bug in acquiring data from g.Nautilus via g.needAccessServer

Obtaining data from various hardware devices
Post Reply
fd301
Posts: 3
Joined: Thu Nov 30, 2017 5:58 pm

bug in acquiring data from g.Nautilus via g.needAccessServer

Post by fd301 »

We have problems with the acquisition of data using gtec nautilus and openvibe.
It seems that openvibe version 1.3 is more stable with the acquisition. With version 2 we have big drift and the number are unrealistically big.
I tried acquiring data on windows 10 (g.NEEDaccess Server V1.16.00) [Intel(R) Core (TM) i7-6700 CPU @ 3.40GHz 3.41GHz, 16GB RAM]. I did some testing with a dry cap and I expected the impedance to be relatively high.
OpenVibe 1.3: I get values mostly around -200mV to 200mV
[img]http://www.doc.ic.ac.uk/~fd301/tmp/gtecOpenvibe1_3.jpg[\img]
with openvibe 2 we got unrealistically big numbers and lots of drift:
[img]http://www.doc.ic.ac.uk/~fd301/tmp/gtecOpenvibe2.jpg[\img]

I tried also windows 7 with (g.NEEDaccess Server V1.14.01 as well as g.NEEDaccess Server V1.16.00).
It seems that internally openvibe doesn't mind and you can move on and try some motor imagery classification scenarios that work. Another possibility is that there is also something wrong with the conversion of the ov files to csv files. However, by looking on the real-time drift feedback (openvibe server) I could tell that things were worse.

I cannot say with certainty that openvibe 1.3 is working because it has similar problems when I tried it elsewhere but openvibe 2 hasn't worked properly neither on windows 7 nor on 10.

Any help would be appreciated.

jtlindgren
Posts: 775
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: bug in acquiring data from g.Nautilus via g.needAccessSe

Post by jtlindgren »

Hello,

lets be sure of the terminology first.

In OV, drift means 'time drift'. It indicates a situation that the amplifier and the acquisition server do not agree on the elapsed time: how many samples/second there should be. In ov 1.3 -> ov2.0, drift correction is no longer enabled by default and hence AS may show more drift. If this is an issue, you can try to enable it again. This should give you the previous behavior.

What I see in your images however is baseline or 'voltage drift', i.e. the signal doesn't oscillate around zero. I'm not sure why the signal scale itself would have changed from 1.3 to 2.0. The baseline drift is normal (at least we've noticed it with Camille Jeunet's gusbamp). The baseline is usually removed by applying temporal bandpass filter to the signal, e.g. [1,40]hz.

These two 'drifts' are not at all the same thing. AS doesn't know, measure or try to correct anything about 'voltage drift'.


Hope this helps,
Jussi

fd301
Posts: 3
Joined: Thu Nov 30, 2017 5:58 pm

Re: bug in acquiring data from g.Nautilus via g.needAccessSe

Post by fd301 »

This terminology makes perfect sense.
So you think that time-drift and voltage-drift are completely independent issues. In my case, seems that when things work I have less time-drift and no voltage drift.
I also enabled the driver drift correction and this helped a bit but as you say the voltage drift hasn't changed.

Is there any case you test g.nautilus drivers with openvibe 2 to get an insight on the voltage drift issue? I believe that the issue is going to be reproduced easily.

Many thanks for your help.

jtlindgren
Posts: 775
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: bug in acquiring data from g.Nautilus via g.needAccessSe

Post by jtlindgren »

We don't have the device here at the lab so we cannot give it a try, but if you hint about the issue to gtec, they might find the time to give it a test run. They contributed the driver.

What we can do however is to test with some other devices that they give us the same behavior with 1.3 and 2.0. I'll get back to you about it next week.

Edit: The Acquisition Servers in 1.3 and 2.0 should be compatible. You should be able to run 1.3 server and send data to 2.0 Designer. It'd be interesting to see if the issues persist this way or not.


Cheers,
Jussi

fd301
Posts: 3
Joined: Thu Nov 30, 2017 5:58 pm

Re: bug in acquiring data from g.Nautilus via g.needAccessSe

Post by fd301 »

thanks for looking into this.

I tried using acquisition server v1.3 with designer v2. It seems ok.
I think the problem is on the server side only.

gtec/openvivo says they don't officially support openvibe. I'm not sure whether I could persuade them to check it out. In mean time, I could try build openvibe from source and have a look.

Best Regards,
Fani

jtlindgren
Posts: 775
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: bug in acquiring data from g.Nautilus via g.needAccessSe

Post by jtlindgren »

Hello again,

one thing that comes to mind is that 1.3 and 2.0 use separate setting files. The settings are not automatically migrated. Have you checked that both of them use the same gusbamp amplifier filters? These settings are in 'Driver properties'. The voltage drift seems more likely if no bandpass filter has been configured in the 2.0 driver. The 'no filter' is the default.


Cheers,
Jussi

Post Reply