Johnson noise thermometry measurements using a quantized voltage noise source for calibration

We describe a new approach to Johnson noise thermometry (JNT) that exploits recent advances in Josephson voltage standards and digital signal processing techniques. Currently, high-precision thermometry using Johnson noise is limited by the nonideal performance of electronic measurement systems. By using the perfectly quantized voltage pulses from a series array of Josephson junctions, any arbitrary broadband waveform can be synthesized and used as a calculable noise source for calibrating the cross-correlation electronics used in JNT systems. With our prototype JNT system, we have found agreement to two parts in 10^3 with a 1(sigma) uncertainty of 1*10^-3 between the voltage noise of a 100-(Omega) resistor in a triple-point Ga cell (T/sub 90/= 302.916 K) and a pseudo-noise waveform with the same average power that is synthesized by a quantized voltage noise source. We estimate the temperature of the resistor to be 302.5 K +- 0.3 K (1(sigma) uncertainty based on the uncertainty from the cross-correlation). With better characterization of our JNT system, we expect to achieve relative accuracies of parts in 10^5 for arbitrary temperatures in the range between 270 and 1000 K.