I am building a differential amplifier to boost a 0-5mV (delta) signal off a load cell wheatstone bridge.

In simulation, the design works here (falstad): enter image description here

(for those who have not used falstad circuit simulator, the "2.53V" line are just voltage readings).

However, when factoring ambient noise, as observed in experimentation, the amplifier is essentially useless. I built the circuit in our lab yesterday to confirm this and unfortunately forgot to capture any scope screenshots. However, it appears that the noise off the wheatstone bridge is not from any direct source (eg 60Hz).

Apart from keeping wires short and ensuring consistent grounding, what else can I do to get rid of this noise? The very small delta of 5mV is observable on the scope when loading the load cell via an "up and down" translation on the scope. However, the channel itself has about +/-50mV of noise in the signal.

Any advice?

EDIT And I should finally add in case it is not clear, this is a DC circuit. Thus, frequency is ideally <1Hz.

  • \$\begingroup\$ Are you applying an ideal voltage source of 2.5V and getting a reading of 2.53V? That would make me suspicious of the simulation. \$\endgroup\$ – Elliot Alderson Apr 12 '19 at 14:11
  • \$\begingroup\$ There are 22 bit ADCs with 5 volt rails, and excellent CMRR, with 5 to 15 conversions per second. Use the differential inputs. \$\endgroup\$ – analogsystemsrf Apr 12 '19 at 14:17
  • \$\begingroup\$ @Elliot The absolute voltage off the wheatstone bridge is 2.528V (no stress) to 2.531V (fully stressed). The delta is approximately 5mV full-swing. \$\endgroup\$ – Matt Apr 12 '19 at 14:20
  • \$\begingroup\$ @analogsystemsrf Ah yes, results have been confirmed using the off-shelf HX711 ADC for strain gauges. The system our team is constructing is going to used for teaching purposes. It is an "exploded" ADC itself used to demonstrate individual components within off-shelf ADCs. Thus, this issue arises. \$\endgroup\$ – Matt Apr 12 '19 at 14:23
  • 1
    \$\begingroup\$ The circuit you have up is NOT an instrumentation amplifier. \$\endgroup\$ – Scott Seidman Apr 12 '19 at 15:04

The simple answer is to buy the right IC for the job. You need an instrumentation amplifier -- or at the very least, a specialized differential amplifier on a chip, with both the positive and negative inputs having the same gain.

Assuming your 60Hz noise is uniformly present on both inputs, this would be common mode noise. Thus, you need a circuit with a good Common Mode Rejection Ratio (CMRR). Your circuit does not have that, by design, as you have different gains for the positive and negative inputs. Even if the inputs had the same gain, it is still fairly difficult to construct circuits with good CMRR's because of tolerances on resistors, but it is very easy to buy them.

I suggest you confirm this for yourself in sim, by putting a 60Hz voltage source on both inputs.

| improve this answer | |
  • \$\begingroup\$ Makes sense. I have not personally dealt with small signal (<100mV) system design. Would the TI INA821 be an appropriate selection for this design? Instrumentation amplifier with a high >100dB CMRR. \$\endgroup\$ – Matt Apr 12 '19 at 18:59
  • \$\begingroup\$ @Matt probably more than good enough, if you can work w/ the SOIC package. Pick up a 2.5V precision voltage reference, as well, for the reference pin. \$\endgroup\$ – Scott Seidman Apr 12 '19 at 20:42

A real instrumentation Op. Amp. is really the simplest option.

If you have to use general purpose amplifiers (which, by the way, you did not mention) the best option would be to implement an instrumentation amplifier with two stages, with the gain divided among them. That way it would be easier to compensate for DC offset at the inputs.

| improve this answer | |
  • \$\begingroup\$ I placed an order for a proper 2-stage instrumentation amplifier. I would like to know, what is your idea for a strategy of DC-offset compensation at the inputs? Because once it arrives, that will be an issue. \$\endgroup\$ – Matt Apr 13 '19 at 1:50
  • \$\begingroup\$ Great. Please kindly provide a full schematic with the INA part number if you still need advice when implementing it. The compensation I meant was referring to the second option, if you had decided to implement an instrumentation amplifier using conventional Amp. Ops. \$\endgroup\$ – vangelo Apr 13 '19 at 16:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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