FAQ

SBART RVs have very different RVs than the CCF

Description: In typical cases, the S-BART algorithm is less prone to numerical errors than the CCF. However, the classical algorithm might be affected by them, under some edge cases. In such cases, the uncertainty of a given spectral orders goes to very small values, having the larger weight in the (classical) RV estimate.

Solution: For such cases, check inside the RV_step/individual_subInst/<instrument>/plots/ for the RV_orderwise_errors.png file, where we can find a plot of the RV uncertainty for each order. If we can clearly see a value much smaller than all others, rejecting that order might improve the overall solution.

SBART RVs have very large uncertainties

Description: It is often the case that a single observation is leading to the rejection of a large number of spectral orders. Recall that we enforce that any given sub-Instrument uses exactly the same spectral orders.

Solution: For such cases, check inside the RV_step/individual_subInst/<instrument>/metrics/ for the DataRejectionSummary.png file, where we provide the reason behind the rejection of each spectral order of each observation. At the end of the file, there is also a short summary of the rejection.