Scaling of GSR data

Post Reply
dkrehbiel
Posts: 1
Joined: Thu Jun 18, 2009 4:19 am
Location: Bethel College, KS, USA

Scaling of GSR data

Post by dkrehbiel »

Hello ActiVIEW/ActiveTwo users,

I have an ActiveTwo system with a 16 Hz GSR module. I am confused about the scaling of the resulting data. I am currently reading the BDF files with EEGLAB (MatLab-based), selecting the GSR channel and saving the data as an ASCII file. My problem is that the resulting values are in the range of 20 some thousand, i.e. a far larger number for values in microsiemens than I ought to have. Any suggestions? Should I be referencing the GSR channel to something else (I have mastoid electrodes that I use for EEG recording at the same time)? Do I need to use some other software to read the data? If so, do you have suggestions (preferably cheap ones)?

Thanks for any clues.

Dwight K.

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Post by Coen »

The GSR data for the 16Hz version is stored in nanoSiemens with a scaling of LSB = 31.25 nS (1/32th uS). This information is found in the file header. So, a value of 20,000 represents 0.625 microSiemens (= 1600 Ohm)

Best regards, Coen (BioSemi)

cmuehl
Posts: 4
Joined: Mon Mar 17, 2008 2:21 pm
Location: Twente (NL)

Post by cmuehl »

Dear Coen,

I believe that you mean 0.625 milliSiemens instead of microSiemens.

You mention that the smallest unit that can be measured (as I underrstand this is the implication of the LSB) is 31.25 nanoSiemens (nS). In my understanding that makes 20000 units = 20000*31.25 nS = 625000 nS = 625 microSiemens (uS) (which are indeed 1600 Ohm).

Please correct me, if I made a mistake at one point, as I try to understand a strange phenomenon in my data:

I am recording GSR during computer games. The signal looks good in its (smooth) form. However, the skin conductivity responses that I see are between 5 to 1000 microSiemens (sometimes higher). In the literature I find the notion of a normal phasic SC response being smaller by a factor of 10 or 100, i.e. between 0.5 and 10. I am using EEGlab for import of the .bdf, but also checked with BDF reader and EDF reader, so I exclude a mistake in conversion of units.

Could differences in the recording equipment used in different labs lead to the higher scaling of skin conductivity responses I observe?

Best,
Christian

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Post by Coen »

Thanks for your comment, "micro" instead of "milli" was indeed a typing mistake.

The absolute values of GSR impedance are depended on many factors such as skin prepreration, electrode material, and frequency of the excitation current (16 Hz in our case). Large variations between subjects and experimental setups are indeed seen. By the way, you can easily check the calibration of the GSR with some resistors of know value. (connect resistor between 2 GSR electrodes, connect CMS and DRL to one of the GSR electrodes)

Best regards, Coen (BioSemi)

erfan.y
Posts: 11
Joined: Wed Jul 06, 2011 11:01 am
Location: Sweden

GSR Value

Post by erfan.y »

Hi
I am having quite same problem but with numbers larger than 20 some thousands.

I am getting the data for GSR with my own data acquisition program. The value that I get from the ring buffer-the original value without any multiplication- is varying from 8 digit number to 9 digit number e.g. 194070016.

I divide this value by 256 and 1000 and multiple by 0.032 to get the us value.

but after all division what i get is around 20 to 50 thousands.

I am using the speed mode 4 which i guess it gives me 2khz, and placing the sensors on my hands.

please tell me what is wrong.

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Post by Coen »

What are the values displayed by ActiView ?

Best regards, Coen

Chanel
Posts: 17
Joined: Thu Dec 08, 2005 11:03 am

Post by Chanel »

Dear Cohen,

I also have a question on the SCR scale. I did import some biosemi bdf data in matlab. But when I check the minimum difference (except 0) between two SCR samples I obtain 0.0321 (although the header indicate nS).

In this case is this safe to consider that my recording/file unit is actually microSiemens ?

Guillaume

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Post by Coen »

Your BDF file uses nS as unit for the GSR with a scaling of LSB = 31.25 picoSiemens. So, the minimum difference between samples ( = 1 LSB) is 0.03125 nanoSiemens instead of 0.0321 nS. It is safe to consider that your Matlab software converts the 24-bit integers from the BDF file to floating point values in nanoSiemens by using a slightly wrong conversion factor (31.15*LSB instead of 32*LSB)

Best regards, Coen (BioSemi)
Last edited by Coen on Wed Nov 23, 2011 2:45 pm, edited 1 time in total.

Anders
Posts: 6
Joined: Tue Oct 11, 2011 2:56 pm
Location: Sweden

Post by Anders »

I also have problems understanding my GSR data and instead of creating a new thread I thought I could post here.

Here http://www.mypicx.com/11232011/bdfRead/ is a picture of the GSR signal as seen using BDFreader. Note the 50 s time interval and 500 "microVolt" y-axis.

I then converted the data using polyrex to .cnt and try to analyze in analyzer. The character of the signal looks the same, but for example in the time interval in the picture the peak value is 12633 and lowest value 12234. As I understand it, the signal is recorded in nanoSiemens. Thus I should divide the signal with 1000. So my signal then varies between 12.6 and 12.2 microSiemens in the first 50 s. The most extreme value I have in the whole recording is 17.2 microSiemens.

My question is, can this be correct?

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Post by Coen »

The change in conductance is approx 0.4 uS. This can be correct when the absolute conductance values are quite low (electrode/skin resistances in the order of 100 kOhm), but would be suspiciously low for more typical conductances (with resistances in the 10 kOhm range)

I believe that Polyrex applies a (optional ?) baseline correction to fit the 24-bit data within the 16-bit CNT format. If this correction is indeed applied, then you have no information about the absolute conductance and resistance in Polyrex. However, the conductance changes in Polyrex seems to match the values displayed in BDF-reader (0.4 uS is displayed as 400 uV in BDF-reader). BDF-reader does not show the absolute values either (but can be easily modified in LabVIEW to do so).

I recommend to make a test file with two known resistors, and compare the (absolute) values displayed by ActiView in the left sidebar during the recording with the values displayed in the various review programs.

Best regards, Coen (BioSemi)

naomi
Posts: 2
Joined: Wed Feb 03, 2016 9:28 am

Re: Scaling of GSR data

Post by naomi »

Just to make sure, is it true as well for the 24Hz version that the GSR data is stored in nanoSiemens with a scaling of LSB = 31.25 nS (1/32th uS)?

Many thanks

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Re: Scaling of GSR data

Post by Coen »

The signal from the 16 Hz GSR is stored in 24-bit BDF files with nS as unit and a scaling of LSB = 31.25 picoSiemens.

Coen (BioSemi)

naomi
Posts: 2
Joined: Wed Feb 03, 2016 9:28 am

Re: Scaling of GSR data

Post by naomi »

Dear all,

I need to score SCR in a software, and I need to score the ones between .05 and 10 microsiemens. However, my software only gives me microVolt. Hence, my question: what is .05 and 10 microsiemens in microvolt? Then I could just use these numbers as thresholds...

Many thanks.

Coen
Site Admin
Posts: 1124
Joined: Fri Mar 26, 2004 7:00 pm
Location: Amsterdam, Netherlands
Contact:

Re: Scaling of GSR data

Post by Coen »

1 microVolt = 1 nanoSiemens

Coen (BioSemi)

Post Reply