Input Current Bias and CMRR degradation

thewhiteflagthewhiteflag Canada
in Ganglion
From the Ganglion design files it appears that the positive and negative inputs of the instrumentation amplifiers are tied to I_BIAS (the midpoint voltage of the INAs) via a 330k resistor.  

I'm trying to understand why this was done, because in my simulations the CMRR is much better without them.

If the electrodes were capacitively coupled this would make sense as this would act as the source for the input bias current of the INA. However in my understanding most electrodes are not capacitively coupled, or am I wrong in this assumption?

Comments

  • bpworkbpwork work
    Yes that would be expected from the sims and from actual performance, particularly if you're (not at all unreasonably) modelling the electrodes as being of unequal impedance, and the higher the impedance the worse it'll be.  Since there are a lot of ways to model & measure various kinds of CMRR, could you post a schematic or text-file showing one or more sims?  

    As for the purpose of the 330K resistors, I've wondered that too.  Unfortunately I've been too swamped with other things and I never did get very far using my board, but I've thought that if/when I do, one of the changes I would make is to increase the value of those resistors by probably a couple orders of magnitude.  Myself, I don't know much about the complex impedance of various EEG electrode-skin interfaces, but the instrumentation amp bias current is only a few hundred picoamps.

          -- Bruce P.

  • wjcroftwjcroft Mount Shasta, CA
    Mentioning Joel Murphy @biomurph.
  • thewhiteflagthewhiteflag Canada
    edited October 2018
    After further investigation I am still stumped by this. I'm starting the think that these bias resistors are redundant, although I'm really not sure if my understanding is correct.

    My understanding is as follows:

    First consider a circuit without a DRL or Ground electrode. Just the two electrodes to the INA. This will be unstable because input bias currents will lead to a charge build up or depletion on the capacitors C1 and C2, cause the inputs to go out of the common mode range for the INA.


    https://ibb.co/eaOqP9

    Adding a “ground electrode” which is driven via an opamp solves this problem. The inputs now stay in the common mode range for the INA as the bias currents have a resistive path to ground. Bonus is that it reduces the effective impedance from the person to ground, lowering the common mode voltage as well.

    https://ibb.co/cgBzHU

    Now looking at the ganglion design, it also adds what I will call the bias circuit. 330k resistors tied to ground just before the input of the INA. If there were no ground electrode this would act as the input bias return path, but we already have that. Secondly it adds resistance in parallel with the INA input, reducing the input impedance and leading to worse quality.


    https://ibb.co/dqrMWp

    Maybe I’m oversimplifying, but I’d really like to know where my understanding is wrong, or if it really is redundant.


  • BillhBillh
    Does this address the problem at all?



  • thewhiteflagthewhiteflag Canada
    edited October 2018
    Not really, as to my understanding the electrodes aren't ac coupled.

    Here are the sims that I ran. (*Note I used a different INA than the ganglion and I omitted the integrator circuit to simplify)

    The electrode impedances are 10k and the other 20k, I’ve got the ground electrode at 10k too. The CM input signal is 400mVpp @ 60Hz.



    When I run the sim with the first setup I get about 20mVpp of the 60Hz noise at the output.

    Using the following setup (removing R4 & R5) I get only 0.002mVpp of the 60Hz noise! Quite a difference!



  • wjcroftwjcroft Mount Shasta, CA
    @thewhiteflag, hi. I edited your 'image' references back to plain https links. Such links don't work with the 'image' rich text button; that requires an URL that ends in png or jpg.

    Regards, William
  • BillhBillh
    Can you simulate any crosstalk in the electrode lead wires with the sim?
  • thewhiteflagthewhiteflag Canada
    ^Thanks William

    @Billh Why do you think a crosstalk simulation would be insightful? 
  • BillhBillh
    edited October 2018
    The crosstalk between wires can cause capacitive coupling. How much coupling between the EEG leads and wires there is in practice would be important in the reasons for the bias tying in the design.

  • thewhiteflagthewhiteflag Canada
    But crosstalk is usually due to coupling between signals, not a capacitance series with the signal itself, which is what I previously mentioned as ac coupling. 
  • thewhiteflagthewhiteflag Canada
    After further research I have come to my own conclusions to this matter. I'm posting them here so that hopefully this can help anyone else who has the same question in the future.

    I believe that my assumption that most electrodes are not capacitively coupled was wrong. Gold cup electrodes such as the ones OpenBCI sells in the store are capacitively coupled (for the most part, they will never be perfectly capacitive or resistive). The 330k resistors are there to accommodate these electrodes such that the amplifiers do not saturate. This is a trade-off, worse performance when you are using silver-silver chloride electrodes, but the ability to use Pt or Au electrodes. 
Sign In or Register to comment.

© Vanilla Keystone Theme 2026