Amplitude of Raw Output
freqamp
Australia
Hi all,
I'm using Ganglion and OpenBCI GUI with wet gold cup electrodes. Generally impedance is around 10 - 20 kΩ.
I'm loading recorded data (the CSV text files of raw data from the GUI) into MATLAB to gather frequency information for small sessions (2 mins) as part of an experiment.
**How do the amplitudes of my outputs correspond to those given in the GUI? **
I looked around the forums and couldn't find an answer, apologies if a repost.
Pics attached:
Here is a small window of recorded session from the GUI. FFT amplitude shows about 4 μV, EEG about 7.49 μVrms (channel 2):
Here is the same session in MATLAB. FFT of whole 2 mins after bandpass and notch filtering (1-50 and 50, same as GUI). Black line is smoothed data (with Savitzky-Golay filter). The peak on the same channel at 8 Hz has an amplitude of 0.542. Power spectrum is also shown. I would also like to understand this better as most of my peaks end up in the negatives...
Comments
Do you know what the units are on the Y axis on the right hand side Matlab plot?
Hi Bill, No, that's part of my issue.
I would assume its Volts. But if that's the case, the amplitude is 542000 uV.
If its mV, the amplitude is 542 uV.
We are expecting about 4 or 5 uV.
Here is a sample of the same data, Raw, after the Matlab filters applied, FFT:
Anyone have any insight?
@wjcroft @retiutut
It's just that you can't compare voltages well if the units do not match, and you lack Matlab units. So I dunno.
According to the matlab forums the fft absolute amplitude should be in the same units as the input data. So I don't know what's going on. Perhaps it's my code, although, I followed the matlab fft example and it gives the same result...
Do you know the units of the input data for OpenBCI and for Matlab then??
OpenBCI GUI should be µV, matlab will be the same. But I don't know if the raw output from the GUI adjusts for the reference data. It must not as how could I get -7.23 for example.
And if it was a straight conversion why do I get 0.5422 as the peak on channel 2 matlab fft above and about 4 from the GUI fft?
If both are microvolts and you are using the same electrodes and references, you may also want to look at the number of frequency bins for the two FFT and make sure that sampling rate and bin numbers match up as the same.
The picture in the first comment looks almost like the electrodes are not attached to a subject. Also, there is no accelerometer data and the text in the filters is not drawing in the correct spot. There is not enough information here to make any conclusions, but I do recommend using the latest GUI, 5.0.8 https://openbci.com/downloads.
This is GUI 5.0.8 on Mac. Text has been drawing that way for the last few updates. I don't normally use the accelerometer widget so I don't notice it much. When I do it seems to work. The board was laying flat on a table in this session, so could they be all neutral? Also, that is a shot of recorded data, not sure if that has something to do with the accelerometer data.
Regarding the EEG, the leads are gold cup with electrode paste and attached with micropore tape. Impedance test at beginning of the session showed about 8-15 kΩ including the reference. When I look at that output, I see some fairly clear alpha bursts, especially at -4 to -3 seconds, and the FFT shows a good peak.
What makes you say the leads are not attached? Should the amplitudes be higher? Do you notice a lot of noise?
I'd love to figure this out, show if you'd like more info, please ask.
@Billh
I think you're onto something. I'll look into that. I know what my matlab values are, but I wouldn't know how to get that info from the GUI
I'm still not ok with some of the details of the picture in the original post. Please share full details on what you are using, the Mac OS X version, and a recording file. Also would appreciate a console log file.
Playing back Ganglion files that only check impedance is generally not valuable. Impedance should be checked live, before recording..
Here is a screenshot of GUI 5.0.8, on my Mac running Catalina. I am playing back a file of Ganglion, sitting upright on my desk, with no electrodes attached. You'll notice everything on the front end is drawing correctly and the Accelerometer values are non-zero.
I don't think the filters buttons drawing incorrectly should have anything to do with you being in Australia, since your Mac would still be using English.
Your GUI looks good. The poor text drawing on mine hasn't concerned me much, and figured it would get fixed with a future update.
Unfortunately the forum won't let me post a text file.
Regarding impedance, the values I stated earlier were checked live before recording.
The accelerometer data is a funny one. The playback above does have some accelerometer data toward the beginning, then all goes zero.
Wow, sorry for duplicate posts above. Cant seem to delete them.
@freqamp Can you share a screenshot of the GUI with no electrodes attached to the Ganglion in Live mode? Make sure to have the FFT widget and the Accelerometer widget open. I'd like to see if the noise is present there.
Good idea, Ive actually never checked baseline noise
So regarding my original question.
Does anyone know what units output are to the txt file as raw output? Why are there negative values in the raw output, are they set from bestline from ref electrodes?
@wjcroft
Freq, hi.
I myself am questioning the Ganglion sample values shown in your screenshot. They all appear to be in the large negative microvolts range of around -3700 to -4200. This normally does not happen on Ganglion. Instead microvolts are usually both plus and minus, and probably in the range under + or - 200 microvolts, max. With quiet channels say below 10 uV.
The fact that you are seeing large offsets like this on Ganglion, may mean that your electrodes are suspect. Gold cup electrodes can 'chip', or crack / flake off the gold plate coating. If this happens, the tin metal underneath forms a 'galvanic', or battery effect with the gold, causing large offsets like this.
It's possible to remove paste from cup electrodes, without any rubbing or abrasion. But instead dipping them in very hot (almost boiling) water. Related threads here,
https://www.google.com/search?as_q=cleaning+cup+electrodes&as_sitesearch=openbci.com
William
With clean, 'new' electrodes, you can use a glob of paste to connect a test channel, reference, and ground (bias) together. This should always produce a channel signal of very close to zero microvolts. If your Ganglion cannot do that with non-chipped electrodes, then something is wrong with the Ganglion board.
Hi William, that is exactly the state of the gold cup electrodes and the ear clips.
These aren't new electrodes, so I'm not sure how they were treated previously. I generally clean them with a paper towel or a soft toothbrush. Occasionally I use contact cleaner.
I'm writing these experiments up for a project and won't have time to get new electrodes. But its important to know whats happening with the data, so I appreciate the info.
Im actually testing some flexible polymer electrodes with a conductive paint. When I sort out the issues with that I'll test with your directions. Thanks again
If you can acquire new cup electrodes, clean them with the forum threads mentioned by the previous link. No paper towels, toothbrush, contact cleaner needed. Just almost boiling water. The paste melts right off.
https://openbci.com/forum/index.php?p=/discussion/219/electrode-cleaning-after-use
https://openbci.com/forum/index.php?p=/discussion/1117/electrode-cleaning-prevent-gold-rub-off
There are also other solid metal cup electrodes that are NOT plated like the gold. These cannot chip or rub off. Typical metals are solid tin or solid silver. Another option are disposable silver chloride electrodes. These ARE plated (plastic, then silver plate, then chloride), but since you only use them a limited number of times, they generally don't have time to degrade. They are relatively inexpensive.
https://www.google.com/search?q=disposible+eeg+electrodes
Is + or - 200 μV, max are generally normal values for raw amplitude, as you said?
If so, is this an indication of quality signal? So if raw amplitude is let's say, + or - 300 μV, then the electrode is not placed correctly, right?
I am talking about dry electrodes, if this makes a difference. And not strictly speaking about the Ganglion, but EEG devices in general.
Thanks!
Hi
Do you mean the raw output into the text files? If its the text file output, it seems depending on the chemistry of your electrodes, offsets can occur. My outputs above had offsets of about -3800, but that didn't affect the graph in the GUI (once the filters kick in). Ive since changed electrodes and the offsets aren't occurring (some are starting to creep back).
I personally don't get amplitudes ±200 µV, unless its EMG. My EEG amplitudes are more like ±10 µV. I have noticed that if impedance is higher filtered EEG amplitudes are lower.
Using dry electrodes does lead to higher impedance, unless they are active, but that's a whole other can of worms.
Yes, I am talking about the raw EEG output (of any EEG device actually), without filtering.
For example, I am streaming raw data from Muse 2 that does not have any impedance check control, so I was wondering if I can indirectly perform a signal quality check just by looking at the raw amplitudes.
I personally get a broadth of values .. ±10, 100, 200, 300, 500, 800.. µV. When the headband is tighed up, these raw values decrease. That's why I was wondering for the "accepted" values.
Thanks for sharing your raw values!
I don't think you can infer the impedance from the raw values. Judging by my recent experiments, the chemistry of the electrodes offset it a lot more than the connection quality. I've been reading papers trying to better understand it, but chemistry is not my strength. Perhaps someone here can explain.
If you do want to check the impedance of your connection with the Muse 2, you may be able to use BrainFlow.
https://brainflow.org
It'll take a little coding, but the community is supportive