Nightly correction differences account for 98 percent of the fitted BERV slope in 138 Ross 128 HARPS exposure pairs

We test whether the DRS minus SERVAL velocity difference for Ross 128 follows barycentric Earth radial velocity (BERV) in HARPS-RVBank version 02. In 138 later exposures, the corrected difference has a fitted slope of 0.05056 m/s per km/s after year offsets. The difference between nightly zero-point columns accounts for 0.04945, or 97.8 percent of this slope. Removing that term leaves 0.00112 with a conditional 95 percent interval of -0.00776 to 0.01000. The raw difference has a slope of 0.00017. A primary 10-day block test gives an adjusted fraction of 0.000977, but 30-day blocks and time terms weaken the evidence for a BERV relation. In 134 older exposures the slope is unresolved, and transfer between 2013 and 2014 gives no error benefit. The column decomposition identifies where the later relation enters the paired products. It does not establish that a correction is wrong, find a new planet, or identify a detector or stellar cause. Both reductions share spectra. The later span is only 91 days. The package includes code, aggregate results, source hashes and retrieval steps. External source records remain outside the baseline package.

astronomy · Ross 128 · GJ447 · HARPS · BERV · nightly zero point · paired velocity differences · data quality