Intercomparison of radiation temperature measurements over the temperature range from 1600 K to 3300 K

INSTITUTE OF PHYSICS PUBLISHING Metrologia 40 (2003) S39 S44 METROLOGIA PII: S0026-1394(03)56056-X Intercomparison of radiation temperature measurements over the temperature range from 1600 K to 3300 K B B Khlevnoy 1, N J Harrison 1, L J Rogers 1,DFPollard 1, NPFox 1, P Sperfeld 2, J Fischer 2, R Friedrich 2,Jürgen Metzdorf 2, Joachim Seidel 2, M L Samoylov 3, R I Stolyarevskaya 3, V B Khromchenko 3, S A Ogarev 3 and V I Sapritsky 3 1 Centre for Optical and Analytical Measurement, National Physical Laboratory (NPL), Teddington, Middlesex TW11 0LW, UK 2 Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany 3 All-Russian Research Institute for Optical and Physical Measurements (VNIIOFI), 46 Ozernaya, Moscow, 119361 Russia Published 7 February 2003 Online at stacks.iop.org/met/40/s39 Abstract An intercomparison of radiation temperature measurements was performed at VNIIOFI during October 2000 using a pyrolytic graphite blackbody operating over the temperature range from 1600 K to 3300 K. A pyrometer and two photometers from VNIIOFI, a pyrometer and four broadband glass filter detectors from PTB, and two narrow-band interference filter based radiometers and a broadband glass filter radiometer from NPL were used to perform the temperature measurements in either radiance or irradiance mode. Across almost the entire temperature range the VNIIOFI, NPL and PTB instruments showed results within the combined standard measurement uncertainties. 1. Introduction The present intercomparison was part of the International Project for Luminous Flux Realization [1], based on the high-temperature blackbody (HTBB) as a standard source. Cavity temperature measurements contribute the dominant uncertainty for the realization of emission scales based on high-temperature blackbody radiation in the visible range. Therefore the project participants have been spending significant effort to develop and improve the methods of blackbody temperature measurement and to compare their thermodynamic temperature realizations. The accurate determination of radiation temperature is important not only for photometry but also more critically for spectroradiometry. Recently spectral emission scales have been established at a number of national metrology institutes that rely on the use of a HTBB to realize the spectral radiance and spectral irradiance standards [2 5]. For these applications radiation temperature is usually determined by measuring the radiance or irradiance from the blackbody using a radiometer that is a filter detector combination calibrated in terms of spectral responsivity over a specific wavelength band. The filter radiometer is calibrated, as a rule, against a cryogenic radiometer, linking emission scales to an absolute detector [6, 7]. The first stage of the intercomparison took part at VNIIOFI in Moscow in 1997 [8] between NPL, PTB and VNIIOFI. Seven radiometers were used: four radiance-mode filter radiometers from NPL, two irradiance-mode filter detectors (FDs) from PTB and a photometer from VNIIOFI. The agreement obtained between participants was around 1 K to 2 K, which corresponds to ±0.5% in terms of spectral radiance of the blackbody at 800 nm. In October 2000 the second stage of the intercomparison took place at VNIIOFI, which is the subject of this paper. The same laboratories participated in the second stage. Eleven instruments were used including two pyrometers, one from VNIIOFI (TSP-2) and one from PTB-Berlin (LP3), calibrated against the national temperature scales of these countries. 2. Equipment and procedure A schematic diagram of the experimental configuration is shown in figure 1. The facility consisted of a blackbody 0026-1394/03/SS0039+06$30.00 2003 BIPM and IOP Publishing Ltd Printed in the UK S39

B B Khlevnoy et al drift was taken into account to work out the corrections for the temperature values measured by each individual radiometer. TSP-2 LP3 Feedback Detector BB3200pg Blackbody NPL Radiometer Aperture PTB FDs VNIIOFI Photometer Figure 1. Schematic diagram of the experimental facility used for the intercomparison of radiation temperature measurement. and a set of radiometers mounted on a translation stage to allow rapid positioning of each instrument. The radiometers used comprised instruments operating in both radiance and irradiance mode. In front of the blackbody there was a precision water-cooled aperture that allowed any detector, aligned along the blackbody cavity axis, to view only the central part of the cavity bottom, which has the highest uniformity. Accurate knowledge of the distance to the aperture and its area were necessary for the irradiancemode radiometers. The aperture diameter of approximately 8 mm had been measured with an uncertainty less than 0.5 µm. The distance between the aperture and the irradiancemode radiometers was approximately 900 mm and had been measured with uncertainty as low as 10 µm. The radiancemode radiometers were focused on the centre of the aperture. 2.1. Blackbody The blackbody was a BB3200pg type blackbody with a pyrographite cavity. It was investigated carefully during the first-stage intercomparison in 1997 [9]. Further investigations of this type had confirmed its high quality [10, 11, 15]. Only the stability of the BB3200pg had been criticized during the 1997 intercomparison. Since then a new blackbody radiation stabilizing system had been developed. A temperaturestabilized filtered detector monitored, as part of a feedback loop, some of the cavity radiation. This feedback detector was situated directly in front of the blackbody between its opening and the aperture. Residual instability was characterized by a slow drift. To estimate the level of the drift during the intercomparison, measurements with TSP-2 and LP3 were done before and after the sequences of other radiometer measurements for each measurement run. For the majority of runs the drift was so low that it was not possible to detect it. For other runs the drift level was around 0.1 K h 1 to 0.2 K h 1 with the total time of each run being about 40 min. In these cases the S40 2.2. Radiometers Eleven radiometers of five different types from three laboratories were used for the intercomparison. All of them were based on the combination of a silicon photodiode and a glass, interference or liquid filter. NPL provided three filter radiometers [12] to measure the radiance of the blackbody. For these a lens was used that focused the image of the blackbody aperture onto the 5 mm filter radiometer aperture with a magnification factor of 1. Two radiometers had narrow-band interference filters and one had a broadband glass filter. The radiometer temperature was stabilized at (20 ± 0.1) C using a water thermostat. The filter radiometer and lens were mounted on the same optical axis with a 10 mm diameter thin film aperture placed immediately behind the lens. The separation distance from lens to radiometer was 600 mm. The filter radiometers had been calibrated in terms of absolute spectral responsivity. The spectral transmittance of the lens had been measured separately using the same geometrical conditions. The signal recorded on the filter radiometer, its spectral responsivity, lens transmission and the measurement geometry define the blackbody radiation flux, and hence the radiation temperature can be calculated according to Planck s radiation formula. The uncertainty (k = 1) of the temperature measurements with the NPL filter radiometers was, depending on temperature region, between 0.11 K and 0.49 K for narrow-band radiometers and between 0.3 K and 1.3 K for the broadband one. Only one NPL radiometer could be set on the translation stage. Therefore they were mounted and aligned and the signals were recorded in turn for each blackbody temperature regime. Two types of radiometer were provided by PTB: a radiance-mode Linearpyrometer LP3 and four irradiancemode FDs. The LP3 [13] used a specially selected silicon photodiode and an interference filter with the effective wavelength of 653 nm. The calibration of the LP3 was performed in two different ways. For both, the HTBB was used as a transfer standard. In the first case the HTBB temperature was measured with respect to the standard PTB gold fixedpoint blackbody, e.g. according to ITS-90. In the second case the temperature of the HTBB was measured by an interference filter radiometer, whose spectral responsivity was traced back to the radiation thermometry cryogenic radiometer. Having determined the HTBB temperature, the LP3 was focused onto the entrance aperture of the blackbody and calibrated. The uncertainty (k = 1) of the LP3 calibration was in the range from 0.25 K to 0.82 K. Several calibration runs of the LP3 had been performed before the comparisons in September 2000. Both calibration methods agreed well within the stated uncertainties. An additional calibration after the comparisons in January 2001 confirmed the values obtained within the stated uncertainties. The LP3 was set at the distance of 690 mm to the blackbody aperture with the target spot diameter of 0.8 mm and the view angle of 0.0025 sr. The PTB FDs were constructed using broadband glass filters with Si photodiodes and high-precision apertures [14, 15].

Intercomparison of radiation temperature measurements Table 1. Summary of filtered detector characteristics used for the radiance temperature intercomparison. Filtered Radiance/ Centre Laboratory detector irradiance mode Identifier wavelength/nm Bandwidth/nm Filter type NPL 710W50S5227B Radiance NPL710 710 80 Glass NPL 647W10H5227F Radiance NPL647 647 10 Interference NPL 800W20THE Radiance NPL800 800 20 Interference PTB Pyrometer Radiance LP3 653 20 Interference PTB FD16 Irradiance FD16 510 165 Glass PTB FD17 Irradiance FD17 558 104 Glass PTB FD20 Irradiance FD20 437 85 Glass PTB FD23 Irradiance FD23 728 56 Glass VNIIOFI Pyrometer Radiance TSP-2 650 22 Interference VNIIOFI Photometers Irradiance Photom. 555 100 Liquid The FDs had been calibrated in terms of absolute spectral irradiance responsivity against the radiation thermometry cryogenic radiometer. The radiometric HTBB temperature could be calculated according to Planck s law from the measured photocurrent by considering the radiometer s responsivity and the geometric irradiance conditions. The uncertainty (k = 1) of the radiometric temperature determinations was between 0.4 K and 3.1 K depending on the temperature and the radiometer. The following radiometers were used: FD16, FD17, FD20 and FD23 with central responsivity wavelengths 510 nm, 558 nm, 437 nm, 728 nm and bandwidths 165 nm, 104 nm, 85 nm, 56 nm respectively. The FDs holder was temperature stabilized to (25±0.1) C using a water thermostat. The holder could contain three FDs at the same time. Therefore FD16 and FD17 remained throughout but FD20 and FD23 were replaced in turn. VNIIOFI provided two photometers and the pyrometer TSP-2. The photometers were irradiance-mode instruments based on the specially designed liquid filter to get the spectral responsivity very close to the V (λ) curve [16]. The photometers were temperature stabilized by a wirewound heater coil at 30 C. They had been calibrated against the VNIIOFI photometric scale to measure illuminance from the blackbody. Temperature could be calculated from the value of measured illuminance, blackbody aperture area and the distance from the aperture to the photometer using the standard tabulation of V (λ). The two photometers had been calibrated as an integrated pair and the resultant measurement of this pair, according to the normal VNIIOFI photometric practice, was a single value, which was worked out from the average of the two photometer readings. The design of the TSP-2 [17] was based on a silicon photodiode/interference FD head with a maximum responsivity at about 650 nm. The head temperature was stabilized using a Peltier cooler. The distance between the pyrometer and the blackbody aperture was about 650 mm with the solid angle of detected flux of 0.0015 sr and the target size of about 1 mm. The calibration of TSP-2 was performed using two procedures: (1) measurement of the relative spectral responsivity of the TSP-2 and absolute calibration against the copper fixed-point blackbody; (2) use of radiance temperature lamps calibrated with respect to the gold point. The uncertainty (k = 1) of the TSP-2 calibration was determined to be between 0.5 K and 0.9 K. The two calibration methods agreed within the uncertainties. Table 1 summarizes the key attributes of all the filtered detectors used in the intercomparison. 3. Results and discussion All the instruments were calibrated before and after the comparison to check their stability. All radiometers showed very good reproducibility, except FD20, which only had to be corrected after the second calibration in the long-wavelength wing, since it showed a much lower responsivity above 600 nm than was expected after the first calibration. Hence the FD20 results presented in this paper are based on the calibration that took place after the intercomparison, while the results for all other radiometers are based on the prior calibration. The results are presented in table 2 and figure 2 in terms of the difference in radiation temperature from the mean values, which was the average values of temperature measured by all radiometers except NPL647, which showed significantly different results from the others. The values of both FD16 and FD20 at a temperature of 1600 K and the value of FD20 at a temperature of 3300 K were also not taken into account when the mean value was calculated. Figure 3 illustrates the same results recalculated in terms of the equivalent difference in spectral radiance at 800 nm. This wavelength was chosen to make it easy to compare the results presented here with the ones obtained during the first stage in 1997 [8]. In table 2 the standard uncertainties (k = 1) of the radiometer calibration are also given. Only the NPL647 radiometer shows a significant discrepancy over the whole temperature range, both with the mean and with the other two NPL radiometers. The radiometer did not have a recent calibration before the intercomparison but was calibrated a short time after. The calibration after the intercomparison did not correct these results. There is no exact understanding why the NPL647 differed so much from the other NPL radiometers. Transportation or uncontrolled conditions between the laboratories may have been the cause. All other radiometers at all temperature points (except the three particular points mentioned above, which were excluded from the mean values) showed differences less than their k = 2 uncertainties. Moreover, 80% of all points agreed within the k = 1 level. In terms of absolute temperature most radiometers showed agreement within ±0.6 K, about three times less than during the 1997 intercomparison. In terms of spectral radiance the agreement improved as the temperature increased: above 2400 K it was around S41

S42 Table 2. Results from radiation temperature measurements relative to the mean value: temperature difference, T, and standard (k = 1) uncertainty of radiometer calibrations. Those in bold have deviations higher than the standard uncertainty. TSP-3 LP3 NPL 710 NPL 647 NPL 800 FD16 FD17 FD20 FD23 Photom. Mean temp. T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K T /K U/K 1600.07 0.59 0.6 0.03 0.25 0.22 0.3 0.67 0.12 0.02 0.11 3.10 1.4 0.21 1.1 6.11 2.8 0.59 0.4 0.73 0.50 1801.03 0.53 0.55 0.13 0.28 0.11 0.38 1.18 0.15 0.01 0.14 0.57 0.8 0.33 0.7 1.17 1.2 0.73 0.5 0.12 0.50 2001.42 0.42 0.5 0.02 0.32 0.02 0.48 1.37 0.18 0.06 0.18 0.02 0.6 0.22 0.6 0.38 0.7 0.32 0.5 0.73 0.55 2002.98 0.58 0.5 0.03 0.32 0.06 0.48 1.34 0.18 0.04 0.18 0.22 0.6 0.28 0.6 0.52 0.4 0.28 0.5 0.42 0.55 2202.62 0.81 0.5 0.17 0.38 0.24 0.57 1.42 0.22 0.32 0.22 0.02 0.5 0.08 0.5 0.68 0.4 0.18 0.6 0.63 0.60 2412.70 0.60 0.5 0.23 0.43 0.24 0.69 1.70 0.26 0.40 0.26 0.30 0.5 0.30 0.6 0.40 0.5 0.50 0.8 0.60 0.60 2600.83 0.36 0.55 0.19 0.5 0.26 0.8 1.58 0.31 0.37 0.3 0.07 0.6 0.68 0.6 1.01 0.5 0.59 0.9 0.60 0.65 2803.58 0.49 0.6 0.06 0.57 0.12 0.93 1.88 0.36 0.49 0.35 0.18 0.7 0.18 0.7 1.12 0.6 0.58 1 0.62 0.70 3001.96 0.38 0.7 0.06 0.65 0.39 1.07 1.93 0.41 0.16 0.4 0.00 0.7 0.75 0.8 1.18 0.7 0.94 1.1 0.59 0.80 3003.71 0.62 0.7 0.23 0.65 0.42 1.07 1.98 0.41 0.64 0.4 0.11 0.7 0.21 0.8 1.49 0.7 0.31 1.1 0.19 0.80 3204.57 0.30 0.8 0.13 0.73 0.63 1.22 2.27 0.46 0.05 0.46 0.57 0.8 0.17 0.9 0.13 0.8 0.13 1.3 0.08 0.90 3294.87 0.38 0.9 0.52 0.82 1.48 1.29 1.68 0.49 0.02 0.48 1.54 0.9 0.43 0.9 2.11 0.8 0.48 1.4 0.34 1.10 B B Khlevnoy et al

Intercomparison of radiation temperature measurements Difference to Mean Value / K 2.5 2.0 1.5 1.0 0.5 0.0-0.5-1.0-1.5-2.0 TSP-2 LP3 NPL710 NPL647 NPL800 FD16 FD17 FD20 FD23 Photom. -2.5 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 3400 Temperature / K Figure 2. Difference in radiation temperature values from the mean values. The solid line represents the uncertainty of the NPL710 (as an example). Difference in radiance at 800 nm 100 0.8 0.6 0.4 0.2 0.0-0.2-0.4-0.6 TSP-2 LP3 NPL 710 NPL 647 NPL 800 FD16 FD17 FD20 FD23 Photom. -0.8 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 3400 Temperature / K Figure 3. Difference in blackbody radiance at 800 nm from the mean, calculated from radiation temperature difference values. ±0.2% and down to 1600 K it increased to ±0.4%. FD16 and FD20 at 1600 K disagreed by 2% and 4%, respectively. The problems at the lower end of the temperature range may reflect the increasing dominance of the red and near-infrared end of the spectrum in the blackbody spectral emission profile. Measurement of such spectral distributions severely tests the out-of-band blocking of any visible wavelength radiometer as the radiance from the blackbody can be several orders of magnitude larger in the near-infrared than in the visible pass band of the filter. Although most of the radiometers lie within their uncertainty limits at 3300 K, all of them show an increased difference from the mean value at this point. There could be several explanations for this behaviour. The likely cause is that the filters were heated by the high radiation level. A signal drop was observed for some radiometers after exposure until it reached a steady state. The alternative explanation is out-ofband blocking. At such blackbody temperatures not only is the long wavelength blocking important, but the short wavelength blocking is also. Another possibility is fluorescence in the filters because at 3300 K there is significant UV radiation. An increase in nonuniformity along the blackbody cavity could also have been the cause. Figure 4 shows a difference between the thermodynamic temperature scales of the participants. VNIIOFI s scale is presented as an average of the TSP-2 and photometer results, NPL s as an average of the NPL710 and NPL800 radiometers, PTB s as an average between the LP3 and the average scale of Difference / K 0.60 0.40 0.20 0.00-0.20 VNIIOFI PTB NPL -0.40 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 3400 Temperature / K Figure 4. Difference between participant s temperature scales (as an average of the radiometers used by each laboratory) and the mean. the FDs. All scales agreed with each other to within ±0.3 K excluding the temperature 3300 K where the discrepancy was about 0.6 K, which lies within the k = 1 uncertainty for each scale. During the intercomparison the determination of two hightemperature fixed points was also carried out. The fixed points were based on the metal carbon eutectics Ir C and Re C. The fixed-point cells were made and provided by the National Metrology Institute of Japan [18]. Three radiancemode radiometers were used for this purpose: TSP-2, LP3 and NPL800. The average values of the measured melting and freezing temperatures of the Ir C fixed point were determined to be 2564.3 K and 2564.6 K, respectively. The details of the determination have been published in [19]. The corresponding values for Re C were 2748.5 K and 2748.2 K, respectively. The fixed points were found to be highly reproducible to around 0.03 K. The participants agreed that this fact makes this kind of fixed point very promising as a reference point for realizing temperature and radiometric scales, or at least as a perfect blackbody radiation source for further intercomparisons of temperature scales. 4. Conclusion Good agreement was obtained between the individual radiometers used for the intercomparison with temperature differences of around ±0.6 K from the mean values, which is about three times less than it was during the 1997 intercomparison. To increase the accuracy of temperature measurement, a group of radiometers should be used instead of a single radiometer. The comparison of the participant s scales calculated as an average for the groups of radiometers provided by each laboratory showed agreement between the scales within ±0.3 K. For the temperature region from 2600 K to 3000 K, which is important for the Lumen Project realization, this agreement is around ±0.15 K, i.e. ±0.04% in terms of luminous flux with the target uncertainty of ±0.05% [1]. The Ir C and Re C fixed-point temperatures were determined. The fixed points appeared very promising for use in further scale realizations and intercomparisons. References [1] Sapristky V I et al 2002/2003 Project to realize the lumen at VNIIOFI (Russia) with the participation of NIST (USA), NPL (UK) and PTB (Germany), presented at NEWRAD 2002, Gaithersburg, MD [2] Sapritski V I 1990 Metrologia 27 53 60 S43

B B Khlevnoy et al [3] White M, Fox N P, Ralph V E and Harrison N J 1995/1996 Metrologia 32 431 4 [4] Sperfeld P, Metzdorf J, Galal Yousef S, Stock KDand Moller W 1998 Metrologia 35 267 71 [5] Friedrich R and Fischer J 2000 Metrologia 37 539 42 [6] Fox N P, Chunnilall C J, Harrison N J and Hartree W S 1996 Proc. SPIE 2815 32 41 [7] Friedrich R, Fischer J and Stok M 1995 Metrologia 32 509 13 [8] Harrison N J et al 1998 Metrologia 35 283 8 [9] Sperfeld P, Metzdorf J, Harrison N J, Fox N P, Khlevnoy B B, Khromchenko V B, Mekhontsev S N, Shapoval V I, Zelener M F and Sapritsky V I 1998 Metrologia 35 419 22 [10] Galal Yousef S, Sperfeld P and Metzdorf J 2000 Metrologia 37 365 8 [11] Woolliams E R, Harrison N J and Fox N P 2000 Metrologia 37 501 4 [12] Anderson V E, Fox N P and Nettleton D H 1992 Appl. Opt. 31 536 45 [13] Fischer J, Neuer J, Schreiber E and Thomas R 2001 Proc. TEMPMEKO 2001 801 6 [14] Sperfeld P, Raatz K-H, Nawo B, Möller W and Metzdorf J 1995/1996 Metrologia 32 435 9 [15] Sperfeld P 1998 Entwicklung einer empfängergestützten spektralen Bestrahlungsstärkeskala http://www.biblio.tubs.de/ediss/data/199990628a/199990628a.html Braunschweig [16] Ivanov V S, Krutikov V N, Sapritsky V I, Stolyarevskaya R I and Khlevnoy B B 2001 Proc. Int. Lighting Congress, Istanbul, 12 14 September vol 11, pp 365 9 [17] Sapritsky V, Khlevnoy B, Samoylov M and Yamada Y 2001 Proc. TEMPMEKO 2001 513 18 [18] Yamada Y, Sakate H, Sakuma F and Ono A 2001 Metrologia 38 213 19 [19] Khlevnoy B, Khromchenko V, Samoylov M, Sapritsky V, Harrison N, Sperfeld P and Fischer J 2001 Proc. TEMPMEKO 2001 845 50 S44