over 126kg before overloading.  I probably could have put it back in the vise to try get it up to 150kg like it's spec'd but didn't feel it was worth the effort.  In the meantime I also ordered another one of the same meters to compare its function with the repaired one (I needed two meters anyway).
The final no-load load cell offset voltage was 2.8mV (was somewhere around 7-9mV before squeezing in vise)
The gap in the aluminum load cell part was:
3.04mm @126kg
2.92mm @0kg

~0.12mm, 120 microns, is a pretty tiny movement and near the center where the sensor is it must be a fraction of that.  Pretty amazing the simple resistive network of the load cell sensor can measure such a minute movement.  Or not, I guess, depending on how you look at it.

I didn't have a ton of weights to use to test it but luckily I had the base parts to a cnc machine I'm building which got me up to ~57kg.


Pics:
1,2 - test rig
3 - max tension I could apply before it beeped to complain of overload
4 - width of slot in aluminum under 126kgf
5 - " " " no load

[eTimesV]

Testing repaired meter compared to new one.
Meter without the piece of tape saying "repaired" is the new one.

Pics:
1 - zero load offset voltage 2.8mV (finally using the mV range )
2 - first test weight
3 - 25.60kg new meter
4 - 25.65kg repair meter
5 - 2nd weight
6 - 43.9kg new meter
7 - 44.0kg repair meter
8 - 3rd weight
9 - 56.55kg new meter
10 - 56.60kg repair meter

[magic]

Good like new

You could reduce testing time by half if you chain the two meters "in series".

[eTimesV]

That's a good point, although it would have been a little difficult to raise the bar I was hanging it/them from to give enough space to the floor.