Project 14A565 is the first science project using SubBeamNod observations. It requires deep observations with flat baselines to search 4GHz of bandwidth for a high-redshift CO source (z~7.5) in Q-band. A similar project was carried out with the spectrometer in early 2014 for another team. The standard sub-beam-node reduction scripts do not work due to the data volume. We processed the VEGAS data one sample at a time. The previous vector-calibration done with the spectrometer yields bad baselines with VEGAS. However, a simple scalar Tsys calibration works well when applying an appropriate high-pass filter to remove large-scale smooth frequency structures (Ta= Tsys_scalar*(ON-OFF)/OFF). The SubBeamNod observations were taken with a nod period of 6 sec and 0.2sec integrations to minimize the blanking. The high-frequency frequency ripples of VEGAS and the spikes calibrate away nicely. Figure shows difference between vector-calibration vs scalar calibration. The red spectrum shows the vector-calibration which has ugly baseline structure that cannot be removed. The black spectrum shows the result of scalar-calibration which yielded a flat baseline and noise that was consistent with the radiometer equation.