[Eeglablist] Weird Patterns in the Data: Bad Power or Ground?

[Malte Anders]

Hello Nathan,

good to see somebody struggling with gTec Equipment and power line issues
as well. I have been struggling in 2017 to get our g.HIamp going as we had
huge power line artifacts. I have spammed the EEGLAB list with walls of
text as well, but then solved the problem by complete coincidence when I
actually touched a ground connector while having the EEG mounted to myself.
Keep in mind that I could not see your EEG-Data as the EEGLAB list does not
allow email attachments, you can however email me the Image directly and
also send an imageshack-link or something like that to the list so
everybody can see.

Here are my 2 cents for your problem: I think your subject is not properly
grounded. The g.USBamp seems to have 4 ground connectors, so with 4 amps
you have 16 ground connectors... Do you have all of those connected to the
subject?
Some suggestions:

-I know that some people might think this is stupid, but try to connect the
subject to the same ground as the amps. You can snitch a ground wire
directly from the power outlet (dangerous!) or maybe from some medical
equipment that is also connected somewhere in the room. For starters, a
bare wire that is connected to a ground connection on the one end and the
blank part of the other end is held by the subject in his/her hand is
sufficient.
Ground connections do look like this:
https://www.mikrocontroller.net/attachment/246107/bild1.jpg
This is the symbol for protective earth inputs:
https://cdn3.iconfinder.com/data/icons/electronic-with-circle/64/ground-512.png
In my case, the EEG recording looked like garbage when I tried recording
without any further grounding measures. As soon as the subject simply
touched any grounding spot in the room, the magic happened and boom, the
EEG looked beautiful. Without proper grounding, common mode rejection fails
and in this case it would be helpful to see your actual recording images.

-Connect only one (!) electrode to each amp, so 4 electrodes in total. Then
connect all the corresponding 4 ground wires from the amps. You should now
have connected 8 electrodes to the patient, 4 regular ones and 4 ground. I
don't know if the g.USBamp allows you to record like this but with the
HIamp its possible, as the ground wire is then used as reference electrode.
If you need to specify another reference electrode, connect 4 more
electrodes that correspond to the same ground wires you have connected.
Either way, try to record with as little electrodes connected to the amps
as possible (8 or 12) and look if the data looks better. The reason for
this: g.Tec hardware gets exremly picky if one electrode fails. In the
g.HIamp, if one electrode is broken, you _seem_ to get good data but the
output is completly nonsense. There is no way to know if one electrode has
failed other than doing impendance measurements from time to time (all
black: please find the on electrode out of 64 that screws up everything,
have fun) or having a close look at the amplitude values. With the g.HIamp,
amplitude values of 1 µV or less (even nV are plotted) are a good indicator
that one of your expensive electrodes went bad.

-Does the g.USBamp provide an interface for impendance measurements? If
yes, what are the results? Do you get any readings or does impendance
measurement fail?

Heres some further reading:
https://en.wikipedia.org/wiki/Ground_loop_(electricity)

Keep me posted, we'll figure this out. I've been through a lot with g.Tec
hardware, the only good news is: once its up and running, you'll get really
good data.

Am Mo., 27. Jan. 2020 um 17:20 Uhr schrieb Nathan Sanders <nesander at ncsu.edu
>:

> Hey Everyone,
>
> I am in the middle of trying to diagnose what I think is some sort of
> hardware issue and I could use the benefit of your experience. I apologize
> for the wall of text but please bear with me!
>
> Backstory: originally I had all four of our 16-channel g.USBamps plugged
> into a single AC adapter with a four-way splitter. I have been able to get
> really nice data in the past, but recently channels have begun to flatline
> one after another until they were all flat. This intermittent problem has
> been getting steadily worse until a few days ago I couldn't get any signal
> at all. So I replaced the sync cables, but that did not solve the problem.
> I then discovered that this flatline problem would *not* occur if I only
> tried to collect data from one amp at a time (16 channels). Interestingly,
> the problem would come back if I turned on another amp (even if I was still
> only collecting data from the first amp), but it would go away again if I
> disconnected the ground jumper between the two (even if the second amp
> remained on). That led me to think that maybe our AC adapter was going bad
> or had a bad ground. We had three spares, so I tried hooking up the master
> amp with the suspected bad adapter, and the other three slave amps with the
> spare adapters. I was able to get a signal in all 64 channels (progress!).
> I recorded 5 minutes of resting state data as a test and this is what it
> looked like (see two attached photos)...
>
> data_filt_noreref.png shows what the data looked like right after
> filtering, *but before re-referencing to average*. All the channels look
> suspiciously correlated.
>
> After re-referencing to average, an interesting pattern emerges. You can
> see that the correlation pattern appears in groups of four channels. These
> channels correspond to the A, B, C, and D groups in each amplifier.
>
> I ran ICA and did not see anything that looked like a brain source.
>
> At first I thought maybe I had the channel locations mapped incorrectly,
> but I double checked and they are correct. I now think it has something to
> do with the AC adapter -- maybe it's failing or there is a bad ground or
> something and it is somehow contaminating the signals via the ground
> jumper. I know that bad power supplies and grounds can cause weird problems
> with computers and ham radio equipment...
>
> Has anyone ever seen anything like this? Any comments or suggestions would
> be greatly appreciated. Maybe they will be useful to someone in the future
> as well.
>
> Thanks as always,
>
> Nathan
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>


-- 
Mit freundlichen Grüßen,

Malte Anders