Units powspctrm

Martin Bleichner bleichner.martin at GMAIL.COM
Mon Aug 16 10:46:15 CEST 2010

```Hi Arjen,

my point is that I am interested not in the relative difference between
signals but in the actual difference in microvolt. I know that this is not
the standard thing to do but that is what I need .

After playing aroud a bit i come to the following conlcusion

Lets say I have two signals (pure sine 20 Hz; sampling freq 512) one with an
amplitude of 100uV and one with an aplitude of 200uV.
Computing a power spectrum using matlab fft function I can eventually  read
out that the difference between my two signals is 100uV.

In ft_freqanalysis_mtmfft the signal is multiplied by a taper
line 340 dum=fft([data.trial{perlop}(sgnindx(sgnlop),:).*tap(taplop,:)

The output can hence not be transformed back right away to the real
microvolt values, but happens to be scaled (in my case ) by a factor of
~0.036 (which is the mean value of the taper)

At least that is what I think happens.

Regards
Martin

2010/8/13 a.stolk at fcdonders.ru.nl <a.stolk at fcdonders.ru.nl>

> If you use the same mtmfft settings for both conditions, in theory it
> should be fine. Make sure that you compute the fourier transformation on
> equal window lenghts.
>
> If you happen to have variable triallengths per condition, my suggestion
> would be to cut the trials in (e.g.) 1 second pieces. This way your
> 'frequency resolution' is the same for condition A and B.
>
> If your experimental setup allows you, you could choose to normalize your
> data (e.g. a baselinecorrection) first before making the A and B comparison.
> This way you correct for (non-)cognitive powerchanges over time (case it's
> not an ABABABAB etc. design). Such as conductivity reducal of the EEG paste
> over time, subject habituation effects and whatelse you have.
>
> Best,
> Arjen
>
>
>
>
> ----- Original Message -----
> From: "Martin Bleichner" <bleichner.martin at GMAIL.COM>
> To: FIELDTRIP at NIC.SURFNET.NL
> Sent: Thursday, August 12, 2010 10:48:14 PM GMT +01:00 Amsterdam / Berlin /
> Bern / Rome / Stockholm / Vienna
> Subject: Re: [FIELDTRIP] Units powspctrm
>
> Yes, typically, but is it the case for the powspctrm you get from the
> mtmfft
> freqanalysis?
> I am asking for the following reason:
>
> Lets say, I have two experimental conditions  A and B and I compute the
> powerspectra for both of them.
> When the units actually are uV^2/Hz i could tell for each frequency by how
> much uV A and B differ.
>
> What confuses me is that in
> ft_freqanalysis_mtmfft<
> http://fieldtrip.fcdonders.nl/reference/ft_freqanalysis_mtmfft>the
> signal is multiplied with a taper (tap) and thereby scaled.
> The resulting powspctrm hence does  not reflect the the true changes in
> microvolt.
> Do I see that correctly, or am I missing something here? Is the
> powerspectrum corrected for that somewhere else?
>
> Regards,
> Martin
>
>
>
> 2010/8/12 a.stolk at fcdonders.ru.nl <a.stolk at fcdonders.ru.nl>
>
> > Hello Martin,
> >
> > A powerspectrum typically has the frequencies on one axis and the power
> of
> > the corresponding frequencies on the other. The power represents how
> strong
> > a certain frequency occurs in a signal, given by the square of the
> voltage
> > per frequency.
> >
> > In your case: square of the microvolts per frequency: uV^2/ Hz
> >
> > Best regards,
> > Arjen
> >
> > ----- Original Message -----
> > From: "Martin Bleichner" <bleichner.martin at GMAIL.COM>
> > To: FIELDTRIP at NIC.SURFNET.NL
> > Sent: Thursday, August 12, 2010 5:52:19 PM GMT +01:00 Amsterdam / Berlin
> /
> > Bern / Rome / Stockholm / Vienna
> > Subject: [FIELDTRIP] Units powspctrm
> >
> > Hello together,
> >
> > What units does the powspctrm have?
> > Starting with an EEG signal in microvolts and computing the powerspectrum
> > using freqanalysis what are the units of powspctrm?
> > I tried to figure it out on my own but I am stuck somehow.
> >
> > Here some example code
> >
> > cfg=[];
> > cfg.method='mtmfft';
> > cfg.output='pow';
> > cfg.foilim=[1 130]
> > cfg.taper='hanning'
> > cfg.channel=1
> > cfg.keeptrials='no'
> >
> > R=ft_freqanalysis(cfg,data);
> >
> > I am interested in the units of R.powspctrm
> >
> > Thanks
> > Martin
> >
> > ----------------------------------
> > The aim of this list is to facilitate the discussion between users of the
> > FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG
> > http://listserv.surfnet.nl/archives/fieldtrip.html and
> > http://www.ru.nl/neuroimaging/fieldtrip.
> >
> > ----------------------------------
> > The aim of this list is to facilitate the discussion between users of the
> > FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG
> > http://listserv.surfnet.nl/archives/fieldtrip.html and
> > http://www.ru.nl/neuroimaging/fieldtrip.
> >
>
> ----------------------------------
> The aim of this list is to facilitate the discussion between users of the
> FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG
> http://listserv.surfnet.nl/archives/fieldtrip.html and
> http://www.ru.nl/neuroimaging/fieldtrip.
>
> ----------------------------------
> The aim of this list is to facilitate the discussion between users of the
> FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG