Comparison of the Josephson Voltage Standards of the CENAM and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.b)

Transcription

1 Comparison of the Josephson Voltage Standards of the CENAM and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.b) S. Solve, R. Chayramy, and M. Stock Bureau International des Poids et Mesures F Sèvres Cedex, France D. Avilés, E. Navarrete and D. Hernández, Centro Nacional de Metrología Km 4.5 Carretera a los Cués El Marqués, Querétaro. C.P , Mexico

2 Comparison of the Josephson Voltage Standards of the CENAM and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.b) S. Solve, R. Chayramy, and M. Stock Bureau International des Poids et Mesures F Sèvres Cedex, France D. Avilés, E. Navarrete and D. Hernández, Centro Nacional de Metrología Km 4.5 Carretera a los Cués El Marqués, Querétaro. C.P , Mexico Abstract. A comparison of the 10 V Josephson array voltage standard of the Bureau International des Poids et Mesures (BIPM) with that of the Centro Nacional de Metrología (CENAM), Mexico, was made in September For this comparison, both options of the BIPM.EM-K10.b comparison protocol were applied (A and B). Option B required the BIPM to provide a reference voltage for measurement by the CENAM using its Josephson standard with its own measuring device. Option A required the CENAM to provide a reference voltage for measurement by the BIPM using its analogue detector and associated measurement loop. In both cases the BIPM and CENAM arrays were kept floating from ground. The final results observed were in excellent agreement with the combined relative standard uncertainty of 6.7 parts in for the nominal voltage of 10 V. 1. Introduction Within the framework of CIPM MRA key comparisons, the BIPM performed a direct Josephson voltage standard (JVS) comparison with that of the CENAM, Mexico, in September The comparison followed the technical protocol of BIPM.EM-K10.b comparisons and followed options A and B of the protocol. This involved the BIPM measuring the voltage of the CENAM JVS using its measurement loop with an analogue voltmeter as a detector for option A while the CENAM measured the voltage of the BIPM transportable Josephson Voltage Standard (BIPM JVS) using its own measurement chain for option B. The BIPM JVS was shipped to CENAM, Querétaro, Mexico where an on-site direct comparison was carried out from 18 to 25 September CENAM/BIPM comparison 2/22

3 For both protocol options, the BIPM and CENAM arrays were kept floating from ground and were biased on the same Shapiro constant voltage step for each polarity, which was necessary to maintain stability during the timeframe required for the measurement acquisition. This article describes the technical details of the experiments which were carried out during the comparison. 2. Comparison equipment 2.1 The BIPM JVS In this comparison the BIPM JVS comprised a cryoprobe with a Hypres 10 V SIS array (S/N: 2538E-7), microwave equipment and the bias source for the array. The Gunn diode frequency was stabilized using an EIP 578B counter, and an ETL/Advantest stabilizer. To visualize the array I-V characteristics, while keeping the array floating from ground, an optical isolation amplifier was placed between the array and the oscilloscope. During the measurements, the array was disconnected from this instrument. The measurements were carried out without monitoring the voltage across the BIPM JVS. The RF biasing frequency was adjusted to minimize the theoretical voltage difference to zero between the two JVS. The series resistance of the measurement leads was less than 4 Ω in total and the value of the thermal electromotive force (EMF) was found to be of the order of 20 nv and was eliminated by polarity reversal of the arrays. The leakage resistance between the measurement leads was greater than Ω for the BIPM JVS. 2.2 The CENAM JVS The CENAM JVS comprised a SIS (Nb/Al 2 O 3 /Nb) 10 V IPHT-PTB array (SN: JA-162/7), a JBS 500 bias source, an EIP 578B frequency counter, a dielectric wave-guide, a Tektronix TAS 465 oscilloscope, a high frequency filter which was included in the cryoprobe, and a Millitech GDM I InP Gunn diode connected to other microwave components. The frequency reference was a HP 5071A caesium clock. The system was controlled by NISTVOLT software connected to a HP 3458A digital voltmeter (DVM) for Zener calibrations. Measurement software for this comparison was developed using Labview, which allowed the use of different detectors: Agilent 3458A, Agilent 34420A, Keithley 2182A, Keithley Nanovolt Preamplifier 1801, and EM N11. For low uncertainty measurements the software calculates the microwave frequency and the step number needed to produce a voltage lower than 100 nv, and so reduces the uncertainties related to detector gain and linearity. CENAM/BIPM comparison 3/22

4 The laboratory was housed in a new building equipped with new facilities. The main characteristics of the new laboratory were: temperature control of 23 C ± 0.1 C; relative humidity control of 45 % ± 10 %; shielding for low frequency electromagnetic noise; low resistance to the ground potential; high quality uninterruptable AC power supply; and LED lighting. The insertion of an additional low frequency filter as part of the JVS and the installation in the new laboratory improved the noise immunity to electromagnetic interferences. The same constant voltage step could be maintained for several hours if the microwave power is well adjusted. The thermal EMF is around 30 nv when the dewar is full of liquid helium. To obtain an appropriate cancellation of the linear evolution of the thermal EMF during the measurements, stabilization was achieved by adding copper blocks and thermal isolation to the more sensitive points of the measurement circuit; in the BIPM s case, the Bendix connector of the cryoprobe and the output voltage connection of the JVS. Some changes to the circuit of the bias source were performed to achieve a finer step selection and a symmetrical bias voltage at both polarities. The insertion of a switch allowed disconnection of the ground from the shield, and therefore, it was possible for the JVS to float from the ground potential. The series resistance of the measurements leads was 0.78 Ω, and the leakage resistance between the measurement leads was Ω [1]. 3. Comparison procedures - Option B at the 10 V level 3.1 First measurements The BIPM JVS was set-up and checked for trapped flux. The array was biased to the same RF frequency as the CENAM JVS, f = 74 GHz, where the voltage stability was found to be appropriate to allow preliminary measurements to be made. The BIPM JVS was connected to the CENAM measurement system and seven measurements were taken following option B of the BIPM protocol ( The result was (U CENAM U BIPM ) / U BIPM = with an experimental standard deviation of the mean of This comparison result shows that the two standards were in very good agreement. The stability achieved on the two pieces of apparatus, even when connected together, was excellent. CENAM/BIPM comparison 4/22

5 Therefore, during the remaining period of the comparison, many experiments and measurement configurations were tested to achieve the lowest voltage difference between the two JVS and the lowest Type A uncertainty. Details of the experiments are described in Appendix A. 3.2 Description of the CENAM measurement procedure The BIPM and CENAM arrays were disconnected from their bias source during the entire data acquisition process. The reference ground of the chassis of the instruments that constituted the BIPM JVS was connected to the laboratory ground potential. The BIPM probe needed to be referred to this reference potential to achieve suitable stability on the JVS voltage. The two arrays were connected in series-opposition via a dedicated CENAM polarity switch. In this comparison scheme (option B), the CENAM JVS measurement set-up was used to measure the JVS voltage as if it were a Zener voltage standard. During the comparison, only the biases of both arrays were reversed (no mechanical switch reversal). This operation was carried out manually on both JVS. The polarity reversal was typically completed in less than 5 s. The normal measurement procedure was changed from the configuration of the preliminary measurements. The main differences between the procedures are described in the following paragraphs Measurement procedure for preliminary measurements The measurement loop was arranged so that the positive polarity of the measurement leads of both arrays were connected together and the voltmeter (HP3458A, 100 mv range) was placed between the two negative leads of the arrays. The High of the voltmeter was connected to the CENAM array. A test measurement of the BIPM JVS voltage was obtained with a DVM before setting the CENAM JVS to 10 V. A measurement point was acquired according to the following procedure: 1 - Positive polarity of the arrays; 2 - Data acquisition of 15 readings at NPLC = 10; 3 - Negative polarity of the arrays; 4 - Data acquisition of 30 readings at NPLC = 10; 5 - Positive polarity of the arrays; 6 - Data acquisition of 15 readings at NPLC = 10; CENAM/BIPM comparison 5/22

6 3.2.2 Measurement procedure for the measurements carried out to compute the final result Option B The measurement loop was modified: both positive polarity measurement leads of the arrays were still connected together but a nanovoltmeter (K2182A, 10 mv range) was placed between the two negative leads of the arrays. The High of the nanovoltmeter was connected to the CENAM array. The measurement software was changed as follows: 1 - Positive polarity of the arrays; 2 - Data acquisition of 10 readings at NPLC = 5; 3 - Negative polarity of the arrays; 4 - Data acquisition of 20 readings at NPLC = 5; 5 - Positive polarity of the arrays; 6 - Data acquisition of 10 readings at NPLC = 5; Neither the analogue filter nor the digital filter of the nanovoltmeter was engaged. The internal temperature of the nanovoltmeter was carefully recorded because of a known associated and rapid change in the gain value [2]. Furthermore, the meter option LSYNC ON was not selected: the data acquisition was not synchronized with the 60 Hz frequency of the mains Option A The measurement loop was modified: both negative polarities of the arrays were connected together but a nanovoltmeter (EM N11, 3 µv range) was placed between the two positive polarities of the arrays. The High of the nanovoltmeter was connected to the CENAM array. The equipment included a voltage divider to prevent the detector from overload if both systems deviate from the same selected steps. The measurement software set-up was changed as follows: 1 - Positive array polarity and reverse position of the detector; data readings acquisition; 3 - Positive array polarity and normal position of the detector; data readings acquisition; 5 - Negative array polarity and reverse position of the detector; data readings acquisition; 7 - Negative array polarity and normal position of the detector; CENAM/BIPM comparison 6/22

7 8-500 data readings acquisition; 9 - Negative array polarity and reverse position of the detector; data readings acquisition; 11 - Negative array polarity and normal position of the detector; data readings acquisition; 13 - Positive array polarity and reverse position of the detector; data readings acquisition; 15 - Positive array polarity and normal position of the detector; data readings acquisition. During the measurement process, the BIPM bias source was adjusted to manually select the same step after each polarity reversal. After each polarity reversal 10 seconds elapsed before beginning the data acquisition to avoid filter capacitor discharge effects. 4. Uncertainties and results 4.1 Option B protocol Final result The result using option B, expressed as the relative difference between the values attributed to the 10 V BIPM JVS (U BIPM ) by the CENAM JVS measurement set-up (U CENAM ) is: (U CENAM U BIPM ) / U BIPM = and u c / U BIPM = , where u c is the total combined standard uncertainty and the relative Type A is u A / U BIPM = All 15 individual measurements computed to calculate the final result are presented in Fig. 1. CENAM/BIPM comparison 7/22

8 U (CENAM) - U (BIPM) /nv Measurement Points Fig. 1: Individual results obtained to calculate the option B comparison result at the level of 10 V: The solid line represents the mean value, the dotted dashed lines ( - - ) represent the experimental standard deviation, and dotted lines (- - -) are the experimental standard deviation of the mean Type B uncertainty components (option B protocol) The sources of Type B uncertainty (Table 1) are: the frequency accuracy of the Gunn diodes; leakage currents; and detector gain and linearity. Most of the effects of detector noise and frequency stability are already contained in the Type A uncertainty. Both array polarities were reversed during the measurements, so the effect of the residual thermal EMF (i.e. non-linear drift) and electromagnetic interferences are already contained in the Type A uncertainty of the measurements. Uncertainty components related to RF power rectification and sloped Shapiro voltage steps are considered negligible because no such physical effect was observed. CENAM/BIPM comparison 8/22

9 Type BIPM Relative uncertainty CENAM Frequency offset (A) B Leakage resistance (B) B Detector (C) (D) B Total (RSS) B Table 1: Estimated Type B relative standard uncertainty components. (A) Both systems referred to the same 10 MHz frequency reference, therefore only a Type B uncertainty from the frequency measured by the EIP is included. The frequency reference used for the comparison was produced by a HP 5071A caesium clock frequency standard. BIPM JVS: It has been demonstrated on many occasions that the EIP-578B is a good frequency locker and the accuracy of the frequency can reach 0.1 Hz. Assuming a rectangular distribution, the relative uncertainty for the offset of the frequency can be calculated from the formula: u f = (1/ 3 ) (0.1/75) 10 9 = CENAM JVS: The Type B uncertainty estimation of the microwave frequency was initially based on the worse case reviewed specifications of the EIP-578B frequency counter which are u f = (1/ 3 ) ((74/20) 2 / 74) 10 9 = [3], but an experiment based on an heterodyne method was carried out after the comparison and lead to the current value. A detailed description is given in [4]. CENAM JVS: (B) Assuming a rectangular statistical distribution, the relative uncertainty contribution of the leakage resistance R L can be calculated from the formula: u f = (1/ 3 ) (r / R L ). The values attributed to resistances were measured during the comparison exercise on both JVS. Note: r = 0.78 Ω and R L = Ω for the CENAM JVS [4]. (C) A large proportion of the detector uncertainty is already contained in the Type A uncertainty of the measurements. This component expresses only the uncertainty on the internal residual thermal electromotive force and possible non-linear internal offset of the K2182A. CENAM/BIPM comparison 9/22

10 U (CENAM) - U (BIPM) / nv (D) The DVM HP 3458A was used as a null detector for the first measurements. All of the initial measurements were less than 30 nv. The uncertainty for the gain and linearity was negligible considering that the gain and linearity errors of this DVM are in the order of parts in A large proportion of the detector uncertainty is already contained in the Type A uncertainty of the measurements. The type B component of this DVM was estimated at 3 nv; it only expresses the uncertainty of the DVM resolution [3]. 4.2 Option A protocol Final result Figure 2 shows the 19 individual measurements used to compute the option A comparison scheme that correspond to the final result of the comparison Individual Measurment Points Fig. 2: Individual measurement points obtained to calculate the final comparison result at the level of 10 V for the option A protocol: The solid line represents the mean value, the dotted dashed lines ( - - ) represent the experimental standard deviation, and dotted lines (- - -) are the experimental standard deviation of the mean. Each individual point is represented with Type A uncertainty bars for k = 1. The standard deviation of the mean of the 19 measurements carried out on 24 September 2011 was considered as the Type A uncertainty. The comparison result is therefore: CENAM/BIPM comparison 10/22

11 (U CENAM U BIPM ) / U BIPM = and u A / U BIPM = , where u A is the Type A uncertainty Type B uncertainty components (option A protocol) The sources of Type B uncertainty (Table 2) are equivalent to those listed in Table 1: the only difference being that the detector was changed. Type BIPM Relative uncertainty CENAM Frequency offset (E) B Leakage resistance B Detector (F) B Total (RSS) B Table 2: Estimated Type B relative standard uncertainty components (option A protocol). (E) As pointed out in the uncertainty budget of the option B protocol, this CENAM component was initially estimated at but an experiment based on an heterodyne method was carried out after the comparison and lead to the current value. A detailed description is given in [4]. (F) A large proportion of the detector uncertainty is already contained in the Type A uncertainty of the measurements. This component only expresses the effect of the uncertainty of the detector non-linearity correction. The uncertainty on the accuracy of the nanovoltmeter is calculated from the difference between the nominal calibration factor and the measured value. This difference is applied to the maximum voltage difference measured by the N11 on the 3 µv range, and is equal to: u D = nv. CENAM/BIPM comparison 11/22

12 5. Discussion and conclusion The results of the comparison are as follows: The preliminary comparison result is: (U CENAM U BIPM ) / U BIPM = and u c / U BIPM = (Type A uncertainty = 8.1 nv). It was obtained using the original CENAM measurement set-up where the detector is a HP3458A. This result supports the CMCs (Calibration and Measurement Capabilities) of the National Metrology Institute. Three different detectors were used during the comparison: HP3458A, K2182A and EM N11. A systematic error of 10 nv was measured with the K2182A detector, after several changes to the configuration the offset was cancelled, but no clear explanation was found. An investigation of the CENAM measurement set-up and the use of a digital nanovoltmeter allowed a significant technical improvement which lead to the following final result for the Option B of the comparison protocol: (U CENAM U BIPM ) / U BIPM = and u c / U BIPM = (Type A = ) The best measurement configuration was obtained using an analogue nanovoltmeter (option A) and allowed the achievement of the best and final result for this comparison: (U CENAM U BIPM ) / U BIPM = with u c / U BIPM = 6.7 x (Type A uncertainty = ). The most significant uncertainty component was initially related to the accuracy of the CENAM frequency source (Type B). An appropriate investigation, carried out after the comparison, has lowered this component and reduce the combined total uncertainty to the sub-nanovolt level [4]. The comparison was carried out in the new CENAM Electricity laboratories where the environmental parameters were excellent: the shielded laboratory, laboratory temperature regulation, and quality of the mains power distribution definitively contributed to the achievement of the very good result. Excellent stability on both JVS was obtained from the start of the comparison. The absence of electromagnetic interferences contributed mainly to stability of the measurement set-up when two JVS (very low output impedance) were connected in series-opposition. During this comparison, two new BIPM devices were evaluated (voltage divider in front of the input of the N11 and a new array DC bias source). The results obtained with the new BIPM devices were in excellent agreement with the results obtained when using the traditional BIPM CENAM/BIPM comparison 12/22

13 transportable JVS. We can therefore conclude that use of the new equipment did not affect the metrological characteristics of the BIPM transportable JVS. References [1] David Avilés Castro, Dionisio Hernández Villaseñor, Enrique Navarrete García, 15 años del patrón de tensión eléctrica continua en base al efecto Josephson en el CENAM, Proceedings of the Symposium of Metrology 2008, Querétaro, Mexico. [2] Solve S., Christian L., et al., Comparison of the Josephson voltage standards of the MSL and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.b) to be published in Metrologia Tech. Suppl. [3], Enrique Navarrete, Dionisio Hernández, David Avilés. Estimación de la incertidumbre del patrón nacional de tensión eléctrica continua basado en el efecto Josephson del CENAM. Proceedings of the Symposium of Metrology 2006, Querétaro, Mexico. [4] David Avilés, Enrique Navarrete, Dionisio Hernández et al.,direct comparison of Josephson Voltage Standards at 10 V between BIPM and CENAM, submitted to IEEE Trans. Instrum. [5] Solve S., Zou Y., et al., Comparison of the Josephson voltage standards of the NMC, A*STAR and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.a and BIPM.EM-K10.b), Metrologia, 2011, 48, Tech. Suppl., [6] Solve S., Šíra M., et al., Comparison of the Josephson voltage standards of the CMI and the BIPM (part of the ongoing BIPM key comparison BIPM.EM-K10.b), Metrologia, 2012, 49, Tech. Suppl., CENAM/BIPM comparison 13/22

14 DISCLAIMER Certain commercial equipment, instruments or materials are identified in this paper in order to adequately specify the environmental and experimental procedures. Such identification does not imply recommendation or endorsement by the BIPM or CENAM, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose. CENAM/BIPM comparison 14/22

15 Appendix A This part of the report describes the measurements performed in chronological order. 19 September 2011 After both JVS systems were connected together and satisfactory stability conditions were achieved, the CENAM procedure for calibration of DC Voltage Standards was run. The measurement set-up is based on the operation of a HP3458A as a detector and NISTVolt software. Both systems were operated at the frequency f = 74 GHz and seven measurement points were performed for which the mean value constitutes the preliminary result. A possible correlation between the measured voltage difference and the laboratory temperature were investigated because the ambient parameters are continuously recorded but no correlation was observed. As the two JVS were biased on the same Shapiro step, the theoretical voltage difference between them was null and the calibration of the gain of the voltmeter was therefore not required. Furthermore, the level of the thermal electromotive force along each JVS was equivalent. As a consequence the voltmeter was measuring a signal which was close to zero volts (less than 30 nv). 20 September 2011 The HP3458A was replaced with a digital nanovoltmeter HP34420A but the CENAM Josephson array lost its stability. A second device of the type was installed but the same behaviour was observed [2]. A high value for the filter inductance (>400 mh) combined with the value of the injection current of the nanovoltmeter are possibly responsible for this behaviour. The HP34420A was changed to a Keithley 2182A nanovoltmeter and the instability problems of the voltage steps disappeared. Five consecutive measurement points were performed with a high level of noise (of the order of 70 nv). For this series of measurements, the BIPM JVS probe was not grounded. It was decided to ground the BIPM JVS probe from the CENAM potential reference point through the shielding of the measurement wires. The CENAM measurement loop configuration allowed the removal of the reference potential of the bias source and therefore allowed the device to be floating from the ground. After the activation of this option, five points were carried out and the level of noise was decreased to 1.4 nv. CENAM/BIPM comparison 15/22

16 Note: When the reference potential of the bias source is removed, the shielding of the lead that connects the nanovoltmeter to the measurement loop is no longer grounded. Even if this new configuration consequently improves the noise level, the mean value of these two series was of the order of 10 nv. However this series allowed a possible offset of the noise to be derived. Note: The software used to perform the measurement was developed using Labview and was operated from a notebook computer. For the first point of the first series a relative instability of the voltage provided by the CENAM array was observed when compared to that observed previously. It was found that the power supply of the computer was still connected to it. Stability was retrieved when the power supply was unplugged from the notebook computer. Five new points were carried out when grounding the BIPM JVS probe from the BIPM JVS equipment (scope). The level of noise increased to 8 nv. Note: The output Z parameter of the BIPM bias source had to be increased compared to the level normally selected for most of the BIPM Josephson comparisons carried out to date. This adjustment is a consequence of the presence of a large inductance (L = 430 mh) in the CENAM measurement lead filter. The BIPM probe was grounded from the CENAM potential reference point again and five more points were carried out to confirm the 10 nv offset. The reason for this increase in the noise level was found in the metallic box containing the switch that connects both JVS in the measurement loop that was no longer grounded. The problem was fixed and 10 new points were carried out. The mean value was 14.6 nv with a standard deviation of the mean of 2 nv. It was decided to investigate the electrical noise of the measurement set-up and 15 points were performed at the level of 0 V (critical current of the arrays) within the following configuration of the detector (10 mv range): 5 points at NPLC = 1 with the analogue filter OFF and the digital filter ON; 5 points at NPLC = 1 with the analogue filter ON and the digital filter OFF; 5 points at NPLC = 1 with the analogue filter ON and the digital filter ON. The 2182A filtering adjustments did not influence the results and gave a voltage difference mean value of 0.6 nv with a standard deviation of the mean of 0.35 nv. This result confirmed that the measurement set-up and the detector in particular, behave as expected. CENAM/BIPM comparison 16/22

17 21 September 2011 The 10 nv offset observed on 20 September was confirmed by a first series of five points which gave a mean value of 11.4 nv with a standard deviation of the mean of 2 nv. It was suspected that the offset might come from a leakage on the CENAM JVS. Effectively U CENAM - U BIPM < 0 implies that CENAM measures a higher value of the BIPM array, which could correspond to the fact that the CENAM JVS provides a voltage output lower than the theoretical voltage output due to resistance leakage. In that case, the level of the leakage might be proportional to the voltage output of the JVS. Therefore, a series of five points were performed at the level of 5 V for which stability problems were encountered that resulted in discrepant measurement values. It was decided to compare the two JVS, at the level of 0 V (0 V Shapiro voltage step), for which similar results were obtained to those of the previous day. A second series of five measurement points at the level of 5 V were carried out and gave a mean value of the voltage difference of 0.05 nv with a standard deviation of the mean of 0.65 nv. A possible leakage resistance was temporarily excluded. In two previous comparisons [5, 6], the bias source was found to be responsible for adding a voltage offset of a few nanovolts leading to a systematic corresponding error in the measurements. In order to verify this hypothesis, five points were performed after physically disconnecting the CENAM bias source after adjustment to the selected voltage step was completed. The results were discrepant. An investigation of the measurement set-up was performed and showed that the connection of the CENAM Dewar to the ground was unsatisfactory. Compared to the previous configuration of the measurement set-up, the position of the biasing cable was changed to allow easy removal by the operator. During this operation the ground connection to the cable was broken. The problem was fixed and five new points were performed at the level of 10 V while disconnecting the biasing after adjustment. The result gave a mean value of 5.85 nv with a standard deviation of the mean of 3 nv. Based on the results of leakage measurement carried out at CENAM before the comparison we excluded the possibility that the offset could originate from the filter on the measurement leads. Further experiments were carried out to determine the origin of the voltage offset. CENAM/BIPM comparison 17/22

18 Six measurements points were made with the negative polarity of the CENAM array forced to the potential reference. The noise level was increased and the offset was still present. The CENAM switch that connected both JVS was replaced with a BIPM switch that had very high leakage resistance to ground (>10 13 ). The offset was still present and the noise level was low because the CENAM array was floating from the ground. 22 September 2011 The detector was reversed. Previously the nanovoltmeter was placed between the two negative leads of the arrays. Five measurements were carried out by connecting the positive lead of the nanovoltmeter on the BIPM side, and five more measurements were made after connecting the positive lead of the nanovoltmeter to the CENAM side. The offset direction changed confirming the existence of a systematic error. When using the Keithley 2182A, it was observed that the resolution decreased by two digits compared to what could be expected when using this device. Furthermore, the device did not fully respond to the IEEE commands sent to it. An investigation found that the nanovoltmeter was configured to emulate a Keithley 182 nanovoltmeter. The corresponding parameter (182 language) was changed to SCPI language (IEEE-488). It was decided to investigate the leakage resistance of each of the four feed-through capacitors (7 nf) mounted on the measurement leads after investigating the structure of the measurement lead filter on the CENAM set-up. Two different measurements were carried out, one using the batterypowered Megaohmeter Keithley 500 to measure the leakage resistance and a second using a RCL meter PM 6304 to measure the impedance of the capacitor at f = 10 khz. The results are presented in Table A1: CENAM/BIPM comparison 18/22

19 Feed-through capacitor identification Leakage resistance ( ) Keithley 500 Impedance measurement (k ) at f = 10 khz #1 Between to #2 Between to #3 Between to #4 Between to Table A1: measurements of the leakage resistance and the impedance at f = 10 khz of the 4 feed-through capacitors mounted on the CENAM measurement leads of the filter. The CENAM probe was re-installed without the feed-through capacitors and the array was cooled. Five measurement points were carried out that showed a significant increase in the noise. This was due to a loss of stability on the CENAM JVS compared to the previous arrangement of the probe (feed-through capacitors included). A BIPM LC filter was installed between the two JVS, but the stability was increased, however, it was much more difficult for the CENAM operator to bias the array on the selected step. It was then decided to remove the BIPM filter and to insert a 10 nf capacitor between the CENAM measurement leads. By doing this, the stability of the CENAM array recovered, however the level of noise was one order of magnitude higher than expected (20 nv instead of 2 nv). Therefore it was not possible to reach a conclusion on the existence or disappearance of the 10 nv offset. Note: As we noticed that the CENAM array often jumps at the time the software was run to read the data, we decided to change the IEEE bus isolator to a simple line connector where the grounding connection was removed. 23 September 2011 The CENAM probe was returned to its initial configuration with the feed-through capacitors. However, a selection of components was performed in such a way that the four capacitors presented an impedance value at f = 10 khz between 160 k and 180 k. During the time required for these technical modifications, the gain of the nanovoltmeter K 2182A was calibrated using the BIPM JVS. CENAM/BIPM comparison 19/22

20 A significant correction of 150 µv/v was needed (mean value of four repeatable runs). After applying the correction to all previous measurements we confirmed that the voltage differences measured between the two JVS was null and there was no impact on the results calculated. After cooling both JVS again (the BIPM JVS needed to be warmed up because its Dewar required refilling with liquid helium). An initial series of five measurement points was carried out soon after the cooling process and gave a mean value of 27 nv with a standard deviation of the mean of 14 nv. The unsatisfactory result was supposed to originate from the relaxing time of the thermal electromotive force which was not long enough. After a suitable time, a new series of 10 points gave a mean value of 2.8 nv with a standard deviation of the mean of 1.1 nv. An adjustment of the RF power on the CENAM array during the acquisition resulted in a discrepant measurement that was considered an outlier and was therefore rejected for computation of the result. A last series of six points was performed and confirmed this good result where the 10 nv offset was no longer present. All 15 individual points contributed to the final result of the Option B comparison. The amount of time spent correcting the observed 10 nv offset meant that no more measurements with the K2182A or with the EM N11 were possible for Option B and it was decided to move to Option A of the comparison. 24 September 2011 A large proportion of the day was devoted to setting up the measurement chain for the Option A of the comparison. An EM N11 was borrowed from the CENAM DC voltage laboratory and a new voltage divider made by the BIPM was installed in front of the input of the nanovoltmeter to prevent it from a long period of overload in case of a quantum voltage jump on one of the arrays. When the new measurement set-up and software showed suitable behavior, three preliminary measurement points were carried out to investigate the stability of both arrays at both polarities. The behaviour was extremely satisfactory and six consecutive measurement points at f = 74 GHz on the 3 µv range of the nanovoltmeter were performed. Note: This range is convenient because it is possible to directly calibrate it by shifting the frequency of the BIPM array to meet the full scale of the detector. For instance if the frequency is moved from f = 74 GHz to f = GHz, this corresponds to a shift of +2.7 µv at 10 V. Within the same scheme, a change in the frequency from f= 74 GHz to f= GHz will CENAM/BIPM comparison 20/22

21 correspond to a shift of 2.7 µv at 10 V. 10 khz is the limit resolution of the EIP counter used to locked the phase of the frequency source. It was then decided to change the frequency of both RF sources to f = GHz and to carry out a series of six new points. At the beginning of the first point the BIPM RF source broke down. The Gunn diode was replaced with a spare which was set to the same frequency. A third and last series of six points was carried out at another frequency (f = GHz). For this series, the BIPM bias source was changed for a new bias source. The idea behind this was to validate the use of this new source in future comparisons with the guarantee that the metrological characteristics of the transportable BIPM JVS are not affected. The result of this series gave a mean value of 0.21 nv with a standard deviation of the mean of 0.48 nv. Nineteen individual results were computed to derive the final result of this comparison. After the comparison: Several measurements were made at CENAM to find out the cause of the 10 nv offset observed during the comparison: - The series and leakage resistances of the CENAM JVS measurement leads were measured again; the results were the same as reported previously. The series resistance is 0.78 Ω, and the leakage resistance is Ω. - The leakage resistances of the feed-through capacitors of the CENAM high frequency filter were measured individually. The measured leakage resistances of all the feed-through capacitors were in the order of 1.7 TΩ. - The K2182A was calibrated at 150 µv level. The error at this level was 23 nv. - A test for microwave induced offset was performed on the CENAM JVS. No microwave induced offset was detected. - An offset of about 7.7 nv was detected when the shield of the K2182A cables was connected to the shield of the CENAM JVS, nevertheless this offset could be cancelled by the polarity inversion technique used during the comparison. - The comparison circuit was simulated using the next measured parameters: series resistance and leakage resistance of the JVSs; bias current; shielding current; offset; and input impedance of the K2182A. When the simulation was run using these parameters, an error of less than 0.1 nv in the output voltage of the JVS was observed when biased at 10 V. CENAM/BIPM comparison 21/22

22 None of these measurements and simulations explained the offset measured during the comparison when the K2182A detector was used. CENAM/BIPM comparison 22/22