Gain and offset for physical values

Post Reply
apresacc
Posts: 4
Joined: Wed Nov 25, 2015 6:16 pm

Gain and offset for physical values

Post by apresacc »

Hello,

Apologies if this question sounds a bit silly.

So I have just realized that when converting the digital data into physical values I was not using the offset. I see that the offset is not mentioned in the page where the structure of the data for Biosemi is described (https://www.biosemi.com/faq/file_format.htm). But the equation for the gain is reported. Now my question is the following: isn't the offset mentioned because it is something superfluous (its value is very small and therefore basically negligible) and doesn't affect the data (and when using any mode, but the special mode 9, you have the reference, so its value should cancel out)? Or is it something that you recommend adding and it is not mentioned in the web-site because it should be common knowledge?

Thank you for your help.

Best,
Alex.

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

Re: Gain and offset for physical values

Post by Coen »

It is a DC amplifier, data is not high-pass filtered. Offset is simply the low-passed value of the data (cut-off frequency < 1 Hz)

ABR (speedmode 9) saves 5 channels in ActiView 813 and 900: 3 DC monopolar (unreferenced) low-frequency channels and 2 AC bipolar (referenced) high-frequency channels. Offsets can be extracted from the 3 LF monopolar channels, also see https://www.biosemi.com/download/ABR.pdf .

Best regards, Coen (BioSemi)

apresacc
Posts: 4
Joined: Wed Nov 25, 2015 6:16 pm

Re: Gain and offset for physical values

Post by apresacc »

Thank you for your quick response. I think I might have not explained correctly my question or maybe I misunderstood your answer.

Bear with me, so I can be on your same page.

So, in my question I was not referring to the DC voltage, which I know is present because the data are not high-passed. I was referring to the offset that I see people are using (see eeglab for instance) with the edf format when converting data from digital to physical values. I don't believe they can be the same thing, because the DC offset is huge, while the offset that they use in this equation is basically negligible.

Basically, I have always used this equation: "physicalValue = gain * digitalValue", but then I realized that I should have added the "offset = physicalMax - gain * digitalMax", which would give me the following value: "physicalValue = gain * digitalValue + offset". I have also checked what the folks from eeglab are doing and they add this offset to the equation. Now, this offset is so small, that you cannot appreciate any difference in the data, but I would still like to know which is the correct equation that I should use with Biosemi data to convert them into a physical value.

As for ABR is speedmode 9, I have the old version of actiview and can only read the 2 AC bipolar (referenced) high-frequency channels + the triggers. But thank you for sending me the documentation, because I have not kept myself updated with the changes made to speemode 9 :-)

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

Re: Gain and offset for physical values

Post by Coen »

The BDF file header defines the gain of each channel with the physical and digital minimum and maximum values. The file format does in principle support minimum and maximum values that are not symmetrical around zero. I am not familiar with the EEGlab routines, but I suspect that the "offset" parameter is added to handle files with asymmetrical physical and/or digital values.

All BDF files generated by BioSemi ActiView always have physical and digital values symmetrical around zero. The "offset" parameter is then close to zero (< 1 uV), with the small value caused by round-off errors. You can ignore the "offset" parameter.

Acquisition of ABR (speedmode 9) is now integrated in standard ActiView (version 813 for RTE 2016, version 900 for RTE 2021). The special ActiABR software that was previously used for ABR is not needed or supported anymore.

Best regards, Coen (BioSemi)

apresacc
Posts: 4
Joined: Wed Nov 25, 2015 6:16 pm

Re: Gain and offset for physical values

Post by apresacc »

Thank you so much for your quick and detailed explanation. The small value of the offset that you mention is exactly in line with what I found (a fraction of uV), so it is negligible. I just wanted to have a confirmation from you guys, because I trust your expert opinion. I occasionally use eeglab when I want to check my code. It is always better to have a "second opinion".

And thank you for the info about the special mode 9.

As usual, I am grateful for all the technical help that you have provided with me since I started using Biosemi in 2013 and want to let you know how much I appreciate the quality of your EEG system, which I consider to be the best one that I have ever used.

Best,
Alex.

Post Reply