Decimator problem

Post Reply
andpoho
Posts: 1
Joined: Tue Feb 09, 2010 9:52 am
Location: Geneva

Decimator problem

Post by andpoho »

Hi,

I try to use the Decimator executable to reduce size of files; because I have memory problems with EEGLab! However, when I decimate a dbf 2056Hz file to 1024Hz or less, the trigger channel is damaged, some trigger numbers are changed and sometimes new triggers appear! I looks like the Decimator 'decimate' the trigger channel like a normal signal channel recreating and consequently destroying the pulse square signal!
Any idea? I do something wrong?
My triggers are very short (1-5ms) square pulses send with the parallel port!
Thanks
A Posada

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

Post by Coen »

When decimating from 2048 to 1048 Hz, an OR function is applied on the triggers each successive group of 2 samples. So, if more than 1 trigger value exist within a group of 2 samples, then the trigger value in the eventual file will not be equal to any of the original trigger values.

When decimating a file, you must be sure that the zero time between triggers is larger than the time between the samples in the decimated file (> 1 millisecond in your example).

Best regards, Coen (BioSemi)

xicaque
Posts: 5
Joined: Fri Apr 25, 2008 5:39 am
Location: Canada

Post by xicaque »

Hi Coen,

We are decimating from 2048Hz to 128Hz and we're encountering a similar problem -- some of our triggers disappear.
How can you make sure that the time between triggers is larger than the time between samples in the decimated file? Is this something you can do in Decimator? If so, how?

Thanks

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

Post by Coen »

The only thing you can do is to choose a output sample rate with the time between samples larger than the minimum zero time between the triggers in the original file. So, in your case, 128 Hz is simply to low. Inspect the trigger repetition rate in the original file and choose a better (higher) output sample rate. By the way, 128 Hz sample rate would give you an analog bandwidth of only approx. 25 Hz. Are you sure that you can do the intended analysis when the data is reduced to such a low bandwidth ?

Best regards, Coen (BioSemi)

Post Reply