skin-impedance measurement of active vs. passive electrodes

Comments

• wjcroft Mount Shasta, CA

  January 2017 edited June 2019

  I would suspect the buffering in the amplifier co-located at the electrode. I'm not sure the Texas Instruments technique for passive electrode impedance measurement is applicable to actives, which are considered superior to passives. But at a huge cost increment.

  
  

  My other hunch is that the value '4K' is suspiciously 1K down from the magic number '5K', which has been touted as the "research standard" for impedance measurements. (From a couple decades in the past, with older amplifiers.) In other words, the buffer amplifier electronics output might be juggled to produce a value that people 'expect' to see for a stellar skin impedance value. It might not reflect reality.
  
  This paper from EGI,
  
  http://wwe.eeginfo.com/research/ElectrodeImpedance.pdf
  
  Makes clear that 40 K is still excellent for almost all applications. As long as the passive amp has high input impedance, which the ADS1299 does. 1 GigaOhm.
  
  William
  
  
• markosullivan

  January 2017 edited January 2017

  Thanks for the reply William.
  Yes, the paper by Ferree is very interesting and I know that even under 100 kOhm is still quite usable due to the 1 GOhm input impedance of the ADS1299.
  
  I'm more just interested in trying to see the benefits of active over passive in reality. Both Signals look almost identical. Only difference I have found so far is the skin impedance measurement. Even the unfiltered 50 Hz noise is roughly the same amplitude in passive vs active.
  
  Should I be observing improved signal quality with active electrodes (without amplification of the signal, just high pass filter and buffering)?
  
  What characteristics should theoretically be better? SNR? 50Hz interference? Motion artefacts? DC offset? Because at the moment I don't see any significant benefit with the active signal I'm receiving.
  
  Thanks
• markosullivan

  January 2017 edited January 2017

  Also, on a side note..

  In Ferree's paper, the equation on page 5 for potential difference of the signal received by the amplifier..

  I seem to come out with much lower signal losses than what Ferree states.

  Ignoring the amp impedance mismatch as he does and ignoring the last (3rd) term as I assume (not sure) he does.. You're left with:

  
  

  (Va - Vb) = ((V1 - V2)/2) * (2 - (Z1 + Z2)/Zin)

  Taking Z1 = Z2 = 50 kOhm and Zin - 200 MOhm, I get the following result..

  (Va - Vb) = ((V1 - V2)/2) * (2 - (50000 + 50000)/200000000) = 0.99975.

  
  

  So a max signal loss of 0.00025%... not 0.025% as in he paper.

  
  

  Any idea?
• wjcroft Mount Shasta, CA

  January 2017

  Mark, hi again.

  
  

  re: benefits of actives over passive. It appears generally that these full cap active systems and amplifiers cost thousands of dollars and are only affordable to clinical and research groups. In that arena, especially in noisy operating rooms or labs, the increased noise immunity of the actives may be the most prominent benefit.

  
  

  It sounds like you are just outputting microvolts from your actives, the same signal levels as passives. Is it possible that some active systems output higher signal levels that are internally converted back down inside the amp? This higher signal level on the cables would likely provide better immunity to noise and movement. Some actives also use shielded cables.

  
  

  Since I have no experience myself with active systems, I would defer to your manufacturer. Can you say what you are using?

  
  

  Actually OpenBCI itself might perform better in terms of noise factors with something like the Electro-cap / gel 10-20 caps. The wet skin interface always is more stable because the electrolyte layer is less pressure sensitive.

  
  

  re: math in Ferree's paper. I believe he is still at EGI or UOregon, you could write him there.

  
  

  Regards,

  
  

  William
• markosullivan

  January 2017

  Yes, same signal levels as passive. The method of transmitting at high voltage over the wire and reducing back to original voltage at the front-end isn't used.

  The cap simply implements a highpass filter and unity gain buffer.

  The active cap is from Cognionics.

  
  

  Below is a image of a few plots from Matlab of the dry active channel vs dry passive channel.

  Taken from Fp1 with reference at A1 and ground at A2.

  
  

  The sections at 11-15 secs, 42-46 secs, 70-74 secs, 92-96 secs are blocks of rapid eye blinks just to give an idea of the amplitude you're looking at.

  There are some slight differences, but probably not enough to warrant the extra costs involved with active.

  
  

  P.s. thanks, I actually found a paper by Ferree on his UOregon site discussing the maths. :-)
  

  
  
• markosullivan

  January 2017

  This is a better view of the FFT data.

  
  
• wjcroft Mount Shasta, CA

  January 2017

  Mark, thanks for those graphs!
  
  Nice to see that OpenBCI compares decently with Cognionics dry headset. Do you have an idea of the cost of the Cognionics gear?
  
  This recent paper by Jeremy Frey compares a g.tec amp with OpenBCI. Both amps connected to the same active cap. So it's not an active / passive comparison like yours. But does offer some insights on amp comparison techniques.
  
  https://hal.inria.fr/hal-01328427/document
  
  William
  
  
• markosullivan

  January 2017

  That's not exactly comparing Cog vs OpenBCI. Just comparing two openBCI channels, one with a standard dry sensor, one with the dry sensor connected to an active cap.

  Thanks for that paper. Nice to see openBCI competing well!

  
  

  Just to clear up the skin-impedance measurement in my head.

  Is the 31.2 Hz, a 6nA signal? Passed through the REF to the skin and then picked up and measured on the "recording" channels? Or am I getting mixed up?

  Looking at the openBCI GUI, when impedance checking, I see something like this:

  
  

  
  

  
  

  Using the equation from EEGHacker page (http://eeghacker.blogspot.ie/2014/04/openbci-measuring-electrode-impedance.html)

  
  

  (V*sqrt(2)) / I = ( 76.4u * 1.41 ) / 6n =  18 kOhm.

  
  

  Is this value, the skin-electrode impedance PLUS the 5 kOhm series resistor? I would ahve thought the OpenBCI GUI would display it without the 5 kOhm?

  
  

  Mark