Reading Nihon files does not give correct voltage

[firejake308]

Describe the bug

I am using the mne.read_raw_nihon() function to read in an EEG file along with its accompanying 21E, LOG, and PNT files. However, when I plot() the raw data, some of the channels have voltages that are 10,000x the voltages of the other channels. I tried using a different tool to convert my Nihon file into an EDF, and after opening this EDF, I saw that the voltage amplitudes were much more reasonable. Without knowing too much about Nihon's file format, I'm guessing that there is some sort of unit_multiplier that MNE isn't reading in that the other tool did notice.

For my personal workflow, I can just use the other tool to generate EDFs and then read them into MNE, but I think it would benefit the project to address this bug.

Steps to reproduce

My file contains patient data, so I cannot share it at the moment. However, I did manually inspect the EEG file, and it looks like the header is EEG-1100C V01.00. Not sure if that helps

Expected results

Actual results

Notice how far off T7, T8, P7, and P8 are. I tried zooming out within MNE, and the signals look fine if you zoom out enough, so I'm pretty sure it's just a scaling issue

Additional information

Platform:      Windows-10-10.0.19041-SP0
Python:        3.6.4 (v3.6.4:d48eceb, Dec 19 2017, 06:54:40) [MSC v.1900 64 bit (AMD64)]
Executable:    c:\Users\seand\eeg_proj\venv\Scripts\python.exe
CPU:           Intel64 Family 6 Model 142 Stepping 9, GenuineIntel: 4 cores
Memory:        7.9 GB

mne:           0.23.4
numpy:         1.19.5 {blas=D:\\a\\1\\s\\numpy\\build\\openblas_info, lapack=D:\\a\\1\\s\\numpy\\build\\openblas_lapack_info}
scipy:         1.5.4
matplotlib:    3.3.4 {backend=Qt5Agg}

sklearn:       0.24.2
numba:         Not found
nibabel:       Not found
nilearn:       Not found
dipy:          Not found
cupy:          Not found
pandas:        1.1.5
mayavi:        Not found
pyvista:       Not found
vtk:           Not found

[welcome[bot]]

Hello! 👋 Thanks for opening your first issue here! ❤️ We will try to get back to you soon. 🚴🏽‍♂️

[firejake308]

Okay, I think I found the problem! The problematic channels are T7, T8, P7, P8, FT9, and FT10. The reason for this is in the definition of the function _map_ch_to_specs, which will use a different digital-to-physical conversion for channel names that are not in _default_chan_labels, and apparently these channels that I have are not default. My proposed solution would be to add these channels to the list of default channels, since I think that these labels will always be associated with EEG channels.

[agramfort]

PR welcome

thx !

[610265158]

I met the same problem. And it is solved by adding "T7, T8.." to the defaut _default_chan_labels. But what is this function "_map_ch_to_specs" doing? What is the meaning? I got no clue. And why there is different params when convert the data from int16 to float64?