filter_length distortion warning

💬 Support & Discussions
, ,
1 
MNE version: 1.0.2
MNE-bids version: 0.10
operating system: Ubuntu 20.04

Hello,

when preprocessing my fNIRS data, I’m applying a bandpass filter and get a warning like the following:

FIR filter parameters

Designing a one-pass, zero-phase, non-causal bandpass filter:

  • Windowed time-domain design (firwin) method
  • Hamming window with 0.0194 passband ripple and 53 dB stopband attenuation
  • Lower passband edge: 0.01
  • Lower transition bandwidth: 0.01 Hz (-6 dB cutoff frequency: 0.01 Hz)
  • Upper passband edge: 0.50 Hz
  • Upper transition bandwidth: 0.20 Hz (-6 dB cutoff frequency: 0.60 Hz)
  • Filter length: 5371 samples (660.096 sec)

/tmp/ipykernel_226266/3535447880.py:112: RuntimeWarning: filter_length (5371) is longer than the signal (4248), distortion is likely. Reduce filter length or filter a longer signal.
raw_haemo_filt = raw_haemo.filter(0.01, 0.5, h_trans_bandwidth=0.2, l_trans_bandwidth=0.005)

While I’ve searched through the forum, my problem is neither that any epoch is too short oder that I have to transpose anything. So I’m a bit stuck on how this problem arises and how I can influence the filter length. Surprisingly, I have a second data set with a different task for which there are no warnings put out, but the prepro code is the same (bad and short channel exclusion, motion correction, OD2haemo, bandpass filter).

I hope you can help me on this. What more info do you need?

Thanks in advance,
Dani

2

Can you show the command that you use to create that filter?

It looks like you require a very steep filter, so its length is very long. Also, since you are designing a bandpass filter it might be worth trying two separate filters (i.e. separate lowpass and highpass filters).

2 Likes
3

I posted the command already, it’s

raw_haemo_filt = raw_haemo.filter(0.01, 0.5, h_trans_bandwidth=0.2, l_trans_bandwidth=0.005)

But I see, now that you mentioned a “steep” filter, I tried adjusting the lower transition bandwidth from 0.005 to 0.0075 and ta da, the filter length got shorter which means it’s working now. Thanks for the hint!

On another note: I also applied two separate filters, but that didn’t change a thing.

I guess my homework will be some signal processing theory, cause I still don’t have a clue how a filter length is determined. But thanks again!

1 Like