Hello and thanks beforehand.

I'm designing a circuit to measure 4 wheatstone bridges using 4 Analog Devices AD8422 in-amps. Now, i'm feeding the AD8422 output to a 3.3V ADC, so i'm offsetting the output 1.65V in order to be able to measure the full scale output of the bridge. However i'm aware that tiny differences in the bridges, opamps and resistors will require slightly different offsets for every bridge (i'm expecting something like +-0.05V would do).

The AD8422 requires the reference to be fed from a low impedance source

AD8422 reference

Now, to create the 1.65V reference, i'm using a 3.3V source, a resistive voltage divider, and a Texas instrument's OPA333 as a voltage follower. However, to individually adjust every AD8422, I would need 4 of those circuits with potentiometers in the voltage divider, each one feeding into one AD8422. The boards that have the AD8422 are very small and integrating another opamp (even a SOT23 one like the OPA333) is pretty difficult (and expensive if we want to expand the system), so, here comes the question:

How would you apply small variations in a reference voltage without altering the low-impedance nature of the output?

I'm aware that the potentiometer aproach would not work, as you would need precise knowledge of the intensity going into the REF pin and pretty high resistances (2k+) which would be like a voltage divider.

Thanks for your help!


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ Can you use CAPACITORS to establish the low impedance? \$\endgroup\$ – analogsystemsrf Aug 17 '18 at 9:46

However i'm aware that tiny differences in the bridges, opamps and resistors will require slightly different offsets for every bridge (i'm expecting something like +-0.05V would do)

I wouldn't bother trying to do this in hardware; I'd just compensate the ADC reading digitally to overcome the slight discrepancies in offsets. I'd produce one solid 1.65 volt reference voltage and feed this to all AD8422 InAmps.

After all, if there are resistor variations that produce voltage offsets, those variations will also produce gain variations that are likely to need compensating in software.

  • \$\begingroup\$ That's the easiest way to do it but I wanted to avoid it because, if this works, we will have many more of those modules which we would plug into a board with a microcontroller in it. Having the calibration on the PCB would greatly simplify the software, as we might have something like 20 boards plugged to the same microcontroller, and keeping track of the software calibration of each one is complicated. If this proves difficult though we'll end up doing it. Thanks for your answer!! \$\endgroup\$ – Guzmanus Aug 17 '18 at 10:14
  • 1
    \$\begingroup\$ Software is there to add functionality and performance beyond the wildest dreams of a simple resistor. How would you keep track of the hardware calibration (given that you state you would need to keep track of the soft calibration)? How will you do gain calibration if not in software? Step up and face reality. \$\endgroup\$ – Andy aka Aug 17 '18 at 10:17

If you take a look at the datasheet there's figure 55:

enter image description here

showing us how the REF input is used.

If the REF input would have an additional impedance in series (like a reference voltage coming straight from a resistive divider, without a buffer, then this impedance adds to the 10 kOhm resistor present on the chip.

Then accuracy suffers and offsets might be introduced. This sort of defeats the purpose of using this Precision amplifier.

You can indeed compromise and tune the offsets as you propose but performance (accuracy) will suffer. What I think is the right way to go forward is to determine how much inaccuracy you can tolerate. From that you should be able to determine how much series resistance can be allowed in your design.

  • \$\begingroup\$ You're right, I had checked that diagram but i'm still not experienced enough to be able to undestand how would it work. As i understand it, A3 is working as an adder, and thus changing the resistance that A3 sees on the REF pin would change the factors of the different voltages added, thus maybe not adding exactly and decreasing accuracy, isn't it? We're working with accuracies of 1-2% so anything over 100ohms would already screw with that. Thanks for your insight! \$\endgroup\$ – Guzmanus Aug 17 '18 at 10:16

Something like this:


simulate this circuit – Schematic created using CircuitLab

You will need one op-amp per ADC to get the best accuracy. The voltages are different, you want them to be ~zero ohms impedance compared to 10K. Depending on your accuracy requirements, it might be better to have one divider and pot per op-amp or maybe you can put several pots in parallel and share the three resistors, which will result in worse tempco typically and a slight interaction.

R1 and R2 (in parallel with the pot element(s)) and R3 set up the limits of your Vref adjustment. R2 is used because the tempco and tolerance of pot elements is typically vastly inferior to that of a precision resistor. So you'd have R2 << R4. And you'll pick an op-amp with low enough Ib that the the source impedance does not cause excessive error or drift.

Calculating the values, worst case range and drift I will leave to you.

Having a single voltage and b*ggering with it is less accurate but might be "good enough" if you're looking for less accuracy. This involves 3 parts per ADC plus two resistors and an op-amp to generate the 1.65V.


simulate this circuit

The added 50 ohms or so will affect the gain accuracy by less than 1%. If you are adjusting each one then you could trim that out, however the two trims will interact slightly so a couple iterations might be required.

Either way you're looking at 9-20 parts or so. Accuracy and power consumption will vary with choice of resistances and op-amps. For example, the circuit #1 with paralleled pots, dual op amps would be 9 parts. With individual dividers and dual op amps 18 parts, Circuit #2 would be 15 parts.

  • \$\begingroup\$ Thanks a lot for your help! I'll try to make a schematic using your advice and check it. Prototypes are cheap at least. \$\endgroup\$ – Guzmanus Aug 20 '18 at 8:12

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.