The ARCADE-2 CMB balloon experiment detected an unexplained low frequency component which was postulated to be a previously unknown population of extragalactic sources. We conducted a deep 2-4 GHz survey with the recently upgraded EVLA to search for this population and to measure the faint radio source counts. Faint source counts constrain the cosmic evolution of the populations represented including sources powered by star formation. Previous attempts at measuring faint source counts produced more scatter in results than are easily accounted for and may reflect the variety of techniques used to correct for various effects. I will describe a deep confusion limited analysis using the "P of D" technique to derive the source counts at, and below the confusion which avoids many of the difficult to correct effects. Results are in excellent agreement with theoretical expectations effectively ruling out the ARCADE-2 anomaly being due to discrete sources. I will also summarize ongoing observations to determine the size distribution of the microJy population.