Inter-laboratory Comparison of Impedance-Type Hygrometer in the Range from 10 % to 95 % at 5 C to 55 C

Transcription

1 DOI /s Inter-laboratory Comparison of Impedance-Type Hygrometer in the Range from 10 % to 95 % at 5 C to 55 C Domen Hudoklin Regina Mnguni Hans Liedberg Igor Pušnik Jovan Bojkovski Received: 30 July 2013 / Accepted: 4 July 2014 Springer Science+Business Media New York 2014 Abstract Traceability in the field of relative humidity (RH) measurements is typically assured indirectly through dew point and temperature scales. Conducting an interlaboratory comparison at the national metrology institute (NMI) level, using a direct approach with a precision RH hygrometer as a transfer standard would, therefore, be of a particular interest, especially if the measurement setups were of a different type. This paper presents an RH comparison at the NMI level between the National Metrology Institute of South Africa (NMISA) and University of Ljubljana, Faculty of Electrical Engineering, Laboratory of Metrology and Quality (MIRS/UL-FE/LMK). In scope of this inter-comparison, calibration of an impedance-type hygrometer in the range from 10 %rh to 95 %rh at air temperatures of 5 C, 25 C, and 55 C, respectively, was performed. It was recommended that the participants use their standard procedure for the calibration of RH sensors and, at the same time, follow the specific criteria of the review protocol for uncertainty estimation accepted by Bureau International des Poids et Mesures (BIPM), marked as BIPM CCT-WG8/CMC-10. An interesting part of the comparison was the two different calibration methods which were used by the two partners and which also have different traceability routes. MIRS/UL-FE/LMK calibrated the sensor in the humidity generator by comparison against the reference chilled mirror hygrometer, which is traceable to the MIRS/UL-FE/LMK primary dewpoint generator. NMISA calibrated the transfer standard against certified salt solutions, which were kept in a temperature-controlled chamber. Results showed acceptable agreement at all 15 calibration points. D. Hudoklin (B) I. Pušnik J. Bojkovski Metrology Institute of the Republic of Slovenia/University of Ljubljana-Faculty of Electrical Engineering/Laboratory of Metrology and Quality (MIRS/UL-FE/LMK), Ljubljana, Slovenia domen.hudoklin@fe.uni-lj.si R. Mnguni H. Liedberg National Metrology Institute of South Africa (NMISA), Pretoria, South Africa

2 Keywords Dew-point measurement Inter-laboratory comparison Relative humidity Salt solutions 1 Introduction Relative humidity (RH), as the most frequently measured humidity parameter, is predominantly measured directly by impedance-based hygrometers. Due to the limited accuracy, these types of hygrometers typically provide traceability indirectly through the dew point and temperature. This is also one of the reasons, why at the level of national metrology institutes (NMI), calibration and measurement capabilities (CMC) are inter-compared for dew-point realizations using dew-point hygrometers as transfer standards. On the other hand, RH calibration introduces additional intrinsic uncertainty components which result from both humidity and temperature measurements. Moreover, the uncertainty contribution of the RH hygrometer as a device under calibration (DUC) adds additional uncertainties, such as reproducibility, long-term instability, etc. For this reason, inter-laboratory comparison in the field of RH using an RH hygrometer as a transfer standard is of great interest, because it gives needed supporting evidence to CMC claims [1 3] of the laboratory and to hygrometry in general (also in terms of [4,5]), particularly because such intercomparisons at the level of NMIs are scarcely performed and reported [6 8]. The aim of this bilateral inter-laboratory comparison was to evaluate the metrological equivalence of NMISA s RH calibration capabilities, over a range of air temperatures from 5 Cto55 C, with those of LMK (short for MIRS/UL-FE/LMK), the NMI of Slovenia. The institutes are accredited according to ISO 17025:2005 and ISO 17025:2010, for humidity calibrations, respectively. NMISA acted as the pilot (coordinating) laboratory, while the LMK provided the comparison reference value. The participating laboratories used their standard procedure for the calibration of RH sensors and at the same time followed the specific criteria of the BIPM CCT-WG8/12-10 CMC review protocol for relative humidity. An interesting part of this comparison was the fact that the two RH realizations were based on two different methods; salt solutions [9] and primary dew-point generation [10 12], while for the transfer standard [13], a precision RH hygrometer with a capacitive sensor was used. 2 Technical Objectives of the Intercomparison The objective of this comparison was to assess the procedures and standards used in calibrating RH hygrometers, particularly at temperatures other than ambient. The transfer standard, Vaisala MI70/HMP77 (Serial Nos. D and D ) was provided by NMISA. It is a precision capacitive RH hygrometer with a known and stable calibration history. During the transport, the transfer standard artifact was packed in the accompanying protective case and shipped via courier. Both labs, first the NMISA followed by the LMK, calibrated the artifact at different RH set points of 10 %rh, 35 %rh, 50 %rh, 75 %rh, and 95 %rh, all at air temperatures other than ambient: 5 C, 25 C, and 55 C.

3 2.1 Calibration Facilities NMISA used unsaturated salt-solution ampoules, traceable to an accredited commercial calibration laboratory, as reference standards for RH measurements. The unsaturated salt solutions were supplied in sets of 5 or 50 in a box. Two ampoules of the same salt solution were poured onto clean fiber pads in a 100 ml stainless-steel chamber, placed in a large temperature-controlled chamber. The transfer standard was passed through the O-ring seals into a 100 ml stainless chamber. The measurements were done at three temperature points, i.e., 5 C, 25 C, and 55 C, however, not in a particular order for a set of ampoules. The ampoules used were for 10 %rh, 35 %rh, 50 %rh, 75 %rh, and 95 %rh salt solutions. The chamber was set to the three temperatures points consecutively, with typically a 5 h stabilization time for each set of measurements, and a total time for all three sets of 15 h. The traceability of the unsaturated salt solutions at different temperatures was provided through the supplier of the ampoules. At LMK, the transfer standards were calibrated in a humidity generator. The reference RH was calculated according to the Sonntag formulae [14] from the measured dew-point temperature and the temperature of the air. Both parameters were directly measured by a chilled mirror hygrometer MBW 373H (Serial No ). Traceability of the dew-point temperature is assured by a LMK primary dew-point generator [10 12]. Air-temperature measurements are traceable to an LMK temperature primary standard, realized by temperature fixed points. At each temperature point, 5 C, 25 C, and 55 C, RH points were set in rising and falling order to check for hysteresis of the instruments. At each set point, stability criteria were set according to the standard deviation of the calculated RH. 3 Method for Comparison Results Analysis Each participating lab determined corrections D lab,η,t of the artifact indication at each nominal measurement point of RH η and temperature t according to D lab,η,t = η ref,η,t η ind,η,t, (1) where η ref,η,t denotes the lab s reference RH at a nominal measurement point and η ind,η,t a RH indicated by the artifact. LMK s corrections served as the intercomparison reference value D RV,η,t for each nominal point D RV,η,t = D LMK,η,t, (2) where the index lab is substituted by LMK. For each correction, both labs provided also the expanded uncertainty (with coverage factor k = 2) of the correction, U(D lab,η,t ). The uncertainties were estimated according to the lab s calibration procedure taking into account also specific acceptance criteria for uncertainty budget (inter-laboratory-evaluation-based criteria) of the BIPM CCT-WG8/12-10 CMC review protocol [15]. The uncertainty calculation included the contribution due to hysteresis of ±0.3%rh[8] for the transfer standard.

4 As the LMK provided the reference value of the comparison, the expanded uncertainty of the comparison reference value U(D RV,η,t ), is taken as a combination of the expanded uncertainty U(D LMK,η,t ), where the index lab is substituted by LMK, and the expanded uncertainty due to the long-term instability of the artifact U drift : U ( D RV,η,t ) = U 2 ( D LMK,η,t ) + U 2 drift, (3) where U drift is estimated from a history of calibration by NMISA to 0.23 %rh by conservatively taking into account the worst case maximum value of differences between two corrections at the set points of 5 C and 75 % [8]. To compare the results of the two NMIs, individual differences D lab,η,t to the comparison reference value were calculated for each nominal point and for each lab D lab,η,t = D lab,η,t D RV,η,t, (4) To show bilateral equivalence of the comparison results, the differences from Eq. 4 were normalized to the expanded (k = 2) uncertainty of the bilateral difference that was calculated to obtain the value E n,η,t. E n,η,t = D NMISA,η,t U 2 ( ) D RV,η,t + U 2 ( ) D NMISA,η,t, (5) where the index lab is substituted only by NMISA, because the LMK provided the reference value and therefore LMK s D n,η,t is meaningful (zero values). 4 Results and Discussion Table 1 shows a summary of comparison results between the LMK and NMISA at different nominal points of RH and temperature. Results are graphically shown also in Figs. 1, 2, and 3. From Table 1, it can be concluded that results at all nominal points of RH and temperature meet the condition E n,η,t 1, which means that the differences to the reference value agree to within the expanded uncertainties with a coverage factor k = 2. From Figs. 1, 2, and 3, it can be seen that the agreement of the results is better at the extremes of the RH range, 10 %rh and 95 %rh at temperatures above ambient, 25 C and 55 C, respectively. At 5 C, however, the agreement is better in the middle of the range at 50 %rh with an E n,η,t value of Except for one point at 5 C and 50 %rh, the deviations do, however, show small systematic behavior. If the reason for that would be a deviation in either the dew-point or temperature measurement of the LMK setup, then the deviation would be higher at a higher relative humidity because of the higher sensitivity coefficient. Additionally, the deviations of the dew-point reference from the reference value obtained in the inter-laboratory comparison [12] are smaller or equal to C, which in the worst case results in a RH deviation of <0.1%rh. This is significantly smaller than the observed deviation in Figs. 1, 2, and 3.

5 Table 1 Summary of comparison results between LMK and NMISA at different nominal points of RH and temperature Nominal Nominal relative temperature, humidity, η (%rh) t ( C) LMK exp. unc., U(D LMK,η,t ) (%rh) NIMSA exp. unc., U(D NIMSA,η,t ) (%rh) NMISA difference, D NMISA,η,t (%rh) Normalized diff., E n,η,t D lab,,t - D RV,,t, %rh lab=nmisa lab=lmk Relative humidity, %rh Fig. 1 Comparison results at nominal temperature t = 5 C; bars on the graph show the expanded uncertainty (k = 2) of bilateral difference between the individual correction and reference value 5 Conclusion This paper presents an inter-laboratory comparison in the field of RH at the level of national metrology institutes. Set-ups of the two participating laboratories, NMISA and MIRS/UL-FE/LMK, used a different type of humidity generation and different traceability routes. The use of a single impedance-based RH hygrometer as a transfer standard showed to be sufficient for this bilateral inter-comparison. It is to be noted,

6 D lab,,t - D RV,,t, %rh lab=nmisa lab=lmk Relative humidity, %rh Fig. 2 Comparison results at nominal temperature t = 25 C; bars on the graph show the expanded uncertainty (k = 2) ofbilateral difference between theindividual correctionandreference value D lab,,t - D RV,,t, %rh lab=nmisa lab=lmk Relative humidity, %rh Fig. 3 Comparison results at nominal temperature t = 55 C; bars on the graph show the expanded uncertainty (k = 2) ofbilateral difference between theindividual correctionandreference value however, that for longer intercomparisons with more participants, it would be wise to use more transfer standards in order to avoid potential long-term instability. The results of this comparison show agreement with all E n,η,t values being <1, which is in accordance also with one of the specific criteria of the BIPM CMC review protocol CCT-WG8/ The participants agreed also to align their uncertainty estimations with other specific criteria of the mentioned protocol. References 1. D. Zvizdic, M. Heinonen, D. Sestan, Int. J. Thermophys. 33, 1536 (2012) 2. V. Fernicola, M. Banfo, L. Rosso, D. Smorgon, Int. J. Thermophys. 29, 1668 (2008) 3. M. Heinonen, D. Zvizdic, D. Sestan, Int. J. Thermophys. 33, 1451 (2012) 4. G. Begeš, J. Drnovšek, L.R. Pendrill, Accredit. Qual. Assur. 15, 147 (2010) 5. G. Begeš, H.J. Dalsgaadr, J. Drnovšek, J. Test. Eval. 36, 345 (2008) 6. M. Stevens, R. Benyon, S.A. Bell, T. Vicente, Int. J. Thermophys. 29, 1685 (2008)

7 7. R.M. Gee, R.S. Farley, P. Sax, R. Benyon, M. Stevens, N. Böse, Int. J. Thermophys. 29, 1696 (2008) 8. D. Jonker, M. Heinonen, H.G. Liedberg, M.R. Mnguni, Int. J. Thermophys. 33, 1458 (2012) 9. D. Jonker, M.R. Mnguni, H.G. Liedberg, Int. J. Thermophys. 29, 1644 (2008) 10. D. Hudoklin, J. Bojkovski, J. Nielsen, J. Drnovšek, Measurement 41, 950 (2008) 11. D. Hudoklin, J. Drnovšek, Int. J. Thermophys. 29, 1652 (2008) 12. M. Heinonen, M. Anagnostou, S. Bell, M. Stevens, R. Benyon, R. Anita Bergerud, J. Bojkovski, R. Bosma, J. Nielsen, N. Böse, P. Cromwell, A. Kartal Dogan, S. Aytekin, A. Uytun, V. Fernicola, K. Flakiewicz, B. Blanquart, D. Hudoklin, P. Jacobson, A. Kentved, I. Lóio, G. Mamontov, A. Masarykova, H. Mitter, R. Mnguni, J. Otych, A. Steiner, N. Szilágyi Zsófia, D. Zvizdic, Int. J. Thermophys. 33, 1422 (2012) 13. A.B. Kentved, M. Heinonen, D. Hudoklin, Int. J. Thermophys. 33, 1408 (2012) 14. D. Sonntag, Meteorol. Z. 3, 51 (1994) 15. BIPM web page.