Final Report on the International Comparison of Luminous Responsivity CCPR-K3.b

Transcription

1 Fnal Report on the Internatonal Comparson of Lmnos Responsvty CCPR-K3.b R. Köhler, M. Stock, C. Garrea Brea Internatonal des Pods et Mesres Pavllon de Bretel 93 Sèvres Cedex France Janary 004 Abstract Drng ts 3 th meetng n 994 the Consltatve Commttee for Photometry and Radometry (CCPR) decded to carry ot a comparson of lmnos responsvty to be organzed by the BIPM. Ths comparson was classfed as a key comparson by the 4 th CCPR n 997 and sbseqently named CCPR-K3.b. Ffteen laboratores took part n ths comparson by sendng calbrated photometers to the BIPM. The photometers were compared wth a common set of reference photometers to dedce the relatve dfferences between the calbratons made n the partcpatng laboratores. In ths report the detals of the procedre and the reslts are presented.

2 Contents ABSTRACT... CONTENTS.... INTRODUCTION THE PARTICIPANTS THE PHOTOMETERS PROTOCOL OF THE COMPARISON PREPARATORY MEASUREMENTS PHOTOMETER MOUNTING SPECTRAL RESPONSIVITY INFLUENCE OF THE LAST APERTURE IN FRONT OF THE PHOTOMETERS LINEARITY SENSITIVITY TO ALIGNMENT MEASUREMENTS FOR THE COMPARISON ALIGNMENT PROCEDURES LAMP POSITION DISTANCE PHOTOMETER ALIGNMENT COMPARISON MEASUREMENTS DATA ANALYSIS RESULTS UNCERTAINTIES OF THE COMPARISON REALIZATION OF THE PHOTOMETRIC QUANTITIES IN THE DIFFERENT LABORATORIES CONCLUSIONS.... ACKNOWLEDGEMENTS REFERENCES... APPENDIX : CORRECTIONS OF CLERICAL ERRORS IN DRAFT A AND DRAFT B... 3 APPENDIX : REVISED RESULTS SUBMITTED AFTER DRAFT A...4 APPENDIX 3: DEGREES OF EQUIVALENCE FOR CCPR-K3.b... 6 APPENDIX 4: FORMULAS FOR DEGREES OF EQUIVALENCE FOR WEIGHTED MEAN WITH CUT-OFF... 9

3 . Introdcton Drng the 3 th meetng of the CCPR n 994 t was decded to carry ot a comparson of lmnos responsvty sng V(λ)-corrected detectors. Ths was the frst tme that the CCPR decded to carry ot an nternatonal comparson of a photometrc qantty sng only detectors as transfer standards. So far n the hstory of the CCPR only lamps were sed to compare photometrc nts, except for a mxed lamp / detector comparson n 978 whch was consdered a tral comparson to assess the performance of lamps and detectors as transfer standards rather than a comparson of the nts themselves []. The prpose of the crrent comparson s not to dentfy a 'best' photometer, bt to compare lmnos responsvty vales determned by the laboratores. The present exercse was planned to gve smlar nformaton abot the coherence of natonal lmnos ntensty nts n parallel wth another comparson (key comparson CCPR-K3.a, []), prepared at the same tme, sng lamps as transfer devces. It was hoped that the otcome of ths comparson wold show whether state-of-the-art photometers cold provde a qalty of transfer standard for photometrc nts whch wold eqal, or even srpass, that of lamps. A workng grop consstng of CSIRO, NPL, OMH, PTB wth the BIPM as the convenor was set p and recommended the parameters for the comparson.. The partcpants The followng table lsts the partcpatng laboratores: acronym laboratory name cty contry BIPM Brea Internatonal des Pods et Mesres Pars BNM-INM Brea Natonal de Métrologe Insttt Natonal de Pars France Métrologe CSIC / IFA Conseo Speror de Investgacones Centfcas / Insttto Madrd Span de Fsca Applcada CSIRO Commonwealth Scentfc and Indstral Research Lndfeld Astrala Organzaton HUT Helsnk Unversty of Technology Espoo Fnland IRL Indstral Research Lmted Lower Htt New Zealand KRISS Korean Research Insttte of Standards and Scence Teaon Rep. of Korea NIM Natonal Insttte of Metrology Beng Chna NIST Natonal Insttte of Standards and Technology Gathersbrg USA NPL Natonal Physcal Laboratory Teddngton U.K. NRC Natonal Research Concl Ottawa Canada OFMET Edgenösssches Amt für Messwesen Wabern Swtzerland OMH Országos Mérésügy Hvatal Bdapest Hngary PTB Physkalsch-Technsche Bndesanstalt Branschweg Germany SMU Slovenský Metrologcký Ustav Bratslava Slowaka VNIIOFI All Rssan Research Insttte for Optophyscal Moscow Rssa Measrements Table : Partcpatng laboratores. IRL : now MSL, Measrement Standards Laboratory OFMET : now METAS, Bndesamt für Metrologe nd Akkredterng / Swss Federal Offce of Metrology and Accredtaton 3

4 3. The photometers The photometers sed were commercally avalable ones. It was decded that they shold be flly fltered and thermally stablzed. After contactng the dfferent manfactrers three photometers were chosen, one from each of the followng manfactrers: LMT, PRC Krochmann and Inphora, for the rest of ths docment they wll be referred to as LMT, PRC and IPR, respectvely. The BIPM prchased one photometer of each type pror to the start of the comparson, henceforth referred to as BIPM-, BIPM- and BIPM-3, servng as a reference grop (Table ). In a later stage for more photometers were prchased (BIPM-4 to BIPM-7). These photometers were modfed to derve the BIPM detector-based lmnos responsvty scale. BIPM-6 faled drng the comparson. Photometer Manfactrer Seral Nmber BIPM- PRC 9506 BIPM- LMT BIPM-3 IPR PO00 BIPM-4 LMT (modfed) BIPM-5 LMT (modfed) BIPM-6 PRC (modfed) BIPM-7 PRC (modfed) Table : Lst of photometers sed at the BIPM. Photometers BIPM-, BIPM- and BIPM-3 formed the reference grop. BIPM-4 to BIPM-7 were sed as transfer standards to represent the BIPM lmnos responsvty nts. 4

5 4. Protocol of the comparson Each partcpant was asked to send two photometers to the BIPM, together wth ther calbraton report ncldng the ncertanty bdget and expermental condtons. The followng table lsts the photometers whch were sent to the BIPM. The photometers are dentfed by the acronym of the laboratory and a rnnng nmber. The table also gves the lmnos responsvty 3 of each photometer as gven by the partcpants before and after the devces were at the BIPM. Three laboratores opted to send three photometers. The BIPM partcpated wth ts for photometers BIPM-4, BIPM-5, BIPM-6 and BIPM-7 (Table ). Photometer Manfactrer Seral Nmber S (before) / (na/lx) S (after) / (na/lx) BNM / INM- LMT BNM / INM- PRC CSIC- IPR 30TH CSIC- IPR PO CSIRO- LMT CSIRO- PRC CSIRO-3 IPR 0696PO HUT- PRC F.73.7 HUT- PRC F.5.4 IRL- PRC IRL- PRC KRISS- LMT KRISS- LMT KRISS-3 LMT NIM- PRC NIM- IPR 30 TH NIST- LMT NIST- LMT NPL- LMT NPL- PRC NRC- PRC NRC- PRC OFMET- LMT OFMET- LMT OFMET-3 LMT OMH- PRC OMH- PRC PTB- LMT PTB- LMT SMU- LMT SMU- LMT VNIOFI- LMT VNIOFI- LMT Table 3: Lst of photometers sed n the comparson. Note: INM- faled shortly after arrval at the BIPM. The lmnos responsvty vales S(before) and S(after) are those the partcpants stated. 3 The lmnos responsvty s defned as the rato of the photocrrent and the llmnance n the plane of the entrance apertre. Its nt s A / lx. 5

6 All photometers were generally measred aganst the reference grop for the frst tme at the BIPM only a few days after ther arrval. They were all measred at least on fve dfferent days. The devces were then retrned to the orgnatng laboratores where the partcpants checked them for drft n responsvty. They then reported the fnal lmnos responsvty vales to the BIPM. Measrements were orgnally schedled to be made at llmnant A and at an llmnance level of approxmately 50 lx or lower. Alternatvely a proposal had been made to perform measrements also wth an llmnance of 500 lx or hgher and at a dstrbton temperatre of 3000 K. Only a few laboratores dd these spplementary measrements. When the meetng of the 4 th CCPR was advanced n date, t was decded to se the occason to retrn the devces to the laboratores and becase of the lack of tme only measrements at llmnant A at abot 40 lx were made at the BIPM. All bt one photometer showed sffcent stablty drng the tme they were kept at the BIPM; the photometer INM- faled some tme after arrval at the BIPM. After the frst two seres of measrements gvng the same vale t sddenly changed ts responsvty twce by several percent. Vsal nspecton showed a frnge pattern, presmably Newton's rngs, n the flter. The spposton s that the cement between the dfferent layers n the flter faled and that the flter sbseqently delamnated. Smlar behavor n ths type of photometers has been reported n two other cases. 5. Preparatory measrements The spectral responsvtes of the photometers BIPM-, BIPM- and BIPM-3 were measred sng an expermental arrangement wth a doble monochromator, descrbed elsewhere [3, 4]. The photometrc bench employed for the comparson of the photometers s descrbed later. 5. Photometer montng One observaton made at the BIPM drng these preparatory measrements was that the temperatre stablzaton of the photometer heads may not work correctly f the devces are clamped on a metal block wth too large a mass. A small layer of thermal nslaton (e.g. Teflon) easly avoded ths problem. Fgre shows the montng sed at the BIPM wth the nslaton ndcated. Ths nformaton was gven to the partcpants pror to the start of the comparson. Teflon screw Photometer Thermal nslaton Fgre : Thermal nslaton of the photometer mont. Montng base 6

7 5. Spectral responsvty The spectral responsvty crve was measred sng a doble monochromator arrangement descrbed elsewhere [3, 4]. The three BIPM photometers formng the reference grop and a reflecton trap detector whch had been calbrated aganst the BIPM cryogenc radometer, were rradated by a monochromator and compared wth each other. The measrements were done n the followng seqence: photometers A-B-C-C-B-A, takng the average of the two reslts for each photometer corrected for the small drft n the sorce radance drng a rn. The relatve spectral responsvty, normalzed to nty at 555 nm s shown n Fgre, compared wth the V(λ) fncton.. Relatve spectral responsvty LMT PRC IPR V(lambda) Wavelength / nm Fgre : Relatve spectral responsvtes of the three BIPM photometers formng the reference grop compared aganst V(λ). All three photometers have a farly small f ' vale as defned n [5]: f ' (BIPM-, PRC) =.5 % f ' (BIPM-, LMT) = 3.0 % f ' (BIPM-3, IPR) =.5 % Partally fltered photometers wth smaller f ' vales are avalable, bt t was dged by the workng grop preparng the comparson that the nformty of response over the entrance plane of the photometer was more mportant than a smaller f ' vale. The photometrc msmatch as a fncton of the dstrbton temperatre of a sorce F(T d ) [5] s shown n Fgre 3a. 7

8 Photometrc msmatch factor Dstrbton temperatre / K LMT PRC IPR Fgre 3a: Spectral msmatch for the BIPM photometers as a fncton of dstrbton temperatre. It s clear from the graph that a small error n the dstrbton temperatre drng the calbraton wll only have a neglgble nflence on the reslts. Ths s especally tre when lookng at the ratos between two photometers (Fgre 3b). Relatve change n photometer sgnal rato Dstrbton temperatre / K PRC/LMT IPR/LMT IPR/PRC Fgre 3b: Change n the measred rato of photometer lmnos responsvty as a fncton of dstrbton temperatre. 5.3 Inflence of the last apertre n front of the photometers Measrements were made to see f the dameter of the apertre closest to the photometer head cold nflence the reslts of the measrements. Ths was necessary becase of the lmted openng of the shtter. An rs daphragm, effectvely the one nclded n the shtter assembly, was placed abot 0 cm from the photometers. The dstance between the 8

9 photometers and the lamp was.5 m. The dameter of the apertre was adsted sch that ts shadow lne was st otsde the largest entrance apertre of the photometers. The ratos between the photometer sgnals were taken and then the daphragm was opened frther so that new ratos cold be measred. Ths seres was contned ntl the largest possble openng of the daphragm was reached. Fgre 4 shows the reslt. The rato of the sgnals LMT/PRC hardly changes as a fncton of the dameter of the daphragm. The only small change vsble occrs when the dameter corresponds closely to the openng of the photometer. Ths s probably de to dffracton. The rato IPR/PRC however, alters dramatcally wth the beam dameter. The steepness of the crve changes twce before reachng a constant vale. Ths behavor can be attrbted to the shape of the front end of the IPR photometer head (see Fgre 5). A seres of concentrc rngs precedes the defnng apertre on the V(λ) flter. Scattered lght from the nsde of these rngs changes as the llmnated area ncreases. For the measrements on the photometrc bench a dameter of mm was chosen for the last daphragm becase the effect dsappears at dameters greater than ths vale Rato of photometer sgnals LMT/PRC IPR/PRC Dameter / mm Fgre 4: Inflence of the apertre dameter on the photometer sgnal rato. Fgre 5: Shape of the front sde of the photometers. 9

10 5.4 Lnearty All measrements at the BIPM were made at approxmately the same llmnance level. Antcpatng that some laboratores cold se dfferent parameters we checked the lnearty of the photometers wth the flx addton method. A lens formed the mage of a lamp on to the photometer to be tested. Close to the plane of the lens an opaqe dsk contanng a crclar hole was nserted. The openng n the dsk had two crossed bars (see Fgre 6). Behnd that dsk, a second one cold be rotated to nsert holes havng the forms shown n the fgre, ths allowng a sbdvson of the flx transmtted throgh the dsk. Dfferent combnatons then allowed a check of the lnearty. Lamp Lens Shtter Photometer Dsc wth apertres Fgre 6: Set-p for the lnearty measrement. Explanaton n the text. Measrements were done wth dfferent sorces and dfferent focal lengths of the lens p to an llmnance level of abot 000 lx. No devaton from lnearty was detected wthn the ncertanty of the measrements for all photometers. The measrements for the comparson at the BIPM were made at an llmnance level of abot 40 lx. 5.5 Senstvty to algnment The nflence of an nclnaton of the photometers wth respect to the optcal axs was stded by ntrodcng sch an nclnaton prposely and recordng the dfference fond n the rato between the dfferent photometers, wth and wthot nclnaton. It was fond that wth a sorce-detector dstance of.5 m an nclnaton of abot 5 reslted n a change of the measred rato from 0. % to 0.4 %, dependng on the photometer type. We estmated an ncertanty of 0.3 n the orentaton of the photometers drng or measrements resltng n a contrbton of 0.03 % to the overall ncertanty. 0

11 6. Measrements for the Comparson 6. Algnment procedres 6.. Lamp poston The lamps sed were Osram W 4 G type lamps, adsted to a dstrbton temperatre of 856 K ± 5 K. Accordng to the reslts shown n Fgre 3b, the ncertanty n the dstrbton temperatre has only a neglgble nflence on the reslts. The lamp was postoned reprodcbly sng the followng technqe (Fgre 7): Fgre 7: Method of lamp algnment. The lamp flament was orented vertcally by vsal comparson of the flament nclnaton wth a plmb-lne. The heght of the flament was adsted by alterng ts poston ntl the mddle of the flament concded wth a marker on the plmb-lne. The lamp was then dsplaced along the axs of the photometrc bench ntl the flament, the plmb-lne and a vertcal lne on a wall n the laboratory, abot.5 m away from the lamp, algned. Fnally the lamp was lt and the lamp holder rotated ntl the weak shadow of the flament became sharpest, ndcatng algnment of the flament perpendclar to the optcal axs.

12 6.. Dstance Three photometers were monted at the same tme on a translaton stage whch allowed dsplacement perpendclar to the optcal axs 4. Ths stage tself was monted on a sldng table whch cold be manally dsplaced along the photometrc bench n the drecton of the optcal axs. The frthermost part of the sldng-table was postoned at the pont normally occped by the lamp by placng t n lne wth the plmb-lne and the vertcal lne on the wall, n a manner smlar to that sed for algnment of the lamp flament. A rler on the photometrc bench was then read sng a mcroscope attached to the table. The table was then dsplaced by.5 m. A laser dode, adsted to be vertcal sng the reflecton of ts beam from a water srface, was sed to 'pnpont' ths poston (Fgre 8, poston A). We estmate the ncertanty n the dstance from ths procedre to be 0.3 mm. Fgre 8: Algnment of the dstance of the defnng apertre. The table was then moved agan ntl one of the photometer front srfaces st toched the vertcal laser beam ndcatng that the photometer srface was then.5 m from the lamp's flament. The fnal algnment for the lmtng apertres was done relatve to ths poston sng small sldng tables ftted wth mcrometer screws (Fgre 8, poston B) Photometer algnment A beam splttng cbe cold be nserted n the optcal path of the photometrc bench. A laser beam reflected off the cbe was then algned parallel to the optcal axs of the photometrc bench (Fgre 9). 4 we defne the optcal axs as the lne between the lamp and the photometer to be measred.

13 Sldng table wth photometers Daphragm Beam-splttng cbe Algnment laser Fgre 9: Lamp Photometer algnment. The prncpal reflecton from the cbe determned the photometer poston, whle the secondary reflecton from the sde of the cbe defned the lamp poston. The cbe was algned once sch that the beam ht the flament of the lamp, the latter havng been monted n the standard way to defne the optcal axs (paragraph 6..). The photometers were centered wth respect to the beam and the frst back reflecton from the photometer head was sed to algn the heads perpendclar to the beam. The daphragm closest to the photometers was the one n the shtter, whch was opened to abot mm dameter and placed approxmately 0 cm from the photometers. A seres of baffles was sed to mnmze stray lght. The choce of the dameter of the last daphragm and ts dstance was made on the bass of the reslts of the preparatory measrements (paragraph 5.3). 3

14 6. Comparson measrements Three photometers cold be monted at the same tme on the translaton stage. Two of the three photometers n each rn were from the BIPM reference grop, the thrd one beng one of the photometers from a partcpant. At least half an hor was allowed for thermal stablzaton after the photometer had been connected to the power spply. The photometers were optcally algned sng the technqe descrbed above. A measrement cycle conssted of fve back and forth dsplacements of the translaton stage, placng the dfferent photometers n trn on the axs of the bench. Measrements were taken at each poston, back and forth. Ths cycle was repeated fve tmes before another photometer was monted on the table. Each partcpant's photometer was measred at least fve tmes on dfferent days. Data were processed off-lne. 7. Data analyss As there were always two of the BIPM photometers on the bench together wth one photometer from a partcpant, the rato between the BIPM photometers can be sed to check the stablty of the reference grop comprsng the three BIPM photometers. Spplementary checks were done reglarly by montng all three BIPM photometers on the comparson bench and comparng them. Fgres 0a and 0b show the ratos of the photometers of the reference grop over for months. BIPM/BIPM Relatve change to frst measrement Nmber of measrement Fgre 0a: Stablty of the reference photometers, normalzed rato BIPM / BIPM 4

15 BIPM/BIPM3 Relatve change to frst measrement Nmber of measrement Fgre 0b: Stablty of the reference photometers, normalzed rato BIPM / BIPM3 Althogh some strctre s present, the relatve standard ncertantes of the ratos montored are relatvely small. Ths relatvely small spread for fltered detectors s smaller than the ncertantes of the scales to be compared. The fact that the ratos of the responsvtes of photometers of dfferent type are stable ndcates that the ndvdal photometers are stable, as t can be assmed that dfferent types wold not exhbt smlar flctatons n ther responsvty. The contrbton to the relatve standard ncertanty of the comparson was estmated as 0.08 % (see Table 5). To derve the dfferences between the partcpants lmnos responsvty nts, all LMT and IPR photometers were normalzed wth respect to BIPM- (PRC type). The PRC photometers were normalzed wth respect to BIPM- (LMT type) and the BIPM- / BIPM- rato was sed to calclate the reslts of the comparson. Ths procedre was adopted owng to the mch larger sze of the PRC photometer heads and becase t avoded contnal remontng of the mechancal parts on the photometrc bench each tme a measrement was made. The nflence of the amplfer gans was stded by nvertng the amplfer connectons. It was fond that the effect was wthn the repeatablty of the reslts (< parts n 0 4 ), so that t was not taken nto accont n the calclaton. To compare the photometer calbratons the followng calclaton was made: s ref, =s y ref / y where s ref, represents the calbraton of the BIPM reference photometer by the -th photometer of the comparson batch, s s the responsvty for that photometer as gven by the partcpant and y ref, y are the sgnals measred from the reference photometer and the partcpant's photometer, respectvely. In that way the BIPM reference photometer was calbrated wth respect to each ndvdal partcpatng photometer. 5

16 8. Reslts The CCPR workng grop on key comparsons decded how the key comparson reference vale (KCRV) shold be calclated and ts decson was approved by the 5 th CCPR. The key comparson reference vale shold be calclated as the weghted mean wth the nverse sqare of the ncertantes as weght. It was consdered that some of the ncertantes stated were possbly too small and that therefore a 'ct-off' vale for the ncertantes shold be sed, whch was fxed to be 0. %. The ct-off was only to be sed n the calclaton of the reference vale bt the orgnally gven ncertantes by the laboratores shold be sed n the tables and graphs. Laboratores concerned by ths rle are marked wth an * n Table 4 and n Fgre. The KRISS declared that they had dentfed a problem wth ther reference photometer and the workng grop decded to exclde the KRISS from the calclaton of the reference vale. For each laboratory the reslts obtaned for the dfferent photometers were frst averaged, then wth the excepton of the KRISS reslt, the weghted mean vale of these averages for all laboratores was calclated sng the nverse sqare of the ncertantes as a weght after the applcaton of a ct-off of 0. % as the mnmm ncertanty. Both the 'before' and 'after' vales were sed for the calclaton of ths key comparson reference vale, except for the BNM/INM vale becase of the falre of the second photometer drng the measrements at the BIPM. The percentage dfference of the mean of each laboratory from the key comparson reference vale s shown n Fgre and n Table 4. Also ndcated n the fgre s the ncertanty of the reference vale of 0.06 % whch s calclated as the standard ncertanty 0 of the weghted mean sng the formla 0 = where the are the ncertantes stated by the partcpants after applcaton of the ct-off. In Fgre the percentage dfference for each ndvdal photometer before and after the BIPM measrements s plotted aganst the same reference vale as n Fgre. As some laboratores sent three photometers, ther reslts wold have entered wth a greater weght than those wth only two photometers f the weghted mean of the comparson had been taken from the ndvdal photometer data as opposed to the mean per laboratory. For clarty n Fgre the ncertantes are ndcated only for the 'before' vales. There have been several correctons of data drng the draft stage. The correctons of clercal errors and of msnterpretatons of some of the reported laboratory reslts by the plot after Draft A are lsted and explaned n Appendx. There were reqests from CSIC/IFA and OFMET to correct ther data after Draft A de to techncal reasons. These revsed data were not accepted for the Fnal Report bt are shown n Appendx for nformaton. Only mnor correctons as below reqested by HUT and NPL were accepted. The HUT has observed a drft n one of ts photometers over a perod of more than one year. The data measred at the BIPM agree wth ths observed drft. Conseqently the HUT asked to correct the 'before' and 'after' vales for ths photometer by 0.7 % and +0. %, respectvely, resltng n a change of the HUT vale by only 0.0 %. 6

17 NPL has corrected an error t made n calclatng the stated ncertanty of the mean vale reported n Draft A. The standard ncertanty vale was corrected from 0.5 % to 0.8 % n Draft B and the Fnal Report. partcpant percentage dfference ncertanty (k=) from key comparson reference vale BNM-INM CSIC CSIRO * HUT IRL (KRISS) NIM * NIST NPL * NRC OFMET OMH PTB * SMU VNIIOFI BIPM BIPM (L85) Table 4: Percentage dfference from the key comparson reference vale (weghted mean wth ct-off of 0. %) for the dfferent partcpants. The ncertantes are those gven by the laboratores. The ncertantes of the laboratores marked wth an * are below the ct-off of 0. % and were replaced by 0. % n calclatng ther weghts. The KRISS reslt was not sed n the calclaton of the weghted mean snce ts reslt was consdered to be an otler. The reslt BIPM (L85) was obtaned by calbratng the BIPM photometers aganst the lmnos ntensty lamps mantanng the mean vale of the 985 comparson of lm. ntensty. It can be seen from Fgre that the overall spread s probably somewhat hgher than cold have been hoped for, bt that most laboratores do agree well wth the reference vale. Twelve laboratores agree wth the reference vale wthn σ. The reference vale agrees well wth the mean vale from the 985 nternatonal comparson of lmnos ntensty sng lamps as t s mantaned n the form of a grop of lamps at the BIPM (see entry BIPM (L85) n Table 4). The standard devaton of the partcpants reslts (wthot KRISS) s 0.47 % and ths smlar to the standard devaton observed n the comparson CCPR-K3.a made wth lamps. 7

18 3.0.5 KCRV = weghted mean wth ct-off 0. % dfference from KCRV / percent standard ncertanty of the weghted mean (0.06 %) BNM-INM CSIC *CSIRO HUT IRL (KRISS) *NIM NIST *NPL NRC OFMET OMH *PTB SMU VNIIOFI BIPM partcpants Fgre : Reslts of the comparson averaged over all photometers per laboratory. The graph shows the relatve dfference from the weghted mean n percent for each laboratory. The ncertantes correspond to the standard ncertantes (k=) stated by the laboratores and do not nclde the ncertanty of the comparson tself. 8

19 BNM / INM BNM / INM CSIC CSIC dfference from the KCRV / percent CSIRO CSIRO CSIRO3 HUT HUT IRL IRL KRISS KRISS KRISS3 NIM NIM NIST NIST NPL NPL NRC NRC OFMET OFMET OFMET3 OMH OMH PTB PTB SMU SMU VNIIOFI VNIIOFI BIPM4 BIPM5 BIPM7 photometer Fgre : Reslts of the comparson for each ndvdal photometer. The graph shows the relatve dfference from the same reference vale as n Fgre n percent for each photometer before (damonds) and after (trangles) the measrements at the BIPM. The ncertantes correspond to the standard ncertantes (k=) stated by the laboratores and do not nclde the ncertanty of the comparson tself. 9

20 9. Uncertantes of the comparson UNCERTAINTY CONTRIBUTION RELATIVE STANDARD UNCERTAINTY 00 Measrement repeatablty (A type) 0.0 Stablty of reference grop 0.08 Lnearty 0.0 Algnment (pvot) 0.03 Algnment (dstance) 0.0 Total relatve standard ncertanty 0.09 Table 5: Uncertanty bdget for the comparson of lmnos responsvty. The largest contrbton to the ncertanty of the comparson s de to the stablty of the reference grop. The spectral msmatch of the reference and the dstrbton temperatre contrbte only a neglgble factor to the ncertantes. The total relatve standard ncertanty of the comparson s 0.09 %, whch s consderably smaller than the calbraton ncertantes of the photometers. 0. Realzaton of the photometrc qanttes n the dfferent laboratores The followng short descrptons abot the realzaton of the photometrc nt and the method of calbraton are taken from the laboratory reports of the calbraton. BIPM: Measrement of the spectral responsvty aganst a trap detector calbrated at a cryogenc radometer and separate determnaton of the apertre area. An addtonal calbraton was made sng the grop of lmnos ntensty lamps mantanng the mean vale of the 985 comparson of lmnos ntensty. The reslt of ths calbraton s shown n the last lne of table 4 as BIPM (L85). BNM/INM: Comparson wth a grop of standard lamps at 800 K. The vale for these lamps was known from the 983 radometrc realzaton of the candela sng ESR's. The BNM-INM has nformed s that they are crrently completng a new realzaton of the nt candela. Prelmnary reslts of that stdy seem to ndcate that the dfference between the old and the new realzaton s of the order of 0.8 %. CSIC / IFA: Comparson wth three standard photometers. The standard photometers were measred n relatve spectral responsvty and then calbrated at Ar laser wavelengths aganst an S detector, whose responsvty s traceable to a cryogenc radometer. CSIRO: Comparson wth a room-temperatre, V(λ)-corrected ESR. 0

21 HUT: Comparson of the photometer wth a reference photometer consstng of a characterzed trap detector ftted wth a V(λ) flter and a precson apertre. The trap responsvty s traceable to the HUT cryogenc radometer. The photometers sed were compared drectly wth the reference photometer. IRL: Comparson wth standard photometers. Ther responsvty was determned sng a trapdetector spectral responsvty scale whch had been derved from an nternal qantm effcency model and slcon photodode reflectance measrements; cryogenc radometer measrements were sed for renormalzaton. KRISS: No nformaton was commncated. NIM: Comparson wth a grop of standard lamps. No nformaton was commncated on how the lamp vales were determned. NIST: A grop of characterzed photometers traceable to the cryogenc radometers form a reference grop. The photometers were compared drectly wth ths grop. NPL: The measrement of responsvty of photometers at NPL s carred ot by comparson wth the NPL scale of lmnos ntensty. The dervaton of the NPL scale of lmnos ntensty, s based on a radometrc realzaton of the candela sng the NPL absolte cryogenc radometer. The lamps sed to dssemnate the scale are calbrated aganst reference photometers whch are calbrated aganst the NPL cryogenc radometer. The realzaton of the candela at NPL s descrbed n the paper: Metrologa 5, 9-40 (988). NRC: Comparson wth a grop of reference lamps. The lmnos ntensty vales of the lamps are obtaned by comparson wth the NRC 985 realzaton of the candela sng V(λ)-corrected ESR's and slcon cells. OFMET: Comparson wth a grop of lamps. The lmnos ntensty vales of the lamps are known from the 985 CCPR comparson addng the recommended correcton of %. These vales were confrmed by provsonal reslts of a radometrc realzaton. After the OFMET had receved draft A and after knowng the provsonal reslts from the lmnos ntensty comparson they commncated revsed vales based on the latter comparson. OMH: The lmnos ntensty nt at OMH s realzed and mantaned by a grop of three V(λ)- corrected standard photometers. The absolte spectral responsvty of the photometers s based on the predctable qantm effcency method of the slcon photodode and some characterzed trap detectors.

22 PTB: At the PTB the lmnos ntensty nt s realzed annally, based on the cryogenc radometers n the clean-room center. V(λ)-corrected photometers are sed as transfer standards between the radometers and the network of photometers and lamps for the realzaton and mantenance of the candela. The stablty of the network s hgher than the ncertanty of a realzaton, so t s sed to average over several realzatons. The PTB photometrc nts have remaned nchanged snce the CCPR comparson n 985. SMU: Calbraton of a spectrally characterzed photometer. The absolte vales were calbrated aganst a QED-00. VNIIOFI: Comparson wth a standard lamp. No nformaton was commncated on how the lamp vale was determned.. Conclsons The frst nternatonal comparson of a photometrc qantty sng only photometers was carred ot. The photometers have generally proven to be sffcently stable and robst for sch a comparson. The agreement of the reslts of the dfferent laboratores s farly good, wth most partcpants agreeng wthn the stated ncertantes. The spread of the reslts, expressed as standard devaton of the reslts, s comparable to the spread observed n a smlar comparson made wth lamps as transfer standards.. Acknowledgements The athors wold lke to thank P. Zabarel who dd a maor contrbton to the settng p and ntal testng of the facltes sed for ths comparson. 3. References [] Comté Consltatf de Photométre et Radométre, 0 th sesson, 98, p75. [] G. Sater, D. Lndner, M. Lndemann, CCPR Key Comparsons K3a of Lmnos Intensty and K4 of Lmnos Flx wth Lamps as Transfer Standards, PTB-Report, PTB-Opt-6, Branschweg, December 999, ISBN [3] R. Köhler, R. Goebel, R. Pello, "Reslts of the nternatonal comparson of spectral responsvty of slcon photodetectors", Metrologa, 3, 996, [4] R. Köhler, R. Goebel, R. Pello, "Report on the nternatonal comparson of spectral responsvty of slcon photodetectors", Report BIPM-94/9. [5] CIE pblcaton 53 / 98.

23 Appendx Correctons of clercal errors n Draft A and Draft B. BNM/INM Followng ther "after" measrement, BNM/INM revsed ther "before" vales and reported the revsed "before" and "after" vales. The laboratory also reqested to se only "before" data for one of the photometers de to a sgnfcant change detected. Draft A, however, dd not make these correctons by mstake of the plot laboratory, and sed only the orgnal "before" vales. Ths error was corrected n Draft B and the Fnal Report.. NPL Followng ts "after" measrement, NPL revsed ther "before" vale and reported the mean of the "after" and the revsed "before" vales to BIPM. The change of the "before" measrement was de to a drft of the workng standard lamps and was on average 0.6 %. The mean vales were erroneosly nterpreted by the plot as reslts of the "after" measrement n Draft A. Ths error was corrected n Draft B and the Fnal Report. 3. NRC The reslts gven n ths Fnal Report for the two NRC photometers are dfferent from those shown n the prevos drafts. Ths s de to a correcton of a msnderstandng of the commncated data by the plot. The error was already dscovered after Draft A bt forgotten to be taken nto accont n Draft B. The effect of ths change on NRC's data s 0.08 %. 3

24 Appendx Revsed reslts sbmtted after Draft A. CSIC/IFA The CSIC/IFA observed a too large nfra-red response of ther photometers. Ths fact was stated n the orgnal calbraton report. The CSIC later decded to apply a correcton of 0.4 % to ther calbraton vales to sbtract the nfra-red contrbton from the lmnos responsvty. In the Draft B of ths report ths correcton had been appled. In the fnal data redcton we removed ths correcton becase the IR response contrbtes to the measred photo-crrent and ths mst be nclded n the lmnos responsvty vale assgned to the photometer.. METAS/OFMET The reslts of METAS/OFMET shown n ths report are based on a grop of lmnos ntensty lamps whch mantaned the nternatonal mean vale of the 985 lmnos ntensty comparson. A the tme of the lmnos responsvty comparson CCPR-K3.b METAS was workng on a new detector-based realzaton. Ths realzaton was not yet avalable at the begnnng of the comparson. However, after the crclaton of the Draft A report n Janary 998 the new realzaton was completed and sed to calclate mproved lmnos responsvty vales for the three photometers. The ncertanty remaned nchanged (0.5 %). On average, the new vales were 0.5 % lower than the orgnal ones. These modfed data were orgnally accepted by the other partcpants and sed n the sbseqent draft reports. The Workng Grop on Key Comparsons of the CCPR dscssed ths sbect agan n ts meetng on 5 Jne 003 and decded that ths data modfcaton - havng occrred after the dstrbton of the Draft A - was aganst the prncples of the Gdelnes for CIPM key comparsons. The offcal reslt gven n ths (second) Fnal Report s therefore based on METAS' orgnal reslts. For nformaton ths appendx shows the reslts based on the new detector-based lmnos responsvty realzaton of METAS. The followng table shows the lmnos responsvtes of the three photometers based on the old and the new realzaton. old realzaton new realzaton photometer "before" "after" "before" "after" OFMET na/lx OFMET na/lx OFMET na/lx Table A.: Sbmtted change of the METAS/OFMET reslts. 4

25 The reslts of all partcpants, calclated sng the new METAS reslts are shown n the table below. partcpant percentage dfference ncertanty (k=) from key comparson reference vale BNM-INM CSIC CSIRO * HUT IRL (KRISS) NIM * NIST NPL * NRC OFMET OMH PTB * SMU VNIIOFI BIPM BIPM (L85) Table A.: Recalclaton of the K3.b reslts sng the revsed METAS/OFMET vales. For frther detals see the capton of Table 4 of the man report. Compared to Table 4 of the Fnal Report, all dfferences from the KCRV (expressed n percent) are ncreased by 0.03 %, beng the dfference between the KCRVs calclated wth the orgnal and the new METAS vales. The METAS reslt mproves from.0 % to 0.53 %. 5

26 Appendx 3 Degrees of Eqvalence for CCPR-K3.b M. Stock, BIPM In the man text of ths Fnal Report the reslts are presented n the followng form (Table 4): partcpant 00 x relatve devaton from key comparson reference vale 00 x relatve ncertanty (k=) BNM-INM CSIC CSIRO * HUT IRL (KRISS) NIM * NIST NPL * NRC OFMET OMH PTB * SMU VNIIOFI BIPM Table A3.: Relatve dfference from the key comparson reference vale (weghted mean wth ct-off for the relatve ncertanty of x0-3 ) for all partcpants. The ncertantes are the calbraton ncertantes gven by the laboratores. The relatve ncertantes of the laboratores marked wth an * are below the ct-off and were replaced by x0-3 n calclatng ther weghts. The KRISS reslt was not sed n the calclaton of the weghted mean snce ts reslt was consdered to be an otler. The reference vale s calclated as the weghted mean sng a ct-off of x0-3. The ncertantes gven n Table A3. are those clamed by the partcpants for ther calbraton. To state the degrees of eqvalence as defned n the MRA, the ncertanty of the devaton from the reference vale s needed, whch ncldes the calbraton ncertanty, the ncertanty of the comparson at the BIPM, the ncertanty of the reference vale and the correlaton between the partcpants ncertantes and that of the reference vale. De to the applcaton of an ncertanty ct-off n the calclaton of the weghted mean x w, the ncertanty of the devaton d cannot be calclated smply by sng the formla whch s vald for the ncertanty of the devaton from the (normal) weghted mean: (d ) = - (x w ). As shown n Appendx 4, the more complcated formlas ( d ) = + ( xw) and ( xw) = 4 6

27 have to be sed, where the are adsted ncertantes defned by = max (, mn ) wth mn beng the ct-off vale. The ncertanty of the reference vale ths calclated s (x w ) = 6.0 x 0-4. To nclde the relatve ncertanty of the comparson at the BIPM of 9x0-4 (Table 5), t wll be added qadratcally to the ncertanty of the devaton defned above. It wold probably have been better to nclde ths contrbton before the calclaton of the weghts needed for the determnaton of the reference vale, bt the choce was made earler drng the preparaton of the Fnal Report not to do so. The ncertantes obtaned by ths calclaton are shown n Table A3.. partcpant 00 x relatve devaton from key comparson reference vale 00 x relatve ncertanty (k=) of devaton 00 x relatve ncertanty (k=) of devaton BNM-INM CSIC CSIRO * HUT IRL (KRISS) NIM * NIST NPL * NRC OFMET OMH PTB * SMU VNIIOFI BIPM Table A3.: Degrees of eqvalence for CCPR-K3.b. The ncertanty ncldes the calbraton ncertanty of the partcpants, the ncertanty of the comparson at the BIPM and that of the reference vale, takng nto accont ts correlaton wth the partcpants ncertantes (except for KRISS, whose reslt dd not contrbte to the reference vale). The blateral dfferences are calclated from the devatons of the two laboratores from the reference vale: d, = d d = ( x xw ) ( x xw ) = x x They do not depend on the reference vale. The ncertantes of the blateral dfferences are calclated as the qadratc sm of the calbraton ncertantes pls two tmes the ncertanty of the comparson at the BIPM. Snce by far the largest contrbton to the latter s the stablty of the reference grop, these contrbtons are assmed to be ncorrelated. The ncertanty of the reference vale does not contrbte to the ncertanty of the blateral dfferences. The matrx of blateral dfferences and the assocated ncertantes s shown on the next page. 7

28 Table A3.3: Degrees of eqvalence (ble) and blateral degrees of eqvalence (yellow) 8

29 Appendx 4 Formlas for Degrees of Eqvalence for Weghted Mean wth Ct-Off M. Stock, BIPM Gven are the reslts x and ther ncertantes for N partcpants : (x, ) for =,.., N. If the key comparson reference vale (KCRV) s calclated as the weghted mean, the laboratory wth the smallest ncertanty wll have the largest weght. If the smallest ncertantes are consdered as mplasble or to lmt the nflence of a sngle laboratory, the KCRV can be calclated as the weghted mean x w wth a ct-off mn (mnmm) appled to the ncertantes n the calclaton of the weghts: xw = w x = x wth w and = max (, mn ). In the case that the ncertanty s smaller than the ct-off, t s replaced by the ct-off mn n the calclaton of the weght of laboratory. The adsted ncertantes are, however, not consdered as the real ncertanty of laboratory n the followng ncertanty calclaton. The ncertanty (x ) s stll. Ths n some sense s a choce, bt t s compatble wth the decson of the CCPR to apply the ct-off only n the calclaton of the reference vale for CCPR-K3.b bt to se the s n the tables and graphs showng the fnal reslts. Under the assmpton that all x are ncorrelated, the ncertanty of the weghted mean s gven by ( x w ) = 4 If all v =, that s f no ct-off s appled, the reslt becomes dentcal to the well known reslt ( xw) =., 9

30 30 The devaton d from the reference vale s defned by = = = k k k k k k w x x x x x x d Agan, f all the x are ncorrelated, the ncertanty of d s + = + = + = k k k w w w k k k x x x d 4 4 ) ( ) ( ) ( ) ( The last term descrbes the effect of the adsted ncertantes. If no ct-off s appled (all v = ), the term n the sqare brackets becomes zero and we obtan ) ( ) ( w x d = whch s the formla already known for the weghted mean wthot ct-off. Smmary: Uncertanty of reference vale: 4 ) ( = w x Uncertanty of devaton: ) ( ) ( w x d + =