[fitsbits] Appropriate BUNIT for ADUs vs electrons?

[James Turner]

We would like to use BUNIT correctly to distinguish between images
stored in ADUs and those which have been multiplied by GAIN to give
units of electrons. From the FITS standard, tables 3-4, the allowed
units for counting events are "count", "photon" and "adu" (but for
some reason not "electron"). The standard doesn't seem to elaborate
on how it's appropriate to use these. Can anyone clarify what would
be the expected unit for representing photoelectrons, please? Maybe
the intention is to use "photon", but of course they're not quite
the same thing -- even if you consider the QE curve as just part of
the optical throughput, I believe that some detectors, such as
superconducting tunnel junctions, can produce many electrons per
incident photon. So it's not 100% obvious to me that "photon" is
appropriate. If, on the other hand, the intention is to use "count",
that would seem rather ambiguous with respect to ADUs.

What do other observatories do?

Thanks for any suggestions!

James.