Avoid cross talk between recording electrodes

Question

I am trying to compare two record signals coming from two INA amplifiers simultaneously. Both amplifiers are actually recording the same signal, just with different recording electrodes. We are interested in the performance comparison of the two electrode types. We are aware of input current biasing and have designed our circuitry thus far in the manner as shown below. However, the two different recording electrode types (WE vs RE) sit naturally at quite a high DC potential difference between each other. Meaning, WE electrodes in solution might be 500-1000 mV above the RE-electrodes in solution due to their Nernst potentials in our recording medium. Note: Both electrode types (Types: WE and RE, 2 each) are in the same solution at the same time.

The circuit below records... the problem is, we do not want a leakage current path connecting the WE electrodes to the RE electrodes, as this is essentially forming a battery between the two. With lower resistor values, we even saw corrosion of one recording electrode and deposition onto the other electrode type. Note: The 10M ohm resistors connecting to AGND are needed to bias the amplifiers to common mode input, but they also create a 20M ohm path between the various electrode types where roughly 100nA of current flows, killing the low leakage current feature of the INAs.

Are there any ideas on how to avoid this or go around this?

Edit for more information:

This is only a snapshot of the analog front end of the board. The power comes from an external supply and is converted to the required voltage supplies (iso+, iso-, AGND) with an isolating DC/DC converter.These power rails are common to the entire analog end.

The outputs of the INAs go to an ADC which then heads to an isolated SPI comms to communicate back to the rest of the system on Digital Ground. The electrodes are different metal types in various conducting solutions and/or acids. One could picture here ion/chemical sensing through potential measurements.

Answer 1

Note: The 10M ohm resistors connecting to AGND are needed to bias the amplifiers to common mode input, but they also create a 20M ohm path between the various electrode types where roughly 100nA of current flows, killing the low leakage current feature of the INAs.

There should be no need to put these 10MΩ resistors from each electrode to the circuit GND to bias the amplifiers to common mode. The solution is to remove these resistors, and find an alternative way to achieve the same outcome.

The solution should be simple: connect the conducting fluid to the AGND of your circuit, and ensure the power rails of the op-amps (INAs) have sufficient range (both positive and negative) to keep all electrode voltages within the INA121 common-mode range at all times. The datasheet of the INA121 states that the maximum supply voltage is 36V, so you could apply power supply rails of, for example, +12V & -12V, or as high as +/-18V; from what you have described, however, it would seem +/-5V should be sufficient.

The schematic below shows this idea; the connection to the conducting fluid could be via a different electrode, or perhaps to the container or pipe, provided it is sufficiently conductive.

Note that I did not have a model of the INA121, the closest I could find in the standard LTspice library is the AD SSM2141, the schematic does not show the feedback components.