Uncertainties And Inter-Laboratory Comparisons of Dry Piston Gas Flow Provers
Harvey Padden
Abstract:
Dry piston provers are similar in nature to conventional mercury-sealed flow provers, with the exception that the viscosity of the gas forms the seal across a very small (< 10 ยต) section. As a result, very small, portable provers can be produced. Calibrations can be performed with the accuracy, primacy and large dynamic range of the earier mercury-sealed provers, but far more rapidly. The transportable nature of these primary instruments also makes them very useful for the inexpensive, ongoing harmonization of laboratories. Our latest uncertainty analysis shows a combined expanded uncertainty of less than 0.08 % over the range of 5 to 50,000 sccm. Although we made every effort to perform an accurate analysis, empirical verification of any uncertainty analysis is necessary. At this level, the only means of verification is through peer-to-peer inter laboratory comparisons. As a result, we have performed informal comparisons of a single pair of provers with a number of national and private laboratories on three continents. We compared for reproducibility with respect to different national laboratories and over time, with transportation and between overlapping cellranges. In addition, and we compared newly manufactured provers with the original pair to establish reproducibility with manufacture. Typically, the provers exhibited discrepancies (within their specified range) of less than 0.1 % in comparison with critical flow venturis at NIST and NMIJ, with the possible exception of the 50,000 scm point. At the highest flow, the original prover exthibited a discrepancy of 0.15 % to 0.2 %. A second prover later showed a discrepancy of 0.035 %. We must conduct further investigations to determine the linearity of the design for flows above 30,000 sccm. Here, we will present an introductory summary of the uncertainty analyses, details of the comparison methodology, data on potential experimental error sources (such as inventory volume), the comparative data, and the methods of data analyses used. We will also discuss our ongoing research at the lower limit of this design's useful range (approximately 1 sccm), where leakage is not necessarily constant and conventional viscosity may not apply.
Keywords:
Gas flow calibration, primary prover, inter laboratory comparison
