Developments towards an industrial Johnson noise thermometer

In this paper we present new developments in our practical, drift-free Johnson noise thermometer as it moves from a proof-of-principle prototype towards a practical implementation to become a viable primary thermometer for industrial applications. We will discuss bandwidth optimisation to obtain the lowest possible uncertainty of temperature measurements. The concept of weighting the cross-correlation frequency bins to improve the uncertainty will be introduced and it will be shown that this can decrease the measurement uncertainty in our system by 28%. Weighting fundamentally changes the optimum bandwidth calculation such that information from higher parts of the measured spectrum can always contribute to a reduction in total measurement noise. We will present the design approach used to provide extreme immunity to electromagnetic interference, a problem that has severely degraded previous efforts. We describe the results of tests by a third-party test laboratory demonstrating immunity up to field strengths of 10 Vm(-1), as required in standards for equipment operating in heavy industrial environments. We have also looked at interference at frequencies that could directly interfere with our thermometer that are not covered in the standards. The latest test results will be presented showing a five-fold improvement in accuracy on an extrapolation to absolute zero compared with the original proof-of-principle prototype with similar measurement parameters. Developments in techniques for providing traceability of the thermometer's calibration to electrical rather than temperature national standards will be presented.