Avoid lowpass=0 in brainvision data

[Lychfindel]

Reference issue

Example: Fixes #10516 .

What does this implement/fix?

If in the header high cutoff is 0, then the lowpass filter is considered as disabled.

[agramfort]

no opinion here. @sappelhoff thoughts?

[sappelhoff]

I have never used a V-Amp or analyzed data produced by it, but the proposed changes look reasonable to me. If @Lychfindel can vouch for the V-Amp data being organized this way, I think it's fine to go ahead.

@Lychfindel did you record the data yourself and made sure the 0 gets entered when you
have a BrainVision Recorder setting to disable software filters?

Note that hardware filters are surely still in place, even if neither the data nor the recording interface (I assume BrainVision Recorder) tells you about that. That's just a necessity to prevent aliasing of the signal when digitizing to some sampling frequency.

I guess we need to distinguish 3 types of filters:

1. hardware --> this filters the data during / prior to digitization of the analogue signal
2. software 1 --> this is applied in the recording interface (e.g. BrainVision recorder) and directly applied to the recorded data, cannot be reversed
3. software 2 --> this is applied e.g., in MNE-Python to already recorded and saved data ... can be changed and reversed at will

Example for 2. --> the Brain Products "Brain AMP DC" always samples at 5kHz -- even when you set the sampling frequency to 1000Hz in the recording interface. "Under the hood", the recording interface just applies an additional software filter and downsampling process prior to saving the data. This has been confirmed to me by Brain Products in a personal Email. :-)

[Lychfindel]

Yes, I recorded the data myself even last week, and I always disable software filters. The schema in the hdr file is always the same:

Brain Vision V-Amp Data Header File Version 1.0
...
Channels
--------
#     Name      Phys. Chn.    Resolution / Unit   Low Cutoff [s]   High Cutoff [Hz]   Notch [Hz]    Gradient         Offset
1     Fz          1          0.0488281 µV             DC                0              Off
2     Cz          2          0.0488281 µV             DC                0              Off
3     C3          3          0.0488281 µV             DC                0              Off
4     C4          4          0.0488281 µV             DC                0              Off
5     T7          5          0.0488281 µV             DC                0              Off
6     T8          6          0.0488281 µV             DC                0              Off
7     Pz          7          0.0488281 µV             DC                0              Off
8     P3          8          0.0488281 µV             DC                0              Off
9     P4          9          0.0488281 µV             DC                0              Off
10    EOGL        10         0.0488281 µV             DC                0              Off
11    EOGR        11         0.0488281 µV             DC                0              Off

[sappelhoff]

cool, then I'd say: please add a detailed changelog entry about this bugfix and we can merge. CI failures seem unrelated

If you wanna go the extra mile, you'd record a new datafile with software filters ON, and check how that looks in the data and whether we can read it in mne.

[Lychfindel]

In the V-Amp is not possible to set hardware filters, but I did three more test recordings: without software filters, with software filters, with software filters except in one channel. All of them are correctly read by mne and in the set with one channel not filtered I receive the expected warnings

<ipython-input-10-4e1c2c12eafd>:1: RuntimeWarning: Online software filter detected. Using software filter settings and ignoring hardware values
  raw = read_raw_brainvision(vhdr_path)
<ipython-input-10-4e1c2c12eafd>:1: RuntimeWarning: Channels contain different highpass filters. Lowest (weakest) filter setting (0.00 Hz) will be stored.
  raw = read_raw_brainvision(vhdr_path)
<ipython-input-10-4e1c2c12eafd>:1: RuntimeWarning: Channels contain different lowpass filters. Highest (weakest) filter setting (250.00 Hz, Nyquist limit) will be stored.
  raw = read_raw_brainvision(vhdr_path)

V-AMP_test.zip

[sappelhoff]

one nitpick, other than that, +1 to merge

[larsoner]

@Lychfindel I used the :newcontrib: above and realized we needed an entry in names.inc, feel free to push to change the URL if this is not the one you want!

[Lychfindel]

Sure thanks a lot, then I will remove it from my other PR!

[welcome]

🎉 Congrats on merging your first pull request! 🥳 Looking forward to seeing more from you in the future! 💪

[larsoner]

Thanks @Lychfindel !