MARVELS 1D Pipeline Development, Optimization, and Performance

We describe the processing pipeline of one-dimensional spectra from the SDSS III Multi-object APO Radial Velocity Exoplanet Large-area Survey (MARVELS). This medium-resolution interferometric spectroscopic survey observed over 3300 stars over the course of four years with the primary goal of detecting and characterizing giant planets (>0.5 M _Jup) from within a large, homogeneous sample of FGK stars. The successful extraction of radial velocities (RVs) from MARVELS is complicated by several instrument effects. The wide field nature of this multi-object spectrograph provides spectra that are initially distorted and require conditioning of the raw images for precise RV extraction. Also, the simultaneous observation of sixty stars per exposure leads to several effects not typically seen in a single-object instrument. For instance, fiber illumination changes over time can easily create the dominant source of RV measurement error when these changes are different for the stellar and calibration optical paths. We present a method for statistically quantifying these instrument effects to combat the difficulty of giant planet detection due to systematic RV errors. We also present an overview of the performance of the entire survey as it stands for the SDSS III DR 12 as well as key results from the very latest improvements. This includes a novel technique, called lucky RV, by which stable regions of spectra can be statistically determined and emphasized during RV extraction, leading to a large reduction of the long-term RV offsets in the MARVELS data. These improved RV data are to be released via NASA Exoplanet Archive in the fall of 2015.