Direct comparison of two NIST PJVS systems at 10?V

Two NIST programmable Josephson voltage standard (PJVS) systems are directly compared at 10 V using different nanovoltmeters. These PJVS systems use arrays of triple-stacked superconducting niobium Josephson junctions with barriers made of niobium silicide. Compared with the voltages produced by Josephson voltage standards based on hysteretic junctions, PJVS systems using damped junctions produce predictable voltage levels. However, in order to guarantee the quantization of the voltages and to minimize the errors at the room-temperature voltage output, additional precautions are required. We report several experimental results of voltage measurements that contain significant systematic errors. The generated voltages appear reproducible but they are, in fact, inaccurate. When proper measurement procedures are followed, the results of a direct comparison using an analogue detector show that the two independent PJVS systems agree within 2.6 parts in 1011 at 10 V with a relative total combined uncertainty of 3.4 parts in 1011 (k = 1). Investigations show that the largest systematic error and most significant contribution to the uncertainty budget is caused by the leakage resistance of each PJVS to ground. This paper describes a measurement procedure to characterize this leakage resistance and one approach to including the resulting voltage error in the uncertainty budget.