BUG?: KIT coils have units FIFF_UNIT_T, not FIFF_UNIT_T_M

[larsoner]

The unit for all channels, including FIFFV_COIL_KIT_GRAD channels (and FIFFV_COIL_KIT_REF_MAG, which makes sense), is FIFF_UNIT_T. Is this a typo? It seems like KIT grads should have units FIFF_UNIT_T_M. Not having the correct units is making the picking functions think they're magnetometers, and thus breaking Maxwell filtering, which needs to distinguish between grad and mag to exclude external components. I can make the change in the KIT code, but then it won't match what MNE-C's conversion routines give. @mshamalainen @teonlamont any ideas?

[larsoner]

From reading a bit about KIT sensors, they do talk about converting the data to Tesla units (not Tesla / m), so it seems correct. But the channels behave like gradiometers according to Maxwell filtering (first degree external components are noisy around zero). I'd still like to hear if you think UNIT_T is the right unit, but now I'm also curious that, if it is, should doing pick_types(meg='grad') also return these channels? I think it probably should, even if the units don't match other grad channels...

[larsoner]

Scratch that, I was reading about CTF sensors...

[agramfort]

I think that KIT are like for CTF axial gradiometers; The second gradiometers are far from the head making the units of the gradiometer very close to a magnetometer. @mshamalainen I learnt my lesson :)

[kingjr]

The second gradiometers

"magnetometers".

I learnt my lesson :)

Close..

[larsoner]

So it's correct for the picking functions to treat these as mags? Any other coil types that have this "feature" that you're aware of?

[agramfort]

yes as the scales for plots etc should use mag defaults.... I was puzzled too a few months ago...

[larsoner]

Thanks @agramfort