Thinkpulse: signal quality, default settings & impedance ?
uditkumar007
United States
Hi Team,
I’m using ThinkPulse active electrodes with an OpenBCI Cyton board (Ultracortex setup). I want to know if there's a default configuration that you recommend for a clean starting point before troubleshooting.
Problem:
My ThinkPulse sensors are connected to the Cyton board correctly, but when the sensors are not touching my scalp (they’re away from my head, still connected to the board). Even in that “no contact” case, the Open BCI GUI is showing low impedance, which is confusing to me. Because in the Open BCI documentation, it says otherwise. Attached the ss.
Questions:
1. Default baseline settings (before troubleshooting)
What are your recommended OpenBCI GUI hardware settings?
PCA Gain (per channel)
Input Type
SRB2 / SRB1
Bias include (on/off) and any guidance about the bias earclip
2. Impedance behavior with ThinkPulse active electrodes
Is the OpenBCI GUI impedance reading reliable with ThinkPulse active sensors?
Is it expected to see “low impedance” even when electrodes are not in contact with the scalp (but still connected and powered)?
If impedance isn’t reliable for active sensors, what should I use instead as the contact-quality indicator?
3. How to verify that we are receiving good quality signals if the impedance test is not viable
What are your recommended practical tests to verify:
each channel is connected correctly,
electrode contact is good,
signal quality is good and not just noise?
If there’s a specific way you recommend verifying channels one-by-one and validating them, please share that procedure.
Thanks a lot.
Best regards,
Udit Kumar Das
Comments
Hi Udit,
re: impedance. See this comment at link below by Julien, the Thinkpulse developer.
https://openbci.com/forum/index.php?p=/discussion/comment/20737/#Comment_20737
re: Cyton settings. See the docs section:
https://docs.openbci.com/ThirdParty/ThinkPulse/ThinkPulse/
https://docs.openbci.com/ThirdParty/ThinkPulse/ThinkPulse/#troubleshooting-the-thinkpulse-sensors-with-openbci-cyton-board
Regards, William
Hi @William,
Thanks for the reply. The links helped, and I previously went through these.
Some follow-up questions since it's still confusing, so I understand the correct settings for my 16 channels setup:
BIAS earclip: Should I remove/disconnect it by default? If yes, what’s the reason? If not, when should I keep it connected?
https://openbci.com/forum/index.php?p=/discussion/2869/thinkpulse-active-electrodes-q-a/p2
This thread says to keep it connected, but the documentation says otherwise.
Ideal “not-railed” signals: what exact settings do you recommend when signals are not railed?
Impedance: since impedance results may not be reliable, what’s the best way to verify
1. Is each channel is connected correctly and contact is actually good? 2. signal quality is good (not just noise)? Because this has happened, when we remove the headset, and are not in contact with anything, we are seeing "not railed" signals, which are just noise. How to differentiate between these?
Thanks again,
Udit Kumar Das
Hi, I suggest you email (contact at openbci.com), which is customer support. The Thinkpulse developer, Julien, previously answered Thinkpulse questions in that 'Q&A' Forum thread. Your best bet currently is customer support.
To clarify on the impedance, Cyton WITHOUT Thinkpulse uses a current injection method to measure skin impedance. This cannot work through an active electrode. The impedance widget displayed impedances only show the impedance between the Cyton and the Thinkpulse, not the skin.
With any type of electrodes, passive or active, if the electrode is not in contact with the skin, the returned samples just reflect environmental noise. Have you tried the suggested tests such as eyes closed alpha, jaw clenching, etc.?
We did the blink/jaw/eyes-closed tests, and we do see the signal changes. But we still see noise though and can't figure out if it's noise or a good signal without an impendence test.
Is there a recommended method for validating each channel and verifying that we are receiving reliable signals? Last question, without impedence, how can we verify the contact is good?
How much noise are you talking about? A few microvolts of noise is the norm with any EEG, active or passive.
Have you tried tests with just the passive electrodes as a point of comparison?
Please email customer support soon for responses from the engineering staff members.
Thank you for the reply.
I have used the suggested settings and am using 50uv to 100uv. The image above is for no contact. See, even with no contact with the head for some of the channels, the signal quality shows "not railed", and in the range 50uv, meaning a good signal.
Yes, we are also seeing "not railed" when we put on our head and test the signals. But it's confusing, if suppose channel 1,2 is not in contact with the scalp and recording just noise, how can we verify this and differentiate between a good signal and noise?
In the discussion above, we saw that Julien mentioned the impedance test should be correct with a x24 gain, but he didn't test it. We tested it with "no contact" with the head and it doesn't work. It shows good conductance. So it's not reliable either. The above images verifies it.
Unfortunately, I emailed the customer service 3 days still didn't get a response.
As you can see from the FFT in your first snapshot, ALL the channels have the same noise spectrum pattern. This is almost certainly a sign of environmental noise from electrical devices in your immediate area.
Have you tried moving to an 'quieter' area of your building? Devices that generate EMF fields include: extension cords, power supplies, LED or CFL lighting, conduits in (walls floors ceiling), cellular towers, wireless equipment, etc. There are EMF meters available as phone apps for certain types of phones (which include magnetometer chip).
https://www.google.com/search?q=phone+apps+for+emf
https://www.amazon.com/s?k=emf+meters [and as standalone devices]
The GUI / Cyton 'impedance widget' WILL NOT work with Thinkpulse, because that widget depends on the ADS1299 chip acting as current source on the passive (non active) electrode having direct skin contact.
Thanks for your suggestions. I just tested the EMF with my phone. I am seeing a range of 45-50 uT in our environment. Is that normal? Can the EMF noise be filtered in post-processing?
Is there any other way, as you mentioned, to differentiate between a good signal and just noise?
Visually judging “good vs noise” feels pretty subjective. Is there an objective metric or checklist to verify a good signal quality per channel? For our experiment, it is imperative that we receive good signals from all the 16 channels.
Your screenshots show CLEAR evidence of massive environmental noise in your room / building / area. The only solution is to address this noise. The phone apps are not as good as the stand-alone EMF meters. The idea with any EMF meter is to locate the 'hot-spots', by moving the meter around the area and see where the peaks are located.
My number one suggestion, previously made, is to try your setup in a completely different area away from all possible EMF sources. Another approach is to go to the power panel in your area and shutoff the circuit breaker for the room(s) in your area to see if that eliminates the noise. If it does, then use an EMF meter to figure out where it is coming from.
Also note that certain noise sources, such as cell towers / transmitting antennas, may be outside your control.
I don't get what you are referencing regarding "good vs noise" metrics. You have plain evidence of massive EMF noise. This is NOT a subtle issue.
I also previously suggested you could try some tests using passive electrodes, such as cups with paste. They will pickup noise as well, but the active electrodes are even more sensitive to EMF noise effects.
Thanks for the reply. I understand the point. We are testing the setup in a different location to verify this.
By Good Signal vs Noise metrics- I meant, is there any objective measure (like impedance) to confirm the contact quality so we can say “this channel contact/setup is OK and we are getting a good signal”?
One constraint for our study is that participants must sit in front of a driving simulator, which has a lot of EMF that we can’t fully remove. I will share the comparison plots by tomorrow, with the passive electrodes and active ones. Setting it up.
If we still have EMF contamination, is there any post-processing you recommend to remove this, or is the data basically not worth using if the noise is that dominant?
Thanks again. I really appreciate you taking the time to reply.
Great.
Best subjective measure is just to confirm that EEG is in 'normal' range. Below 100uV, not railed, no unusual patterns in the FFT, etc.
EMF is hard to 'shield', but can be reduced by using appropriate EE best practices.
https://www.google.com/search?q=best+practices+to+avoid+electronics+designs+emitting+excessive+EMF+noise?
https://www.allpcb.com/allelectrohub/pcb-layout-design-for-emiemc-compliance-best-practices
Etc.
There are no filters that can remove the WIDE-BAND noise you showed. If the noise is narrow band it could be filtered, at the expense of losing that portion of your EEG information.
Hi William,
Happy New Year! Hope you had a good break.
I need some suggestions. Unfortunately, we don't have all the passive electrodes right now. So I tested the thinkpulse active electrode setup in two environments - near the driving simulator (60-70uT EMF) and away from the simulator (40-50 uT EMF) in a separate room.
Near the simulator
Away from the simulator
Does the signals look okay? We do notice a little bit of lift in the FFT plot when we are near the simulator.
Also, I figured, if we wear the reference ear pins, then we don't see any signals in the GUI. Only when reference electrodes are not in contact with the ear, then we tend to get noise from the environment. That clears out the confusion regarding good signal vs noise.
With the passive electrodes, we did collect some data near the simulator previously. This is how the FFT plot looks for the passive ones.
Hi Udit,
Your screenshots now look much better than the original ones you posted that had huge noise spikes. As you say the noise floor is a bit higher when you are near the simulator. But that is probably to be expected, given the proximity to the noisy electronics.
You ask: "do the signals look okay?" Only you can determine that based on what you are trying to analyze from the EEG. What are you expecting to detect? Do you see the expected alpha increase with eyes closed in channels towards the back of the head? Jaw clenches?
Since you say this is a 'driving simulator', are you trying to possibly measure the 'stress level' of the driver? Generally I would expect bands beta (15-20) and highbeta (20-30) to increase under stress. But there are many more EEG stress metrics:
https://www.google.com/search?q=measuring+stress+level+with+EEG
Regards,
Thanks William. Yes, we are trying to measure the cognitive workload from the signals.
Yes, I am seeing the alpha increase and when jaws are clenched, the differences in signal. It goes high when jaws are clenched.
Unfortunately, we ran into a different issue now. The battery. From the discussion forum I saw, the cyton board draws 50ma. Is it the same for Think Pulse electrodes?
We have a 500mah battery from adafruit which is out of stock now, unfortunately. We have noticed when they battery is low, the signals get messed up, and the whole head plot is red. We have a 4-hour-long experiment. And the battery ran out in the middle of the experiment after approx. 2 hours.
I ordered 3 batteries from Amazon with the same voltage rating 3.7v. I have charged them to full. But none of them work with the cyton board. It doesn't even power up. What might be the issue? I have attached the link.
The charger also started malfunctioning from today. Both the lights are on. Which shouldn't be the case. I have also charged the batteries with a different charger. Doesn't work.
https://www.amazon.com/dp/B07BTQKK91?ref=ppx_yo2ov_dt_b_fed_asin_title&th=1
Unfortunately the JST-PH plug wiring (which pins are positive vs negative), is NOT standardized across the entire industry. Your Amazon seller probably uses the reversed polarity than Adafruit. Adafruit polarity matches what OpenBCI uses. You can always cut and splice the Amazon battery leads to reverse the polarity. But wise to label what you have done to avoid confusion. May be possible to pull the connector pin/socket out of the connector body and reverse with the proper tool.
The Cyton board contains a diode in the power input circuit, to prevent damage from such situations. I'm not sure about the charger, check to see if it still works with your original battery.
As far as power draw of the active Thinkpulse amplifiers, you will just have to experiment. You can also measure the milliamp draw by using a DVM meter. I suggest just getting a large enough Adafruit battery so you cover the hours you need.
https://www.adafruit.com/category/574
Regards,