Active Electrode Circuit Design

biomurphbiomurph Brooklyn, NY
in Electrodes
Hey, 
Starting a thread here about the best way to go about designing active electrode circuitry. From what I understand, placing an active (op amp or other) device in very close physical proximity to the contact site improves noise reduction and allows for a dry contact between the electrode sensor (gold, silver, Ag-Cl, etc) and the subject scalp.
My basic reference for this is the active electrode circuit published by Olimex


I've used these, and they perform quite well with OpenBCI (5V differential between AVDD and AVSS to power).
The basic shape of the thing has 
TVS and Diode suppression on both ends
Power supply decoupling and zener protection on the V input
Some impedance added to the input and output
High Frequency rejection filter on the op amp

but beyond that, it is just a unity gain follower.
Of course, the specs of the op amp are going to play a very important role:
high impedance inputs
high CMRR
low equivalent input noise at low frequencies

My question is, what about the basic circuit, besides the op amp selection ('tho that could enter this discussion) stands to be improved? Or discarded?


«1

Comments

  • wjcroftwjcroft Mount Shasta, CA
    edited January 2015
    Joel, found a really cool paper from Imec.

    S    ection 3.3.1 in this paper I posted on the Dry Electrode thread, they are using a modified     OpenEEG     circuit. And this is Imec no less, one of the leaders in EEG tech.

    http://www.mdpi.com/1424-8220/14/12/23758/htm

    The schematic is there as well as the small circular board layout. They put the tiny circuit board on the back of their EPDM rubber combs. Dang, this looks pretty easy!

    Here's a high rez image of their circuit board you can zoom into. Open in another tab in your browser, then zoom in.

    http://www.mdpi.com/sensors/sensors-14-23758/article_deploy/html/images/sensors-14-23758f4.png

    The TI chip on that tiny board is, http://www.ti.com/product/tlc2272

    It has very good EMI immunity as well, eliminating extra components,

    http://www.ti.com/lit/an/sbot014/sbot014.pdf

    ---

    A few more:

    Here's an old paper on BioSemi's actives,


    And one from KnightLab, UCB,


    UCSD Body Sensor Network conference 2009


    ---

  • wjcroftwjcroft Mount Shasta, CA
    edited January 2015
    More complete board schematic is on Pedro Ortega's page,

    http://users.dcc.uchile.cl/~peortega/ae/

    I see @jarek had a hand in this as well. Jarek do you have any comments on the circuit board that Imec is using (image link previous post), or the TI 2272 op amp? This all looks pretty sweet and doable without custom ICs. The Imec design seems more compact than the Olimex.

  • biomurphbiomurph Brooklyn, NY
    The Olimex AE design uses TL062 op Amp. There is a graph of equivalent input noise on page 11 of the DS
    Appears that the input noise is down at 85nV per root Hz.

    The TLC227 that you link to says on page 9 that the equivalent input noise is 50nV per root Hz.

    The CMRR are about the same, Input resistance for both is the same 10^12 ohm.
  • jarekjarek
    William,

    yes, I spent a long time researching AE after I built my first EEG device (OpenEEG). Looks like Imec's circuit is very similar to my first prototype.

    I am not an electrical engineer, so I can't comment here on any electrical properties. But those AE I built worked A LOT better than passive electrodes used with my OpenEEG device (old version I soldered myself, without SMT) which required shielded wires.

    Since then however I found that other (commercial) amplifiers offered quite comparable (to what I measured) results using just passive electrodes (even without skip preparation, placed on forehead under a headband).

    So I am not using AE in my training. They are bulkier (harder to setup) and very sensitive to movement.

    Also I think the AE I built may have broken eventually. So the improvement (proposed by Joerg and now sold by Olimex) may be well justified. As I recall its primary purpose was ESD protection.

    I remember there was one more idea quite recently which I found absolutely fascinating - capacitive AE (it was discussed on openeeg email list). But I am not sure that is doable. Some people (like Stefan Jung) spent some time on that with no luck.
  • biomurphbiomurph Brooklyn, NY
    I did some research into capacitive (non-contact) electrodes. Here's some papers 
    That fellow, Chi, is all over this stuff.
    I made some that use the PCB soldermask as a capacitive connection to the skin, and they work well for ECG. I was able to see some interesting EEG signals, but they were very susceptible to noise... The circuit he's using is very interesting. It appears that the entire thing is 'floating' in a capacitive bubble. Not sure how it works.

  • wjcroftwjcroft Mount Shasta, CA
    Joel, Jarek, thanks.

    Jarek, I absolutely agree that modern amps like the ADS1299 produce excellent data from passive sensors. Whether used with paste, gel, or saline solution pads.

    Where all the consumer device direction seems to be going these days are dry + active sensors. The dry part is almost a requirement if you want to have a headset that you don like a cap. And since dry sensors have a more tenuous and varying impedance connection to the skin, they are frequently paired with some kind of active circuitry for best signal quality.

    Over on this other thread I posted some photos of the Imec EPDM conductive rubber comb sensors. That their active board mounts on the back of.

    http://openbci.com/forum/index.php?p=/discussion/138/buying-dry-active-electrodes#latest

    Chip will be receiving soon some Ag-AgCl small passive comb sensors. (Links are on that page.) It will be interesting to see how those perform purely passively. And then to imagine or try out some kind of active augmentation if it seems beneficial.

    The Imec paper cited above seems to conclude that the rubber comb must be used with the active, otherwise they don't get the results similar to wet passive sensors. Maybe this is just a characteristic of these comb style sensors, I don't know.

    The rubber combs I'm sure will be more comfortable than pointy metal combs pressed against the skin. But the comb approach seems a popular one, being used in a number of commercial designs.

    Then again there are the dry active polymer sensors used in the new Emotiv Insight. This is not a comb at all, just a smooth rounded bump. But their headset is designed to be more or less "slid" into the hair as you would a comb. Not pressed on from outside like more typical headsets.

    William
  • wjcroftwjcroft Mount Shasta, CA
    Looking more closely at the g.tec actives,

    http://www.gtec.at/Products/Electrodes-and-Sensors/g.Electrodes-Specs-Features

    Notice how all their actives just use TWO leads(!) [Touchproof connector has two contacts.] This supplies both power going to tiny active board -- AND must also be the return for the amplified signal. In other words, they must be doing some kind of modulation on top of the DC to get that information back to amp.

    @jfrey , do you have any idea what they are doing here, since you have these amps in your lab. Karl @kzurn is testing an active board of his FRI design, but I think that is using 4 leads.

    William
  • jfreyjfrey
    edited February 2015
    I do not know how they do it -- I have yet to dismantle a broken electrode to see what's inside -- but it also startled me a bit so I searched the Internet and found some references about similar 2-wires amplification:

    http://www.ncbi.nlm.nih.gov/pubmed/16485759

    I'm bad at electronics so I don't know what it's worth...
  • wjcroftwjcroft Mount Shasta, CA
    Jeremy, nice find. Seems likely that g.tec is using something like this.

    Found the pdf file as well. Cool.

    William
  • biomurphbiomurph Brooklyn, NY
    That is really cool. Looks like they are providing a constant current to the active element that then changes voltage according to the input signal (?) 
    I'm still learning this stuff!
  • wjcroftwjcroft Mount Shasta, CA
    Joel, here's my impression:

    The active sensor tiny board is supplied with a constant voltage over the 2 wire DC supply pair. THEN the sensor board adjusts it's current draw based on an translation / amplification of the microvolts measured.

    Meanwhile, back at the power supply, the changes in current draw are translated back into voltage changes, which are then relayed to the ADC / amp.

    A clever dance that results in 2 less wires. But significantly more complexity at the amp side.

    If you look at the way g.tec did their g.SAHARA active dry combs, they have a separate box (g.SAHARAbox) that does this translation. The output of that box goes into their amp which can then be either passive or active sensors.

    William


  • jfreyjfrey
    Yes, you do need additional "boxes" on top of the g.tec amplifier, which is troublesome when you want to use their electrodes -- the g.saharabox or g.gammabox is twice the size of the OpenBCI board. At least, if you use a "regular" number of cable, it should be easier to plug active electrodes into other system...
  • wjcroftwjcroft Mount Shasta, CA
    With any active system, some type of adapter / power supply board is necessary between the sensors and the amp. Here is the image again of the Imec 4 wire system adapter, from the Dry Sensor thread.

    Admittedly, the Imec board is large because their use of the 16 channels of touchproof connectors. One possible additional benefit of the two wire current modulation, is that it would be impervious to wire motion artifacts. As current flow is not affected the same way that passive sensor wires are.
  • jfreyjfrey
    You'd need a power supply, of course, but then it would not require much electronic parts besides a battery, no? Like, if the current draw is not too important, could the OpenBCI board be directly attached to a "regular" active electrode using some of the arduino/chipkit pins?
  • biomurphbiomurph Brooklyn, NY
    The OpenBCI Boards (8bit and 32bit) break out AVDD and AVSS (+2.5V and -2.5v respectively) and GND. They are on the input bus at the 'top' of the board (input side). Each supply has two pins available to tap.
    This was done to provide power for active electrodes if desired. 
    We have used the Olimex active electrodes with great success by powering them this way. 

    OpenBCI will likely no be able to use a two wire electrode as described above. 

    The 8bit board has pins broken out from the ATmega328. They run at 5V, and they are on the digital side of the star ground, and not recommended for use in any analog application.

    The 32bit board has pins broken out from the PIC32. They run at 3.3V, and they are on the digital side of the star ground. not recommended to include in anything on the analog side.

    Then again, it's all open source, so go for it!
  • jfreyjfrey
    edited March 2015
    Seems promising regarding active electrodes :)

    For extra analog sensors, why is it not recommended to use  5V/3.3 with the analog pin of the microcontrollers? I have one pulse sensor that awaits to be plugged!
  • biomurphbiomurph Brooklyn, NY
    Oh, when I said it's not a good idea to use the digital power pins, I meant for powering active electrodes that would send signals to the ADS1299 inputs. the analog powersupply (+2.5/-2.5) is super quiet and much better for powering active electrodes.

    For extra analog sensors, like the Pulse Sensor, it would be best to power with the digital power supply, and then use the analog input on the micro (ATmega or PIC32) to read in the analog values. I think you will have enough time to read the analog pin, and include that value in the data stream as an auxiliary value.

    Then again, It would be fun to try plugging the Pulse Sensor into the analog power and send the pulse signal to one of the ADS1299 inputs. 


  • wjcroftwjcroft Mount Shasta, CA
    edited March 2015
    Here's another type of 'active' electronics I've seen in the past. Several manufacturers are using this: Cognionics, Nexus and ANT amps.. It's called active shielding. Not just 'passive' style shielding where the shield is tied to ground. But they put some type of signal on the shield (the Bias?). In the case of Nexus and Cognionics, they actually sell passive systems (no active amplification at the sensor) -- and get results comparable to the much more expensive active sensor approach(!) Cognionics has packages where they offer both options: passive sensors + active shielding; or active sensors + active shielding.

    Here is the ANT page describing how that works, including before and after spectrograms showing ZERO mains noise:


    Key benefits of active shielding:
    * Movement of electrode cables do not interfere with EEG recording.
    * Good signals, even when your impedances are high.
    * 50/60 Hz environmental noise is shielded
    * No need for expensive and fragile pre-amplifiers at the cap.

    The picture below shows the clear reduction of 50Hz between unshielded vs. shielded recordings. With active shielding the feedback signal from the amplifier on the shield results in 'zero-capacitance' between core and shield. This makes the wires (thus whole cap) quite insusceptible to outside noise, e.g., caused by mains interference or movement of the cables.

    I'm looking around on the web now, trying to see if there are any papers describing this active shielding approach. Maybe it's just as simple as putting the bias signal on the shield. Or some variation of that, such as the Bias signal reduced in amplitude.
  • jfreyjfrey
    In the months before the OpenBCI board arrived I did some literature research about active electrodes and this is what I got for active shielding/guarding:

    Matsuzaka, Y., Ichihara, T., Abe, T., & Mushiake, H. (2012). Bio-amplifier with Driven Shield Inputs to Reduce Electrical Noise and its Application to Laboratory Teaching of Electrophysiology. J Undergrad Neurosci Educ., 10(2), 118–124. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/3598091

    Gargiulo, G., Bifulco, P., Calvo, R. A., Cesarelli, M., Jin, C., & van Schaik, A. (2008). A mobile EEG system with dry electrodes. In 2008 IEEE Biomedical Circuits and Systems Conference (pp. 273–276). IEEE. doi:10.1109/BIOCAS.2008.4696927

    Gargiulo, G., Bifulco, P., Calvo, R. A., Romano, M., Ruffo, M., & Shephard, R. (2011). Giga-ohm high impedance FET input amplifiers for dry electrode biosensor circuits and systems. In K. Iniewski (Ed.), Integrated Microsystems: Electronics, Photonics, and Biotechnology (pp. 165–194). CRC press. Retrieved from http://www.researchgate.net/publication/255994293_Giga-ohm_high_impedance_FET_input_amplifiers_for_dry_electrode_biosensor_circuits_and_systems

    Usakli, A. B. (2010). Improvement of EEG signal acquisition: an electrical aspect for state of the art of front end. Computational Intelligence and Neuroscience, 2010, 630649. doi:10.1155/2010/630649

    It is something I *also* planned to try out, but then I spent more time on the software than on the hardware. It seems to require "only" a bit of soldering and to choose the right cable (coax?).

  • wjcroftwjcroft Mount Shasta, CA
    Jeremy, very cool. Here's the full text on your first reference,

    http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3598091/

    It looks like the "driven shield" is a feature of the Burr Brown instrumentation amp they chose, I'll do some more looking into that part number. This idea is also mentioned in the Fraden text (below).

    For the first stage of amplification, we chose Burr-Brown’s INA116. This
    integrated circuit (IC) is a differential amplifier: i.e., it takes
    inputs from the reference and the indifferent electrodes (pins 6 and 3,
    respectively) and amplifies their difference, thus subtracting out the
    noise common to both inputs. A notable feature of this IC is the “driven
    shield” inputs: i.e., it holds the shield of the input coaxial cable at
    the same voltage as the electrodes connected to the input through the
    buffered guard drive pins (pins 2 and 4 for negative input pin, pins 5
    and 7 for positive input pin). As a result, the capacitance between the
    electrode and the shield is cancelled, thus preventing the electrostatic
    interference through the capacitive coupling between them (Fraden, 2003).

    This lines up exactly with what ANT said on their page about capacitance cancellation(!)

    William



  • wjcroftwjcroft Mount Shasta, CA
    edited October 2016
    Here is the Driven Shield schematic from Fraden's book:

    https://drive.google.com/file/d/0B8Z-FI2dBPKKU3hzX3lXc09yNW8/view?usp=sharing

    And the text describing the figure. Preceding text mentions active shielding board traces; then that is generalized to driving the sensor coax shield:

    The so-called driven shield is also highly effective. Here, the input circuit is surrounded by a conductive trace that is connected to a low- impedance point at the same potential as the input. The guard absorbs the leakage from other points on the board, drastically reducing currents that may reach the input terminal. To be completely effective, there should be guard rings on both sides of the printed circuit board. As an example, an amplifier is shown with a guard ring, driven by a relatively low impedance of the amplifier’s inverting input.

    Figure 5.4b shows a voltage follower connected to the inverting input of an amplifier. The follower drives the shield of the cable, thus reducing the cable capacitance, the leakage and spurious voltages resulting from cable flexing. A small capacitance at the follower’s noninverting input improves its stability.

    More background info,

    https://www.google.com/search?q=driven shield

    http://en.wikipedia.org/wiki/Driven_guard

  • jfreyjfrey
    From what I understood, compared to active electrodes, with active shielding the cumbersomeness (ie the complexity of the electronics) shifts from the electrodes to a circuit that is positioned right next to the amplifier (minus correctly craft the shielded wires).

    PS: this page tells a different story: http://www.biosemi.com/faq/shielding vs active electrodes.htm

    The thing is, I'm not entirely sure we need *that* badly those "active things". One of the strengths of a wireless board as lightweight as OpenBCI comes from the fact that it could (should?) be positioned right next to the head, as with the spiderclaw prototypes. As you wrote, active shielding prevent noise to arise when cables moves, but with cables cut at the perfect size, there is almost no "dangling", hence mush less noise that what you'd get with passive electrodes and "regular" systems. And to prevent external noise... why coudn't we shield the *headset* instead, as a whole or just the parts that cover the electrodes cables? Yes, I *do* think about a foil hat when I write that; dibs on the first foil-EEG-hat :D (seriously!)



  • jfreyjfrey
    PS: I also this document to be useful to comprehend how (active) shielding works, and how *not* to do it: http://www.analog.com/static/imported-files/application_notes/41727248AN_347.pdf
  • biomurphbiomurph Brooklyn, NY
    i agree with your sentiment jfrey. The whole idea of the OpenBCI SpiderClaw is that it will keep all of the wires and connections tidy so that there is 'no' or 'little' movement artifact from wires.

    with the accelerating development of 3D printable materials with some usable ohm-cm values, shielding, and potentially 'wires' and electrodes could be printed right along with the SpiderClaw. 


  • wjcroftwjcroft Mount Shasta, CA
    Recently found this company in Taiwan making active electrode leads, I'm in the process of getting some more info from them, such as spec sheets, interface requirements, etc. Since they are in China, possibly could be low priced.

    http://www.nvk.com.tw/eeg-lead-wires/active-eeg-electrodes.html

    image

    William

  • jamakerjamaker
    William, I was wondering if you found out more about those nvk active electrodes? Will they work with the openBCI?
  • wjcroftwjcroft Mount Shasta, CA
    Jamaker, hi.

    I did contact NVK in March 2016. Unfortunately they informed me that this active electrode assembly is NOT a product. But an example of the OEM work that they do for customers. In other words it was designed by the customer and manufactured by NVK. 

    I just sent another email to NVK today. Asking if we could purchase through the company who designed the assembly. Will let you know what I find out.

    William

  • jamakerjamaker
    Ok, thanks. There are no other ready-made active electrodes available that are somehow compatible with OpenBCI?
  • wjcroftwjcroft Mount Shasta, CA
    I believe Joel @biomurph or someone in the lab did experiment with the Olimex AE's with OpenBCI; and got that to work. But needs 5v. And form factor is a bit clunky.


  • nekrodezynfekatornekrodezynfekator Poznan, Poland
    So, has anyone tried to use DIY dry, active electrodes with OpenBCI??
Sign In or Register to comment.

© Vanilla Keystone Theme 2025