ThinkPulse active electrodes, Q&A

124

Comments

  • Hi George3d6,
    indeed abnormal signal.
    I suggest you refer to tutorial video numer 4 "Cable management and connection guide" on the OpenBCI ThinkPulse documentation page. Here's a screenshot from it, all the connections are explained

    However, I guess that if you had it working at some point, you managed to do it successfully so maybe try looking at what might have changed since the last time it worked.
    If not powered, the ThinkPulse sensors do not measure any brain signal so what you would get is just noise.
    Best,
    Julien

  • hirohiro Japan
    edited May 16

    Ultracortex with ThinkPulse, need BIAS? small amplitude

    Hello all,

    I used Ultracortex with ThinkPulse. I followed the instruction is this page. https://docs.openbci.com/ThirdParty/ThinkPulse/ThinkPulse/
    I turned off BIAS and decrease the PGA Gain from 24 to 8. I tried to gather data during meditation. As a result, I got obviously poor waves like below.

    I doubted something clashed. I did impedance-check. Electrodes had around 5~50Ω impedance. Are these high? I checked other thread in this forum like this.
    https://openbci.com/forum/index.php?p=/discussion/comment/20337#Comment_20337
    I think my impedance were not significantly large.
    I also read this thread.
    https://openbci.com/forum/index.php?p=/discussion/comment/20338#Comment_20338
    I doubted my ear clips are disconnected but these are fine as a result of testing by multimeter. I strived to reduce environmental noise but it did not make better though I turned off light, removed AC cables and cleaned electrodes.
    Additionally, I read this thread and imitated Jason's setting.
    https://openbci.com/forum/index.php?p=/discussion/2869/thinkpulse-active-electrodes-q-a/p2
    I ADDED BIAS and got data during meditation like below.

    Even though the signal is weak. It seems that better result than data without BIAS. Thus, I have mainly two questions.

    1. Should I include BIAS though I ignore the instruction?
    2. Amplitude of my wave is low. Do you know why? (Although reducing noise contribute to make shape of wave clear, amplitude is not changed dramatically. Are some settings wrong?)

    Hiro

  • hirohiro Japan
    edited May 16

    Hi all,

    I forgot if I save data to microSD, amplitude would decrease. I said the similar issue in the past thread.
    https://openbci.com/forum/index.php?p=/discussion/comment/19540#Comment_19540
    I did a simple experience again to get data with PGA Gain setting to 2. Data was saved both of microSD and PC via wireless connection at the same time. I compared the amplitude between saved in microSD and PC. The amplitude of PC is roughly 12 times of microSD.(I just checked the value on GUI like below. The left is data on PC, The right is microSD.)

    In addition to, I calculated the difference of amplitude when the gain is 8, I found that the amplitude of PC was preciously 3 times of microSD. I believe that the amplitude decrease to (the value of current PGA gain)/24. 24 is the default value of the gain. So, if I would like to modify this gap, I should multiply current eeg data on microSD and the reciprocal number of it, isn't it? For example, the current PGA is 8, and thus, I could multiply my data and 24/8=3. Then I would obtain data has correct amplitude.

    Hiro

  • wjcroftwjcroft Mount Shasta, CA

    Hi @hiro,

    I merged your new thread into this existing thread on Thinkpulse Q&A. This way the developer, @julienConscious sees your new question.

    One remark I would have is that your images all show cross-contamination across all channels. In other words, all channels show same wave patterns. This is usually an issue with the reference electrode, since it is shared across all channels.

    William

  • hirohiro Japan

    One remark I would have is that your images all show cross-contamination across all channels. In other words, all channels show same wave patterns. This is usually an issue with the reference electrode, since it is shared across all channels.

    I think so too. I saw the similar wave pattern when the reference electrode was disconnected. Yet, I checked the reference ear probe carefully. I never get proper wave without BIAS. What is the solution? Should I change the place of the reference electrode from ear?

  • wjcroftwjcroft Mount Shasta, CA

    I am hoping Julian will have suggestions. I would immediately suspect your reference / ear electrode. That includes cable, connectors, electrode metal, age of electrode etc. If the electrode is silver chloride plated (grey coating), check that it is the same uniform grey color. If it wears off, then less effective. In your case some more serious issue may be lack of conductance continuity due to failure of cable or connectors. A DVM digital voltmeter resistance check may show issues between the female header connector and the ear electrode.

  • hirohiro Japan
    edited May 17

    A DVM digital voltmeter resistance check may show issues between the female header connector and the ear electrode.

    I have checked these.

    I doubted my ear clips are disconnected but these are fine as a result of testing by multimeter.

    My two ear probes and cables have conductivity. Rather, if my reference electrodes and their cables are broken, wave always keeps it's shape like the first image but I noticed that the wave changed like the second image just adding BIAS. I thought this wave was relatively fine but the settings contradicted an instruction from OpanBCI. I was confused. So, I thought that it could be the issue of settings and thrown a question.

  • Hi Hiro,
    just reading your messages (thanks William for the notification!).
    Regarding bias, you can totally include the bias in your settings, that is how we use the ThinkPulse at Conscious Labs, it really helps get a better signal.
    Regarding the gain, the use of ThinkPulse with ear clips usually creates a higher offset which prevents you from using the maximum gain x24 without railing (saturation). That is why it is good to use x8 as you did. I suspect very few people are actually using a gain that is not x24 and sometimes it is assumed in some programs or configurations that the gain is 24 by default. I believe that is the case for the SD card conversion: the ADU value (arbitrary unit from the analog to digital converter) is by default divided by 24 and converted to microV, but this formula is not true when the gain is not 24. As you guessed, you have to multiply the value by 24/RealGain to recover the correct value in microV.
    Regarding your setup, the first signal is noise indeed. The second screen shot looks much better but the amplitude (2 or 3 microV) is definitely not right. Is it a recording coming from the SD card at Gain X8 ? In that case, if you multiply by 3, this becomes acceptable amplitude values. Otherwise, something is wrong.
    The third image, left panel, shows a correct 12ish microV amplitude, although it also depends on your configuration (which positions on the scalp, flat or spiky, lots of hair or bald, EM environment,...). I suggest when you check the signal quality that you keep the 50microV vertical scale instead of Auto which can be misleading.
    Anyway, if you are not sure, follow the basics steps (on the GUI directely, not SD):

    • deactivate all channels except the 2 first ones,
    • to check your acquisition system and ear clips, come back to gold cups electrodes with gel, put them on the forehead Fp1/Fp2 approximately (with sport headband for instance), use both Ref and Bias ear clips, include bias, use Gain=X8,
    • check the signal: amplitude around 15microV, clear blinks wave, clear increase in amplitude and freq when jaw clenching
    • if it is ok, do the same with 2 flat ThinkPulse sensors on the forehead. Make sure the 3 wires (signal, power +, power -) are correctly connected to the Cyton for each sensor.
    • if everything is ok, add 2 occipital ThinkPulse sensors (O1/O2) and check the Alpha power increase when closing your eyes.
    • once previous steps are validated, you do get neurophysiological signal, you can come back to your full initial config.
      Best,
      Julien
  • hirohiro Japan

    Hi Julian,
    Thank you for reply. Your suggestion help me so much.

    Regarding bias, you can totally include the bias in your settings

    That's good.

    Is it a recording coming from the SD card at Gain X8 ?

    Yes.

    The second screen shot looks much better but the amplitude (2 or 3 microV) is definitely not right.

    As you say, quality of my signal is sometimes poor. I think the left side of the third image is a good case. I got often wave, which includes up to 10 microV and around 7 microV on average in the past experience before I bought ThinkPulse. It is not always but I frequently observe such a weak signal.

    As you guessed, you have to multiply the value by 24/RealGain to recover the correct value in microV.

    Nice. Now, my main 2 questions are solved but the reason of low amplitude still exists not because of active electrodes. I would like to find it. I will follow your steps.

    to check your acquisition system and ear clips, come back to gold cups electrodes with gel,

    By the way, I do not have gold cups electrodes and gel. Should I buy these? Or, could I did experiences with default electrodes or others? I have EMG/ECG Snap Electrode Cables and paste.

    And wjcroft,
    Thank you for your help though I guess you may be busy.

    Hiro

  • Hi Hiro,
    if you do not have gold cups, no need to buy them. The idea was just to come back to a configuration that you knew and that was working well to discard any other issues. If you have other electrodes like sticky pads for instance, you can always use them on the forehead. Otherwise, you can directly use the ThinkPulse sensors but make sure they are correctly wired, powered, that the electrode tips is tighly screwed, that the contact with the skin is good and that the unused channels are deactivated
    Best,
    Julien

  • hirohiro Japan

    Hi Julien,
    Thank you for your reply. I understand.

  • AbdullahAbdullah Germany

    Hello everyone,
    Recently I have bought the OPenbci ALL-IN-ONE ACTIVE EEG ELECTRODE BUNDLE. When _ connected the 16 channels and started to test the kit, almost all the channels were railed 100%.
    I have read alomst all the suggestions and the other similar issues by the other customers. However, everything was set and placed carefully, including the refrence and the ear padds anf the connections.

    I do not know what is the problem and I would like to ask for some help

  • wjcroftwjcroft Mount Shasta, CA

    @Abdullah, hi.

    I merged your new thread regarding railing into this existing Q&A on Thinkpulse. Julien will see this post and hopefully respond.

    William

  • Hi Abdullah,
    it could be a Gain problem. Did you set it to X8 in the Hardware settings? If you keep the default X24 Gain, together with the ear clips, the offset is too high and most channels get saturated. Include bias, make sure the settings are sent to the board and check that the power supply boards are well connected. Otherwise, if this is your first use of the 16 channels ThinkPulse Ultra Cortex and you still experience issues, I suggest you start without Daisy and just with 2 pre frontal channels (remember to deactivate all the unused channels). Once you have a working configuration with a limited number of channels, you can add up all the other channels. Let us know how it goes. Best,
    Julien

  • George3d6George3d6 Remote

    @julienConscious
    Actually I abandoned using the think pulse electrodes and switched to the normal ones, because I was out of time.
    Now that I have plenty I am trying them again and getting roughly the same pattern OR if I have the biases on I get this:

    (Note, I only plugged in 8 electrodes to save time, since the setup is likely wrong.

    Could wiring lead to this amount of noise ? I do have the signal wire going in the correct place and the ground & power too. Might it be that this is a "silent failure" on behalf of the power circuit?

    I'm kind of at a loss as to where to even start investigating.

  • George3d6George3d6 Remote

    Maybe more specific questions:

    Is this the kind of pattern I could get from the electrodes not getting current properly ?
    Could I easily replace the thinkpulse power gadget with just a parallel connector (a write) joining the + and - of the electrodes to my openbci power supply ?
    Could I just use a separate 5v power source for the electrodes ?

  • George3d6George3d6 Remote

    Also, one thing that's unclear from the video is if it matters which of the power pins (there are two on each side) the electrodes are plugged into or soldered to

  • George3d6George3d6 Remote

    Maybe one last question to ask here is: Does the order in which the electrodes are plugged into the power supply matter? That's the last free variable I can think of here.

  • George3d6George3d6 Remote

    I tried wiring the electrodes in an 8 channel configuration without the extension as to keep them all in order and to have the power supply stick directly into the openbci. I am still getting the following patterns:


    setting the gain to 8 results in this instead:

    Which is not railed but also doesn't look like signal (?)

    If I disconnect the biases and turn SRB1 off and bias to 'no" I get the following:


    My only remaining solution is to make the cables mix up less but I doubt tiny changes in SNR on the cable should matter?

  • Hi George3d6,
    the first before the last pic looks decent for channels 1,2 7, 8 and possibly 5. This is probably the right settings. Contact of the sensors with the scalp (through hair?) on channels 3,4 and 6 is probably wrong and needs to be readjusted. Looks like you might have a lot of 50Hz from mains that is not fully attenuated by the notch: try if possible to move away from possible outlets, computer screens, chargers. In your first picture, there is most probably no power. You just need to power them with +/- 2.5V (you can double check the voltage for each electrode with a multimeter), which of the 2 pins you connect to and in which order on the power supply board does not matter, external 5V source is possible but not recommended since you need the same ground. Start with the 2 pre-frontal and add them one by one. Deactivate the bad or not used channels. Check that you have fresh batteries on the Cyton and that you are not too far away from the Dongle (you can check the packet loss tab in the GUI). In my understanding, SRB1 is the ref for positive pins: it should be off if you use the negative pins (bottom ones) which is what is recommended, independantly of using ThinkPulse sensors. I suggest you use the exact same configuration as when you use the gold cups or other working configuration (except for the gain that you need to set at X8). Good luck with your experiments. Best, Julien.

  • edited August 2

    I upgraded my Mark IV headset to ThinkPulse Active, but the signal's accuracy and reliability are terrible.
    Here is the view I captured with the following hardware settings:

    Also here is a snapshot without wearing the headset:

    and here is immediately after wearing it

    and sometime is like this:
    :
    Also, I turned off any devices to reduce the EMF as much as possible.
    I contacted openbci support and said to mention the problem here.
    @retitut @richard @openbci.com @wjcroft
    I will greatly appreciate your support and consideration.

    Thanks.

  • wjcroftwjcroft Mount Shasta, CA

    Mentioning @julienConscious, the ThinkPulse developer in Paris, who may well be attending the Summer 2024 Olympics there(!)

  • Hi Soroushzrzz,
    indeed the signal does not look good. It looks like the values in microV are very high overall. The hardware settings seem ok (providing you are indeed using the bottom negative pins). I do not see your filters but make sure you have a 50Hz notch (or 60Hz in the US) and a band pass with a lower bound over 1Hz (5hz is better to start with).
    Have you ever had good signal with this set up or does it always look like that?
    Using 16 channels with UltraCortex equipped with ThinkPulse can be a little tricky at first. Make sure you followed the instructions about wiring: the Y connector for the bias, the double power supply board for the ThinkPulse that you have to solder to the AVDD/AVSS and connect together with the provide wires (everything is explained in the tutorial videos). If your setup has never provided good signal with ThinkPulse, I would definitely start double checking the wiring.
    My suggestion is to start with a simpler configuration first. Please consider the following configuration:
    Use just one Cyton board, plug-in one single power supply board, and only connect the 2 pre-frontal sensors Fp1 and Fp2 (flat on the forehead), the bias and ref earclips. Deactivate all the other sensors. Take some time to adjust the headset so that the sensors are nicely sitting on the forehead. I suggest not to use AUTO setting for the scale since it can be misleading and hard to read. Prefer using the 50 microV or 100microV setting. A good signal is under 50microV so checking a 2000microV signal does not make sense. I suggest not proceeding to any further step before you have satisfactory signal with this setting, which implies seeing clear eye blinks.
    Let us now how it goes.
    Best,
    Julien

  • edited August 12

    Hi Julien,

    Thank you so much for the advices. I followed the same steps and now here is my two forehead EEG signals, and from my POV it is fine, but how can I find if it is valid or not. Alway I should consider the signal to be below 100/3?
    Because I set the PGA to 8 and from this forum, I found the actual signal is time 3 because the default is 24 and I set to 8, right?

    Also what will be the range of Impedance that I should look for.

    Also I greatly appreciate that what should do next step? Just attached all the 16 Channels?

    Best,

    Soroush

  • Hi Soroush,
    thanks for getting back to us with these results. They do look fine to me. The gain should have no influence on the readings in microV, it only affect the ADU (arbitrary unit) measure: what you see in the GUI is the converted measure in microV which does should not be affected by a gain change (only the vertical resolution increases with the gain).
    So yes, a first kind of validation of acceptable signal is the range in amplitude which should mostly be under 50microV in absolute values. Then you can check that you clearly see jaw clenching (big increase in amplitude and higher frequencies), the typical eye blinking artefact, and finally alpha increase in the spectrum when closing your eyes (although mostly visible in occipital positions and depends on subjects).
    The impedance can not really be reliably used with ThinkPulse sensors (since they are active sensors, they can not inject a current into the scalp), but this gives you an idea of the quality of contact: yours look good. Also, I have not checked recently but the value of the impedance used to be correct with gain=24 only. Anyway, clinical grade values are under 5kOhms but a couple hundreds kOhms impedance is still totally acceptable considering the high input impedance of the Cyton ADS1299 AFE.
    As for next step, I suggest you try to activate one by one each channel (or 2 by 2) up to 8 on the main board (don't forget to adjust the position of the sensor on the scalp) and repeat all the checks until you are satisfied. Start ideally with a subject with short easy hair so ThinkPulse pins can nicely reach the scalp and bad contact is not an issue. In case one particular channel is consistently bad despite all your efforts, just deactivate it for now and keep going. Conversly, you can of course go faster if everything goes super smoothly but always consider coming back to the last working configuration when things get bad.
    Once you're happy with all the channels of the main board (hopefully 8), add the Daisy board, set up the 2 power supply boards, and start over with activating only the 2 first channels of the Daisy board (9&10), connect Fp1 and Fp2 sensors (that used to be on 1&2) to these 2 channels 9&10 and focus your attention on getting similar results to the ones that you got for the main board. If you get stuck there, it means that there is a problem with your Cyton+Daisy set up which needs to be fixed (power supply board wiring, bias). Double check the +/-/signal of each ThinkPulse sensor, it gets quickly crowded and confusing when using Cyton+Daisy+ThinkPulse. Once you're happy with the signal of Fp1&Fp2, keep activating more channels until you reach the 16 channels with good signal hopefully. Again, don't get stuck on a particular channel. After putting reasonable effort to make it work, if it does not show good signal, deactivate it and move on: it will always be time to look at this particular channel/sensor later (and often, problem is naturally solved when you get back to it after all the other channels are correctly set). You normally have to do that whole procedure once. Then, once you are confident with your set up, the next times you will get good signal consistently, or it will be easy to detect what sensor does not show good signal and correct the individual problem.
    Best,
    Julien

  • edited August 14

    Hi @julienConscious,

    Thank you so much for the information. I moved forward and here is my recording for all 16 channels. To me all of them seem ok, I want to ask your opinion.
    Also in this topic I found you mentioned that "As you guessed, you have to multiply the value by 24/RealGain to recover the correct value in microV."

    I am recording the data using python and sending the command of setting for gain 8 and then record the data using mne library.

    BoardShim.enable_dev_board_logger()
    params = BrainFlowInputParams()
    params.serial_port = '/dev/cu.usbserial-D200Q62O'
    board = BoardShim(BoardIds.CYTON_DAISY_BOARD.value, params)
    board.prepare_session()
    commands = [ "x1040110X", "x2040110X", "x3040110X", "x4040110X", "x5040110X", "x6040110X", "x7040110X", "x8040110X", "xQ040110X", "xW040110X", "xE040110X", "xR040110X", "xT040110X", "xY040110X", "xU040110X", "xI040110X"]
    for command in commands:
    board.config_board(command)
    time.sleep(0.1)

    Also her is my code to save the data:
    board.start_stream()
    time.sleep(10)
    data = board.get_board_data()
    board.stop_stream()
    eeg_channels = BoardShim.get_eeg_channels(BoardIds.CYTON_DAISY_BOARD.value)
    eeg_data = data[eeg_channels, :]
    eeg_data = 3 * eeg_data / 1000000 # BrainFlow returns uV, convert to V for MNE
    ch_types = ['eeg'] * len(eeg_channels)
    ch_names = BoardShim.get_eeg_names(BoardIds.CYTON_DAISY_BOARD.value)
    sfreq = BoardShim.get_sampling_rate(BoardIds.CYTON_DAISY_BOARD.value)
    info = mne.create_info(ch_names=ch_names, sfreq=sfreq, ch_types=ch_types)
    raw = mne.io.RawArray(eeg_data, info)
    raw.filter(l_freq=0.5, h_freq=30) # Filter before downsampling
    raw.notch_filter(50)
    raw.plot()

    So in this case should I multiply the value of eeg by 3 to get the real EEG value?

    I’m using a 3.7V, 500mAh battery, which is the default one, but after 15 minutes, the signals degrade significantly, and I suspect the battery needs recharging. I want to confirm if this is the issue or if there could be another problem. Additionally, how can I estimate how long a fully charged battery will last when used with the Cyton-Daisy board with 16 channels ThinkPulse active electrodes?

    I greatly appreciate your support and consideration.
    Best,

    Soroush

  • Hi Soroush,
    thanks for following up. The signal looks good indeed, glad it worked out.
    Would you mind sharing with other users what you think could have prevented you from getting proper signal in the first place, if you figured it out?
    Regarding Gain and microV: the OpenBCI GUI does the right conversion when you read the values (times, series, fft plot, recorded text file, etc) so you do not have to worry about the change of gain (except if you use the SD tho). This being said, if you use a third party software, it happens that the default gain X24 is hard coded in the OpenBCI driver. I'm not sure about brain flow and MNE, I would assume if they propose values in microV (or V), they do the right conversion but I'm not completely sure. One way to check that is to short-circuit one particular channel with the bias channel (with a simple wire, or you just clip the ear clip to a particular sensor), it causes saturation and you see the maximum possible value which is 4.5V/Gain (in your case it should display +/-562 500microV, if you get +/-187 500 it means the conversion is not correct and you do have to multiply values by 3).
    Regarding the battery, I do not have the exact numbers at hand but Cyton+Daisy+16 active ThinkPulse should consume around 200mA (probably less). You should have at least 2 hours of battery life so 15min does not look right. You can check the battery level when it starts going wrong (you find tables that convert the voltage between 3.7 and 4.2 to the approximate percentage, or maybe you have a battery tester). Anyway, the max current they deliver can be a parameter too. I suggest you upgrade to a bigger one 1000 or 2000mAh, they should still fit in the case ;-). This being said, a weak battery can have an influence on the signal, especially on the wireless connection.
    Best,
    Julien

  • Good morning,

    I've decided to switch to using the ThinkPulse Active Electrodes for my project involving the third-party software suite BCI2000. Currently, I'm running into some issues w/ setting gains and would appreciate any input on the ThinkPulse side of the house (currently in comms w/ BCI2000 people to help sort this out).

    The issue is that, when using the active electrodes and setting the channel gains to 1x through the OpenBCI GUI, I receive totally normal signals as shown:

    However, on the BCI2000 side of things, I get completely anomalous signals, even after I ostensibly set the channel gains to 1x (I'm still not convinced BCI2000 is setting it correctly):

    I noticed that after allowing the electrodes to settle down a bit more, Channels 3 - 8 start to show signals, albeit in that mV range. Channels 1 and 2 however, remain in that fV range and appear dead.

    Could you please offer any guidance/insight?

  • wjcroftwjcroft Mount Shasta, CA

    Hi @parguello,

    Your BCI2000 graphs show what appears to be tons of mains noise, The data stream from Cyton is UNFILTERED, raw. But the GUI windows show filtered streams. Normally you will want to do something like a bandpass from .5 Hz to say 45 Hz before any processing. A notch at 60 Hz is also recommended.

    William

  • parguelloparguello USA
    edited August 23

    W/ the help of the folks at BCI2000 forums, I was able to determine that the author of the OpenBCIADC Source module erroneously named one of his parameters, resulting in some confusion of setting the channel gains. It seems as if the channel gains are now being set correctly through BCI2000 to the Cyton board, w/ the exception of channels 1 and 2 seen here:

    @wjcroft said:
    Hi @parguello,

    Your BCI2000 graphs show what appears to be tons of mains noise, The data stream from Cyton is UNFILTERED, raw. But the GUI windows show filtered streams. Normally you will want to do something like a bandpass from .5 Hz to say 45 Hz before any processing. A notch at 60 Hz is also recommended.

    William

    Notice that the scale is in femtovolts. The screenshot I provided above better demonstrates just how small the signal is. Wouldn't this suggest that the electrodes are 'dead'? Though this wouldn't make sense given that they seem to be working just fine in the OpenBCI GUI. Contact is solid as Channels 1 and 2 are my Fp1 and Fp2 electrodes; w/ no hair getting in the way these should be just about the most robust electrodes in my setup. In BCI2000, filtering is already implemented, w/ a notch at 60 Hz and bandpass from 1 to 40 Hz. Additionally, I was wondering how impedance testing is done w/ these active electrodes. I used the impedance testing in the OpenBCI GUI a couple of times before I realized that injecting current into the active circuit couldn't possibly be very good for it :#

Sign In or Register to comment.