Very strong noise in the Delta waves range

SagiSagi Israel
in Build-it-yourself

Abstract- I've built the EEG device with 3 dry electrode made from stainless steel that sit on the forehead.
The problem-The only problem is the very strong noise in the Delta waves range. The second harmonic of these noises can also be in the Theta and Alpha bands.
All other problems with TBR and TAR are only consequences of this problem.
See Figure 4 for real EEG noise and Figure 5 for its spectrum.
These graphs also show that the noise of one of the channels can be significantly larger than the other two. If I manually move the corresponding electrode, it’s possible to find the point where these noises are normalized.

DC offset.
One of the reasons that can cause these noises can be a large displacement between the reference electrode and the rest of the electrodes. The problem is not even the magnitude of this offset, but its change over time. The rate of this change can be fast enough to generate noise in the Delta range.
See Figure 7 for real EEG DC offset and Figure 8 for emulator DC offset.

Analysis using fragmentation.
The idea is that the entire data array is divided into small fragments.
Then, for each fragment, the FFT is performed and the TBR is calculated.
Then all fragments are sorted in ascending order of TBR and presented in graphical form.
In the figures below, you can see that for "good" records, only a small number of fragments have an abnormally high TBR and a clear boundary is visible between them and normal fragments. For "bad" records, such a boundary does not exist.

I really appreciate your help!
[email protected]

Comments

  • wjcroftwjcroft Mount Shasta, CA

    Sagi, hi.

    Steel metal makes for very POOR EEG electrode material. That is because of the high noise levels. There are some published papers on comparing various EEG electrode materials. I don't have a link handy at the moment. In any event, the highest quality comes from Ag-AgCl. Other good metals are gold, silver, tin. Although the gold are going to be plated material; and when this wears can cause galvanic effects.

    William

  • wjcroftwjcroft Mount Shasta, CA

    Here is one research paper comparing electrode metals:

    https://drive.google.com/file/d/1lxXs8DAlh3zHI7j5ds3qQd56RVIxH5lL/view?usp=sharing

    "Evaluation of commercially available electrodes and gels"

    Abstract
    Objective: To test the applicability of different types of commercially available electrodes and electrode gels or pastes for recording of slow EEG potentials.
    Methods: Experiments were carried out on six types of reusable electrodes (silver, tin and gold cup electrodes, sintered silver–silver chloride (AgjAgCl), platinum, stainless steel), six disposable AgjAgCl electrode models, and nine gels or pastes. We studied the parameters, which are critical in slow-potential recording, such as polarization, initial and long-term stability and low-frequency noise.
    Results: The best results were obtained with the reusable sintered AgjAgCl electrodes. The six disposable AgjAgCl electrode models also proved to have appropriate electrical properties. Other types of reusable electrodes suffered from diverse degrees of polarization, baseline drift, low-frequency noise, high resistance, and changes in properties due to wear and tear. Seven out of nine gels or pastes contained a significant amount of chloride, which is a prerequisite for DC stability of AgjAgCl electrodes, whereas the absolute concentration of chloride had little effect.
    Conclusions: Direct current (DC) coupled recording of EEG is critically dependent on the choice of electrode and gel.
    Significance: Our results provide rigorous criteria for choosing DC-stable electrodes and gels for DC-coupled or long time-constant AC-coupled recordings of slow EEG potentials.

  • SagiSagi Israel

    William, thank you very much for the detailed answer and for attaching the article!

    1. Do you have a recommendation link from where I can buy/order AgjAgCl or silver dry electrodes?

    2. From your experience, can noises in the Delta wave and problems as I mentioned with the DC OFFSET be caused by other factors that you can think of?

  • wjcroftwjcroft Mount Shasta, CA

    Hi Sagi,

    There are many online stores where you can purchase electrodes and other EEG supplies. These typically have 'touchproof' connectors instead of Berg / Dupont connectors. But the Shop has an adapter you can use. Shop also has gold cups and AgCl dry electrodes.

    https://www.google.com/search?q=eeg+electrodes
    https://shop.openbci.com/products/touch-proof-electrode-cable-adapter
    https://shop.openbci.com/products/openbci-gold-cup-electrodes
    https://shop.openbci.com/collections/frontpage
    https://fri-fl-shop.com/collections/electrodes-eeg-electrodes

    You must replace your steel electrodes, they are the primary noise source. See graphs in the paper. Noise can also come from motion artifacts, EMG from jaw / scalp muscles, etc.

    DC offset is present for all electrode types, including AgCl. See,

    https://openbci.com/forum/index.php?p=/discussion/201/large-millivolt-data-values-fbeeg-full-band-eeg

    William

  • SagiSagi Israel

    Thank you so much ????????????

  • wjcroftwjcroft Mount Shasta, CA

    Also note that the normal EEG spectrum distribution (amplitude) is considered '1/f' in shape. That is, delta is largest, then theta, alpha, beta. Amplitude is inversely related to frequency.

    https://www.google.com/search?q=1/f+eeg+spectrum

    But this is for normal electrode materials. Steel will add additional low frequency noise.

  • wjcroftwjcroft Mount Shasta, CA

    https://www.sciencedirect.com/science/article/pii/S105381191930895X

  • wjcroftwjcroft Mount Shasta, CA
    edited April 2023

    https://behavioralandbrainfunctions.biomedcentral.com/articles/10.1186/1744-9081-3-62
    "Distinguishing low frequency oscillations within the 1/f spectral behaviour of electromagnetic brain signals"

    The Origin of the 1/f spectral behaviour
    The inverse relation of the power density of the EEG with frequency in the mammalian cortex is a result of the physical structure of neuronal networks and the limited speed of neuronal communication arising from axon conduction and synaptic delays [9]. A large cluster of neurons, each generating a unit activity, forms a functional network which is held together by the neurons' synchronisation that ensures activity control [18]. Such synchronised behaviour seems to attract further neurons and causes the oscillation amplitude to increase. Moreover, the period of oscillation is determined by the size of this neuronal cluster that constitutes a given cycle [9]. Thus, large neuronal areas are associated with slow, high amplitude oscillations whereas a small, localised concentration of neurons gives rise to higher frequency, low amplitude signals [8]. This explains why most of the power of the EEG signals is concentrated in the low frequency spectrum.

  • wjcroftwjcroft Mount Shasta, CA

    Your Figure 9 showing "Pulse like noise" is not normal, and is artifact. Either from electromagnetic interference, or other sources. Are you using OpenBCI board? Which?

  • SagiSagi Israel

    Thanks a lot for for the detailed answer. No, I'm using a device I built myself.
    I have a few more questions if you may answer me,

    1. Should the 1/f component be removed from the signal spectrum before TAR and TBR are calculated? Since this component is a general property of noise and does not relate to the work of the brain, it seems logical to remove it.
    2. How is it customary in EEG practice to calculate TAR and TBR - by the ratio of the energy density (per Hertz) or by the ratio of the total energies of the corresponding frequency ranges?
    3. What Beta range is used to calculate TBR? Full up to 30 Hz or low only up to 20 Hz.
  • SagiSagi Israel

    More questions.

    It is clear that DC offset always exists, but in our device it is abnormally high.
    When checking with an emulator, it is measured in microvolts, but when experimenting with a real EEG, it can be from a few millivolts to 100 millivolts. And what's worse, it changes over time.
    What could be the reason other than the material of the electrodes?
    Perhaps the attachment of the electrodes should be more flexible to provide a more constant pressure to the skin?
    Or the electrodes must be provided with additional degrees of freedom so that they can move with the skin.
    Are electrodes made of silver with Ag/Cl galvanic coating applicable?
    How good are electrodes coated with a special paste consisting of Ag AgCl powders and conductive glue?
    Does the noise depend on the substrate material on which this paste is applied?
    The device has 3 electrodes located on the forehead and a reference electrode behind the ear.
    Are there any preferred electrode placement points on the forehead?
    Is it acceptable to place the reference electrode behind the ear?

  • wjcroftwjcroft Mount Shasta, CA

    @Sagi said:
    Should the 1/f component be removed from the signal spectrum before TAR and TBR are calculated? Since this component is a general property of noise and does not relate to the work of the brain, it seems logical to remove it.

    NO, normal signal processing of EEG does NOT remove the 1/f component. The paper I cited and quoted is just an example of special processing for a specific application.

    How is it customary in EEG practice to calculate TAR and TBR - by the ratio of the energy density (per Hertz) or by the ratio of the total energies of the corresponding frequency ranges?

    https://www.google.com/search?q=theta+beta+ratio+calculation

    What Beta range is used to calculate TBR? Full up to 30 Hz or low only up to 20 Hz.

    Normal beta is considered the 15 to 20 Hz range. 20 - 30 is frequently associated with hypervigilance or over arousal. Many neurofeedback protocols set inhibits or alert signals for activity in that band.

  • wjcroftwjcroft Mount Shasta, CA

    @Sagi said:
    It is clear that DC offset always exists, but in our device it is abnormally high.
    When checking with an emulator, it is measured in microvolts, but when experimenting with a real EEG, it can be from a few millivolts to 100 millivolts. And what's worse, it changes over time.

    Did you read the fbEEG Forum link I gave previously. DC coupled amplifiers such as ADS1299 have this DC offset as a normal feature. You remove it as described in that link, with either a bandpass or highpass filter.

    What could be the reason other than the material of the electrodes?

    Your steel will likely make the offset worse.

    Perhaps the attachment of the electrodes should be more flexible to provide a more constant pressure to the skin?
    Or the electrodes must be provided with additional degrees of freedom so that they can move with the skin.
    Are electrodes made of silver with Ag/Cl galvanic coating applicable?

    Did you look at the electrode links provided? The silver chloride electrodes can be used either wet or dry. In the Shop you can see headbands using velcro to apply constant pressure.

    Silver chloride electrodes are made by starting with a plastic base, then silver plated, then chlorided. A special type of silver chloride electrode is made by 'sintering' silver powder into a solid form, then chloriding. These are the type used in infra low frequency neurofeedback, but generally are expensive and not needed for most applications.

    How good are electrodes coated with a special paste consisting of Ag AgCl powders and conductive glue?
    Does the noise depend on the substrate material on which this paste is applied?

    See previous comment.

    The device has 3 electrodes located on the forehead and a reference electrode behind the ear.
    Are there any preferred electrode placement points on the forehead?

    10-20 placements are at Fp1, Fp2 or Fpz.

    Is it acceptable to place the reference electrode behind the ear?

    Yes, that's fine.

  • wjcroftwjcroft Mount Shasta, CA

    Very interesting paper comparing signal processing methods for TBR:

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7391403/

    This figure from the paper, emphasizes their conclusion that TBR is NOT appropriate for diagnosing ADHD.

  • SagiSagi Israel

    Dear William, I changed the electrodes according to your suggestion - and indeed the results are much better! But there are still problems with the DC offset.
    What other factors besides the electrodes material can affect the DC offset? It must be borne in mind that when working with the emulator, the DC offset is very small.

  • wjcroftwjcroft Mount Shasta, CA

    @Sagi said:
    Dear William, I changed the electrodes according to your suggestion - and indeed the results are much better! But there are still problems with the DC offset.

    Are you using an appropriate filter to remove the offset? For example either a highpass at say .5 Hz, or a bandpass, say from .5 Hz to 45 Hz?

    DC offset is always present with DC coupled amplifiers such as the ADS1299. As you see it is very easy to remove it. Posted this link below and again referred to it in above comments.

    https://openbci.com/forum/index.php?p=/discussion/201/large-millivolt-data-values-fbeeg-full-band-eeg

    Regards, William

  • SagiSagi Israel

    Dear William ,I did use HPF, once again thank you very much for the detailed and clear explanation

  • SagiSagi Israel

    Dear William , Now that my system is working, could you advise me where I can get databases with which I can compare and draw conclusions? My goal is to build an automation that will compare the measurements from my device to the databases and give distinctions such as - this person has ADHD or this person's concentration level is currently increasing, etc.

  • wjcroftwjcroft Mount Shasta, CA

    https://github.com/meagmohit/EEG-Datasets#eeg-datasets

    https://www.researchgate.net/post/Are-there-any-public-EEG-data-sets-that-one-can-try-their-hands-on

  • SagiSagi Israel

    Dear William, I really have no words to thank you for the wonderful help and clear explanations you provide!

    Thank you.

  • SagiSagi Israel

    Dear william , I am interested in learning about the topic of HEG (diagnosis and treatment) using a camera - FlIR / pIR /Fnir. I saw some explanations by Jeffrey Carmen, but I could not get enough information. Can you advise me in this area as well? Thanks

  • wjcroftwjcroft Mount Shasta, CA

    There are at least two types of HEG biofeedback: nirHEG and pirHEG. Carmen is doing the passive infrared style using a camera. The nirHEG style uses headband with emitters and sensors. Check out Josh's offerings. These used to be listed in the Shop on OpenBCI, but I think primary channel now is Crowdsupply.

    https://www.crowdsupply.com/alaskit/hegduino-v2
    https://www.crowdsupply.com/alaskit/hegduino
    https://github.com/joshbrew
    https://web.archive.org/web/20220118083946/https://www.hegalpha.com/ [unclear why hegalpha.com disappeared]

    Before ordering anything, I suggest joining the Brains At Play Discord channel, and asking him directly about what is recommended for ordering, what is stocked, etc.

    https://brainsatplay.com/
    https://discord.gg/tQ8P79tw8j

  • SagiSagi Israel

    Dear William , My system worked up to now via BT transmission, I'm trying to move the system to work with WIFI module but noises are created that affect the EEG (I don't know exactly which frequency/frequencies) Have you encountered a similar problem? Do you have a recommendation on how to deal with this problem? Thanks a lot for your help!

  • wjcroftwjcroft Mount Shasta, CA

    Sagi, hi.

    So called 'mixed signal' circuit boards that combine digital logic, microprocessors, wireless, etc. -- with highly sensitive analog components that measure amplitudes in microvolts -- must be carefully designed to avoid the digital elements inducing noise in the analog sections. The TI ADS1299 datasheet has documentation on how you can achieve this.

    It sounds likely that your old Bluetooth board was 'lucky' in not inducing significant noise. Wifi wireless EMF amplitudes can be much higher than Bluetooth. You might check your Wifi chip settings for signal output levels, it may be adjustable to lower levels. I'm not saying the noise is due to Wifi EMF. Just that it might be a factor. More likely you have circuit board layout issues with the mixed analog / digital pathways on the same board. Read the ADS1299 datasheet about 'star grounding'.

    https://www.google.com/search?q=mixed+signal+star+grounding

    William

Sign In or Register to comment.

© Vanilla Keystone Theme 2026