I am doing a half bridge strain gauge application and it is connect to INA125P and with gain about 185 with Rg 330 ohm.below is my theoretical circuit diagram, my circuit is base on this picture to connect. enter image description here

This circuit will connect to arduino uno for take the result. below is the result from the adruino. enter image description here

base on the graph above above to saw that the voltage output is very unstable, this cause my unable to take accurate result.

My question is: -the output voltage will fluctuate like this is that because of the unstable Wheatstone bridge? i am using 120 ohm resistor and with 5% tolerance. -is that any way to let the output voltage more stable? -i using this application for measuring low weight object (0 - 10g) -is that anyway to increase the sensitivity of the application to able to measure object with 10mg.

Thank you

  • \$\begingroup\$ What is the range of those values on the left scale; are they raw ADC values? Why are you not using at least 1% resistors? How much gain is that amplifier giving? \$\endgroup\$
    – pjc50
    Commented Apr 6, 2015 at 10:23
  • \$\begingroup\$ Try using the Vref output from the amplifier to provide the bridge voltage \$\endgroup\$
    – Chu
    Commented Apr 6, 2015 at 10:54
  • \$\begingroup\$ Isn't it more common to put the 2 strain gauges together in one leg of the bridge to make a 'half-bridge' \$\endgroup\$
    – brhans
    Commented Apr 6, 2015 at 15:05
  • \$\begingroup\$ What is the specified SNR of an arduino's ADC? Whay are there not any decoupling caps on the IA? Is the bridge sharing the same supply as the IA and arduino? What does the layout look like? Are you operating ratiometrically? \$\endgroup\$
    – Andy aka
    Commented Apr 6, 2015 at 16:05

1 Answer 1


You are seeing noise. Exactly where the noise is coming from depends on your circuit layout and what the source of the noise is. (Sorry, bad definition but you know what I mean I hope). It may be difficult, or even impossible to eliminate, depending. Diagnosing it could be harder still, without trial and error or suitably advanced equipment and technique.

Instead, try taking a number of values and calculating the average. Up to a point, the more samples you take, the more accurate the reading. I had a similar problem where I took 10,000 samples each time I wanted to get a good reading, and took the average. This change increased the accuracy by a factor of about 1000! However this will not necessarily solve all your problems. The final accuracy will depend, for example, on the distribution of the noise that is present.

In your code, remember to convert integers to floating point variables before these calculations.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.