Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
m a e t smetrology and accreditation switzerland
EUROMET Project 666 EUROMET.PR-S1
Final Report
Inter-comparison of Chromatic Dispersion Reference Fibres
Bern-Wabern, March 2005 Jacques Morel Swiss Federal Office of Metrology and Accreditation (metas) Lindenweg 50 3003 Bern-Wabern Switzerland [email protected] Phone: +41 31 32 33 350
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 2 of 40
m a e t smetrology and accreditation switzerland
Table of Contents 1 Introduction ...........................................................................................................................3 2 List of participants .................................................................................................................3 3 Technical part........................................................................................................................3
3.1 Measured quantities......................................................................................................3 3.2 Measurement methods and data processing.................................................................4 3.3 Reporting of the calibration results ................................................................................4 3.4 Uncertainty budget........................................................................................................5 3.5 Chromatic dispersion results .........................................................................................5
3.5.1 Reference fibre 1 (G652)...........................................................................................6 3.5.2 Reference fibre 2 (G653)...........................................................................................7 3.5.3 Reference fibre 3 (G655 TeraLight)...........................................................................8 3.5.4 Reference fibre 4 (G655 Leaf)...................................................................................9
3.6 Analysis of the deviation of the chromatic dispersion results .......................................10 3.7 Validity of the analysis.................................................................................................11
4 Deviation of the chromatic dispersion results.......................................................................11 4.1 Reference fibre 1 (G652) ............................................................................................12
4.1.1 CSIC .......................................................................................................................12 4.1.2 METAS ...................................................................................................................13 4.1.3 NIST........................................................................................................................14 4.1.4 NPL.........................................................................................................................15 4.1.5 HUT ........................................................................................................................16
4.2 Reference fibre 2 (G653) ............................................................................................17 4.2.1 CSIC .......................................................................................................................17 4.2.2 METAS ...................................................................................................................18 4.2.3 NIST........................................................................................................................19 4.2.4 NPL.........................................................................................................................20 4.2.5 HUT ........................................................................................................................21
4.3 Reference fibre 3 (G655 TeraLight).............................................................................22 4.3.1 CSIC .......................................................................................................................22 4.3.2 METAS ...................................................................................................................23 4.3.3 NIST........................................................................................................................24 4.3.4 NPL.........................................................................................................................25 4.3.5 HUT ........................................................................................................................26
4.4 Reference fibre 4 (G655 Leaf).....................................................................................27 4.4.1 CSIC .......................................................................................................................27 4.4.2 METAS ...................................................................................................................28 4.4.3 NIST........................................................................................................................29 4.4.4 NPL.........................................................................................................................30 4.4.5 HUT ........................................................................................................................31
5 Zero dispersion wavelength.................................................................................................32 5.1 Zero dispersion wavelength results .............................................................................33 5.2 Deviation from mean zero dispersion wavelength .......................................................33
6 Dispersion slope..................................................................................................................34 6.1 Dispersion slope results ..............................................................................................35 6.2 Deviation from mean dispersion slope.........................................................................36
7 Conclusions.........................................................................................................................37 8 References..........................................................................................................................38 9 Annex A. Discussion of HUT results ....................................................................................39
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 3 of 40
m a e t smetrology and accreditation switzerland
1 Introduction The calibration of the chromatic dispersion properties of singlemode fibres is of critical importance for a proper optimisation of optical fibre communication systems. Although extensive analyses of the commonly used calibration techniques were already performed by some NMIs [6], [9], no large scale inter-laboratory comparisons that could be used as a basis for the MRA existed at this time. The aim of this project was to perform a comparison of chromatic dispersion measurements that were carried out on four of the most commonly used fibre types, namely G652, G653, G655-TeraLight and G655-Leaf. This inter-comparison has also been registered as a supplementary comparison (Euromet-PR-S1) that will serve as a basis for the review of the CMC entries on chromatic dispersion. 2 List of participants Laboratory Contact person email Instituto de Fisica Aplicada, CSIC, Spain Pedro Corredera [email protected]
Helsinki University of technology, HUT, Finland
Hanne Ludvigsen [email protected]
National Physical Laboratory, NPL, United Kingdom
Martin Wicks [email protected]
National Institute of Standards, NIST, United States
Tasshi Dennis [email protected]
Swiss Federal Office of Metrology and Accreditation, METAS, Switzerland, pilot laboratory.
Jacques Morel [email protected]
NIST was invited to participate to this project in a common agreement between all participants. METAS managed the inter-comparison and provided four reference fibres that were circulated among the participating laboratories, according to the rules and time schedules as defined in the “Technical document for the EUROMET Project 666” [5]. A detailed description of the artefacts and of their properties is also presented in the same document. 3 Technical part 3.1 Measured quantities Each laboratory was asked to calibrate the three main quantities that are commonly used to represent the chromatic dispersion properties of a fibre, as specified in Table 1. Quantity Symbol Units
Overall chromatic dispersion D ps/nm Zero dispersion wavelength λ0 nm Dispersion slope at λ0 S0 ps/nm2
Table 1. List of the calibrated quantities. No normalisation of the calibrated quantities to the fibre length was considered for this inter-comparison.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 4 of 40
m a e t smetrology and accreditation switzerland
3.2 Measurement methods and data processing Each laboratory was allowed to use one or several of the standard measurement techniques, namely 1. Phase shift 2. Differential phase shift 3. Spectral group delay in the time domain 4. Non linear (4 wave mixing) 5. Interferometric.
The measurement technique(s) used by each participating laboratory are summarised in Table 2.
Participant Method CSIC Phase shift and
four wave mixing, for a supplementary determination of λ o HUT Phase shift NPL Phase shift NIST Phase shift (broadband data and narrowband analysis) METAS Phase shift
Table 2. Measurement techniques used by the participating laboratory. Most of the above mentioned calibration techniques involve a curve fitting (least squares) of the differential group delay data. For these cases, one of the polynomial functions as given in Table 3 was recommended.
Fibre type
Wavelength domain Model
Equation
G652 1310 nm (around λ o) Sellmeier 3 terms τ(λ) = aλ2 + b λ−2 + c Wider range Sellmeier 5 terms τ(λ) = aλ4 + bλ2 + cλ−2 + d λ−4 + eG653 Around λ o= 1550 nm Parabolic τ(λ) = aλ2+ bλ+ c Wider range Sellmeier 5 terms τ(λ) = aλ4 + bλ2 + cλ−2 + d λ−4 + eG655 Sellmeier 5 terms τ(λ) = aλ4 + bλ2 + cλ−2 + d λ−4 + e
Table 3. List of the standard fitting functions.
Other curve fitting models were allowed, when proved that they would significantly improve the quality of the fit. 3.3 Reporting of the calibration results The calibration of the chromatic dispersion was performed by each laboratory within the largest wavelength range as possible. Depending on the properties of the measurement system and on the applied data processing technique, the calibrations were performed in one or in several disjoined spectral segments. The calibration was performed, whenever possible, within both the 1310 nm and 1550 nm spectral domains. The zero dispersion wavelength λ o and the dispersion slope So around λ o were only reported, when obtained from a measurement scan that included the zero dispersion wavelength itself; i.e. that λ o wasn’t obtained from an extrapolation of the measured dispersion data. The chromatic dispersion D was reported for even integer wavelength values only. The spectral domain that was covered by each laboratory and the applied fitting functions are summarized for each reference fibre in Table 4.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 5 of 40
m a e t smetrology and accreditation switzerland
Parti-cipant
Ref. 1 (G652) Ref. 2 (G653)
Ref. 3 (G655), Teralight
Ref. 4 (G655), Leaf
CSIC 1260 nm – 1640 nm Sellmeier 5 terms for D. Sellmeier 3 terms between 1294 nm and 1340 nm for λo and So)
1260 nm – 1640 nm Sellmeier 5 terms for D. Parabolic between 1530 nm and 1578 nm for λo and So)
1260 nm – 1640 nm Sellmeier 5 terms for D, λo and So.
1260 nm – 1640 nm Sellmeier 5 terms for D, λo and So.
HUT 1500 nm – 1600 nm Sellmeier 5 terms for D.
1490 nm – 1600 nm Sellmeier 5 terms for D; Sellmeier 3 terms for λo and So.
1480 nm – 1600 nm Sellmeier 5 terms for D.
1482 nm – 1600 nm Sellmeier 5 terms for D; Sellmeier 3 terms for λo and So.
NPL 1270 nm – 1340 nm 1490 nm – 1620 nm Sellmeier 5 terms for D, λo and So.
1270 nm – 1340 nm 1490 nm – 1620 nm Sellmeier 5 terms for D, λo and So.
1270 nm – 1340 nm 1410 nm – 1620 nm Sellmeier 5 terms for D, λo and So.
1270 nm – 1340 nm 1410 nm – 1620 nm Sellmeier 5 terms for D, λo and So.
NIST 1284 nm – 1338 nm 1482 nm – 1620 nm D from repeated fitting of a second order polyn. to 12 nm subsets of data. Sellmeier 3 terms for , λo and So over 30 nm subsets
1288 nm – 1336 nm 1482 nm – 1620 nm D from repeated fitting of a second order polyn. to 12 nm subsets of data. Second order polynomial for , λo and So over 30 nm subsets
1288 nm – 1338 nm 1442 nm – 1620 nm D from repeated fitting of a second order polynomial to 12 nm subsets of data. Sellmeier 5 terms for , λo and So over 30 nm subsets
1290 nm – 1338 nm 1482 nm – 1620 nm D from repeated fitting of a second order polynomial to 12 nm subsets of data. Sellmeier 5 terms for , λo and So over 30 nm subsets
METAS 1254 nm – 1366 nm Sellmeier 3 terms 1436 nm – 1638 nm Sellmeier 5 terms
1254 nm – 1368 nm 1436 nm – 1640 nm Sellmeier 5 terms
1254 nm – 1368 nm 1436 nm – 1640 nm Sellmeier 5 terms for D Parabolic between 1435 nm and 1590 nm for λo and So.
1254 nm – 1366 nm 1436 nm – 1640 nm Sellmeier 5 terms
Table 4. Spectral domains covered by each laboratory and curve fitting functions used for the data processing. 3.4 Uncertainty budget Relevant parameters for the calculation of the uncertainty budget strongly depend on the measurement technique and on the applied data processing (curve fitting) methods. Some of the most relevant influence factors to the uncertainty budget of D, So and λo are given in Table 5. Quantity Description uτ Uncertainty in the determination of the differential group delay due to the
measurement system uT Uncertainty due to thermal drifts ufit Uncertainty due to the curve fitting UPMD2 Uncertainty due to the 2nd order PMD uλ Uncertainty in the determination of the wavelength associated to each
measurement point
Table 5. Most relevant parameters for the calculation of the uncertainty budget. Each laboratory developed very different methods for the calculation of the uncertainty budgets, which makes a detailed comparison of the different contributing quantities almost impossible. Nevertheless, the uncertainty of each quantity was reported as the combined standard uncertainty multiplied by a coverage factor k = 2, estimated according to the ISO guide [4]. The reported measurement uncertainty contained contributions originating from the measurement standards, from the calibration method, from the environmental conditions and from the artefacts being calibrated.
3.5 Chromatic dispersion results The calibration results of all laboratories are shown for each reference fibre in the upper graphs of Figs (1), (2), (3) and (4). A visual analysis proved a good agreement between the results of all
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 6 of 40
m a e t smetrology and accreditation switzerland
laboratories but HUT, where larger discrepancies were observed. The expanded combined uncertainties claimed by the participants are shown in the lower graphs of the same figures. The uncertainty of the unweighted mean
sdevsdev DD uU ⋅= 2 was calculated from the standard deviation of the reported values by using Eq. (7) and is shown on the same graphs. A detailed explanation of this analysis is given in Sect. 3.6. The unweighted mean of the dispersion values was calculated, because of the large discrepancy of the results reported by HUT, by considering the results from CSIC, METAS, NIST and NPL only. This first analysis showed a large spread of the uncertainties reported by the participants. The calibration results of reference fibre 1 (G652) give a good example of this fact. A ratio larger than 180 was found between the uncertainties reported by NPL and NIST at 1550 nm. 3.5.1 Reference fibre 1 (G652)
400
300
200
100
0
-100
D (p
s/nm
)
1600155015001450140013501300Wavelength (nm)
CSICHUTMETASNISTNPL
Chromatic DispersionRef. 1. G652 Fibre
7
6
5
4
3
2
1
0
U_D
(ps/
nm)
1600155015001450140013501300Wavelength (nm)
CSIC HUT METAS NIST NPL UDsdev
Chromatic Dispersion Uncertainty (ps/nm)Ref. 1 G652 Fibre
Fig. 1. (Upper): Chromatic Dispersion of reference fibre 1 measured by all participants. A very
good agreement was observed between all results. HUT values showed a larger deviation around 1500 nm and 1600 nm. (Lower): Dispersion uncertainties (k=2) claimed by each laboratory.
sdevDU is the uncertainty of the unweighted mean value, which was calculated from the dispersion results by using Eq. (7).
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 7 of 40
m a e t smetrology and accreditation switzerland
3.5.2 Reference fibre 2 (G653)
-300
-250
-200
-150
-100
-50
0
50
D (p
s/nm
)
1600155015001450140013501300Wavelength (nm)
Chromatic DispersionRef. 2. G653 Fibre
CSICHUTMETASNISTNPL
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Unc
erta
inty
(ps/
nm)
1600155015001450140013501300Wavelength (nm)
CSIC HUT METAS NIST NPL UDsdev
Chromatic Dispersion Uncertainty (ps/nm)Ref. 2. G653 Fibre
Fig. 2. (Upper): Chromatic Dispersion of reference fibre 2 measured by all participants. A very
good agreement was observed between all results. HUT values showed a larger deviation around 1500 nm and 1600 nm. (Lower): Dispersion uncertainties (k=2) claimed by each laboratory.
sdevDU is the uncertainty of the unweighted mean value, which was calculated from the dispersion results by using Eq. (7).
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 8 of 40
m a e t smetrology and accreditation switzerland
3.5.3 Reference fibre 3 (G655 TeraLight)
-140-120-100-80-60-40-20
020406080
100120
D (p
s/nm
)
1600155015001450140013501300Wavelength (nm)
Chromatic DispersionRef. 3. G655 TeraLight Fibre
CSICHUTMETASNISTNPL
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Unc
erta
inty
(ps/
nm)
1600155015001450140013501300Wavelength (nm)
CSIC HUT METAS NIST NPL UDsdev
Chromatic Dispersion Uncertainty (ps/nm)Ref. 3. G655 TeraLight Fibre
Fig. 3. (Upper): Chromatic Dispersion of reference fibre 3 measured by all participants. A very
good agreement was observed between all results. HUT values showed a larger deviation around 1500 nm and 1600 nm. (Lower): Dispersion uncertainties (k=2) claimed by each laboratory.
sdevDU is the uncertainty of the unweighted mean value, which was calculated from the dispersion results by using Eq. (7).
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 9 of 40
m a e t smetrology and accreditation switzerland
3.5.4 Reference fibre 4 (G655 Leaf)
-200
-150
-100
-50
0
50
100
D (p
s/nm
)
1600155015001450140013501300Wavelength (nm)
Chromatic DispersionRef. 4. G655 Leaf Fibre
CSICHUTMETASNISTNPL
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Unc
erta
inty
(ps/
nm)
1600155015001450140013501300Wavelength (nm)
CSIC HUT METAS NIST NPL UDsdev
Chromatic Dispersion Uncertainty (ps/nm)Ref. 4. G655 Leaf Fibre
Fig. 4. (Upper): Chromatic dispersion of reference fibre 4 measured by all participants. A very
good agreement was observed between all results. HUT values showed a larger deviation around 1500 nm and 1600 nm. (Lower): Dispersion uncertainties (k=2) claimed by each laboratory.
sdevDU is the uncertainty of the unweighted mean value, which was calculated from the dispersion results by using Eq. (7).
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 10 of 40
m a e t smetrology and accreditation switzerland
3.6 Analysis of the deviation of the chromatic dispersion results Although many different methods may be considered for the evaluation of the reported results [8], none of them will be based on sound statistics, since the ensemble (amount of participants) is simply too small. Therefore, the simplest approach has been chosen. Commonly used techniques involve the determination of the deviation between the reported results and a mean value, which can be either a weighted or an unweighted (arithmetic) mean. The arithmetic mean was chosen, because of the very large spread of the reported uncertainties. This choice prevents from giving too much weight to the results of the laboratories claiming the smallest uncertainties. As already explained in Sect. 3.5, the mean of the dispersion values was calculated, because of the large discrepancy of the results reported by HUT, by considering the results from CSIC, METAS, NIST and NPL only. The unweighted mean was calculated for each measurement point according to Eq. (1),
∑=
⋅=n
iimean D
nD
1
1. (1)
The uncertainty of the mean value (confidence level k = 1) was then calculated as follows:
∑=
⋅=n
iD iDmean
un
u1
21, where (2)
iDu was the measurement uncertainty claimed by each participant.
The deviation from the mean of each measurement point measured by laboratory i was then given by
meanii DDD −=∆ (3)
The uncertainty of the deviation of each point was then calculated for CSIC, METAS, NIST and NPL by applying Eq. (4), namely
22 21imeani DDD u
nuu ⋅⎟
⎠⎞
⎜⎝⎛ −+=∆ (4)
This equation takes into account the correlation between the mean value and the sample data [1], [2], [3], since the mean value was directly calculated from the Di data according to Eq. (1). The uncertainty of the deviation was calculated for the HUT results according to Eq. (5). In this case, no correlation exists between the mean value meanD and the sample data 2
iDu . Therefore, the uncertainty of the deviation was given by
22HUTmeanHUT DDD uuu +=∆ . (5)
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 11 of 40
m a e t smetrology and accreditation switzerland
3.7 Validity of the analysis
One drawback of this analysis is that the uncertainty of the mean valuemeanDu , and consequently
the uncertainty of the deviationiDu∆ may be overestimated, especially when one laboratory claims
much larger uncertainties than the other ones. In order to investigate this point, the uncertainty of the mean
SiDu∆ was also calculated by replacing the uncertainty of the meanmeanDu by the standard
deviation of the mean of the reported values Du in Eq. (4), namely
22 21isdeviS DDD u
nuu ⋅⎟
⎠⎞
⎜⎝⎛ −+=∆ ,with (6)
DD un
usdev
⋅=1
, where (7)
Du was the standard deviation of each measurement sample, namely
( )2
011 ∑
=
−⋅−
=n
imeaniD DD
nu , where i was the index of each participant. (8)
A direct comparison of
iDu∆ andSiDu∆ gives a good indication of the consistency of the
uncertainties reported by the participants, as shown in Section 4. 4 Deviation of the chromatic dispersion results The deviation of the chromatic dispersion from the unweighted mean meanii DDD −=∆ and the related uncertainties
ii DD uU ∆∆ ⋅= 2 and iSiS DD uU ∆∆ ⋅= 2 were calculated for each reference fibre
and for each laboratory (index i) by using Eqs. (3) to (8). This analysis was performed at each wavelength that was commonly measured by all participants. The results are shown in Figs (5) to (24). The 1300 nm and 1500 nm spectral domains were split in two separate graphs in order to give a better view of the results. The reported dispersion values and their deviations to the unweighted mean show a fairly good agreement between all participants but HUT, where the previously mentioned large deviations are clearly visible. This particular case is more extensively discussed in Annex A of this report.
Significantly dissimilar values of the uncertainties ii DD uU ∆∆ ⋅= 2 and of
iSiS DD uU ∆∆ ⋅= 2 arise
when the claimed uncertainties iDU and the uncertainty of the mean
sdevDU , as calculated by using Eqs. (7) and (8), are dissimilar. This is especially the case when one laboratory claims much larger uncertainties than all the others. In this case, Eq. (6) helps to get a more realistic information about the consistency of the calibration results.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 12 of 40
m a e t smetrology and accreditation switzerland
4.1 Reference fibre 1 (G652) 4.1.1 CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
-2
-1
0
1
2
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
Fig. 5. Deviation from the arithmetic mean of CSIC results for fibre 1.The uncertainty bars show
the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs) and
iSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 13 of 40
m a e t smetrology and accreditation switzerland
4.1.2 METAS
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013301320131013001290
Wavelength (nm)
Ref. 1 G652 Fibre / METAS
-3
-2
-1
0
1
2
3
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / METAS
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / METAS
-2
-1
0
1
2
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / METAS
Fig. 6. Deviation from the arithmetic mean of METAS results for reference fibre 1.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 14 of 40
m a e t smetrology and accreditation switzerland
4.1.3 NIST
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / NIST
-2
-1
0
1
2
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / NIST
-0.4
-0.2
0.0
0.2
0.4
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / NIST
0.6
0.4
0.2
0.0
-0.2
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / NIST
Fig. 7. Deviation from the arithmetic mean of NIST results for reference fibre 1.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). A good consistency of the reported results was
observed for both analyses. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 15 of 40
m a e t smetrology and accreditation switzerland
4.1.4 NPL
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / NPL
-4
-2
0
2
4
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / NPL
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013301320131013001290
Wavelength (nm)
Ref. 1 G652 Fibre / NPL
-4
-2
0
2
4
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / NPL
Fig. 8. Deviation from the arithmetic mean of NPL results for reference fibre 1.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). A good consistency of the reported results was
observed for both analyses. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 16 of 40
m a e t smetrology and accreditation switzerland
4.1.5 HUT
6
4
2
0
-2
-4
-6
D –
Dm
ean
(ps/
nm)
160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / HUT
Fig. 9. Deviation from arithmetic mean of HUT results for reference fibre 1. The uncertainty bars
show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (coverage factor k = 2). Deviations
larger than their uncertainties were observed at one extremity of the spectral domain.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 17 of 40
m a e t smetrology and accreditation switzerland
4.2 Reference fibre 2 (G653) This fibre was especially chosen with a larger PMD (mean DGD of about 1.96 ps measured between 1435 and 1592 nm), in order to investigate the effects of the second order PMD on the calibration of the chromatic dispersion. No particular difficulties were encountered by the participating laboratories. 4.2.1 CSIC
-2
-1
0
1
2
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 1 G652 Fibre / CSIC
Fig. 10. Deviation from the arithmetic mean of CSIC results for reference fibre 2.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 18 of 40
m a e t smetrology and accreditation switzerland
4.2.2 METAS
-2
-1
0
1
2
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / METAS
-0.4
-0.2
0.0
0.2
0.4
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / METAS
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / METAS
-0.4
-0.2
0.0
0.2
0.4
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / METAS
Fig. 11. Deviation from the arithmetic mean of METAS results for reference fibre 2.The
uncertainty bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 19 of 40
m a e t smetrology and accreditation switzerland
4.2.3 NIST
-2
-1
0
1
2
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / NIST
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / NIST
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / NIST
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / NIST
Fig. 12. Deviation from the arithmetic mean of NIST results for reference fibre 2.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 20 of 40
m a e t smetrology and accreditation switzerland
4.2.4 NPL
-4
-2
0
2
4
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / NPL
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre Fibre / NPL
-4
-2
0
2
4
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 2 G653 Fibre / NPL
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / NPL
Fig. 13. Deviation from the arithmetic mean of NPL results for reference fibre 2.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 21 of 40
m a e t smetrology and accreditation switzerland
4.2.5 HUT
3
2
1
0
-1
D –
Dm
ean
(ps/
nm)
160015801560154015201500Wavelength (nm)
Ref. 2 G653 Fibre / HUT
Fig. 14. Deviation from arithmetic mean of HUT results for reference fibre 2. The uncertainty bars
show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (coverage factor k = 2). Deviations
much larger than their uncertainties were observed.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 22 of 40
m a e t smetrology and accreditation switzerland
4.3 Reference fibre 3 (G655 TeraLight) 4.3.1 CSIC
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3 .TeraLight Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / CSIC
Fig. 15. Deviation from the arithmetic mean of CSIC results for reference fibre 3.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 23 of 40
m a e t smetrology and accreditation switzerland
4.3.2 METAS
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / METAS
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / METAS
-0.4
-0.2
0.0
0.2
0.4
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / METAS
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / METAS
Fig. 16. Deviation from the arithmetic mean of METAS results for reference fibre 3.The uncertainty bars show the uncertainty of the deviation
ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the calibration
results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 24 of 40
m a e t smetrology and accreditation switzerland
4.3.3 NIST
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / NIST
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / NIST
0.2
0.1
0.0
-0.1
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / NIST
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / NIST
Fig. 17. Deviation from the arithmetic mean of NIST results for reference fibre 3.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 25 of 40
m a e t smetrology and accreditation switzerland
4.3.4 NPL
-2
-1
0
1
2
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / NPL
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / NPL
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 3.TeraLight Fibre / NPL
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean
(ps/
nm)
1600155015001450Wavelength (nm)
Ref. 3 TeraLight Fibre / NPL
Fig. 18. Deviation from the arithmetic mean of NPL results for reference fibre 3. The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 26 of 40
m a e t smetrology and accreditation switzerland
4.3.5 HUT
8
6
4
2
0
-2
D –
Dm
ean
(ps/
nm)
1600158015601540152015001480Wavelength (nm)
Ref. 3 G655 TeraLight Fibre / HUT
Fig. 19. Deviation from arithmetic mean of HUT results for reference fibre 3. The uncertainty bars
show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (coverage factor k = 2). Deviations
larger than their uncertainties were observed in the whole spectral domain.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 27 of 40
m a e t smetrology and accreditation switzerland
4.4 Reference fibre 4 (G655 Leaf) 4.4.1 CSIC
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / CSIC
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / CSIC
Fig. 20. Deviation from the arithmetic mean of CSIC results for reference fibre 4.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 28 of 40
m a e t smetrology and accreditation switzerland
4.4.2 METAS
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / METAS
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / METAS
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / METAS
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / METAS
Fig. 21. Deviation from the arithmetic mean of METAS results for reference fibre 4.The
uncertainty bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 29 of 40
m a e t smetrology and accreditation switzerland
4.4.3 NIST
-1.0
-0.5
0.0
0.5
1.0
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / NIST
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / NIST
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / NIST
-0.3
-0.2
-0.1
0.0
0.1
0.2
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / NIST
Fig. 22. Deviation from the arithmetic mean of NIST results for reference fibre 4.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 30 of 40
m a e t smetrology and accreditation switzerland
4.4.4 NPL
-4
-2
0
2
4
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / NPL
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / NPL
-3
-2
-1
0
1
2
3
D –
Dm
ean (
ps/n
m)
134013201300
Wavelength (nm)
Ref. 4. Leaf Fibre / NPL
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
D –
Dm
ean
(ps/
nm)
1620160015801560154015201500Wavelength (nm)
Ref. 4. Leaf Fibre / NPL
Fig. 23. Deviation from the arithmetic mean of NPL results for reference fibre 4.The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (upper graphs)
andiSiS DD uU ∆∆ ⋅= 2 (lower graphs). Both analyses show a good consistency of the
calibration results. The coverage factor was k =2.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 31 of 40
m a e t smetrology and accreditation switzerland
4.4.5 HUT
-5
-4
-3
-2
-1
0
1
D –
Dm
ean
(ps/
nm)
1600158015601540152015001480Wavelength (nm)
Ref. 4 G655 Leaf Fibre / HUT
Fig. (24). Deviation from the arithmetic mean of HUT results for reference fibre 4. The uncertainty
bars show the uncertainty of the deviation ii DD uU ∆∆ ⋅= 2 (coverage factor k = 2).
Deviation values significantly larger than their uncertainties were observed.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 32 of 40
m a e t smetrology and accreditation switzerland
5 Zero dispersion wavelength
The zero dispersion wavelength λo was calibrated by measuring the chromatic dispersion around the expected zero dispersion wavelength and by applying a curve fit to the measured data. According to the rules defined in the technical protocol, λo was only reported when obtained from a wavelength scan including the zero dispersion wavelength itself; i.e. that λo wasn’t obtained from an extrapolation of the measured dispersion data. CSIC delivered a second set of λo values that were obtained from a Four Wave mixing (FWM) measurement. The analysis of the calibration results was performed by applying the same principles as defined in Section 3.6. The mean zero dispersion wavelength mean0λ was calculated by considering the results of CSIC (phase shift method only), METAS, NPL and NIST. The results provided by HUT have not been considered, because of the too large deviation of the chromatic dispersion values. The second set of zero dispersion wavelength values provided by CSIC (FWM Method) was only considered as a comparative result and has consequently not been used for the determination of mean0λ . The mean zero dispersion wavelength was given by
∑=
⋅=n
iimean n 1
01 λλ (6)
The uncertainty of the mean value (confidence level k = 1) was then calculated as follows:
∑=
⋅=n
iimean
un
u1
200
1λλ , where (7)
iu
0λ was the measurement uncertainty claimed by each participant.
The deviation from the mean was then given by
meanii 000 λλλ −=∆ (8)
The uncertainty of the deviation was calculated for CSIC (phase shift method), METAS, NIST and NPL by applying Eq. (4), namely
22000
21imeani
un
uu λλλ ⋅⎟⎠⎞
⎜⎝⎛ −+=∆ (9)
This equation takes into account the correlation between the mean value and the sample data [1], [2], since the mean value was directly calculated from the i0λ data according to Eq. (6). The uncertainty of the deviation was calculated for the HUT and CSIC (FWM method) results according to Eq. (10). In this case, no correlation exists between the mean value mean0λ and the sample data i0λ . Therefore, the uncertainty of the deviation was straightforwardly given by 22
000 imeaniuuu λλλ +=∆ . (10)
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 33 of 40
m a e t smetrology and accreditation switzerland
5.1 Zero dispersion wavelength results The zero dispersion wavelength results and their claimed uncertainties are shown in Fig. (29). CSIC performed an extra calibration of this quantity by using a four-wave mixing (FWM) technique. These results are also shown here.
Ref.1. G652
CSIC
CSIC / FWM
NIST
NPL
METAS
1316.9
1316.95
1317
1317.05
1317.1
1317.15
1317.2
1317.25
λο (n
m)
Ref. 2. G653
CSIC
CSIC / FWM
HUT
METAS
NIST
NPL
1552.5
1553
1553.5
1554
1554.5
1555
λο (n
m)
Ref. 3. TeraLight
METAS
CSIC
CSIC / FWM
NIST
NPL
1447.61447.71447.81447.9
1448
1448.11448.21448.31448.41448.5
λο (n
m)
Ref. 4. Leaf
CSIC
CSIC / FWM
HUT
METAS
NISTNPL
1500.8
1501
1501.2
1501.4
1501.6
1501.8
1502
1502.2
1502.4
λο (n
m)
Fig. 29. Zero dispersion wavelength of the four reference fibres measured by all participants. The uncertainty bars show the reported uncertainty (coverage factor k=2). 5.2 Deviation from mean zero dispersion wavelength The mean value was calculated by excluding the results of HUT and of CSIC-FWM (four wave mixing). Most of the reported values were in range, except for reference fibre 3 (G 655 TeraLight), where both CSIC measurements (phase shift and four-wave mixing) showed a deviation larger than the uncertainty of the deviation. According to CSIC comments, possible explanations for this discrepancy may be that the measured zero dispersion wavelength was situated at the boundary between two spectral domains, where different measurement settings were applied. This may have influenced the quality of the results. Another concern was with the influence of the symmetry of the dispersion curve around λo, which is mandatory for an accurate determination of this quantity, and which was not fulfilled in this particular case. The FWM
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 34 of 40
m a e t smetrology and accreditation switzerland
techniques used by CSIC gave good results for reference fibres 1, 2 and 4. The deviation of HUT results was systematically larger than the uncertainty of the deviation.
Ref. 1. G652
NPL
NIST
METASCSIC / FWMCSIC
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
∆λο
(nm
)
Ref. 2. G653
CSIC
CSIC / FWM
HUT
METAS
NIST
NPL
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
∆λο
(nm
)
Ref. 3. TeraLight
CSIC
CSIC / FWM
METAS NISTNPL
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
∆λο
(nm
)
Ref. 4. Leaf
CSIC
CSIC / FWM
HUT
METAS NIST
NPL
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
∆λο
(nm
)
Fig. 30. Deviation from the arithmetic mean of the zero dispersion wavelength. The uncertainty
bars show the uncertainty of the deviation ii
uU00
2 λλ ∆∆ ⋅= (coverage factor k = 2). 6 Dispersion slope
The dispersion slope So around λo was derived from the same measurements that were used for the determination of λo. The analysis of the calibration results was performed by rigorously applying the same principles as defined in Section 5. The mean dispersion slope meanS0 was calculated by considering the results of CSIC, METAS, NPL and NIST. The results provided by HUT haven’t been considered, because of the too large deviation of the chromatic dispersion values.
∑=
⋅=n
iimean S
nS
100
1 (12)
The uncertainty of the mean value (confidence level k = 1) was then calculated as follows:
∑=
⋅=n
iSS imean
un
u1
200
1, where (13)
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 35 of 40
m a e t smetrology and accreditation switzerland
iSu0
was the measurement uncertainty claimed by each participant.
The deviation from the mean was then given by
meanii SSS 000 −=∆ (14)
The uncertainty of the deviation was calculated for CSIC, METAS, NIST and NPL by applying Eq. (15), namely
22000
21imeani SSS u
nuu ⋅⎟
⎠⎞
⎜⎝⎛ −+=∆ (15)
This equation takes into account the correlation between the mean value and the sample data [1], [2], since the mean value was directly calculated from the iS0 data according to Eq. (12). The uncertainty of the deviation was calculated for the HUT results according to Eq. (16). In this case, no correlation exists between the mean value meanS0 and the sample data iS0 . Therefore, the uncertainty of the deviation was straightforwardly given by 22
000 imeani SSS uuu +=∆ . (16)
6.1 Dispersion slope results
Ref. 1. G652
CSIC
NIST
NPL
METAS
1.670
1.680
1.690
1.700
1.710
1.720
1.730
1.740
So (p
s/nm
^2)
Ref. 2. G653
NPL
NIST
METAS
HUTCSIC
0.8200.8250.8300.8350.8400.8450.8500.8550.8600.8650.8700.875
So
(ps/
nm^2
)
Ref. 3. G655 TeraLight
CSIC
NIST
NPL
METAS
0.680
0.685
0.690
0.695
0.700
0.705
0.710
0.715
So
(ps/
nm^2
)
Ref. 4. G655 Leaf
CSIC
HUT
METAS NIST NPL
0.850
0.900
0.950
1.000
1.050
1.100
1.150
So
(ps/
nm^2
)
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 36 of 40
m a e t smetrology and accreditation switzerland
Fig. 31. Dispersion slope measured by all participants on the four reference fibres. The uncertainty bars show the reported uncertainty (coverage factor k=2). 6.2 Deviation from mean dispersion slope The deviations of HUT results were systematically larger than the corresponding deviation uncertainty. Most of the values reported by the other participants were in range, except for reference fibre 2, where somewhat larger discrepancies appeared by two other laboratories (CSIC and NIST). Reference fibre 2 was especially chosen with a larger PMD (mean DGD of about 1.96 ps measured between 1435 and 1592 nm) in order to investigate the influence of the second order PMD on the calibration process. This may have contributed to the larger spread of the reported results, although no first order dependency of the dispersion slope to the second order PMD should be theoretically expected.
Ref. 1. G652
CSCIC METASNIST
NPL
-0.025-0.02
-0.015-0.01
-0.005
00.0050.01
0.0150.02
∆S
o
Ref. 2. G653
NPLNISTMETAS
HUT
CSCIC
-0.03-0.025-0.02
-0.015-0.01
-0.0050
0.0050.01
0.0150.02
∆So
Ref. 3. G655 TeraLight
CSCIC
METAS
NISTNPL
-0.015
-0.01
-0.005
0
0.005
0.01
0.015
0.02
∆So
Ref. 4. G655 Leaf
CSCICHUT
METAS NIST NPL
-0.05
0
0.05
0.1
0.15
0.2
∆So
Fig. 32. Deviation of the dispersion slope from the arithmetic mean. The uncertainty bars show
the uncertainty of the deviation ii SS uU
002 ∆∆ ⋅= (coverage factor k = 2).
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 37 of 40
m a e t smetrology and accreditation switzerland
7 Conclusions This project was the first large scale supplementary inter-comparison on chromatic dispersion reference fibres that simultaneously addressed the calibration of all relevant quantities, namely the chromatic dispersion, the dispersion slope and the zero dispersion wavelength. It took only one year since the decision of starting this project to finalize the first report. This very short schedule was only possible thanks to the excellent collaboration of the participating laboratories. All laboratories used quite similar equipments, but applied very different measurement strategies and data processing techniques. This lead to very different approaches for the determination of the uncertainty budgets. A consequence was a sometimes large spread of the reported uncertainties. A ratio larger than 180 was for example found between the smallest and the largest claimed uncertainties. One laboratory delivered results, which significantly departed from the values reported by the other participants. A closer analysis showed that this deviation may be mostly due to a too large amount of Amplified Spontaneous Emission (ASE) of the laser source used for the measurements. The results of this laboratory have been analyzed in this report, but haven’t been used for the determination of the mean value of the reported quantities. The calibration of the overall chromatic dispersion of the four different types of singlemode fibres proved a very good agreement between all participants (above mentioned laboratory excluded). Most of the reported uncertainties were found to be larger than the standard deviation of the mean of the measured quantities. This very good agreement was obtained, despite of the very different data processing techniques used by the different participants. The measurement of the zero dispersion wavelength and of the dispersion slope showed a rather good consistency between all participants (above mentioned laboratory excluded). Some larger deviations were nevertheless observed at some specific points. This was mainly due to a non optimum choice of the data processing parameters, such as the symmetry of the group delay around the zero dispersion wavelength or the consistency of the group delay values at the boundary between two spectral domains. One laboratory also measured the zero dispersion wavelength by using a four wave mixing technique, which gave comparable good results. One reference fibre (G653) was especially chosen with a large PMD (mean DGD of 1.96 ps), in order to investigate the influence of the PMD and of the second order PMD on the calibration of the chromatic dispersion. No clearly evident influence was observed on the reported results.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 38 of 40
m a e t smetrology and accreditation switzerland
8 References [1] K. Beissner, “On a measure of consistency in comparison measurements”, Metrologia, 39,
pp. 59-63 (2002). [2] M. Neugebauer, F. Lüdicke, “EUROMET comparison: diameter of small ring gauges”,
Metrologia, 38, pp. 259-267 (2001). [3] R. Thalmann, “CCL Key Comparison CCL-K1, Calibration of gauge blocks by interferometry,
Final Report”, (2001). [4] Guide to the Expression of Uncertainty in Measurement, ISO, (1995). [5] J. Morel, “Technical protocol EUROMET Project 666, Inter-comparison of Chromatic
Dispersion Reference Fibres”. [6] Franzen, D. L. Mechels, S. E. Schlager, J. B. (OPTOELECTRONICS DIVISION - 815),
“Accurate Measurement of the Zero-Dispersion Wavelength in Optical Fibers”, Journal of Research of the National Institute of Standards and Technology , May 01, (1997).
[7] T. Dennis and P. A. Williams, “Chromatic dispersion measurement error caused by source
amplified spontaneous emission”, IEEE Photon. Tech. Lett., Vol. 16, 11, pp. 2532 - 2534, (2004).
[8] M. G. Cox and P. M. Harris, “Towards an objective approach to key comparison reference
values “, NPL SSfM Publication. [9] S. E: Mechels, “International Comparison: Zero Dispersion Wavelength in Single-Mode
Optical Fibres (Wavelength); Final Report”, Metrologia, 34, (1997), 449.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 39 of 40
m a e t smetrology and accreditation switzerland
9 Annex A. Discussion of HUT results HUT performed a subsequent analysis of its calibration results in order to better identify the origin of the observed deviations. Most of the group delay data reported by HUT showed some kind of periodic modulation of the residual when fitted with the relevant Sellmeier functions. Typical examples are shown in Fig. (33) for Ref. fibre 1 (G652) and for Ref. Fibre 3 (G655 TeraLight). The data were fitted by using a 5-term Sellmeier function. The origin of this modulation may be either a large PMD of the reference fibre or the influence of the Amplified Spontaneous Emission (ASE) of the laser source used for the measurements [7].
1500 1520 1540 1560 1580 1600-20000
-15000
-10000
-5000
0
5000
10000
15000
20000
-75
-50
-25
0
25
50
75
Gro
up d
elay
[ps]
Wavelength [nm]
Measurement Fit
Res
idua
l [ps
]
1480 1500 1520 1540 1560 1580 1600-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
-50
-40
-30
-20
-10
0
10
20
30
40
50
Measurement Fit
Res
idua
l [ps
]
Gro
up d
elay
[ps]
Wavelength [nm]
Fig. (33) Group delay data and residual obtained by HUT for Reference Fibres 1 (G652) and 3 (G655, TeraLight). Both residuals show a periodical structure. A significant influence of the second order PMD of reference fibres 1 and 4 is unlikely to happen, since the mean DGD of both fibres were shown to be smaller than 0.3 ps within the whole spectral range. As demonstrated in [7], the influence of the ASE of the laser source is the most probable explanation of the observed deviations. The parasitic modulation contributes in disturbing the curve fitting process when using high order Sellmeier functions. This was highlighted by comparing the deviation between the METAS and HUT results when fitting the HUT group delay data with both, a 5-term Sellmeier and a parabolic function. METAS data were fitted with a 5-term Sellmeier function. Figure (34) shows the results of this experiment for Reference fibre 4 (G655 Leaf). The right vertical scale shows the difference between METAS and HUT values, when using the 5-term Sellmeier (circles) and the parabolic (triangles) functions. The parabolic fit significantly reduces the difference between HUT and METAS results, since it tends to average the effect of the parasitic modulation of the group delay data. This tentative explanation is only given as an indication for further analysis and should be supported by more experimental evidences.
1480 1500 1520 1540 1560 1580 1600-40
-20
0
20
40
60
80
100
-1
0
1
2
3
4
5
Dis
pers
ion
[ps/
nm]
W avelength [nm]
METAS 5-term Sellmeier HUT 5-term Sellmeier HUT 2. order polynomial
Dispersion
Diff
eren
ce [p
s/nm
]
Fig. (34) Left scale: chromatic dispersion of reference fibre 4 (G655 Leaf) measured by HUT and by METAS. Right scale: difference between HUT and METAS results for both, a 5-term Sellmeier and a 2nd order fit of HUT data.
EUROMET Project 666, EUROMET-PR.S1 Final Report Page 40 of 40
m a e t smetrology and accreditation switzerland
HUT applied the same second order (parabolic) fit to the calibration data of Ref. Fibre 1 (G652) (see Fig. 35) and of Ref. Fibre 3 (G655, Teralight) (see Fig. 36). Reduction of the difference between HUT and METAS results was observed also in these fits.
1500 1510 1520 1530 1540 1550 1560 1570 1580 1590 1600
260
280
300
320
340
360
380
400
-5
-4
-3
-2
-1
0
1
2
3
4
5
Diff
eren
ce [p
s/nm
]
Dis
pers
ion
[ps/
nm]
Wavelength [nm]
METAS 5-term Sellmeier HUT 5-term Sellmeier HUT 2. order Polynomial
Dispersion
Fig. (35) Left scale: chromatic dispersion of reference fibre 1(G652) measured by HUT and by METAS. Right scale: difference between HUT and METAS results for both, a 5-term Sellmeier and a 2nd order fit of HUT data.
1475 1500 1525 1550 1575 1600 1625 1650
20
40
60
80
100
120
140
-10
-5
0
5
10
15
20
Diff
eren
ce [p
s/nm
]
METAS 5-term Sellmeier HUT 5-term Sellmeier HUT 2.order polynomial
Dis
pers
ion
[ps/
nm]
Wavelength [nm]
Fig. (36) Left scale: chromatic dispersion of reference fibre 3 (G655, TeraLight) measured by HUT and by METAS. Right scale: difference between HUT and METAS results for both, a 5-term Sellmeier and a 2nd order fit of HUT data.