instrumentation-amplifier
I'm having a hard time figuring out how the following circuit works.
I studied the theory behind instrumentation amplifiers and from what I learned they are used to measure a difference between two voltage inputs and amplify the signal.
Now in the following configuration I don't understand the roles of the three amplifiers (for example, why is A3 buffering Vref?).
I'm not looking for a solution, but just an explanation of how it works.
Why don't you get an ac signal source, attenuate it 500:1 with a 499 ohm resistor and a 1 ohm resistor and insert this into one grounded element of the bridge. DC offset introduced can be countered with another 1 ohm resistor in the bottom leg of the other grounded element.
Also shown is a method that doesn't break-into the bridge - it applies a high value resistor across one of the grounded bridge elements in order to simulate a shift in bridge output. This can be controlled with a fet fed with a squarewave as shown. It's even possible to use an opto-coupler in this type of application should the return line of the bridge be very sensitive.
Now you can inject (say) 1kHz at a known measurable amplitude and see what comes out from the instrumentation amplifier. If the AC gain is fine then you have to look at the dc issues that may be causing your problem BUT until you rid yourself of the doubts about ac gain you might keep going round in circles.
I don't see any obvious reasons here, but, you talked about low amplitude and high gain amplification and this is where good decoupling, correct grounding comes into account. It could be that your problem doesn't come from a schematic (or principle) level but from a layout problem.
If this is done on a breadborad and not a PCB, make sure that your decoupling caps are very close to every power pins of your amplifiers.
The observed coupling between the INA could come from the power lines or the ground line.
The added capacitor will decrease the impedance of the signal that enters the ADC, which is good. But the value chosen is HUGE and may present very high ESR. IF your ADC samples at 500 KHz for instance, your capacitors won't be very effective at that speed. This could explain the noise you see.
Another source of the noise at your ADC is the wiring. You have to ensure that the signal and the GND are very short between your board and the arduino.
Best Answer
Vref is produced by a simple potential divider consisting of R6 and R7. The problem with this simple approach is that altering the load current from the potential divider will alter its voltage.
If A3 were not present, and R6/R7 were connected directly to R3, this is exactly what would happen if the V- input changes, because the output of A1 will change and the feedback look will alter the current through R3 to make its - input match its + input. The same is true if the V+ input changes, where feedback through R5 would alter the current through R3.
So A3 provides a buffer that takes a minimal, and constant load from the potential divider and keeps Vref at a constant value regardless of the input signal. It will, however, still be subject to change if the supply voltage changes, which is why alternative methods such as Zener diodes or current sources are often used instead.