HARPS optics upgrade

[megbedell]

HARPS had a major upgrade in July 2015 where fibers were replaced (see this Messenger article for details). In the derived RVs, both from the HARPS pipeline and from wobble, this manifests as an offset of about -15 m/s after the upgrade.

Long-term I think this means we should treat the pre- and post-upgrade spectra as coming from different instruments with different stellar (and telluric??) models. This does mean the post-upgrade data won't be very useful because we don't have many epochs to use. A short-term hack might be to just build an effective post-upgrade RV offset as a free parameter in the model we use to get the final time-series RVs. Or we could just side-step the problem for now and only use pre-upgrade spectra in Paper 1.

[davidwhogg]

I agree we should treat these as different instruments (for now). Eventually, when we have a method for permitting tellurics and spectra to change with time, we might be able to do everything at once, but that time is in the future.