Using multiple strain gauges for digital scale

[jBernard]

Hi all, New user here, althought i THOUGHT i had an account here for a while. must have just been trolling around for the time being


I've been working on a project lately and running into some hiccups on the latest revision. Basically we are trying to reverse engineer a digital scale. And have made 1 prototype using a load cell (beam type) with a 4 wire interface. This was easy to hook up into an INA125P instrument amplifier and works great.
However, they are very bulky, so we are looking at using a much thinner style 3 wire strain gauge sensor like this for our next version:
https://www.sparkfun.com/products/10245

Very simple in design, and easy to use by itself, but we are trying to use 4 of them (1 for each foot) like in this scale system:
http://www.amazon.com/American-Weigh-ONYX-5K-BK-11-Pound-0-1-Ounce/dp/B001S0ZJUM

Thats where the hangup on how it's supposed to be wired is. As much info as I have gathered, the 4 need to be wired up to make a complete wheatstone bridge. By tying some of the 4 wires together, then taking + and - readings off the others.
The best thread I've found so far regarding wiring 4 of the sensors is this one, but still unclear on how it's done as it seems some of the sensors have to be wired to give opposite readings as the others so they don't cancel each other out? (that was a horrible explanation!)

http://www.nerdkits.com/forum/thread/900/

Also unsure if that feed goes into 1 opamp, or 2.

I'm very frustrated at this point, as i know the device is simple, and we already have a working prototype with a similar device, just the lack of documentation on using multiple of these gauges is becoming tiresome. You would think there would be tons of info out there and tutorials. As digital scales, even with 4 sensors, are found everywhere.

Any help is greatly appreciated. My electronics knowledge is definitely mid pack beginner so if I could find a hookup diagram that would be awesome.

Thanks for the help and look forward to reading thru more of the forums here and learning a lot.
Taylor

[ejeffrey]

I think each three-wire unit is two strain gauges with supply lines at the top and bottom and a sense wire in the middle.  You can figure out which is which with an ohmmeter.  You make a wheatstone bridge out of them by wiring the supply leads in anti-parallel, then connecting the sense leads to your instrumentation amplifier.  The output voltage will be proportional to the sum of the load on the two cells.  To use four gauges you can either use two amplifiers and sum the result, or just wire pairs of sensors in parallel.

[jBernard]

still trying to get this working.
had a bit of help from ejeffrey but hung up on using multiple sensors.
cant figure out how to wire it into the inst amp on the - sense input.
tried doing this with 2 sensors and there's no output from the amp showing any voltage fluctuations, so obviously its wrong.



any ideas?

[mjkuwp]

I think those sparkfun sensors are half-bridge each.

black and white are the high and low, I don't recall the polarity.
red = sense


To make a single reading you would wire two of those sensors in parallel and leave one unloaded.  Amplify the difference between the two red lines.

You can also hook two sensors like this;
black to white
white to black

read the difference between the red wires and if you load both load cells you get double the reading.


I * think *... to use four of these:

Connect the four sensors in a circle (series) using black and white wires.  Two of the (leftover) red wires will be + excitation and - excitation.

The other two red wires will be sense + and sense -

Would have to draw it out for it to all make sense and then puzzle out which polarity to use for each section.  You only need one amplifier.

[Rerouter]

if they are half bridged, you only need to use a single op amp in a summing configuration, all the sense lines off of the measuring element get summed, and pick your favorite of the other side, as it is only being used as a convenient ratiometric divider (2 resistor divider whos output is a ratio of the supply voltage,)

you then scale the gain to 1/4, and you have the whole group acting as one, add weight to one corner, it increases, to the opposing corner and it increases some more, etc, just measuring the difference between the sensors would only give you a way to measure the corner that a weight is loaded on,

[jBernard]

thanks for the reply guys. much appreciated.
ive included a new drawing from best i understand the replies.

something like this?

[retrolefty]

I don't think wiring the the gauge outputs together is proper, instead the they should each go through a series resistor to form an adding function input to the opamp?

[mjkuwp]

darnit, realize my response above does not make sense anymore the more I think about it   - sorry!... but I thought I saw it somewhere before.  A person had reverse-engineered the connections on a bathroom scale and found it went to one amplifier.    search in this forum does not seem to work well so I often have a hard time tracking down things I have seen before.

check out this link

http://electronics.stackexchange.com/questions/92946/combination-of-four-half-bridge-load-cells

[Rerouter]

More like the attachment

[jBernard]

this any closer? hopefully i read it correctly!

[mjkuwp]

To me the circuit in the stackexchange link I mentioned above seems correct.   You just need to make sure each load cell is wired with the correct polarity so that the load combines in the correct way.  You  could say that the circuit is re-combining the resistors to make a full bridge.

That circuit with 4 100k resistors looks like one for 4 sensors with low impedance output.  These half-bridge strain gauges are not low-impedance output.... however I am open to learning something.

Any of these ideas could be modeled in a circuit simulation program and it could help show what works and what doesn't.

[jBernard]

mjkuwup,
the stackexchange schematic is a bit confusing to me, mainly how load cells are expressed in them to begin with. so i'm missing how to correctly hook them up.
could you give some easy directions on how to interpret that drawing?
thanks!

[Andy Watson]

Do you have a multimeter with which you can identify the leads from the strain gauges? (We could take a educated guess but then, if it were wrong, you would be even more confused.)

[retrolefty]

Do you have a multimeter with which you can identify the leads from the strain gauges? (We could take a educated guess but then, if it were wrong, you would be even more confused.)

No, it would measure out on an ohm-meter just like a pot with the wiper at it's center position, easy.

[kfitch42]

jBernard,

In your most recent schematic I think you connected the 25K feedback resistor to the wrong place. It needs to be connected to the 'far side' of all the 100K resistors. If I am reading it right, you connected the 25K in between load cell 3 and its 100K output resistor.

That circuit with 4 100k resistors looks like one for 4 sensors with low impedance output.  These half-bridge strain gauges are not low-impedance output.

The spec sheet for the spark fun load sensors lists "Output Resistance" as 1000+/-20 Ohms.  So, you MIGHT be OK if you just wire the 4 together into the OpAmp input, since they effectively already have a 1K(+/- 2%) resistor. Just change your feedback to 250Ohm (assuming I did my mental math right), and everything else stays the same. Or, you can add your own output resistors, and just remember that there is an extra 1Kohm on each output.

P.S. I am a software guy who likes to pretend he knows something about electronics... so feel free to ignore my ramblings.

[Andy Watson]

Do you have a multimeter with which you can identify the leads from the strain gauges? (We could take a educated guess but then, if it were wrong, you would be even more confused.)

No, it would measure out on an ohm-meter just like a pot with the wiper at it's center position, easy.

Unless I've miss it, identifying the centre of the gauge would be a step forward and that is what I'm suggesting. If I were to guess I would say that the white wire would be the centre, but the "schematics" presented so far indicate that the OP doesn't have a clue which wires are which.

[jBernard]

hey,
we've already decided which wires are which. blk - gnd , wht - ref V , red - sense/wiper

resistance readings give me 1K ohm between blk/wht
500 ohm between red/wht
500 ohm between blk/red

[Andy Watson]

hey,
we've already decided which wires are which. blk - gnd , wht - ref V , red - sense/wiper

resistance readings give me 1K ohm between blk/wht
500 ohm between red/wht
500 ohm between blk/red

Not what I would have guessed - but that's what happens when you make assumptions For two gauges, you need to connect black-1 to ground, white-1 to ref/bridge supply, red-1 to amp +ve input, white-2 to ground, black-2 to ref, red-2 to amplifier -ve input.  That should get something sensible happening.

[Rerouter]

ok, had a look over your the actual datasheet, and that 25K resistor can go, your device's gain is set between 2 pins, (the way its draw didnt match up in my head)

The numbers refer to the pin numbers of the chips, and there is a %error due to the 100K resistors but this is easy to fix in software, otherwise bump the values up another decade and the error will be 1/10th,

to adjust the gain put a resistor between pins 8 and 9, according to page 10 of this datasheet,
http://www.ti.com/lit/ds/symlink/ina125.pdf

[jBernard]

Andy,
just wired your version up with 2 sensors.
works great!
i'm instead of taking the sensor voltage off of the V+ output driving the amp, im taking it off the Vref (pin 4) of the amp chip. i saw this in another post and they say it will be more stable this way. its an odd voltage (2.206V or something similar) but voltage doesnt matter to me because i will calibrate the sensor on weight ratio alone.

now any ideas on how to implement the other 2?
thanks again for all the help guys, this is progressing nicely. im honestly surprised theres not tons of tutorials and info out there on these, figuring how many digital scales are in production! so maybe my post will be able to help people in the future.
taylor

[mjkuwp]

mjkuwup,
the stackexchange schematic is a bit confusing to me, mainly how load cells are expressed in them to begin with. so i'm missing how to correctly hook them up.
could you give some easy directions on how to interpret that drawing?
thanks!

I should warn you I am not 100% sure about that circuit.  I have some of these sparkfun load cells and am using them daily in a project.. but I am using two of them, loading only one.  The idea of four load cells is something I had considered but not tested yet.

That drawing has node numbers 1-12.

123 is one load cell, 2 is the red wire
456 is another, 5 is the red wire
789 is another, 8 is the red wire
10 11 12 is the final load cell, 11 is the red wire


-------

I made a circuitlab circuit - what do you all think of this configuration.

link to a picture
https://www.circuitlab.com/circuit/h394q8/screenshot/1024x768/

link to the circuitlab circuit.  shows loaded cells having 999 ohm resistance on one end, 1001 ohm on the other end. - to simulate them being loaded.

https://www.circuitlab.com/circuit/h394q8/4-half-bridge-sparkfun-load-cells/

[jBernard]

ok, had a look over your the actual datasheet, and that 25K resistor can go, your device's gain is set between 2 pins, (the way its draw didnt match up in my head)

The numbers refer to the pin numbers of the chips, and there is a %error due to the 100K resistors but this is easy to fix in software, otherwise bump the values up another decade and the error will be 1/10th,

to adjust the gain put a resistor between pins 8 and 9, according to page 10 of this datasheet,
http://www.ti.com/lit/ds/symlink/ina125.pdf

trying to hook this idea up and having a few issues.
in your schematic, it looks like youve got several resistors (the 10k and 30k's) tied between a few of the pins. these seem to be in the datasheet on the internal circuitry of the chip? or do i need to add these in on my breadboard.
looks like you have the sense ( - ) input going from the V+ rail, thru a 100k, into pin 6? is this correct?
when i have it hooked up on all the sensors summed together and coming in on pin7 like the drawing, and the 100k from hot rail to pin 6 like you had drawn, i get .60 V, on output, with no detection when i press on a sensor. with that circuitry removed and pin 6 empty, i get 2.555V.

also note that my input voltage (from an Electric Imp, which is basically an arduino) ouputs 3.27 (3.3V). so thats my input to the INA125 as well as the sensors voltage input.

let me know what you think. thanks!
taylor

[Rerouter]

Treat pin 6 as a zero adjustment and stick a pot in there if its not outputting a nice 0V,

yes the 10 and 30K resistors are internal circuitry

as for no detection, you will likely need to fit a gain resistor as per the datasheet, as the signal of you pressing it may be quite small in comparison of the maximum deflection of the strain gauge

[jBernard]

thanks for the reply.
ive updated my drawing here, and i'm still using a 60ohm for gain as i did on my old sensor setup, which should be 1000x gain and giving a large amplification.


with pin 6 unhooked i'm getting about .300V from it, but still no readings from any of the sensors when pressed. Still just 2.550V, which i think is reference voltage from the INA125?

let me know if you spot something i'm goofing up on.
thanks for the continued help

taylor

[fcb]

INA125 differential inputs are on pins 6&7.

You need the top two strain gauges into pin6 (or 7) and the bottom two into pin7 (or 6).