“…For each quasar, the extracted flux from all exposures was combined with variance weighting using _ (Murphy 2016, version 0.73) to form a final, 1-dimensional "master spectrum", including a 1σ uncertainty spectrum. Evans et al (2014) and Murphy et al (2016) describe the details of this process. Briefly, for our HDS spectra, the most important aspects are: (i) each echelle order, of each exposure, was redispersed onto a common, log-linear vacuumheliocentric wavelength grid with a 1.75 km s −1 pix −1 dispersion; (ii) after initial, automatic 'cleaning' of the spectra, artefacts in the flux arrays, such as 'cosmic rays', 'ghosts' (internal reflections within the spectrograph), spurious or outlying pixels, and residual blaze removal effects, were masked and/or normalised out manually; these steps were recorded and are fully reproducible; and (iii) an initial, automatically set, polynomial continuum is manually refitted, if required, in regions where metal absorption is present, to ensure profile fits can be established as accurately as possible.…”