Should we link to https://mne.tools/stable/overview/implementation.html?highlight=units#id20 from the docstr?

Yes, feel free to add a proper RST target name to that header and link to it with :ref:. And also maybe link to

:term:`data_channels`

here https://mne.tools/dev/glossary.html since we also talk about SI units there a bit.

These are the units we use in MNE (are these all the units we use?) (ref)

    Volts (eeg, eog, seeg, emg, ecg, bio, ecog)

    Teslas (magnetometers)

    Teslas/meter (gradiometers)

    Amperes*meter (dipole fits, minimum-norm estimates, etc.)

    Moles/liter (“molar”; fNIRS data: oxyhemoglobin (hbo), deoxyhemoglobin (hbr))

    Arbitrary units (various derived unitless quantities)

I am a bit confused how we would handle derived units like Teslas/meter, or Moles/liter

would we allow scaling to Teslas/millimeter? If yes, how?

finally, what "multiples" and "submultiples" should we allow? All those that are in this list?

https://bids-specification.readthedocs.io/en/stable/99-appendices/05-units.html#prefixes

That'd be based on the CMIXF-12 format for representing units in ASCII

would we allow scaling to Teslas/millimeter? If yes, how?

Yes. Somewhere we already parse fractional units and figure out what the scale factors are to get to SI units. Let me dig it up...

These are the units we use in MNE (are these all the units we use?) (ref)

No, but it's probably "good enough" for most people as a reference. The complete list is probably provided by the FIF spec at the moment, because we expand it as we need to:

https://github.com/mne-tools/mne-python/blob/master/mne/io/constants.py#L786-L836

I would not worry about supporting all of these. Let's just get the data units we're likely to need. We can always expand later.

would we allow scaling to Teslas/millimeter? If yes, how?

Refactor code from here, put it in mne/utils/numerics.py or something similar, reuse in the test and in your code:

https://github.com/mne-tools/mne-python/blob/master/mne/tests/test_defaults.py#L25

I would even just start with this list of prefixes and known_SI. It should have 99% of what you need (hopefully).

Just wondering if we can also allow µV in addition to uV, or would that become messy?

Just wondering if we can also allow µV in addition to uV, or would that become messy?

I would suggest to allow u as a common ASCII abbreviation of micro, as well as the two (different!) unicode characters µ (U+00B5) and μ (U+03BC)

Also brings up this comment again where I was a little too slow to chime in that U+03BC is preferred over U+00B5 by the unicode consortium: #7497 (comment)

Is there any way to easily figure out which unicode characterI'm producing on macOS via Option-M (µ)?

See screenshot:

The first one is the micro sign, the second one is small greek mu

wow… thanks!

I thought I'd get to this PR this weekend, but unfortunately that won't happen.

I'll come back to this in a few weeks - if in the meantime somebody wants to pick it up, please be my guest.

EDIT: I won't have time for this for now. Please take it over if you care about this!

This may be a nice lead: https://github.com/mne-tools/mne-python/pull/8760/files#r559649851

closing in favor of #9136