Sky subtraction with the HPF sky fiber is a little subtle for two reasons:
The target and sky fibers have slightly different wavelength solutions. For non-PRV applications those differences may be negligible.
Subtracting a noise-free array from a Spectrum1D object does not propagate uncertainty, whereas subtracting a Spectrum1D from another Spectrum1Ddoes propagate the uncertainty.
The new .sky_subtract() method introduced today solves this Issue, at least for that one method. We will continue to look for places where error propagation falls short.
Sky subtraction with the HPF sky fiber is a little subtle for two reasons:
Spectrum1D
object does not propagate uncertainty, whereas subtracting aSpectrum1D
from anotherSpectrum1D
does propagate the uncertainty.I think the bottom line accounts for both effects, but I'm not sure about how it does the wavelength interpolation.