Hi Rik,
This has been discussed several times on the list before and I even
fixed the manual recently to discourage the use of bandpass filters
with this kind of specifications. The thing is Matlab actually tries
do design one filter with these properties so it needs to have wide
pass band on the one hand and very close to zero low cut-off. I don't
have the exact explanation with poles and zeros but it seems that what
tends to happen is that at some point these filters become unstable
and that's when you get those large values. So my recommendation is
only use bandpass filters to pass a narrow band (like alpha).
Otherwise use high-pass and low-pass sequentially (preferably in this
order to remove DC first and avoid low-pass filter ringing at the
edges).
Vladimir
On Tue, Jul 13, 2010 at 11:38 AM, Rik Henson
<[log in to unmask]> wrote:
>
> Excuse my ignorance, but what is the theoretical reason why bandpass
> filtering (whether implemented via matlab in SPM8rel, or via ft_preproc
> functions in SPM8devel) a 250Hz-sampled continuous file with bandpass [0.1
> 20] Hz produces near-infinite values, whereas the same data with a bandpass
> of [0.1 40] produces reasonable values – is this something to do with the
> profile (roll-on/off) of the bandpass filter?
>
> Thanks
> Rik
>
> -------------------------------------------------------
> Dr Richard Henson
> Assistant Director, Neuroimaging
> MRC Cognition & Brain Sciences Unit
> 15 Chaucer Road
> Cambridge, CB2 7EF, UK
>
> Office: +44 (0)1223 355 294 x522
> Mob: +44 (0)794 1377 345
> Fax: +44 (0)1223 359 062
>
> http://www.mrc-cbu.cam.ac.uk/people/rik.henson/personal
> -------------------------------------------------------
>
>
>
|