I need to measure tensile strains between 0 and 0.01%, with an accuracy between 1 to 10% within that range.

It seems like all systems that read strain gauges rely on a Wheatstone bridge to amplify small variations around the balanced point of the bridge using a diff amp (which makes sense) so I've made a simple model in excel which gives the output voltage of the bridge VS strain (equations checked online thrice), then the amplification of the differential voltage with half-range offset for a single-ended ADC.

The model basically says I need to measure 0 to 0.7mV with an accuracy of 7 to 70uV, which screams for a gain of 1000. But this means any matched resistor of the bridge can only vary by about 0.1 Ohm before the amplifier saturates...

I can do that for the two dummy resistors (1k in my case) using a matched pair IC accurate to 0.01%, but the best I can do for the strain gauge "mirror" is to trim for each board for a few fractions of ohm with a pot until the bridge outputs zero - but I am sure this is not reliable since the temperature alone will certainly make the strain gauge vary by one or more 0.1's Ohm... I thought of using the active-dummy method and replace that "mirror" resistor with another identical strain gauge roughly at the same temperature, but I do not know how much I can expect another unit to be different in its unstrained resistance to be able to trim it to balance. I don't even know if that will guarantee the conditioner will not saturate.

Is this feasible? Is there better? How is it usually done? I don't think that so few people need to measure these sorts of strain, given that we have to stay clear of the yield strength.

  • 1
    \$\begingroup\$ Strain gauges are supposed to be linear aren't they? So putting tuning pot in series with each strain gauge should let you zero them relative to one another? Am I missing something? But why would you make the additional strain gauges dummies? That sounds like a waste. Put them on the DUT as well. Heck make the entire wheatstone bridge nothing but strain gauges. michsci.com/what-is-a-strain-gauge \$\endgroup\$
    – DKNguyen
    Jan 26, 2022 at 6:16
  • \$\begingroup\$ There are ADCs for that, Analog Devices AD7730 is already a mature product, but still a way to go. There are newer alternatives, even more accurate and faster. \$\endgroup\$ Jan 26, 2022 at 10:41

1 Answer 1


You use a Wheatstone bridge to null out the unchanging part of the resistance, and then an amplifier to make the remaining part measurable by a conventional ADC. With some sufficiently sensitive ADCs like HX711, you don't even need the amplifier, but still need the bridge.

It's straightforward to trim a Wheatstone bridge to one part in 1000, either with a potentiometer, or more stably by selecting resistors. The trick is to use a trimmer that's much lower resistance than the bridge itself. Say you have four nominally 1k resistors forming the bridge, a 10 ohm pot on the reference mid point will have plenty of resolution for setting the balance.

Balance is not a problem per se. If your bridge is going to be stable enough to make measurements over time, then it will be stable enough to balance adequately.

When setting up a strain gauge bridge, the obvious way to do it is to use one strain gauge and three stable resistors. This leaves you open to thermal coefficient mismatches.

A better way is to make the measurement leg from two strain gauges, one unstressed as a resistor. Tempcos will now match in both measurement and reference legs.

An even better way is to actually use the 'compensation' strain gauge, strained in the opposite direction. This doubles your bridge output for no increase in cost. It may also encourage you to build a more balanced mechanical arrangement, putting one gauge on each side of your spring member.

Why stop there? You can double the bridge output yet again by using two more strain gauges for the 'reference' leg as well, though this of course costs two more gauges. This is how the standard 'four foot' kitchen or bathroom type scale is connected, so that the force registered by each foot is summed in the Wheatstone bridge to make the whole scale relatively independent of the load position on the top.

  • 1
    \$\begingroup\$ Note that with an ADC like the hx711, you generally have enough dynamic range to null out offsets numerically in your code. \$\endgroup\$
    – Kartman
    Jan 26, 2022 at 9:23
  • \$\begingroup\$ Thanks for your answer. I am actually a bit worried about imbalances post-calibration - in my model, 1°C difference between the two strain gauges (active and dummy) is the maximum possible without saturating -after all the dummy one cannot be loaded so cannot be in the same thermal node. This is assuming that the temp sensitivity of the strain gauges' resistance is what they call the K factor, which I set to 0.01%/K. Without the dummy strain gauge, it basically says the temperature of the strain gauge cannot vary by more than 1°C compared to the calibration, which is impossible in our case. \$\endgroup\$ Jan 31, 2022 at 19:46
  • \$\begingroup\$ @MisterMystère Why not arrange the dummy one to be loaded oppositely to the active one? \$\endgroup\$
    – Neil_UK
    Jan 31, 2022 at 19:50
  • \$\begingroup\$ Because it only works in bending, we're measuring tensile \$\endgroup\$ Jan 31, 2022 at 20:10

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.