Fifth Oxford Conference on Spectrometry
Fifth Oxford Conference
on Spectrometry
Fifth Oxford Conference on Spectrometry
The Fifth Oxford Conference on Spectrometry was organised by the NPL Optical Radiation
Measurement Club (ORM Club) and the Council for Optical Radiation Measurement (CORM), and
was held at Bushy House, National Physical Laboratory, Teddington, UK from 26th June to
28th June 2006. This was the latest in a series of conferences on Optical Spectrometry, which began
in 1986 when the first conference was held at Keble College, Oxford, UK, organised jointly by the
Ultraviolet Spectrometry Group (UVSG) and CORM. This proved so successful that a second
‘Oxford’ Conference was held at Franklin Pierce College in New Hampshire, USA in 1994,
followed thereafter by the third conference at Royal Holloway College, Egham, UK in 1998 and the
fourth at Davidson College, Davidson, North Carolina, US in 2002. There is thus an established
regular four yearly pattern, with alternate conferences in the USA and the UK.
The programme for the Fifth Oxford Conference focused on the themes of fluorescence,
spectrophotometry, and advances in technology and instrumentation. As with previous meetings
prominence was given to contributions from national measurement laboratories, with representation
on this occasion from Europe, USA, Canada and Singapore. More than 35 papers and posters were
presented, covering a range of applications of optical spectrometry for the characterisation and
measurement of the properties of materials in the ultraviolet, visible and infrared spectral regions, as
well as more fundamental issues such as the establishment of basic spectrophotometric scales and
the validation of measurement instrumentation. These proceedings comprise of written versions of
the papers presented at the conference.
NPL would like to take this opportunity to thank the Scientific Programme Selection Committee
(Miss Teresa Goodman, NPL; Mrs Fiona Jones, NPL; Dr Mike Pointer, NPL; and Dr Art
Springsteen, Avian Technologies) and the NPL Organising Committee (Mrs Gill Coggins, NPL; Mr
Roger Hughes, NPL; Mrs Fiona Jones, NPL; and Ms Stacy Skangos, NPL) for all their hard work in
making this conference such a success.
ISBN 978-0-946754-48-9
© Crown copyright 2006.
Reproduced with the permission of the Controller of HMSO and the Queen’s printer for Scotland
Day 1: Fluorescence
Measuring and Certifying True Fluorescence Spectra with a Qualified Fluorescence Spectrometer.
P. DeRose, E. Early and G. Kramer.
Biochemical Science Division, National Institute of Standards and Technology, Gaithersburg, USA..............1
Linking Fluorometry to Radiometry. U. Resch-Genger1, D. Pfeifer1, C. Monte1, A. Hoffmann1,
W. Bremser1, D.R. Taubert2.
Federal Institute for Materials Research and Testing (BAM), Germany
Physikalisch-Technische Bundesanstalt (PTB), Berlin, Germany ....................................................................9
Validation of Fluorescence Lifetime Apparatus. P Miller.
Optical Radiation Measurement Team, National Physical Laboratory, UK ...................................................11
The Determination of Fluorescence Quantum Yields Using a Fluorimeter and Integrating Sphere.
A. Beeby 1, L.O. Palsson2, L. Porres1, A.P. Monkman2.
Department of Chemistry, University of Durham, UK
Department of Physics, University of Durham, UK........................................................................................13
Goniofluorometer for Spectral Quantum Yield Measurements. S. Holopainen1, F. Manoocheri1,
M. Laurila1 and E. Ikonen1, 2.
Metrology Research Institute, Helsinki University of Technology (TKK), Finland
Centre for Metrology and Accredidation (MIKES), Finland ..........................................................................15
Extension of the NRC Reference Spectrofluorimeter to Volume Fluorescence Measurements.
J. Zwinkels and F. Gauthier.
Institute for National Measurement Standards, National Research Council of Canada, Canada...................21
Standardization of Fluorescence Techniques: Where Do We Stand and What Do We Need?
U. Resch-Genger, D. Pfeifer, K. Hoffmann, A. Hoffmann and C. Monte.
Federal Institute for Materials Research and Testing (BAM), Germany .........................................................35
Day 2: Spectrophotometry
Spectral Regular Transmittance Scale at SPRING Singapore. LIU Yuanjie Liu and XU Gan.
National Metrology Centre (SPRING), Singapore...........................................................................................37
Comparison of NRC and NIST Infrared Diffuse Reflectance Scales from 2 µm to 18 µm. L.
Hanssen1 and N. Rowell2.
Optical Technology Division, National Institute of Standards and Technology, USA
Institute for National Measurement Standards, National Research Council, Canada ..................................39
Effects of Aging and Degradations on Regular Transmittance of Interference Filters. A.
Lamminpää1, S. Holopainen1, F. Manoocheri1, and E. Ikonen1,2.
Metrology Research Institute, Helsinki University of Technology (TKK), Finland
Centre for Metrology and Accreditation (MIKES), Finland ...........................................................................41
Development of a Multispectral Texture Measurement Facility for Use in an EU-Funded Study of
the Naturalness of Surfaces. R. Montgomery1, M. Pointer1, T. Goodman1 and A. Harvey2.
Optical Radiation Measurement Team, National Physical Laboratory, UK
School of Engineering and Physical Sciences, Heriot-Watt University, UK ..................................................43
Laser Techniques for High Accuracy Spectrometric Measurements of Parallel-Sided Samples.
J. Cheung, E. Usadi and C. Chunnilall.
Optical Radiation Measurement Team, National Physical Laboratory, UK ...................................................55
Robot-Based Gonioreflectometry at PTB. A. Höpe and D. Hünerhoff.
Physikalisch-Technische Bundesanstalt(PTB), Braunschweig, Germany........................................................65
Characterization of Printed Textile Fabrics. F. Leloup, S. Forment, J. Versluys, P. Hanselaer.
KaHo Sint-Lieven, Belgium ..............................................................................................................................71
Measurement of Bidirectional Reflectance in the Visible and Infrared. P. Raven.
QinetiQ, Malvern Technology Centre, UK.......................................................................................................81
Day 3: Advances in Technology and Instrumentation
The Imaging Sphere – the First Appearance Meter? R. Yeo, R. Rykowski, D. Kreysar and
K. Chittim.
Radiant Imaging, USA......................................................................................................................................87
The Latest Revision of European Pharmacopoeia to 5.2, and its Effect on the Qualification of UVVisible Spectrophotometers. J. Hammond.
Optiglass Ltd, UK...........................................................................................................................................105
Infrared Spectrophotometry; Some of the Pitfalls. E. Theocharous.
Optical Radiation Measurement Team, National Physical Laboratory, UK .................................................107
Integrated Sensors for Point of Care Detection. J. de Mello.
Department of Chemistry, Imperial College, London, UK.............................................................................119
The Multi-Mode Optical Spectrometer: Towards the Simultaneous Measurement of Absorption,
Transmission, Turbidity, Linear Dichroism, Optical Activity, Fluorescence (Normal, Linearly and
Circularly Polarised) and Light Scattering. A Drake.
King’s College, London, UK ..........................................................................................................................121
List of posters presented .................................................................................................................................123
Measuring and Certifying True Fluorescence Spectra with a Qualified
Fluorescence Spectrometer
P. DeRose, E. Early and G. Kramer.
Biochemical Science Division, National Institute of Standards and Technology, 100
Bureau Drive, Stop 8300, Gaithersburg, MD 20899-8300, USA
The demand for fluorescence standards has increased greatly in recent years [1, 2] due
to the development of clinical, biochemical, pharmaceutical and environmental assays
that use fluorescence detection to quantify analyte concentrations and amounts. Many of
these assays are also being used in highly regulated areas where it must be demonstrated
that instruments are qualified to take measurements with a specified degree of accuracy
and precision as part of a method validation process. Conventional physical transfer
(PT) standards, such as calibrated light sources and calibrated detectors, have
traditionally been used by spectroscopists to calibrate the spectral responsivity of the
detection system and the excitation intensity of fluorescence spectrometers [3-17].
These types of standards require a level of expertise to implement that is beyond the
experience and knowledge of most users. Many newer compact fluorescence
instruments, such as portable fluorometers and microwell plate and microarray scanners,
are incompatible with such bulky standards. Their high initial cost and the added
periodic (typically once a year) cost of having them recalibrated make conventional PT
standards even less attractive.
Certified reference materials (CRMs), artifacts supplied with values and uncertainties
established by a national metrology institute (NMI), can be a more practical and easyto-use alternative, especially when they are made in conventional formats, so that they
can be measured just like a sample. CRM artifact standards can also be made to be
portable, relatively inexpensive and not requiring recertification. Several NMIorganized workshops [18-21] have aided in establishing which fluorescence standards
are most urgently needed. Among these, spectral correction, day-to-day instrument
performance and intensity standards were the first CRMs to be developed [22-26]. The
qualification of a research-grade fluorescence spectrometer for measuring true spectra,
i.e., spectrally corrected with a high, known accuracy, and its subsequent use to certify
NIST Standard Reference Materials (SRMs) 2940 and 2941 are presented here.
Fluorescence spectra were taken on a SPEX Fluorolog 3 [27] (Jobin Yvon, Edison, NJ)
using double monochromators with 1200 grooves / mm gratings blazed at 330 nm and
500 nm for excitation and emission selection, respectively. A Hamamatsu R928-P
photon-counting photomultiplier tube (PMT) was set at 950 V for emission detection
(referred to as the signal or S channel), and a 450 W continuous Xe lamp was used as an
excitation source. A small fraction of the post-monochromator excitation beam is
diverted to a “reference” photodiode just before the sample (referred to as the reference
or R channel) to monitor the relative excitation intensity as a function of time and
excitation wavelength. A line of the Xe source lamp and a Hg pen lamp line were used
to calibrate the wavelength of the excitation and emission monochromators,
Light from a calibrated source (CS), a tungsten halogen lamp mounted onto one
integrating sphere that is attached to another integrating sphere, was directed into the
Fig. 1. Schematics of the instrumental setups used for determining relative spectral correction of
a) emission, b) excitation, and c) absolute spectral correction (in combination with setup a or b).
emission detection system by a calibrated diffuse reflector (CR), and the emission
monochromator was scanned to calibrate the relative spectral responsivity of the
detection system (see Fig. 1a). The flux of the excitation beam was measured as a
function of wavelength by collecting the output from a calibrated detector (CD), placed
at the sample position, and scanning the excitation monochromator (see Fig. 1b). Each
set of data, i.e., CS-based and CD-based, was used independently to determine the
absolute spectral responsivity of the detection system, with the aid of a third set of data
that was collected by putting the CR at the sample position at a 45° angle relative to the
incident light from the excitation beam, thereby, reflecting it into the detection system
(see Fig. 1c). The two monochromators were then synchronously scanned.
Results and Discussion
Relative Spectral Responsivity
Neither the excitation intensity at the sample (see Fig. 2) nor the relative spectral
responsivity of the detection system (see Fig. 3) were flat with wavelength, which is
typical for conventional fluorometers.
Power / μW
Wavelength / nm
Fig. 2. Flux of the excitation beam at the
sample position.
Wavelength / nm
Fig. 3. Relative spectral responsivity of the
detection system.
Therefore, multiplicative correction factors for the former and the latter were applied to
excitation and emission spectra, respectively, to obtain the correct spectral shape. Since
the spectral profiles for excitation intensity and detection system responsivity will be
different for every instrument, such correction factors must also be used to allow spectra
comparisons between instruments. The correction factor for emission spectra (CS) is
simply the inverse of the relative spectral responsivity of the detection system and the
corrected emission signal Scor = CS S, where S is the measured emission signal. Since
the intensity of common excitation sources, such as Xe lamps, can fluctuate
significantly over relatively small excitation wavelength regions and with time, a
reference detector is supplied with most instruments to correct S for excitation intensity
fluctuations in real time. These reference detectors, which measure an excitation signal
(R), can adequately correct for such fluctuations by calculating the ratio S/R, but they
are not perfect, particularly over a wide excitation wavelength range [28]. The reference
channel correction factor (CR) for our instrument, which uses a silicon photodiode as a
reference detector, was calculated as the ratio of the CD and reference channel signals
(see Fig. 4), such that the corrected reference channel signal Rcor = CR R. The
instrument independent spectrum, which reflects the true spectral shape of the
fluorescence, is equal to Scor /Rcor.
R C o rre c tio n
EX Wavelength (nm)
Fig. 4. The relative R-channel correction factor as a function of excitation (EX) wavelength.
The y-axis is normalized to one at 414 nm for both Corrected and Reference curves.
The total uncertainty (k = 2) in the relative intensity correction factor CS for an emission
spectrum using the CS-based calibration method was about 5 % through most of the
visible region, but increased significantly below 400 nm. The corresponding
uncertainties in the absolute intensity correction factor KS for an emission spectrum
using the CS-based calibration method was about twice that of CS. The total uncertainty
in both CS and KS using the CD-based calibration method was about 5 % larger than the
corresponding values using the CS-based calibration method. A more detailed
description of these calibration methods and their corresponding uncertainties will be
presented elsewhere. [29]
Excitation emission matrices (EEMs) of tryptophan in aqueous solution are shown here
to demonstrate the importance of spectral correction. The uncorrected signal contour
(see Fig. 5a) can be seen to be qualitatively incorrect, when compared to the instrument
independent spectrum (see Fig. 5c), showing two peaks instead of one and an emission
wavelength shift of about 100 nm. The application of the reference channel with
correction (see Fig 5b) gives a qualitatively correct contour, but the peak position is still
shifted in EM wavelength by 14 nm. Note that a comparison of S/R versus S/Rcor
exhibited an insignificant difference in this case and is therefore not shown.
EX Wavelength / nm
EX Wavelength / nm
EM Wavelength / nm
EX Wavelength / nm
EM Wavelength / nm
EM Wavelength / nm
Fig. 5. Excitation emission matrices (EEMs) of tryptophan in aqueous solution.
Relative Fluorescence Intensity
SRM 2940 (Orange Emission) [22] and SRM 2941 (Green Emission) [23] Relative
Intensity Correction Standards for Fluorescence Spectroscopy are solid, cuvette-shaped
pieces of borate glass doped with 0.11 % MnO2 and 0.01 % U3O8 , respectively. Three
of their long faces are polished, while the fourth is frosted. The certification
measurements were taken with the frosted face turned 180° from the detection system.
The fluorescence was collected at 90° relative to the excitation beam. Excitation and
emission bandwidths of 3 nm were used for all fluorescence measurements. The SRMs
were certified using the CS-based correction factors, due to the smaller uncertainties
related with this method. When combined with SRM 936a Quinine Sulfate Dihydrate
[26], these three standards enable the entire visible region of a fluorometer’s detection
system to be spectrally corrected (see Fig. 6). The two glass SRMs do not photodegrade,
making them useful as performance validation standards for intensity, unlike organic
dyes, such as SRM 936a. The glass SRMs are robust with an estimated shelf life of 10
years or more, when handled as recommended. These SRMs are supplied with certified
values and uncertainties [22,23].
SRM 2940
SRM 2941
SRM 936a
Wavelength / nm
Fig. 6 Instrument independent emission spectra for fluorescence SRMs.
A research-grade fluorescence spectrometer was qualified to measure relative and
absolute intensity corrected spectra that are instrument independent. A relative spectral
correction method for emission using a calibrated light source yielded correction factors
with total uncertainties of about 5 %. Two cuvette-shaped, solid glass SRMs 2940 and
2941 have been released for use as both spectral correction standards for emission and
performance validation standards for intensity.
DeRose, P.C. “NIST Workshop on Luminescence Standards for Chemical Analysis”.
J.Res.Natl.Inst.Stand. Technol., 105, 631 (2000).
Resch-Genger, U., Hoffmann, K., Nietfeld, W., Engel, A., Neukammer, J., Nitschke, R., Ebert, B.
and Macdonald, R., “How to Improve Quality Assurance in Fluorometry: Fluorescence-Inherent
Sources of Error and Suited Fluorescence Standards”. J.Fluoresc., 15, 337 (2005).
Roberts, G.C.K., Chapter 7, “Correction of Excitation and Emission Spectra”. In Techniques in
Visible and Ultraviolet Spectrometry, Vol. 2, Standards in Fluorescence Spectrometry, J.N.Miller,
Ed. (Chapman and Hall, New York, 1981), p. 54-61.
Costa, L.F., Mielenz, K.D. and Grum, F., Chapter 4, “Correction of Emission Spectra”. In Optical
Radiation Measurements, Vol. 3, Measurement of Photoluminescence, K.D.Mielenz, Ed. (Academic
Press, New York, 1982), p. 139-174.
Hofstraat, J.W. and Latuhihin, M.J., “Correction of Fluorescence Spectra”. Applied Spec., 48, 436
a) Zwinkels, J.C., Gignac, D.S., Nevins, M., Powell, I. and Bewsher, A., “Design and testing of a
two-monochromator reference spectrofluorimeter for high-accuracy total radiance factor
measurements”. Applied Optics, 36, 892 (1997). b) Zwinkels, J.C. and Gauthier, F.,
“Instrumentation, standards, and procedures used at the National Research Council of Canada for
high-accuracy fluorescence measurements”. Anal.Chim.Acta, 380, 193 (1999). c) Zwinkels, J.C. and
Gigac, D.S., “Development of a New Reference Spectrofluorimeter”. In Burgess, C. and Jones, D.G.
(eds.) Spectrophotometry, Luminescence, and Color; Science and Compliance, Elsevier,
Amsterdam, p. 97 (1995).
Hollandt, J., Taubert,R.D., Seidel, J., Resch-Genger, U., Gugg-Helminger, A., Pfeifer, D., Monte, C.
and Pilz, W., “Traceability in Fluorometry-Part I: Physical Standards”. J.Fluoresc., 15, 301 (2005).
Nanjo, H.M.M. and Nayatani, Y., “Colorimetry and its Accuracy in the Measurement of Fluorescent
Materials by the Two-Monochromator Method,” Color Res. and Appl., 10 (2) 84 (1985).
Grum, F., “Instrumentation in Fluorescence Measurements”. Journal of Color and Appear., 1 (5) 18
Velapoldi, R.A. and Epstein, M.S. Chapter 7 “Luminescence Standards for Macro-and
Microspectrofluorometry”. In Goldberg, M.C. (ed.) Luminescence Applications in Biological,
Chemical, Environmental, and Hydrological Sciences, American Chemical Society, Washington,
DC, 97 ff (1989).
Melhuish, W.H., “Absolute Spectrofluorometry”. J. Res. Nat. Bur. Stand., 76A, 547 (1972).
Eaton, D.F., “Reference Materials for Fluorescence Measurement”. Pure and Appl. Chem., 60 (7),
1107 (1988).
Mielenz, K.D., “Fluorescence Spectrometry in Analytical Chemistry and Color Science”. In
Burgess, C. and Mielenz, K.D. (eds.) Advances in Standards and Methodology in
Spectrophotometry, Elsevier, Amsterdam, 49 ff (1987).
Mielenz, K.D. (ed.), “Measurement of Photoluminescence”. Volume 3 in Optical Radiation
Measurements, Academic Press, New York, (1982).
Parker, C.A., Photoluminescence of Solutions, Elsevier, Amsterdam (1968).
Miller, J.N, (ed.), Techniques in Visible and Ultraviolet Spectrometry, Vol.2. Standards in
Fluorescence Spectrometry, Chapman and Hall, London, (1981).
Verrill, J.F. and Williams, D.C., “The Development of a New Reference Spectrofluorimeter at the
National Physical Laboratory”. In Burgess, C. and Jones, D.G. (eds.) Spectrophotometry,
Luminescence, and Color; Science and Compliance, Elsevier, Amsterdam, 111 ff (1995).
DeRose PC (2000) NIST workshop on luminescence standards for chemical analysis, Sept 1999. J
Res Natl Inst Stand Technol 105:631
a) Workshop (Jan 1998) Fluorescence intensity standards. NIST. b) Workshop (June 2000) New
directions in fluorescence intensity standards. NIST. c) Workshop (March 2005) Towards national
traceability in fluorescence intensity measurements. NIST. d) Workshop (Feb 2006) Improved
antibody-based metrology in flow cytometry, NIST (comment: a ref. article should be available
soon, the other 3 workshops (a-c) were the precursors to this one)
Workshop (Dec 2002) Fluorescence standards for microarray assays. NIST
21. Workshop (June 2003) Bioanalytical and biomedical applications of fluorescence techniques:
instrument characterization and validation, traceability and need for reference materials. U. ReschGenger (BAM), R. Macdonald (PTB), BERM-9
22. NIST Certificate of Analysis, SRM 2940 Relative Intensity Correction Standard for Fluorescence
Spectroscopy: Orange Emission (2006).
23. NIST Certificate of Analysis, SRM 2941 Relative Intensity Correction Standard for Fluorescence
Spectroscopy: Green Emission (2006).
24. Certificate of analysis, Certified reference materials BAM-F001 - BAM-F005, Calibration kit,
Spectral fluorescence standards for the determination of the relative spectral responsivity of
fluorescence instruments. Federal Institute for Materials Research and Testing (BAM) (2006).
25. NIST Certificate of Analysis, SRM 1932 Fluorescein Solution (2004).
26. NIST Certificate of Analysis, SRM 936a Quinine Sulfate Dihydrate (1994).
27. Certain commercial equipment, instruments, or materials are identified in this paper to foster
understanding. Such identification does not imply recommendation or endorsement by the National
Institute of Standards and Technology, nor does it imply that the materials or equipment identified
are necessarily the best available for the purpose.
28. Holbrook, R.D., DeRose, P.C., Leigh, S.D., Rukhin, A.L. and Heckert, N.A., “Excitation-Emission
Matrix Fluorescence Spectroscopy for Natural Organic Matter Characterization: A Quantitative
Evaluation of Calibration and Spectral Correction Procedures”. Appl.Spec., 60, 791 (2006).
29. DeRose, P.C., Early, E.A. and Kramer, G.W., “Qualification of a Fluorescence Spectrometer for
Measuring True Fluorescence Spectra”. In preparation.
Linking Fluorometry to Radiometry
U. Resch-Genger1, D. Pfeifer1, C. Monte1, A. Hoffmann1, W. Bremser1, D.R. Taubert2
and J. Hollandt2
Federal Institute for Materials Research and Testing (BAM), Richard-WillstätterStraße 11, D-12489 Berlin, Germany
Physikalisch-Technische Bundesanstalt (PTB), Abbéstraße 2-12, D-10587 Berlin,
Fluorescence techniques are amongst the most widely used analytical techniques in
material sciences, environmental analysis, biology, clinical chemistry, and medical
diagnostics. Drawbacks of these techniques are time-dependent instrument-specific
contributions to otherwise analyte-specific signals that limit the comparability of
luminescence data across instruments and render quantification from measurements of
fluorescence intensities difficult [1,2] as well as the lack of reliable and purpose-fit
standards for instrument characterization and performance validation [3].
Employing physical and chemical transfer standards, different approaches for the
determination of the spectral characteristics of the emission channel of fluorescence
instruments under application-relevant conditions are presented. The aim is here to link
fluorescence measurements to the spectral radiance or the spectral responsivity scale via
relative and eventually absolute measurements of fluorescence spectra and a newly built
reference fluorometer. For these different calibration strategies, state-of-the art
uncertainties are provided and the state-of-the art of the comparability of fluorescence
measurements is discussed based on different interlaboratory comparisons.
U. Resch-Genger, D. Pfeifer, C. Monte, W. Pilz, A. Hoffmann, M. Spieles, K. Rurack, J. Hollandt,
D. Taubert, B. Schönenberger , P. Nording, J. Fluoresc., 15, 325, 2005.
J. Hollandt, R. D. Taubert, J. Seidel, U. Resch-Genger, A. Gugg-Helminger, D. Pfeifer, C. Monte, W.
Pilz, J. Fluoresc., 15, 311, 2005.
U. Resch-Genger, K. Hoffmann, W. Nietfeld, A. Engel, J. Neukammer, R. Nitschke, B. Ebert, R.
Macdonald, J. Fluoresc., 15, 347, 2005.
Validation of Fluorescence Lifetime Apparatus
Paul Miller
National Physical Laboratory, Hampton Road, Teddington, TW11 0LW, UK
During the past 20 years there has been a rapid growth in the use of fluorescence
spectroscopy. Key applications include medical imaging, healthcare and more recently
nanotechnology. Whilst steady-state (time-integrated) fluorescence measurements have
been used to characterise diverse samples in terms of emission strength, peak
wavelength and spectral shape, limitations to their effectiveness include reduced
discrimination of spectrally overlapping fluorophores due to their relatively broad
emission bands.
Time-dependent measurements provide additional information about the underlying
fluorescence dynamics and have the advantages of enhanced discrimination among
fluorophores, especially for those with overlapping emission spectra. Fluorescence
lifetime measurements are also sensitive to various environmental parameters such as
pH, temperature and solvation. Another major advantage of lifetime-based sensing is
the fact that fluorescence lifetime is independent of signal strength.
Recently instrumentation costs have decreased dramatically. This, combined with the
size and complexity of operation, has opened fluorescence lifetime measurements to a
wider range of applications and expertise.
The well-established time-domain technique of time-correlated single-photon counting
(TCSPC) can resolve picosecond lifetime with high sensitivity and dynamic range. In
the work reported here a commercial instrument is characterised and calibrated. The
performance of the apparatus was then validated using measurements of fluorescence
lifetime standards and the data analysis tested with simulated data. Further tests were
performed on important molecular probes, including quantum dots and Enhanced Green
Fluorescent Proteins (EGFP).
Non-linearity of the time measurement is the most important source of errors in TCSPC
measurements. Traditionally timing accuracy, stability and linearity is tested using pulse
generators or electrically calibrated delays. Here we develop a novel method based on
the optical delay of a laser pulse. The linearity of the TCSPC system and PMT detector
can also affect the measurement of the fluorescence lifetime and these were investigated
using superposed laser pulses.
Using this system it will then be possible to make traceable measurements of candidate
fluorescence lifetime standards, which in turn can be used to validate measurement
systems in diverse fields such as clinical sensing and diagnostics and fluorescence
lifetime imaging.
© Crown copyright 2007. Reproduced by permission of the Controller of HMSO and Queen’s Printer for
This work was supported by the National Measurement System Policy Unit of the Department of Trade
and Industry.
The Determination of Fluorescence Quantum yields using a Fluorimeter and
Integrating Sphere
A. Beeby1, L.O. Palsson2, L. Porres1, A.P. Monkman2
Department of Chemistry, University of Durham, Durham, DH1 3LE, UK
Department of Physics, University of Durham, Durham, DH1 3LE, UK
Fluorescence quantum yields are an important photophysical parameter, and are the
starting point for the calculation of many other photophysical parameters. The
experimental determination of fluorescence quantum yield from materials in dilute
solution is normally done by comparison to a ‘standard’ material. Even so the method
requires laborious and careful work in order to get reliable values. However, the same
methods cannot be applied to scattering or solid state samples, which require more
complex methods. We will present a facile method for the determination of quantum
yields from both solution and solid state samples. Our method utilises an integrating
sphere in a commercial spectrofluorimeter and provides a rapid measurement. Also, our
method does not rely upon the use of often poorly characterised reference materials.
Goniofluorometer for Spectral Quantum Yield Measurements
Silja Holopainen1, Farshid Manoocheri1, Marko Laurila1 and Erkki Ikonen1,2
Metrology Research Institute, Helsinki University of Technology (TKK), POB 3000,
FI-02015, Finland
Centre for Metrology and Accreditation (MIKES), POB 9, FI-02151 Espoo, Finland
Fluorescent brightening agents (FBA) are used extensively in several branches of
industry to brighten the appearance of material. For example paper industry uses FBA:s
that are excited in the UV region and emit in the blue region of the visible spectrum to
compensate the yellowness normally found in paper and produce a whitening effect.
Accurate knowledge of spectral fluorescent yield of a certain FBA in given environment
provides the knowledge of how much FBA is needed to produce a certain kind of
brightening effect, thus lowering expenses and saving the environment.
In order to measure the spectral fluorescent yield accurately, reliable measurement
facilities for characterization of fluorescent reference standards are needed. Most
measurement facilities providing such a service use measurement geometry of 0°/45° or
45°/0°. However, various fields of industry have different needs for measurement
geometry such as e.g. 0°/10° in biochemical analyzer industry. This paper introduces a
goniofluorometer facility, built at the Metrology Research Institute, Helsinki University
of Technology (TKK). The facility can measure spectral fluorescent yield and bispectral
luminescent radiance factors in the wavelength range of 250 – 830 nm in various
measurement geometries.
The spectral radiance factors for a fluorescent material can not be determined simply by
measuring reflectance from the sample. For such a material the radiance factor is always
a combination of the reflected radiance factor and the luminescent radiance factor. An
explanation of the radiance factors can be found in reference [1]. In order to calculate
the luminescent radiance factor for a given light source, one needs to know the
bispectral luminescent radiance factors βLλ(μ). The bispectral luminescent radiance
factor is defined as the ratio of the fluorescent flux per unit emission band pass dλ from
the sample to the flux from a perfect diffuser identically irradiated and viewed and it
can be calculated with equation (1).
β Lλ ( μ ) =
f (λ ) β r ( μ )
f s (λ )
⋅K = s
rr ( μ )
rr ( μ )
β r (μ )
where fs(λ) is the fluorescent flux per unit emission band pass dλ from the sample at
emission wavelength λ, rr(μ) is reflectance from a white non-fluorescent reference
standard at excitation wavelength μ when irradiated and viewed exactly as the
fluorescent sample, βr(μ) is the known radiance factor of the reference standard and K is
a correction factor taking into account the possible differences in measurement
geometry when measuring the reflectance and the known radiance factors of the
Another important characterization quantity of fluorescent material is the spectral
fluorescent yield γ(μ). It is defined as the ratio of emitted fluorescent power from the
sample to the incident optical power absorbed by the sample and it can be calculated
with equation (2).
f s (λ )
p f ( λ ) dλ
p f ( λ ) dλ
EMS (λ )
γ (μ ) =
pa (μ )
pi ( μ ) − p r ( μ ) I ref ( μ )
rs ( μ )
Rref ( μ ) R EMS ( μ )
where pf(λ) is the fluorescent power at emission wavelength λ when irradiated at
excitation wavelength μ, pa(μ) is the absorbed power at excitation wavelength μ, pi(μ) is
the incident power on the sample and pr(μ) is the reflected power from the sample.
REMS(λ) and Rref(μ) are the responsivity of the emission spectrometer used to measure
fluorescent signal fs(λ) and reflectance rs(μ) from the sample and the responsivity of the
reference detector used to measure full intensity Iref(μ).
Measurement system
The schematic of the measurement facility is presented in Fig. 1. The facility comprises
a light source unit, a sample holder unit and a detection system. It has been integrated
inside a gonioreflectometer built at the Metrology Research Institute. A careful
description of the gonioreflectometer can be found in reference [2].
A. Light source unit
The light source unit comprises of two lamps, a diffraction-grating monochromator and
reflecting optics for imaging and steering the beam. A Xenon arc lamp is used between
250 nm and approximately 460 nm and a quartz-tungsten-halogen lamp is used at
wavelengths longer than ~460 nm. The monochromator is a double grating
monochromator which uses toroidal mirrors in order to minimize astigmatism. The
grating turrets can hold three gratings: a UV, a visible and a NIR grating. The light from
the lamp is focused on the input slit of the monochromator and a small aperture (2.5 mm
of diameter) is used to limit the beam after the output slit. Order sorting filters are used
to cut out the second order diffractions, an off-axis parabolic mirror collimates the beam
and a flat mirror steers the beam into the optical axis of the measurement system. Sheet
polarizes are used to polarize the beam and two irises are placed on the beam path to
produce a uniform circular illumination of ~13 mm diameter at the sample surface.
Fig. 1. Schematic of the goniofluorometer setup: OSF, order sorting filter; A, aperture; OPM,
off-axis parabolic mirror; M, flat mirror; BS, beam splitter; MD, monitor detector; EMC,
emission monochromator; CCD, CCD detector.
B. Sample holder unit
The sample holder unit comprises of two sample holders and some empty space for
measuring full intensity. The two holders are called the reference holder (for a white
non-fluorescent reflectance standard) and the sample holder (for the fluorescent
sample). Both holders are positioned on a sample turn table which can be used to select
the illumination angle. The sample turn table is positioned on a linear translator which
can be used to select which holder is illuminated or to remove both holders out of the
beam. The linear translator in turn sits on a table above a detector turn table which is
used to rotate the detection system. The axis of rotation of both turn tables is parallel to
the plane of the reference sample surface when the reference is brought into the beam
path. In addition, the rotation axes of the turn tables are orthogonal to the optical axis of
the system. The holder for the fluorescent sample is positioned on a small linear
translator which is used to push the sample back to the rotation axis of the detector turn
table if the sample turn table is used to change the illumination angle.
C. Detection system
The detection system comprises of a reference detector, a light collector, a light guide
and an emission spectrometer. The reference detector is used to measure full intensity of
the beam and for the responsivity calibration of the emission spectrometer. The
responsivity of the reference detector has been measured at the Metrology Research
Institute. The detector is positioned on a cantilever on the detector turn table and used at
0° incidence. The light collector comprises a 35 mm aperture and a lens (f/# = 2) after
the aperture. The lens focuses emitted light incident on the aperture into a fiber output.
The collector is mounted on the same cantilever as the reference detector but at the
opposite side of the sample holder unit. When the reference detector is facing the
incident beam at 0° incidence, the collector faces the sample at 10° relative to the
incoming beam. The detector turn table can be used to change the viewing angle of the
collector between 10° – 90°. A UV-VIS light guide with wavelength range of about
250 – 830 nm and inner diameter of 4 mm is mounted into the light collector output.
The other end of the light guide has the shape of a slit (10 mm x 1.25 mm) and is
mounted on the input of the emission spectrometer producing uniform illumination of
the input slit.
The emission spectrometer comprises a single grating monochromator and a CCD
detector (DV420A-OE). The grating turret of the emission monochromator (EMC) can
hold two gratings. A grating with blaze wavelength at 300 nm and dispersion of
8 nm/mm is used. The CCD surface is attached at the focal point of the output of the
monochromator and the output slit is removed. The CCD detector views a 200 nm wide
spectrum, but only 100 nm can be used reliably because of angular dispersion of the
monochromator. The wavelength calibration of the emission spectrometer is done
separately for each EMC position by setting the EMC wavelength, scanning the
excitation monochromator (the double monochromator in the source system) and
recording corresponding pixel positions. The responsivity calibration is done by
measuring full intensity at the reference detector, for which the responsivity is known,
and comparing it to the signal from the emission spectrometer when illuminating a
known reflectance standard. Both calibrations are done separately for each measurement
D. Measurement procedure
In the beginning of the measurement both the excitation and the emission
monocromator are set at the same wavelength. The linear translator is used to move the
sample and reference out of the beam and full intensity is measured by the reference
detector and dark is measured for the CCD. Then the reference and the sample are in
turn moved into the beam, dark is measured for the reference detector and reflectance is
measured from the sample and the reference by the emission spectrometer. After this the
emission monochromator is set 100 nm above the excitation wavelength and emitted
signal is measured from the sample. The emission monochromator is used to scan the
whole emission spectrum with 100 nm steps. This procedure is repeated for all desired
excitation wavelengths.
Fig. 2 presents an example of calculated bispectral luminescent radiance factors βLλ(μ)
for a Spectralon fluorescent standard. The bispectral luminescent radiance factors are
calculated with equation (1). The
required quantities: fluorescent flux fs(λ) and
reflectance from the reference rr(μ) are acquired from the measurement as explained in
the previous section. The fluxes in the equation are corrected by the responsivity of the
emission spectrometer. The radiance factors for the reference are measured with the
gonioreflectometer [2].
Fig. 2. Calculated bispectral luminescent radiance factors per emission band pass of 5 nm for
a Spectralon fluorescent standard.
A goniofluorometer capable of measuring bispectral luminescent radiance factors and
spectral fluorescent yield has been built at the Metrology Research Institute, TKK. The
measurement range is 250 – 830 nm and the measurement facility can be used to
measure solid samples of about 30 – 70 mm diameter in various measurement
Silja Holopainen appreciates the support of TES and KAUTE foundations.
J. C. Zwinkels, F. Gauthier, “Instrumentation, standards, and procedures used at the National
Research Council of Canada for high-accuracy fluorescence measurements”. Analytica Chimica
Acta, 380, p. 193-209, 1999
S. Nevas, F. Manoocheri and E. Ikonen, “Gonioreflectometer for measuring spectral diffuse
reflectance”. Applied Optics, 35, p. 6391-6399, 2004
Extension of the NRC Reference Spectrofluorimeter to Volume Fluorescence
Joanne C. Zwinkels* and François Gauthier
National Research Council of Canada, Institute for National Measurement Standards,
Ottawa, Ontario, Canada, K1A 0R6
Corresponding author
The Reference Spectrofluorimeter developed at the National Research Council of
Canada (NRC) has been modified to enable volume fluorescence measurements of
liquid samples. This work was motivated by NRC’s participation in a recent
exploratory fluorescence study organized by the Federal Institute for Materials Research
and Testing (BAM) and the National Institute for Standards and Technology (NIST) to
assess the state-of-the-art accuracy in correcting emission spectra of volume
fluorescence standards.
The NRC spectrofluorimeter is based on the twomonochromator method and was originally designed for surface fluorescence
measurements of opaque materials, such as paper, plastics, paint and textiles, in
accordance with CIE and ASTM colorimetry standards. The instrument has been wellcharacterized for this application and has been providing high-accuracy total radiance
factor calibrations of fluorescent reflecting materials for almost ten years. There were
several challenges in measuring fluorescent liquid samples, including problems of nonstandard instrument geometry and low signal-to-noise. The paper discusses these
challenges and describes, in detail, the changes that were made to the instrument
operating procedures to improve measurement sensitivity by more than a factor of 20.
Representative measurements on candidate fluorescent dye standards are reported, as
well as a preliminary estimate of the uncertainty of measurements on the NRC
spectrofluorimeter for this new application. Preliminary results show that the modified
instrument is in good agreement with state-of-the-art fluorimeters designed specifically
for volume fluorescence applications.
colorimetry, surface fluorescence, reflection
fluorimetry, volume fluorescence, calibration standards
In the early 1990s, the National Research Council of Canada (NRC) Institute for
National Measurement Standards developed a reference spectrofluorimeter for highaccuracy measurement of surface fluorescent materials [1,2]. This instrument is based
on the two-monochromator method using a well-characterized monochromator in each
of the excitation and detections paths. In this arrangement, the first monochromator is
used to irradiate the fluorescent sample with monochromatic irradiation and the second
monochromator is used to independently analyze the reflected and emitted radiation
leaving the sample; thus it allows a complete separation of the reflected and fluorescent
components. The measurement geometry is 45 degree annular illumination, normal
viewing in accordance with International Commission on Illumination (CIE) [3] and
American Society of Testing and Materials (ASTM) [4] colorimetric geometric
specifications and thus provides an accurate measurement of surface colour, i.e. the
colour appearance of fluorescent reflecting materials for specified illuminant conditions.
The instrument uses a 300 W continuum xenon source and a thermoelectrically cooled
InGaAs PMT for a wide spectral range of operation from 200 nm to 1040 nm and its
optical design has been optimized for 5 nm bandpass for both excitation and emission
monochromators. It has been fully characterized for sources of uncertainty and has
been providing calibration services for surface reflecting samples for the past 10 years.
The types of fluorescent samples that are routinely measured include fluorescently
whitened paper and textiles, fluorescently brightened plastics and paints used in safety
goods and, more recently, fluorescent semiconductor powders for applications in solid
state light sources. The measured quantities are usually the reflected and total spectral
radiance factors for specified illuminant conditions that include the CIE standard
illuminants but also customer-specified illuminant conditions. These facilities are also
used for the calibration of spectral quantum yields. An optical schematic of the NRC
Reference Spectrofluorimeter is shown in Fig. 1.
Fig. 1. Optical schematic of NRC Reference Spectrofluorimeter designed for 45/0 geometry
measurements of surface colours.
A Glan-Taylor polarizing prism in the excitation path produces a known state of
polarization and the samples and reference standard are mounted on precision x-y
translation stages on a computer-controlled rotary table. The reflected radiance factor is
determined by setting both monochromators to the same wavelength setting and
scanning synchronously the entire spectrum. The luminescent radiance factor is
determined by setting the first monochromator (excitation monochromator) to a
wavelength μ in the excitation band of the fluorescent material, and by scanning the
second monochromator (emission monochromator) through all wavelengths λ in the
emission spectrum. This process is repeated for each incident wavelength in the
spectrum. The result is a two-dimensional array of bispectral luminescent radiance
factor data, βLλ (μ) shown in Fig. 2, which have been corrected for the spectral and
polarization effects of the instrument and so are source-independent.
Fig. 2. Corrected bispectral luminescent radiance factor data source-independent) for a red fluorescent
paint sample measured on the NRC Reference Spectrofluorimeter.
The work in extending the capabilities of the Reference Spectrofluorimeter to volume
fluorescence measurements was motivated by a desire to participate in a pilot
comparison of fluorescence measurements. The key objectives of this pilot study were
to determine the state-of-the-art accuracies in spectral correction of emission spectra
and to compare different procedures in the calibration of fluorescence measuring
instruments i.e. comparing a physical standards approach and the use of dye-based
fluorescent standards. This pilot study was notable in this it was to be the first
comparison of national measurement institutes (NMIs) on spectral emission correction
linking radiometry/colorimetry and conventional fluorimetry. The final participants of
this pilot study were four different NMIs, NIST and BAM Analytical Chemistry
Groups, who served as the co-pilots, and PTB and NRC. Since NRC uses a different
measurement geometry (45/0) compared with the other participants that use a 0/90
geometry, this comparison would also provide valuable information on the influence of
measurement geometry for these fluorescence standards.
Experimental Procedures
The calibration of the instrument involves both spectrophotometric and radiometric
calibration procedures. The spectrophotometric procedures include wavelength and
photometric scale, stray light and degree of polarization; the radiometric calibration
procedures include spectral irradiance of the excitation unit, spectral responsivity of the
emission unit and characterization of the instrument slit scattering function. The
calibration procedures that are used at NRC for measuring total spectral radiance factors
of surface colours were described in detail at the Oxford III conference so the focus in
this paper is a description of those procedures that have been modified to enable
extension of the instrument to volume fluorescence measurements. The modified
procedures include a more thorough characterization of the wavelength and photometric
scales, the optimization of the detector sensitivity, a reduction in the system noise and
an increase in spectral resolution.
The wavelength scale of the excitation and emission unit systems of the instrument for
both sets of gratings were calibrated independently using 22 spectral lines of Hg, Cs, He
and Cd spectrum lamps to cover the wavelength range 200 nm to 1040 nm. For
calibration of the excitation unit, the monitor detector was used to record the data,
whereas for the calibration of the emission unit, the spectrum lamp was mounted on the
sample stage and the photomultiplier was used as the analyzing detector. Fig. 3 shows
representative results for the calibration of the emission unit with Grating B. The dots
indicate the measured wavelength versus the true wavelength fitted with a least-squares
line. The magenta points indicate the residual differences from this first order fit which
is better than 0.1 nm for wavelengths up to about 650 nm but increases to 0.3 nm for
longer wavelengths. These residuals were then curve-fitted with a polynomial function
to give a second-order correction and the yellow dots shown here show the residuals
after this second order correction. It can be seen that the average wavelength error is
now 0.05 nm with a maximum error of 0.1 nm over the entire spectral range of
Fig. 3. Wavelength calibration of emission unit of Reference Spectrofluorimeter with Grating A and
residual differences the after first order (magenta) and second order (yellow) correction
The analyzing detector on the Reference Spectrofluorimeter is a thermoelectricallycooled InGaAs PMT (Hamamatsu R6872) which is situated behind a ground Suprasil
diffuser. The linearity of the detector plus diffuser system was characterized using the
NRC-designed high precision variable aperture device which is based on the double
aperture method [5]. This result is shown in Fig. 4, which shows that the linearity of
this PMT is better than 0.2% over 4.5 decades corresponding to the photocurrent range
of use of this device for both surface and volume fluorescence and reflectance
Fig. 4. Measured non-linearity and reproducibility of Hamamatsu R6872 PMT plus quartz diffuser used
on NRC Reference Spectrofluorimeter.
The traceability of the NRC Reference Spectrofluorimeter for total spectral radiance
factor measurements of surface colours is based on NRC physical transfer standards.
These include detector-based standards of spectral responsivity traceable to the NRC
cryogenic radiometer, source-based standards of spectral irradiance traceable to NRC
primary lamps and material-based standards of spectral radiance factor of a white nonfluorescent tablet of pressed PTFE that is traceable to NIST. These transfer standards
are used to calibrate the excitation unit, the emission unit and the instrument slit
scattering function. For this pilot study, the critical instrumental calibration is the
relative spectral responsivity of the emission unit.
The experimental procedure for calibrating the emission unit of the NRC Reference
Spectrofluorimeter involves recording the analyzing detector signal, isw(λ)for a source
of known spectral radiance at the sample position. In our case, the calibrated spectral
radiance source consists of a lambertian reflecting diffuser of known spectral radiance
factor, βstd(λ) which is illuminated at 45° incidence by a spectral irradiance lamp/plane
mirror combination. The spectral irradiance lamp is a 200 W quartz-halogen lamp,
which is mounted in a housing with a baffle tube limiting the source aperture and the
incident light beam is reflected by a plane mirror turned 22.5° from the normal position.
This lamp/ mirror arrangement is fixed on a kinematic mount and has been calibrated as
a unit for spectral irradiance, giving Eref(λ). This source unit is then mounted in the
sample compartment and aligned to give 45° incidence at the sample position. A mask
measuring 10x10 mm is centered at the diffuser position, so that the area of the
calibration source being measured is exactly the same as that for the fluorescent sample.
The spectral responsivity of the detection unit is given by:
Rsys (λ ) =
isw (λ )
k2 Eref (λ ) β std (λ )
where k2 is an instrument-specific constant dependent on the bandpass of the detection
unit. The spectral responsivity of the detection unit depends on the spectral
transmittance of the detection optics including monochromator, and the spectral
responsivity of the analyzing detector. The accuracy of the spectral responsivity
determination is dependent on the transfer uncertainties in the calibration of the standard
lamp and reflectance standards and the uncertainties in the measurement of the
photocurrent and the bandpass-dependent instrument function. The measured emission
spectra are corrected by the corresponding spectral responsivity of the detection unit to
obtain spectrally-corrected fluorescence spectra. Thus, the key to extending the NRC
Spectrofluorimeter to the measurement of weak forward-scattered fluorescent signals
from liquid samples was to reduce the experimental uncertainties in the spectral
responsivity calibration.
The artifacts for this comparison were a set of seven fluorescent dyes at two different
concentrations supplied by BAM , a sample of NIST SRM 936a quinine sulphate
dihydrate in perchloric acid solution, along with solvents for blank determinations. The
participating labs were responsible for providing suitable fused silica sample cuvettes.
Five of the BAM fluorescent dyes, designated A, B, C, D, and E, have broadly
overlapping emission bands, and were also being evaluated by BAM for their suitability
as fluorescent chemical transfer standards for fluorimeter calibration. The two other
BAM dyes, designated X and Y, and the NIST SRM, designated QS, were being used to
check the details of the lab’s emission correction functions; in particular, the
wavelength and spectral bandpass correction. The details of this technical protocol are
given in Table 1 including the wavelengths of excitation, λex, and the spectral range of
the emission bands. It was recommended to use a 5 nm bandpass for both the excitation
and emission monochromator and a sampling interval of 1 or 2 nm.
Table 1. Technical protocol for pilot fluorescence study
Dyes for checking
the emission
correction function
λex/ nm
Emission Range/ nm
300 - 450
330 – 530
390 - 600
450 - 700
570 - 760
HClO4 (aq)
295 - 460
375 - 675
490 - 780
The measurement protocol specified that each sample and associated solvent were to be
measured three times each on a given day using the same cuvette for each dye-solvent
pair. The raw emission data, both with and without background correction, and the
spectrally-corrected emission data using the laboratory’s normal spectral correction
procedures, were to be submitted to the pilot labs.
This pilot study and technical protocol posed a number of challenges for carrying out
these measurements on the NRC Reference Spectrofluorimeter. Firstly, its sample
compartment was designed to handle solid samples with a 45/0 geometry, not liquid
samples with a 0/90 geometry. Secondly, the signals from these fluorescent liquid
samples are much weaker than those typically encountered when measuring fluorescent
surface colours so measurement sensitivity was a critical issue. Finally, the comparison
protocol specified a measurement interval of 1 or 2 nm. This is much narrower than
that required for colorimetric applications, where 5 nm is typically used.
We first addressed the latter challenge. Since our previous calibration of the spectral
responsivity of the excitation and emission units had been carried out only at 5 nm
intervals, this calibration was repeated at a 2 nm interval for both Grating A and Grating
B. It was found that this re-calibration was in excellent agreement with the previous
results with a relative change of less than 0.04% per year.
Fig. 5 poignantly illustrates the magnitude of the challenges that we faced when using
our standard procedures for measuring these liquid samples. In preparation for this
comparison, representative samples of the NIST SRM 936 (quinine sulphate dihydrate
solution) were sent to NRC in order to evaluate our measurement sensitivity. As can be
seen from this figure, the initial results on this sample were less than spectacular with a
signal-to-noise of about 5:1. However, the spectrally-corrected emission spectrum
using our standard procedure for calibrating the emission unit appears to be in
reasonably good agreement with the expected spectral shape for this reference material.
With this encouraging result, we began a systematic optimization of our experimental
conditions to improve measurement sensitivity, starting with optimizing the system
throughput, decreasing the sources of noise and stray light, and then enhancing the
measurement detectivity. For ease of comparison, the net effect of these optimization
procedures is shown in Fig. 6.
Fig. 5. The initial measurement of quinine sulphate dihydrate using routine experimental
conditions for surface fluorescence measurements. The S/N is about 5:1
Fig. 6. The repeat measurement of quinine sulphate dihydrate after optimizing experimental
conditions of NRC Spectrofluorimeter (see text). The S/N has increased to about 200:1
The first choice was selecting an appropriate cuvette for these measurements. Because
of the 45/0 geometry of the NRC Spectrofluorimeter, it was not possible to use standard
rectangular cuvettes. To enable volume fluorescence measurements with this
instrument, we could use cylindrical or tubular-shaped cuvettes. Two different types of
Hellma fluorescence cells were evaluated for this purpose: a cylindrical cell of 22 mm
path length and a tubular cell of 10 mm path length. It was found that the tubular cell
consistently gave a better sensitivity, presumably because the measurement geometry
was closer to the ideal 90/0 scattering geometry. It was also found that the position of
the cuvette in the beam had a significant effect on the measured signal level. By trialand-error, a maximum signal level was achieved with the tubular cuvette positioned
with the front surface recessed by 4 mm with respect to the reflectance plane of the
usual PTFE reflectance standard. It was possible to reproduce this optimum sample
position to within 0.04 mm by using our usual diode laser alignment procedure that had
been developed for measuring opaque reflecting samples with the Reference
To increase the measured detector signal, the high voltage (HV) setting of the PMT was
optimized. For surface fluorescence measurements, the HV setting of the PMT is
typically set at 1000 V. However, this PMT has a maximum recommended operating
HV specification of 1250 V so we began systematically increasing the HV setting and
monitoring the change in signal. It was found that a HV setting of 1200 V increased the
signal by more than a factor of 3. It was verified that the spectral responsivity
calibration of the emission unit was not affected by this change in HV setting.
For surface fluorescence measurements on the NRC spectrofluorimeter, a Glan-Taylor
polarizing prism is placed in the incident beam to obtain a well-defined state of
polarization and the monitor detector is calibrated for s- and p-polarized conditions.
The luminescent and reflected sample data are then calculated by averaging the results
for the two polarizations. However, for this pilot study, the fluorescent samples were
known to have fairly isotropic emission so it was decided that the measurements could
be performed without polarizers in the beam. The elimination of these polarizers helped
to increase the signal by another factor of 2 to 3. The only caveat is that the
wavelengths selected for the bandpass normalization procedure need to be far removed
from the wavelengths of the Wood’s anomalies for this grating instrument to avoid
increased sensitivity to wavelength errors. The instrument degree of polarization of the
NRC Spectrofluorimeter has been well-characterized and the positions of these Wood
anomalies occur near 450 nm and 700 nm so the uncertainty due to polarization is
greatest near these wavelengths. The spectral correction of the comparison samples for
this pilot study was carried out using the calculated average of the s- and p-polarized
spectral responsivity calibration data of the emission unit with the selected grating.
To further enhance the measurement sensitivity, possible noise sources were
investigated. One source of noise was found to be due to ground loop noise. This was
easily eliminated by using a new grounded cable from the PMT amplifier to the DAC
board. By monitoring the time dependence of the PMT sample and dark signals, it was
also found that both these signals had a periodic structure. This effect is illustrated in
Figure 7. The PMT is normally operated so that it is cooled at –15 °C and the cooler
cycles on and off to maintain this temperature. It was noted that the period of this
structure was consistent with the cycling period of the thermoelectric (TE) cooler. This
periodic noise structure was effectively eliminated by forcing the TE cooler to remain
on, i.e. by setting the operating temperature to its maximum value of –20 °C. In this
mode of operation, the cooler did not cycle on and off. This result is shown here in the
red trace of Fig. 7.
Fig.7. Measured signal and noise of the analyzing PMT as a function of the operation of the
thermoelectric cooler. Blue trace is with the cooler (-15 °C) cycling on and off. Red trace is
with the cooler always on (-20 °C).
Results and Discussion
The preceding discussion describes the modified procedures that were developed for the
NRC Reference Spectrofluorimeter to improve its S/N performance for the volume
fluorescence measurements of this pilot study. Fig. 6 illustrates the impact of these
modified procedures for the measurement of the quinine sulphate dihydrate solution
The blue curve is the raw measured data which is has a sensitivity of better than 100:1
or an improvement of more than a factor of 20 from the preliminary measurements (see
Fig. 5). The red curve shows the spectrally corrected emission curve after applying the
standard NRC calibration procedures. With this highly satisfactory result, the NRC
measurements were then carried out on the set of 5 calibrant dyes and 3 unknowns for
this pilot study.
During the course of these measurements, it was found that the Dye A and X samples,
which are excited in the UV region near 280 nm, exhibited a very large signal;
apparently coming from the solvent. However, the ethanol solvent, that was common to
all of the dye samples, had not produced this strong emission when excited at longer
wavelengths. A search of the literature found no published reports of UV-excited
fluorescence in ethanol, so the sample cuvette was emptied and the measurements
repeated. This result showed that that the majority of the emission observed at longer
wavelengths was due to the cuvette itself. Fortunately, this feature was largely
eliminated by subtracting the blank spectrum from the dye-plus-solvent measurement.
As mentioned earlier, the objectives of this pilot study were to compare the
measurement capabilities of the participating NMIs using the lab’s routine calibration
procedures using physical transfer standards but also to investigate the suitability of the
set of five fluorescent dyes developed by BAM for use as fluorescence standards for
instrument calibration. When the pilot labs analyzed the NRC data, it was found that
there was good agreement with the other standardizing laboratories using the lab’s
routine calibration procedures. However, it was found that when using the dye-based
calibration procedure, good agreement was obtained for all samples, with the exception
of the quinine sulphate dihydrate. In the case of this sample, an anomalous feature was
observed in the corrected emission spectrum in the vicinity of 470 nm. According to the
dye based calibration procedure used by BAM for the quinine sulphate (QS) sample, the
spectral correction function is largely derived from the fluorescence measurement data
for dye B. From Table 1, the spectral range of emission of this dye sample is from 330
nm to 530 nm whereas for the QS sample, the spectral range is from 375 nm to 675 nm.
For the NRC measurements for this pilot study, a common grating was not used for all
samples, but was selected for optimum S/N. Thus, for dye B, Grating A was used and
for the QS sample, Grating B was used.
Fig. 8a compares the uncorrected raw NRC data for dye B measured with Grating A
shown in blue. The dashed curve shows the calibrated spectral responsivity of the
emission unit with this particular grating and the red curve shows the corrected spectral
emission of dye B taking into account this instrument calibration function. Since dye B
and the QS spectra had been measured using different gratings, it was suspected that the
change in experimental spectral conditions caused the anomalous effect when using the
dye-based correction procedure.
To confirm this suspicion, a fresh sample of the dye B solution was sent by the pilot lab
to NRC and the spectral emission curve was measured using Grating B, to have
common experimental conditions with the previous NRC measurements of the QS
sample. These data are shown on Fig. 8b, where the blue curve again shows the raw
measured data, the dotted orange curve gives the calibrated spectral responsivity of the
emission unit with Grating B and the blue curve is the corrected spectral emission curve
taking account of the spectral characteristics of the emission unit. The corrected data
only go down to 350 nm which is the lower limit of the calibrated spectral responsivity
data for this particular grating. Comparison of the corrected spectral emission curves
using these two different gratings shows excellent overall agreement. This result also
provides independent validation of this NRC radiometric standard based method of
spectral correction. The increased noise level observed with the Grating B result is to
be expected because of its lower sensitivity in this spectral range. Since the dye B
spectral data is used in the dye-based correction method to determine the instrument
correction function from about 350 nm up to a cross-over wavelength of 420 nm, it
appears that the spectral distortion below 420 nm that is found in the QS dye-corrected
spectrum is due to the change in grating condition from the measurements on the
calibrant dye B to the unknown QS sample.
Fig. 8. Dye B measured with Grating A (8a) and Grating B (8b) on NRC Reference
Spectrofluorimeter comparing raw and corrected data. The dashed curves show the
calibrated spectral responsivity of the emission unit for these two different grating
Table 2 provides a preliminary estimate of the uncertainty of these NRC volume
fluorescence measurements. The main components are the measurement repeatability
which is largely limited by the sensitivity and noise of the NRC Reference
Spectrofluorimeter, its wavelength uncertainty, detector non-linearity, the uncertainty in
the spectral radiance standard used to calibrate the emission unit which is a combination
of a spectral irradiance lamp and PTFE reflector, and the uncertainty in transferring this
relative spectral correction function to measured emission data. It is estimated that the
expanded (k=2) uncertainty in these measurements varies from 1.2 to 6.4% dependent
upon wavelength. This estimate is consistent with the results of this pilot study which
showed a general relative standard deviation between the four NMIs of 4% for the
physical transfer based correction of the spectral emission spectra.
Table 2. Preliminary uncertainty budget for NRC measurements of volume fluorescence emission
Fig. 9 shows the final NRC results for this pilot study of volume fluorescence
measurements over the spectral range 350 nm to 800 nm using the routine NRC
physical transfer standard calibration procedure. These curves are the results for the five
calibrant dyes and three unknowns (see Table 1). In the case of the QS sample,
measurements were also performed at two different concentration levels (0.04 and
0.08A although all the other samples were measured at the higher concentration level).
The data from this comparison have been analyzed by the pilot labs and a preliminary
report assessing the fluorescence measurement capabilities has been prepared.
Comparison of the NRC corrected spectral emission curves with the corresponding
intercomparison reference function shows very good agreement. The comparability of
these results is particularly encouraging for two reasons. Firstly, it confirms that the
spectral emission curves of these candidate fluorescence standards are geometryindependent which is a desirable characteristic of a transfer fluorescent standard.
Secondly, it also provides convincing evidence that fluorescence instrumentation
designed in accordance with CIE and ASTM colorimetric standards is suitable for
performing conventional fluorimetric analysis of spectral emission curves.
Fig. 9. Final NRC results for pilot volume fluorescence study of state-of-the-art measurement
of emission spectra. For description of samples, see Table 1.
J.C. Zwinkels and F. Gauthier, “Instrumentation, standards, and procedures used at the National
Research Council of Canada for high-accuracy fluorescence measurements”. Analytica Chimica
Acta, 380, pp. 193-209, 1999.
J.C. Zwinkels, D.S. Gignac, M. Nevins, I. Powell, and A. Bewsher, “Design and testing of a twomonochromator reference spectrofluorimeter for high-accuracy total radiance factor
measurements”. Appl. Opt., 36, pp. 892-902, 1997.
CIE Publication 15:2004, Colorimetry, 3rd edition (Central Bureau of the CIE, Kegelgasse 27, A1033, Vienna, Austria, 2004).
ASTM Standards on Color and Appearance Measurement, 6th ed., (1916 Race St., Philadelphia, PA
J.C Zwinkels and D.S. Gignac, “Automated high precision variable aperture for spectrophotometer
linearity testing”. Appl. Opt., 30, pp. 1678-1687, 1991.
Standardization of Fluorescence Techniques: Where Do We Stand and What Do
We Need?
U. Resch-Genger, D. Pfeifer, K. Hoffmann, A. Hoffmann and C. Monte
Federal Institute for Materials Research and Testing, Richard-Willstätter-Straße 11,
12489 Berlin, Germany
The use of fluorescence techniques is been ever increasing in the life and material
sciences with new instrumentation and promising techniques quickly evolving. The
comparability of luminescence data across instruments is, however, hampered by
instrument-specific contributions to measured signals [1,2] that are time-dependent due
to aging of instrument components. Moreover, for unexperienced users, often complex
instrumentation favors erroneous results as multiple instrumental parameters influence
the quality and reproducibility of the acquired data [3]. To rule out instrumentation as
major source of variability and to improve the comparability of fluorescence data,
reliable, yet simple chemical and physical standards in combination with tested
protocols for instrument characterization and performance validation are required,
thereby also meeting the increasing desire for quantification from measurements of
fluorescence intensities [1-4]. This eventually provides the basis for the application of
fluorescence techniques in strongly regulated areas like e.g. medical diagnostics.
Here, easy-to-operate liquid and solid fluorescence standards are presented that enable
the determination of the spectral characteristics and long-term performance of different
types of fluorescence instruments like e.g. spectrofluorometers and confocal laser
scanning fluorescence microscopes [2,5] thereby linking fluorescence measurements to
radiometric units.
U. Resch-Genger, D. Pfeifer, C. Monte, W. Pilz, A. Hoffmann, M. Spieles, K. Rurack, J. Hollandt,
D. Taubert, B. Schönenberger , P. Nording, J. Fluoresc., 15, 325, 2005.
J. Hollandt, R. D. Taubert, J. Seidel, U. Resch-Genger, A. Gugg-Helminger, D. Pfeifer, C. Monte, W.
Pilz, J. Fluoresc., 15, 311, 2005.
U. Resch-Genger, K. Hoffmann, W. Nietfeld, A. Engel, J. Neukammer, R. Nitschke, B. Ebert, R.
Macdonald, J. Fluoresc., 15, 347, 2005.
L. Wang, A. K. Gaigalas, P. DeRose, G. W. Cramer, Biophotonics Int., 42, 2005.
K. Hoffmann, R. Nitschke, U.Resch-Genger, manuscript in preparation.
Spectral Regular Transmittance Scale at SPRING Singapore
LIU Yuanjie Liu and XU Gan
Standards Productivity and Innovation Board, National Metrology Centre (a Division
of SPRING Singapore), 1 Science Park Drive, #02-27, Singapore 118221
The scale of spectral regular transmittance at SPRING Singapore was established by
using a specially designed double-grating monochromator based reference spectrometer
with a tungsten ribbon lamp over the wavelength range of 350nm to 1100 nm. The
output monochromatic beam from the spectrometer of about 25mm x 35 mm in size
with a collimation angle of 0.013 radian is used as the input beam for the measurement.
The transmitted beam from the sample is collected by a silicon photodiode or a
photomultiplier through an integrating sphere. The spectrometer was fully characterised
for all major uncertainty components including reproducibility, non-linearity,
wavelength setting, stray light, beam displacement and other beam geometry related
uncertainties, drift, etc. Cascading (step-down) technique was employed for high optical
density samples for which a practical model for analysis of wavelength setting induced
uncertainty was developed. Samples with optical densities up to 5 D can be calibrated
with an expanded measurement uncertainty (k=2) ranging from 0.03% to 3% depending
on the wavelength and optical density of the sample.
Comparison of NRC and NIST Infrared Diffuse Reflectance Scales from 2 µm to
18 µm
Leonard M. Hanssen1 and Nelson L. Rowell2
Optical Technology Division, National Institute of Standards and Technology,
Gaithersburg, MD, USA 20899-8442
Institute for National Measurement Standards, National Research Council, Ottawa,
Ontario K1A 0R6, Canada
The National Institute of Standards and Technology (NIST) and the National Research
Council (NRC) realize scales independently for diffuse (directional-hemispherical)
reflectance in the infrared spectral range. Comparisons of these scales have recently
been completed over the spectral range of 2 µm to 18 µm, and the results are reported
here. The agreement was excellent, lying within the quadrature combined uncertainties
for the great majority of values measured and demonstrating the level of equivalence of
the NIST and NRC diffuse reflectance scales.
Effects of Aging and Degradations on Regular Transmittance of Interference
Antti Lamminpää1, Silja Holopainen1, Farshid Manoocheri1, and Erkki Ikonen1,2
Metrology Research Institute, Helsinki University of Technology (TKK), P.O. Box
3000, FI-02015 TKK, Finland
Centre for Metrology and Accreditation (MIKES), P.O. Box 9, FI-02151 Espoo,
We have studied the long-term stability of bandpass interference filters for the timeperiod of about ten years. The filters have nominal central-wavelengths varying from
313 nm to 900 nm. Our measurements indicate that degradation due to the aging can
affect regular spectral transmittance of the filters by adding a diffuse transmittance
component or modifying their absorption level. In a case of clearly visible degradations
on a filter surface (Fig. 1), an increase of about 3 degrees in the light gathering angle
can result in ~3 % relative change in the spectral transmittance. The study demonstrates
that the effects such as oxidation or penetration of moisture (Fig. 2) into inner layers of
interference filters can limit their long-term use in applications of filter radiometry.
Fig. 2. Visible degradations inside interference
filter. Filter is back illuminated using an
integrating sphere.
Fig. 1. Oxidized surface of a UV glass filter
cemented to an interference filter (passband
313 nm).
Development of a Multispectral Texture Measurement Facility for Use in an EU
Funded Study of the Naturalness of Surfaces
Montgomery Ruth1, Pointer Michael1, Goodman Teresa1, Harvey Andy2
National Physical Laboratory, Hampton Road, Teddington, TW11 0LW, UK.
School of Engineering and Physical Sciences, Heriot -Watt University.
The measurement of the appearance of a textured surface requires the precise
evaluation of a number of physical attributes, such as gloss, reflectance and
translucency, within a multi-angled goniometric system. Robust characterisation of
the illumination, optical and geometric set-up and of the detector characteristics is
required to ensure the results are validated and have a rigorous metrological basis.
The system described incorporates a multi-spectral imaging system, conceptually
related to a Lyot filter, with 28 discrete spectral bands in the range from 370 nm to
800 nm. This facilitates characterisation of the target surface in both a spectral and
spatial sense. The development of a series of algorithms tailored to the class of sample
studied, e.g. wood, stone, fabric etc., coupled to a consistent taxonomy will allow an
accurate description of natural and non-natural surfaces. The theoretical modelling
and tolerances of this spectrometer are presented, together with experimental results
demonstrating excellent agreement between the modelled spectral band profiles and
those actually achieved.
The characterisation of surface textures will provide input to an EU-funded study
measuring ‘naturalness’. The ‘Measurement of Naturalness’ or ‘MONAT’ project will
use the latest technology and scientific understanding in areas of metrology,
instrumentation, cognitive neuroscience, psychology and mathematical modelling to
give a cross-sector approach to understanding the relevant perceptual processes. The
results of an outline study into the perception of naturalness for samples of natural
and synthetic wood is discussed briefly in Section 4 and more comprehensively in [1].
Keywords: Lyot, spectral, goniometric, appearance, naturalness
Applications of multispectral texture measurements
Multispectral analysis of textured surfaces is important in many sectors, including
earth observation, recycling, paper manufacturing, mineralogy, medicine, biology,
cosmetic and other personal health-care industries. As an example, appearance is one
of the most critical parameters affecting consumer choice in the retail food industry.
Food processors are eager to develop systems that can accurately assess the quality or
freshness of their products, and thereby discard those products that fall short of their
quality control standards. Currently this job is done manually by factory workers and
is usually a totally subjective process. Accurate appearance measurement scales that
combine textural and spectral information would be a valuable aid in the assessment
of the freshness of fruit, vegetables and meat [2], the texture of food surfaces, and the
non-uniformity in the colouration of confectionary products (which is often a result of
In medical imaging, images recorded in a variety of spectral bands can give
information about the amount of oxygen in the blood and about blood perfusion in
various parts of the body, including the extremities or organs such as the eyes [3].
This can be important when monitoring the health of patients with conditions such as
heart disease, diabetes or blue-baby syndrome. Furthermore, in the field of
ophthalmology, an ability to measure haemoglobin oxygenation using retinal images
provides useful information for the diagnosis of conditions such as diabetic
retinopathy, age-related macular degeneration and glaucoma.
1.2 The EC MONAT project
The EU MONAT project is part funded under the ‘Measuring the Impossible’ Theme
of the Framework 6 New and Emerging Science and Technology (NEST) Programme.
The project focuses on the measurement of the perceived naturalness of surfaces: a
property of a material that is reasonably easy to define and that has minimal cultural
or gender bias. A key aspect of the study will be to develop a basic understanding of
how the human visual system (HVS) perceives and categorises natural or non-natural
surfaces. Many ‘data reduction’ processes are performed within the HVS in order to
make the transition from the physical response (nerve impulses generated when light
from the object falls on the retina) to the final visual perception and categorisation in
the brain. The HVS also copes transparently with factors such as changes in the
illumination, which alter the reflectance, colour and gloss, thus changing the
appearance of the surface. The development of automated instrumentation that is able
to interpret a scene not just in terms of its physical characteristics, but also in terms of
how it would be perceived and categorised by a human observer, is thus a complex
and challenging task. The MONAT project aims to address some of these issues by
seeking to identify some of the sensory and cognitive processes associated with the
perception of naturalness and by developing and using leading edge measurement
equipment to perform accurate physical measurements of the materials being studied.
1.3 Spectral measurements
Analysing surfaces and objects in terms of their spectral characteristics,
spectrometer, reveals information that would be otherwise inaccessible
broadband response functions of the human visual system. This has led
advances in many areas of science and technology, such as remote
astronomy, medicine, production and manufacturing [4-10].
using a
via the
to huge
Spectrometers use various methods to acquire spatial and spectral information
including: whiskbroom (zero-dimensional); pushbroom-scanning (one-dimensional
with FOV movement) [11-13]; staring (two-dimensional with stationary FOV) [14];
and windowing (two-dimensional with FOV movement) [15]. Staring instruments are
most common and employ a two–dimensional FOV, which remains fixed on the
object; the scene and image do not move with respect to one another. Spectral
discrimination is performed using tuneable filters or imaging interferometers,
including filter wheels, acousto-optic techniques, LCTF, IRIS, BIREFTIS or foveal
systems, [14, 16, 17, 18]. The advantages (Fellget, Jacquinot) and disadvantages of
Fourier Transform (FT) devices are discussed in [19].
Information on the spectral variation across a non-uniform surface can be obtained
using many of these approaches, by either spatial or spectral multiplexing of time
sequential images, although this limits the application to situations in which the data
can be reconstructed faster than the scene varies. To avoid this problem, the system
being developed at NPL uses a ‘snapshot’ approach, in which spatial and spectral
information is obtained simultaneously (see Section 2). This has the advantage that no
temporal mis-registration is introduced between capture of the spatial and spectral
information. The disadvantage is that there is the trade-off between the number of
simultaneously viewed spectral images and the spatial resolution of those images.
2. Instrument Theory
2.1 The LYOT filter
The multi-spectral texture measurement system being developed for the MONAT
project is based on a Lyot filter [9]. This was originally developed to enable
movements of the solar atmosphere to be observed and to allow photography of the
corona of the sun at times other than during a solar eclipse. It permits simultaneous
observation of different points of an extended luminous source in narrow spectral
regions, with minimal loss of light. The original device operated from 380 nm to
2000 nm, with an optimum resolution of 0.1nm for green light.
The Lyot filter is composed of multiple waveplates sandwiched between co-aligned
linear polarisers, which are aligned to pass light polarised at 45 ° to the fast axis of
each waveplate. The waveplates are cut parallel to the optic axis and are made from
birefringent material, which introduces a phase delay between the orthogonally
polarised components.
The relative phase difference introduced between the x and y components of the wave
energy, after traversing a waveplate of thickness d1, is given by:
Δφ =
( no − ne )
where λ0 is the primary wavelength, and no and ne are the refractive indices of the
ordinary and extraordinary waves. (no-ne) is referred to as the birefringence. Thus
each of the polariser-waveplate pairs results in a high intensity central maxima, with
low intensity side lobes. The transmission function for each polariser-waveplate pair
is proportional to:
⎛ n − no ⎞
cos 2 [d1π ⎜ e
⎟] .
⎝ λ ⎠
In the Lyot filter, each of the successive waveplates has a thickness which is double
that of the preceding one. As the number of waveplate-polariser pairs increases, so the
separation between the central maxima also increases. The system can be designed
such that only one bright maximum is formed when the system is used with a defined,
limited, input spectrum. Fig. 1 demonstrates the reduction in the FWHM and the
increase in the number of side lobes as the waveplate-polariser pairs is increased from
one to three.
Cos2 x
Cos2 xCos2 x
Cos2 xCos2 xCos2 x
Transmission %
Wavelength in microns
Fig. 1. Demonstration of polariser retarder pairs in a Lyot filter; increase in number of
pairs reduces FWHM of cyclic maxima.
2.2 The IRIS system of Harvey and Fletcher-Holmes
Harvey and Fletcher-Holmes [14] further developed the Lyot filter to produce an
Image Replication Imaging Spectrometer or IRIS system (Fig. 2). They replaced the
polarisers of the Lyot filter with Wollaston polarising beam splitters, which introduce
a spatial separation of the input image, creating pairs of replicated images (in all other
respects, the theory and physical mechanism of the spectral discrimination remain
2N spectral
images at
Exploded view of
Wollaston prisms
wave plates
cos 2 (πνΔ)
cos 2 (2πνΔ)
cos 2 (4πνΔ)
sin 2 (πνΔ)
sin 2 (2πνΔ)
sin 2 (4πνΔ)
Fig. 2. Image Replication Imaging Spectrometer (IRI S) with filter wheel addition.
Harvey and Fletcher-Holmes also deviated from the two times thickness ratio
described by Lyot. This results in a breakdown in the cyclic nature of the system, but
since they only required eight spectral bands, within a region prescribed by their
passsband filters, neighbouring maxima were irrelevant. Instead they wished to
orthogonalise these eight peak bands, additionally maximising the maxima and
minimising the side lobes.
2.3 The NPL system
The system being developed at NPL uses Wollaston polarising beam splitters but,
unlike the Harvey and Fletcher-Holmes system, in this case the two times thickness
ratio of the waveplates has been re-instated. The system uses three waveplate-prism
pairs, generating transmittance profiles with the general form:
⎛ n − no ⎞
⎛ n − no ⎞
⎛ n − no ⎞
T1 = cos 2 [d1π ⎜ e
⎟] cos 2 [d 2π ⎜ e
⎟] cos 2 [ d 3π ⎜ e
⎝ λ ⎠
⎝ λ ⎠
⎝ λ ⎠
Thus a series of cyclically-repeating transmittance profiles is generated, as described
in the original Lyot paper (note the ‘folded over’ rather than ‘sequentially repeated’
characteristic – see Fig. 3) but, unlike Lyot, each of these is spatially separated to
form eight distinct images. Thus, for example, the transmittance curve represented by
the blue line in Fig. 3 forms a separate image to that represented by the yellow line,
and so on. Each of the images contains several transmittance peaks but, by
introducing cut-on and cut-off filters, the NPL system allows blocks of eight peaks
(one in each image) to be isolated sequentially.
Transmission %
Wavelength in microns
Cos Cos Sin
Cos Sin Sin
Cos Sin Cos
Sin Cos Cos
Sin Cos Sin
Fig. 3. Spectral transmission profiles of the NPL system using Quartz waveplates of
thickness 135, 270 and 540 microns.
The waveplate thicknesses in the NPL system have been chosen such that four sets of
transmittance maxima fall within the visible region of the spectrum. A double filter
wheel with specifically selected cut-on and cut-off filters is used to isolate, in
sequence, one of four contiguous spectral bands, each containing eight bell-shaped
maxima (one from each waveplate-polariser pair) - see Fig. 4. The system is designed
such that each successive region contains one peak from the previous one, to allow
normalisation between the regions.
Intensity value
Wavelength in Microns
0.5 Cos Cos Cos
Sin Sin Sin
LWF 500
Cos Cos Sin
Sin Sin Cos
SWF 600
Cos Sin Sin
0.5 Sin Cos Cos
Cos Sin Cos
Sin Cos Sin
Fig. 4. Transmission profiles of long and short wave pass filters overlaid on IRIS
spectral transmission profiles in the range 480-620 nm.
2.4 Illumination and filtering
The illumination source is a tungsten halogen lamp. This has a spectrum that is a
close approximation to a Planckian Black body radiator, with a colour temperature of
approximately 3100 K. It is of a type frequently used as a transfer standard for
spectral irradiance measurements [20].
Spectral filters used for metrological applications should ideally be:
• Spatially uniform
• Spectrally smoothly varying bandpass
• Stable with time
• Insensitive to temperature
• Insensitive to angle of incidence
• Fully opaque to wavelengths outside of desired band pass
In designing our filter system we have tried to optimise based on the above.
In order that a single image contains only one transmittance peak at any one time, a
series of calibrated cut-on/long-wave pass and cut-off /short-wave pass filters are
placed, in pairs, between the light source and the prisms. The chosen filters have a
hard dielectric coating and interference effects rather than absorption to isolate the
spectral bands. Any misalignment of the filters with respect to the normal to the
incident radiation will lead to a shift to shorter wavelengths; this is allowed for as part
of the calibration uncertainty budget.
2.5 Modelling of the waveplates
The system has been designed such that twenty-eight transmittance maxima are
contained within the spectral range from 370 nm to 800 nm. Many birefringent
materials are capable of achieving this, but the additional factors of availability, cost,
manufacturing tolerances and manipulation also needed to be considered as part of the
design. The spectral transmission profiles were therefore calculated for a range of
materials, using equation 3, to allow the optimum solution to be found. The modelling
involved varying the thicknesses of each of the three waveplates for each material, but
maintaining the two-times thickness ratio throughout.
⎛ n − no ⎞
⎛ n − no ⎞
⎛ n − no ⎞
T1 = cos 2 [d 1π ⎜ e
⎟] cos 2 [ d 2π ⎜ e
⎟] cos 2 [ d 3π ⎜ e
⎝ λ ⎠
⎝ λ ⎠
⎝ λ ⎠
⎛ n − no ⎞
⎛ n − no ⎞
⎛ n − no ⎞
T2 = cos 2 [ d 1π ⎜ e
⎟] cos 2 [ d 2π ⎜ e
⎟] sin 2 [ d 3π ⎜ e
⎝ λ ⎠
⎝ λ ⎠
⎝ λ ⎠
⎛ n − no ⎞
⎛ n − no ⎞
⎛ n − no ⎞
T3 = cos 2 [ d 1π ⎜ e
⎟] cos [ d 3π ⎜ e
⎟] sin [d 2π ⎜ e
⎝ λ ⎠
⎝ λ ⎠
⎝ λ ⎠
etc …….
The most up-to-date coefficient values were used; for Calcite and Quartz this was
from G. Ghosh, using Sellmeier equation 4 [21]. Calcite was discounted because the
4.95, 9.9 and 19.8-micron thickness required for the plates would have proved
extremely difficult to manipulate (it would have been possible to cement the
waveplates onto optical blanks to increase their thickness, but this was not the
preferred option since it was desirable to minimise the number components that would
be integrated into the final system). MgF and SrMoO4 were rejected due to lack of
availability. Finally 135, 270, 540 micron waveplates using quartz were chosen,
without the use of any substrate.
By altering the waveplate thicknesses from 135, 270 and 540 microns, adding and
subtracting 1 or 2 microns in either direction, tolerance graphs were plotted, as in Fig.
5. Introducing such a deviation from the strict two-times ratio clearly leads to a loss in
the inherent symmetry of the graphs. Several of the maxima lose energy, which is
redistributed to the side lobes, thereby reducing the dynamic range of the system. A
tolerance of ±0.5 microns was therefore set on the waveplate thickness. The flatness
of the waveplates was specified in terms of the wavefront distortion (λ/8 peak to
valley at 632.8 nm) and the parallelism was specified as 3 arc seconds, which equates
to a tolerance in the flatness of 0.291 microns for the 20 mm square waveplates used,
or λ/2 at 600 nm.
Transmission %
Wavelength in microns
Cos Cos Sin
Cos Cos Sin Tol 2
Cos Sin Sin
Cos Sin Sin Tol 2
Cos Sin Cos
Cos Sin Cos Tol 2
Sin Sin Sin
Sin Sin Sin Tol 2
Fig. 5. Effect of introducing a deviation in the waveplates from a strict two-times
thickness ratio. Two-micron tolerances are shown.
2.6 Zemax modelling of the prisms
The required dimensions for the prisms (thickness and angular cut) were modelled
using Zemax-EE optical design software. A separate configuration, based on tracing
an ordinary or extra-ordinary ray through the prisms, and use of Jones matrices [22],
was required for each of the transmission functions. Fig. 6 shows all eight
configurations, each highlighted with a different colour. Although prisms are typically
used as dispersive devices, the chromatic aberration of the images in this case was
kept to acceptable levels because (a) only a small range of wavelengths contained in
the spectral peak is transferred along any one path and (b) the system uses prismprism pairs, which goes some way to cancelling the dispersion. Inter-reflections
between the quartz and calcite sandwiched optics can significantly affect the quality
of the final images. Various methods are being considered in order to minimise these
effects, including the use of index-matching fluid and cementing the optics together.
Fig. 6. Zemax modelling of IRIS device (left) and predicted spatial distribution at the
image plane (right).
The calculation of the prism angles was performed such that the space-bandwidth
product was optimised i.e. to optimise the trade-off between the number of
simultaneously viewed spectral images and the resolution of those images. When an
increased image resolution is required and a reduction in the number of spectral bands
can be tolerated, then the system can be configured to isolate just four of the eight
2.7 The Camera
The camera chosen for use in the system is a SPOT, Insight Firewire 4 Megapixel
monochrome device, designed for use in machine vision, metrology and other
industrial applications. This has a 4-megapixel CCD with a 15.1x15.1 mm2 area,
which allows a wide field of view, coupled to high resolution. The 14 bit-depth in real
time capture mode is at the upper end of the required specification and allows image
3. Theory Versus Practical Results
An Ocean Optics S1000 fibre optic spectrometer was used to carry out preliminary
verification tests of the spectral transmission of the NPL IRIS system. A fibre-optic
light-pipe coupled to a diffuser was imaged through the polariser, prisms and
waveplates, forming an image in free space. The fibre-optic coupled to the
spectrometer was suspended in the image plane and the spectral content of the light
forming each of the eight images was recorded. The measurements were repeated at
several spatially separated points across each image. The results were normalised
against the source radiance and all values were dark corrected. Fig. 7 shows two
graphs showing the theoretical and measured individual transmission spectra.
Expt 5 Sin Cos Sin
Expt 5a Sin Cos Sin
Sin Cos Sin THEORY
Percentage transmission comparison
Wavelength nm
Expt 8 Cos Sin Cos
Expt 8a Cos Sin Cos
Cos Sin Cos THEORY
Percentage transmission comparison
Wavelength nm
Fig. 7. Agreement between theory and practical results for spectral transmission profiles.
Although these results have not been wavelength corrected and cannot be considered
a calibrated measurement, their fidelity with the theoretical profiles is very
encouraging. Any drift of wavelength position or loss of dynamic range due to
increased energy in the side lobes at the expense of the central maxima, as a result of
sensitivity to deviation from the two-times thickness ratio, lack of parallelism, or out
of tolerance thickness, has been avoided. It appears, therefore, that the manufacturing
of the optics has been well within specified tolerances.
As the next stage in the development of a system for texture measurements, this IRIS
device will be placed on the NPL’s Gonio Apparent Spectro Photometer (GASP)
stage, to allow full goniometric spatial and spectral characterisation of surfaces.
4. Naturalness Experiments
The results of a preliminary study in preparation for the EU MONAT project
(‘MONAT Zero’) are presented in [1]. The paper describes the visual and touch
assessment of twenty-two wood and wood-effect samples from a variety of different
manufacturers. The samples were mounted in identical sealed containers with a grey
surround and an 8x8 cm2 viewing window. Thirty-eight observers completed a series
of four timed experiments to place randomised samples in each of five categories:
Definitely natural
Probably natural
Probably not natural
Definitely not natural
A weighted scale was then used to assess each participant’s degree of success in
correctly categorising the samples. The results were analysed to identify any bias
based on time of day of experiment, age, gender, scientific background, educational
achievement etc. No such biases were observed, although the authors would like to
emphasise that the data set was rather small. The full MONAT study will involve a
much larger number of observers and a much wider range of experimental
Other experimental observations from the MONAT Zero study included:
It is important to impose a time restriction: observers should be allocated
enough time to make a judgement based on a first impression, but not so long
that an emotional response of preference is introduced.
It is essential to present all the samples to the observers in the same way and to
remove any additional cues, such as being able to see the edges or backing of
the samples or feel differences in their weights.
The understanding of the term ‘naturalness’ varied between the observers. This
highlights the importance of providing a clear taxonomy for the EU MONAT
Observers stated that the key visual factors influencing their decisions were
grain texture and pattern regularity. Several identified ‘randomness’ as an
important property of natural materials. The physical measurement and
modelling of these parameters will therefore form a strong element of the EU
MONAT project.
R. Montgomery, T. Goodman, M. Pointer and A. Harvey. “The measurement of multispectral
texture and its contribution to an EU funded study of the naturalness of surfaces”. CIE
Appearance Symposium 2006, Paris. (In press).
S. Kong, Y. Chen, I. Kim and M. Kim. “Analysis of hyperspectral fluorescence images for poultry
skin tumour inspection”. Applied optics, 43, No4, 2004
S. Ramachandran, N. Taylor, A. McNaught and A Harvey. “Modelling of light propagation in
retinal tissue”. SPIE, Vol. 5486, p 48-60, July 2004
D. Torr, M. Zukic, C. Feng, A. Ahmed, W. Swift. “Miniaturised high-resolution NUV-VIS-NIR
imaging spectrometer array for FAST SAT applications”. Proc. of SPIE, Vol. 2266, p206-220,
M. Descour, C. Volin, E. Dereniak, K. Thome, A. Schumaker, D. Wilson and P. Maker.
“Demonstration of a high speed non-scanning imaging spectrometer”. Optics Letters, Vol. 22,
No.16, 1997
J. Graham. “IFIRS: An Imaging Fourier transform spectrometer for the next generation space
telescope”. ArXiv:astro-ph/9910442v1, 1999
D. Naylor, B. Gom, I. Schofield, G. Tompkins and G. Davis. Mach-Zender “Fourier transform
spectrometer for astronomical spectroscopy at submillimeter wavelengths”. Proc. of SPIE, Vol.
4855, pp. 540-551, 2003
F. Grandmont, L. Drissen, G. Joncas. “Development of an imaging Fourier transform
spectrometer for astronomy”. Proc. of SPIE, 2002
B. Lyot. “Un Monochromateur EO champ utilisant les interference en lumiere polarisee”. Compt
Rend. Hebd. Seanc. Acad. Sci., Vol. 197, p1593-1595
J. Leroy. “Solc elements in Lyot-Ohman filters”. J. optics, Vol.1, No.5. pp293-304, 1980
P. Mouroulis, J. Shea and D. Thomas. “Design, tolerancing and alignment for pushbroom imaging
spectrometers for high response uniformity”. Proc. SPIE, Vol. 4441, pp106-117, 2001
Compact High Resolution Imaging Spectrometer. (CHRIS). Developed by Sira and ESA.
Airborne Visible/Infrared Imaging Spectrometer (AVIRIS).
A. Harvey and D. Fletcher -Holmes. “High throughput snapshot spectral imaging in two
dimensions”. Proc of SPIE, Vol. 4959, pp46-54, 2003
R. Sellar, G. Boreman, L. Kirkland. “Comparison of signal collection abilities of different classes
of imaging spectrometers”. Proceedings of SPIE, Vol. 4816, pp.389-396, 2002
D. Fletcher -Holmes and A. Harvey. “Real time imaging with a hyperspectral fovea”. J.Opt, A.
Pure Appl. Opt., 7, S298-S302, 2005
A. Harvey and D. Fletcher–Holmes. “Birefringent Fourier transform imaging spectrometer”.
Optics Express Vol. 12, No.22, 2004
R. Sellar, G. Boreman, L. Kirkland. Comparison of signal collection abilities of different classes
of imaging spectrometers. Proceedings of the SPIE, Vol. 4816, pp. 389-396. (2002).
E.R. Woolliams, N.P.Fox, M.G. Cox, P.M. Harris, N.J. Harrison, “Final report on CCPR K1-a:
Spectral irradiance from 250 nm to 2500 nm”. Metrologia, 43, Tech. Suppl., 02003, 2006.
Ghosh, “Dispersion-equation coefficients for the refractive index and birefringence of calcite and
quartz crystals”. Optics Communications, Vol. 163, pp. 95-102, 1999
Zemax optical design program user’s guide. Version 10.0.
The authors would like to thank Julie Taylor, Nigel Fox, Neil Harrison and Eric Usadi
of NPL for useful discussions relating to the development of the IRIS system.
© Crown copyright 2007. Reproduced by permission of the Controller of HMSO and Queen’s Printer
for Scotland.
This work was supported by the National Measurement System Policy Unit of the Department of Trade
and Industry.
Laser techniques for high accuracy spectrometric measurements of parallel-sided
Jessica Cheung, Eric Usadi and Christopher Chunnilall
National Physical Laboratory, Hampton Road, Teddington, TW11 0LW
A laser technique for measuring the reflectance of small cross-sections (~ 100 μm) of
samples with parallel-sided surfaces is described. The measurement of the reflectance or
transmittance of parallel-sided samples can suffer from inter-reflections. This paper
describes how a knife-edge can be used to remove interference effects and the
considerations that must be taken into account in order to determine the uncertainty of
the measurement. The overall uncertainty in the reflectance measurement of a low
reflectance (0.14%) surface is 0.014% (k=1).
The measurement of transmittance and reflectance of small samples can be very
challenging. By small we mean samples of cross-sectional dimensions < 1 cm and
thickness on the order of mm and less. Two factors that make such samples challenging
to characterise are:
Small beam cross-section required to characterise sample can lead to non-linear
effects in the detector;
Inter-reflections between parallel-sided surfaces.
The sample may be a thin window used on a cryostat, crystals, scintillators, or small
optical components such as lenses and filters. We describe the characterisation of a nonlinear crystal which generates correlated photons used in the measurement of the
quantum efficiency of photon counting detectors [1, 2]. Absorption in the non-linear
crystal itself can prevent the correlated photons from reaching the detectors. It is
therefore essential to measure these losses, which in turn require measurement of the
crystal transmittance and the reflectances at the crystal boundaries. There are the
complications that the two surfaces of the crystal have different anti-reflection coatings
and the correlated photons are produced at an angle to the crystal surface. The
requirement was to ascertain whether the loss due to the crystal could be measured with
an uncertainty of less than 0.1%.
One transmittance measurement technique that can be used to avoid the effects of interreflections is through the use of a mode-locked laser. Hartree et al. [3] demonstrated the
improved transmittance measurements of a glass plate (3 mm thick) using a modelocked laser with pulse width 300 μm long.
The correlated photon application requires measurement of each face of the crystal in
order to take account of the anti-reflection coatings of this particular crystal. This paper
describes an experimental set up to measure the transmittance and reflectance of such a
crystal. In particular, the paper demonstrates that it is possible to remove inter-reflection
effects through the use of a knife-edge. In this way, it was possible to measure the front
surface reflectance of each surface of the crystal.
Experimental details
Fig. 1. Schematic of measurement set up for reflectance and transmittance measurements
from BBO crystal.
Fig. 1 shows the experimental set up. The sample is a BBO (beta-barium borate) nonlinear crystal, see Fig. 2, polished face area (7 mm x 7 mm), thickness 5 mm. The
crystal has been anti-reflection coated on both surfaces S1 and S2.
S1: AR coat (Broad band: 345 – 370 nm)
S2: AR coat (Broad band: 345 – 370 nm and 680 – 800 nm)
Fig. 2. Photograph of crystal
The following sections describe the alignment of the laser, crystal, knife-edge and
detectors and the measurements necessary to assess the uncertainties of the system.
Ti-sapphire laser
A Ti-sapphire laser (Fig. 1) was used to produce high power monochromatic radiation
at 702.2 nm. The laser power was measured to have a standard deviation of 0.0026%
over a period of 1.5 hours. A beam splitter was used to pick off part of the beam to
monitor the wavelength. The Pockels cell was used to stabilise the intensity of the laser
using the feedback from the silicon photodiode that sits just after the spatial filter
assembly. The spatial filter was used to clean up the beam and the resultant image is an
Airy disc. The iris was opened up such that it only allows the bright central fringe to
pass through. A long focal length lens was then used to produce a beam waist of
100 μm diameter. A CCD camera mounted in place of the crystal was translated along
the optic axis to assess the size of the beam waist. The measurements indicated that the
beam waist was within the upper and lower limits set by the calculation below.
Laser beam waist
In order for the front and rear face reflections from the crystal to be spatially distinct,
the laser beam diameter must be smaller than the separation between these two reflected
beams. For a Gaussian beam with 1/e2 intensity radius of x, 99.85% of the total beam
energy is contained within a radius of 1.8 x, and 99.97% within a radius of 2.0 x.
Therefore to measure the single face reflectance to better than the 0.1% level requires
that each beam can be measured over a radius of 2x, and that therefore the separation, h,
between the beams needs to satisfy:
h ≥ 4x0
Here xo is the beam waist, i.e. the radius at the narrowest point in the beam. The value
of h is determined from simple ray tracing of a ray incident at a particular angle on a
parallel face crystal.
A further restriction comes from considering that the second face reflectance will
diverge into the path of the first face reflectance unless the Rayleigh range (i.e. the
distance from the waist over which the beam radius expands by 2 ) is longer than one
return trip of the beam through the crystal. If λ is the wavelength, n is the crystal
refractive index, and β is the angle of refraction within the crystal, then 2nL/cosβ is the
optical length of this return trip. If ZR is the Rayleigh range, then this condition requires
Z R = π ⋅ x0 λ ≥ 2nL cos β
Combining Equations (1) and (2) yields the acceptable beam waist range:
h 4 ≥ x0 ≥ 2nλ L / π cos β
In our setup, where the angle of incidence is 6° and no = 1.666, equation (3) requires the
beam waist to be between 61μm and 157μm. This shows that there is scope to trade off
beam size and divergence.
The crystal
The crystal was mounted on a linear stage on top of a rotation stage so that the first
surface was above the centre of the rotation of the rotation stage. The rotation stage and
crystal were turned through 90° and the linear stage was adjusted until the beam was
just skimming across the front face. This positioned the first surface above the centre of
rotation. Small rotations away from normal shows that the laser beam does not move
across the crystal surface, demonstrating that the first surface is indeed above the centre
of rotation. Another linear stage at the bottom of the whole assembly moves the crystal
in and out of the beam. All the measurements are carried out with the crystal rotated
such that the incident beam is at 6° to the normal (this is the angle of interest with
respect to the quantum efficiency measurement application described in the
Before mounting the trap detector for reflectance a CCD camera is first mounted in its
place. This is used to aid alignment of the knife-edge. As the crystal is tilted a small
amount of the transmitted beam will be further reflected at the second surface (see Fig.
3). It is possible to see both of these spots if the incident beam itself is around 100 μm
and the CCD camera is used to check beam size. The reflected spots diverge the further
away they are from the crystal and this will result in an interference pattern. The knifeedge is mounted on a motorised linear stage (resolution 1 μm) and set close to the
crystal surface where the spots are still seen to be separate. The knife-edge is used to
block the spot reflected off the second surface and the CCD is used to assess how far the
knife-edge must be moved in order to completely block the back reflection. The
motorised linear stage aids the alignment process. LabVIEW software that controls the
CCD and stage is used to collect the data. The stage is zeroed when the knife-edge
blocks the second beam. The knife-edge is a scalpel blade that has been sprayed with
matt black paint to reduce scatter effects.
knife-edge out
knife-edge in
Fig. 3. (a): Schematic of knife-edge measurements, (b) image of reflectance on CCD camera showing
interference between beams reflected from both surfaces of the crystal. As the knife edge moves in to
block the spot reflected from the furthest surface the fringes gradually disappear until it is totally
blocked, as the image shows is (c). (He-Ne light at 633 nm was used for (b) and (c) since the front
and back surface reflectances were similar, leading to high visibility fringes).
Mounting the trap detectors
The trap detectors were of the NPL three-element design using Hamamatsu S1337-1010
photodiodes. The trap detector was positioned such that the beam was sitting in its
plateau region. All trap detectors suffer a small amount of reflectance loss (~ 0.3% of
incident beam [4]). This reflected beam can aid alignment but once the trap is aligned so
that the beam is normal the trap must be tilted slightly away from normal so that the
reflected spot does not fall back on the sample. The trap detector should not be rotated
more than 5° from the normal position, but to be on the safe side, not more than 1°. This
is so that the reflected beam does not hit the trap housing. The trap detector can suffer
from small polarisation effects, depending on the trap assembly. However, so long as
the orientation does not change between measurements this does not matter. A small
score was made on the trap detector aluminium housing to register the orientation.
Correction factor between the two traps
Each trap detector and its trans-impedance amplifier were taken as a single unit. One
unit was set to measure the reflected beam and the other the incident and transmitted
beam. In order to determine the correction necessary between readings taken from these
two traps, each unit was set up to measure the incident beam (Io). Three sets of
consecutive data for each trap were taken, i.e. trap A would be mounted, 5
measurements made, then trap B would be mounted, 5 measurements taken, then trap A
mounted again and so forth. In both cases the data was corrected for background
All the data collected on the trap measuring reflectance were divided by this correction
factor. The correction factor (‘reflectance’ trap response ÷ ‘transmittance’ trap
response) is 1.0024 ± 0.0031. The uncertainty in this value incorporates the spatial nonuniformity of the trap detectors and the laser instability.
Trap uniformity
A trap detector should have a response non-uniformity of less than 0.02% for a 4 mm
diameter beam over ±2 mm about the centre [4]. In these measurements both the
reflected and transmitted beams had diameters of about 2 mm. The non-uniformity of
each trap was measured using the reflected and transmitted beams respectively.
Laser stability
The stability of the laser was measured using the one of the traps. Over one hour, a
periodic trend is clearly seen, with a period of 15 minutes. This coincides with the air
conditioning cycle in that lab. The measurement could be improved by building an
enclosure around the measurement area. Figure 4 shows that the standard deviation of
the data points is 0.000026. The measurement sequence for the crystal takes about 20
minutes in total.
Laser stability over 1 hour (corrected for background)
stdev = 0.000026
15 mins
DVM reading
time (minutes)
Fig. 4. Laser stability over 1 hour
Measurement procedure for crystal
Once aligned the following measurements were made:
Trap detector in the transmittance position:
I0 = Straight through beam
ITj = Transmittance measurement with crystal in the beam, laser incident on side j.
Trap detector in the reflectance position:
IR2tot = Reflectances off surfaces 2 and 1 of crystal (knife edge is out);
= Reflectance off surface 2 of crystal with knife edge blocking back reflection;
IR1tot = Reflectances off surfaces 1 and 2 of crystal (knife edge is out);
= Reflectance off surface 1 of crystal with knife edge blocking back reflection;
Five measurements, each being the average of 50 readings on the DVM were taken on
both traps. The knife-edge would then be moved in to block the second reflected spot
and another 5 measurements taken. This would be repeated. The crystal was also rotated
and realigned so that the reflectance from the surface 1 can be measured as well with the
knife-edge in and out of the beam. All measurements were dark corrected.
Ideally, the measurements should agree with equation 4:
I0 = IT j + IR j tot + A + S
where A is the absorption in the crystal and S accounts for any losses due to scatter at
each crystal surface and within the crystal which propagates outside the capture region
of the trap detectors.
The following equations were used to calculate the reflectance and transmittance. IR is
the measured reflected signal, It is the measured transmitted signal and I0 is the
measured signal of the straight through beam. Table 1 summarises the results of the
Reflectance =
I 0.Ccorrection factor
Transmittance =
Ccorrection factor = Ratio of measurements of I0 of reflectance trap relative to 7
transmittance trap
Table 1. Uncertainties
mean value
final value
unc (k=1)
R trap signal
R trap uniformity
T trap signal
T trap uniformity
laser stability
trap correction factor
The signal values and stand deviations are calculated from the series of repeat
measurements. The final uncertainty for the trap signals are the combination of the
signal, uniformity and laser stability standard deviations. These uncertainties then
combine to give the uncertainty in the trap correction factor. The uncertainties for the
transmittances I1_tot and I2_tot are the combination of the signal and trap uniformity
standard deviations. Any effects due to laser drift are taken care of by the signal
standard deviations. The uncertainties for the reflectances are the combination of the
signal standard deviations and the trap correction factor. The two traps are triggered
virtually simultaneously, so there are no effects due to laser drift. Any variation in
moving the knife edge in and out are taken care of by the signal standard deviations.
Subsequent to carrying out the measurements it was found that the ‘R’ trap was close to
the edge of its plateau region, which explains the higher than expected non uniformity
of this trap.
Other uncertainty components that were considered are summarised in Table 2.
Table 2. Other uncertainties that were also considered
Beam position on
Angle of incidence
into trap detectors
Trap detectors are insensitive to position as long as the signal beam
sits on the response plateau and is not a strongly divergent beam
During the series of measurements the angle of incidence did not
change as the traps were not moved. There may be a small
difference between the angle at which the reflectance trap was
compared to the transmittance trap and the angle at which it was
used to measure reflectance. However, traps are insensitive to angle
of incidence so long as the beam sits on the plateau.
Laser speckle is only an issue if the beam profile overfills the
Trap was set far away enough so that non-linearities due to tight
focussing would not be an issue, the beam diameter was 2mm
Accounted for in detector uniformities
Accounted for in the ratio calculations, and only relevant in the
reflectance calculations
Trap detectors were not moved within their mounts and therefore
always at the same orientation with respect to the beam
Detector linearity
Laser pointing drift
DVM non-linearity
Polarisation of
incident beam
From Table 1, it can be seen that R2 << R2_tot, whereas R1 ≈ R1_tot. This is to be
expected, given that surface 2 is AR coated at 702 nm, but surface 1 is not. The value of
R2 cannot be taken as being due solely to the first surface reflectance, since the intensity
of the back surface reflectance is ~ 50 times more intense, and therefore the wings of
the back surface reflectance will significantly overlap with the front surface reflectance
when the knife edge is in. However, for the application of this crystal we are
particularly interested in the reflectance losses at surface 2 as this boundary sits between
the point of downconversion and the detector.
If we consider the data for the laser incident on face 1,
R1= r1
R1tot = r1 + ∑ R1 j = r1 +
R1 j = t 2 (1 − r1 ) r2 t 2 r1 r2
t (1 − r1 ) r2
1 − t 2 r1 r2
j −1
and r1, r2, and t are the reflectances at surfaces 1 and 2, and the single pass internal
Ttot =
t (1 − r1 )(1 − r2 )
1 − t 2 r1 r2
We make the following approximations:
since r1 ~ 0.07 and r2 ~ 0.0014, we shall truncate the expression for Rtot after
j=2, since the ratio t2r1r2 ~ t2 (0.0001);
taking r2 ~ 0 leads to Ttot ~ t (1-r1).
R1tot = r1 + (1 − r1 ) t 2 r2 = r1 + Ttot r2
r2 =
R1tot − r1
Since we have confidence in the data obtained from surface 1, we can use (13) to
estimate r2. The difference term in the numerator is likely to lead to a significant
increase in the relative uncertainty of the result.
As a consistency check, we can then use the estimated value for r2 to calculate R2tot
using (14) which is derived in a similar manner to (12):
R 2 tot = r2 + (1 − r2 ) t r1 = r2 + (1 − r2 )
2 2
(1 − r1 )2
From equations (13) and (14) we find:
= 0.00135 ± 0.00014 (k=1)
R2_tot = 0.0718 ± 0.0026 (k=1)
where account has been taken of the correlations in the trap correction factor
The calculated value of R2_tot compares very well with the experimentally measured
value of 0.07244 ± 0.00024 (k=1). The crystal suffers from a non-uniformity of 0.8%
of the mean transmittance in its central 2 mm x 2 mm area, which is believed to be due
to the AR coatings. While all measurements – reflectance and transmittance – on one
face of the crystal were taken with the laser beam incident on the same spot (within the
repeatability of the measurement sequence), all that can be said about the full set of
measurements is that the laser was set incident as close to the centre of either side of the
crystal as could be judged by eye, and therefore there is the likelihood that the
measurements taken on opposite sides of the crystal sample slightly different sections of
the crystal.
We observe that these measurements can in principle lead to a set of over-determined
equations, and that therefore a fuller statistical analysis may lead to more consistent
Transmittance and reflectance measurements from small parallel-sided samples suffer
from interference effects and, coupled with anti-symmetric AR coatings it may be
necessary to measure the reflectance of each surface individually. The crystal studied in
this paper is such a sample. Knowledge of the reflectance losses at each surface was
required with a target uncertainty of better than 0.1%.
This paper demonstrates how it is possible to use a knife-edge to block the reflectance
off the second surface of the crystal through monitoring the interference fringes on a
CCD camera. Measurement considerations necessary for analysing the uncertainties
have been discussed. This technique complements other techniques which have been
developed to carry out spectrometric measurements on small samples [3, 5].
The sampled area of the crystal was of the order of 100 μm in diameter, and the overall
uncertainty in calculated reflectance 0.0014 of surface 2 was 0.00014, while the
uncertainty in the measured reflectance 0.00711 of surface 1 was 0.00024. These values
represent an initial exploration of the technique, and further improvements may be
This work was funded by the UK DTI National Measurement Directorate’s Programme
on Optical Radiation. The authors would like to thank W S Hartree and J Wang for their
helpful advice.
Cheung, J.Y., C.J. Chunnilall, and J. Wang. "Radiometric applications of correlated photon
metrology". Proceedings of SPIE, 5551, p. 220-230, 2004
Migdall, A.L., et al. "Absolute quantum efficiency measurements using correlated photons".
Metrologia,. 32, 479-483, 1995/96
Hartree, W.S., E. Theocharous, and N.P. Fox. "A wavelength tunable, quasi-cw laser source for high
accuracy spectrometric measurement in the 200 nm to 500 nm region". Proc. SPIE,. 4826, p. 104112, 2003 (in press).
Fox, N.P. "Trap detectors and their properties". Metrologia, 28, p. 197-202, 1991
Cheung, J.Y., et al. "High accuracy dual lens transmittance measurements". 2007. In preparation.
© Crown copyright 2007. Reproduced by permission of the Controller of HMSO and Queen’s Printer for
This work was supported by the National Measurement System Policy Unit of the Department of Trade
and Industry.
Robot-Based Gonioreflectometry at PTB
Andreas Höpe, Dirk Hünerhoff
Physikalisch-Technische Bundesanstalt, Section: 4.52 Reflectometry, Bundesallee 100,
38116 Braunschweig, Germany
At PTB a robot-based gonioreflectometer for measuring radiance factor and BRDF has
been developed [1]. The facility enables measurements of the directed reflection
characteristics of materials with arbitrary angles of irradiation and detection relative to
the surface normal.
1. Description of the robot-based gonioreflectometer system
The gonioreflectometer system consists of three major parts, as shown in Fig. 1: a large
rotation stage carrying the irradiating homogeneous sphere radiator, the 5-axisindustrial-robot as a holder for the sample under test, and a monochromator with a
mirror-based imaging system for the spectrally selected detection of the reflected
radiance of the sample. In the following, these three major parts are presented in more
Large rotation
Fig. 1. Picture of the gonioreflectometer showing the 5-axis robot and the large rotation stage
with the homogeneous sphere radiator
1.1 The large rotation stage with homogeneous sphere radiator
The facility uses broadband irradiation of the sample with spectrally selected detection
of the reflected radiation. The unfiltered broadband irradiation is delivered by a special
sphere radiator [2] with an internal 250 W quartz tungsten halogen lamp. It is located on
a large rotation stage with a diameter of 1500 mm and can be rotated 360° around the
5-axis-robot serving as the sample holder.
The combined adjustment of the rotation stage and the robot allows full angular control
of the directed beams of incident and reflected radiation within the full half space above
the surface of the sample, accomplishing also the measurement of out-of-plane
reflection. The aperture angle of the source is 1.50° (solid angle 2.16 × 10-3 sr), and the
aperture angle of the detector 0.32° (solid angle 96.45 × 10-6 sr).
In order to verify the homogeneity of our sphere radiator, a special facility for
measuring the homogeneity was built. The radiator was mounted on a two-dimensional
xy-translation stage to scan over the 40 mm wide precision aperture of the radiator.
Imaging optics for the radiance with a 1:1 reproduction scale were chosen, resulting in a
2 mm resolution in the plane of the aperture. Fig. 2 shows a typical homogeneity plot of
the beam profile over the whole opening with a variation in radiance in the order of
± 0.2 %, only.
y [mm]
x [mm]
Fig. 2. Homogeneity plot of the sphere radiator showing variations in the range of ± 0.2 %
1.2 The 5-axis-robot
The main part of the new gonioreflectometer is the small 5-axis-industrial-robot with an
acromial height of only 550 mm. This robot serves as the specimen holder for the
reflection standard under calibration. The flexibility of the robot arm gives one the
ability not only to measure in plane reflection characteristics of materials, also out of
plane configurations with arbitrary irradiating and detection angles are possible. The
robot has three internal coordinate systems with the capability of making direct
coordinate transformations between them. Due to these internal coordinate systems,
there is no need for the operator to go into the details of coordinate transformations, like
Euler angles, etc. This makes the programming of the movements, translations and
rotations of the robot system relatively simple. The different axes of the robot are
explained in Fig. 3.
For the measurements, the robot is operated in a slightly unusual position for an
industrial robot, with the hand flange above the basic platform collinear with the J1 base
axis (see robot in Fig. 1). The other robot axes (J2 to J4) are positioned in such a way as
to ensure that the surface of the calibration sample is always located within the common
rotation axis of the rotation table and the J1 axis. The robot is able to carry and position
large samples with an outer diameter of up to 500 mm and a maximum weight of up to
2 kg. The position accuracy is 0.02 mm for arbitrary movements within the whole
operating range. The rotation accuracy of the J1 base axis is Δϕ = ± 0.002°.
Fig. 3. Diagram showing the different axes of the 5-axis-robot
1.3 The detection system
The direction of the detection path is fixed due to the fact that a triple grating half-meter
monochromator for spectral selection of the reflected radiation is used. The current
wavelength range for measurements of diffuse reflection is 250 nm to 1700 nm. It is
planned to extend this range up to 2500 nm within the near future.
The facility uses two-stage mirror-based 10:1 imaging optics to map a 20 mm circular
area on the sample onto the 2 mm wide entrance slit of the monochromator. This results
in a 3 nm bandpass within the spectral range 250 nm to 900 nm and a 6 nm bandpass
within the 900 nm to 1700 nm range, depending on the gratings used.
Four different detectors behind the monochromator are used for detecting the radiance
signals of the incident and reflected beams: a solar blind Channel Photomultiplier
(CPM) for the measurements between 250 nm and 350 nm, a yellow enhanced CPM
between 300 nm and 450 nm, a silicon photodiode between 400 nm and 1100 nm, and a
cooled InGaAs photodiode (at - 25 °C) between 1000 nm and 1700 nm. All the signals
are detected with a picoampere-meter and transferred to a computer for data storage and
During calibration it is necessary to detect the reflected radiance signal of the sample
and to look directly into the irradiating lamp. For our setup, the ratio of these two
signals is about 1.5 × 103, making it necessary to have a linear dynamic range of 4
orders of magnitude for the detection. The magnitude of the reflected beam signal
measured with the picoamperemeter depends on the wavelength and the utilized
detector and is in the range of approx. 10 pA. The direct signal of the lamp is in the
range of approx. 100 nA.
The uncertainty budget (k = 2) for calibrations of the radiance factor of highly reflecting
samples (β ≈ 1) ranges from 0.002 in the visible and adjacent spectrum to 0.005 in the
IR at 1700 nm and to 0.035 in the deep UV at 250 nm.
2. Examples for out of plane measurements
The flexibility of the robot arm gives one the ability to measure out of plane reflection
characteristics of materials. It is possible to make measurements with arbitrary incident
and reflection angle within the whole half space above the sample surface. Fig. 4 shows
two measurements of the radiance factor of an opal glass (glossy and matt side). The
incident angle is varied from 0° to 85° in the xz-plane perpendicular to the yz-plane
where the reflection angle (fixed at θr = 25°) is located.
Wavelength: λ = 1000 nm
Radiance factor β(θi,θr = 25°)
matt side
glossy side
Incident angle θi [deg]
Fig. 4. Indicatrix of an off-axis measurement of the radiance factor of an opal glass at a
wavelength of 1000 nm. Incident angle varied from 0° to 85° in a plane perpendicular to the
plane of the reflection angle (fixed at r = 25°), see also sketch of the geometry
D. Hünerhoff, U. Grusemann and A. Höpe, “New robot-based gonioreflectometer for measuring
spectral diffuse reflection”. Metrologia, 43, S11-S16, 2006
W. Erb, Phys-Techn. Bundesanstalt, Jahresbericht 1979, 170-171, 1979
Characterization of Printed Textile Fabrics
F. Leloup, S. Forment, J. Versluys, P. Hanselaer
KaHo Sint-Lieven, Laboratorium voor Lichttechnologie
Gebroeders Desmetstraat 1, B-9000 Gent, Belgium
To expedite colour and gloss measurements, the CIE has recommended some basic
geometries regarding the angle of illumination and viewing. These specific measuring
geometries are implemented in commercial colour and gloss meters which are widely
available. During the measurement, close contact between the sample and the measuring
head is required.
However, there are many surfaces that cannot be adequately measured using the
recommended conditions. The determination of the colour and gloss of textile fabrics
using these instruments is not straightforward. Particularly, fabrics produced by airembossed techniques are very soft and compressible. The pile direction creates a
complex 3D texture resulting in a typical appearance. In order to avoid compression of
the nylon in contact with the instrument measurement head, a non-contact measurement
method is needed.
This paper reports on the difficulties that arise when characterising textile fabrics.
Lightness directionality, detection limits when measuring dark and low gloss samples
and colour differences are considered. Non-contact measurements on fabrics using a full
3D BSDF goniospectroradiometer are presented.
When an object is illuminated, the incident light will be reflected, transmitted, refracted,
absorbed and scattered. These processes are responsible for the optical properties of the
material, described as colour (spectral information), gloss (spatial information),
translucency and surface texture. It is recognised that these measures are not necessarily
independent, which is one of the main difficulties to characterize the overall appearance
of an object.
To expedite colour and gloss measurements, the CIE and ASTM have recommended
some basic geometries regarding the angle of illumination and viewing [1,2]. These
specific measuring geometries are implemented in commercial colour and gloss meters
which are widely available. During the measurement, close contact between the sample
and the measuring head is required.
However, there are many surfaces that cannot be adequately measured using these
industrial instruments [3]. Textile fabrics belong to this category. With the advent of
new synthetic fibre manufacturing technologies, polymer modification, and finishing
methods, the lustre properties or optical characteristics of fabrics have become very
subtle in their variations. Reflectance is dependent on twist level, softeners [4] etc. and
the reflectance properties of the fabrics are highly angle-dependent.
Fabrics produced by air-embossed techniques are very soft and compressible. The pile
direction creates a surface directionality, engendering a typical appearance. In order to
avoid compression of the nylon in contact with the instrument measurement head, a
non-contact measurement method is needed. Furthermore, gloss values of very matt
fabrics are often too low to be measured accurately by industrial gloss meters. In this
respect, the lustre "quality" of a cotton fabric may be perceived as different from that of
a cotton-like textured synthetic filament fabric, even if the gloss meter readings of the
two are the same.
An American company, with European headquarters located in Belgium, offers a
variety of nylon-velvet products, including piece-dyed, fibre-dyed, printed and
embossed fabrics. All products feature a 100% nylon face, backed by a poly/cotton
blended substrate. All of these are available in a variety of colours, designs and textures.
The resulting design and appeal of the fabrics is the main driver for consumers to
purchase the products. Until now, the company has not found an appropriate method to
characterize colour and gloss. Consequently, colour and gloss discrimination is
performed on a visual basis.
A standard grey fabric, which serves as a starting material for all products, and two
coloured printed fabrics showing colour differences, are studied. The influence of the
pile direction on the perceived appearance and the colour and gloss differences on
carpets are examined using a full 3D non-contact BSDF goniospectroradiometer. A
comparison between these results and the results obtained with industrial colour and
gloss meters is presented.
Any gloss and colour measurement can be related to some particular value of the
general spectral Bidirectional Scatter Distribution Function (BSDF). The spectral BSDF
qe ,λ can be defined as the differential spectral radiance dLe,λ of a sample observed in a
specific observation angle characterized by spherical coordinates (θ s ,ϕ s ) due to the
scattering of incident radiation characterized by the differential spectral irradiance dEe,λ
received from a specific angle of incidence (θ i ,ϕ i ) :
qe ,λ =
dLe ,λ
According to ASTM E1392 [5], the practical formula used to determine the absolute
BSDF under the condition that the field of view of the receiver field stop is sufficiently
large to include the entire illuminated area for all angles of interest, can be written as
q e ,λ =
Φ e,λ ,s
Φ e,λ ,i .Ωs . cosθs
where Φ e,λ ,i and Φ e,λ ,s are the spectral distribution of radiant flux received by the
sample and the detector respectively, and Ω s is the solid angle subtended by the receiver
aperture stop from the sample origin. The illuminated area need not be known. Because
the incident and scattered flux is measured with the same detector head, the flux ratio is
equal to the ratio of the detector responses at each wavelength.
In order to measure the spectral BSDF, a goniospectroradiometer was used. A picture of
the instrument is shown in Fig. 1. A 300W Xe light source is used for the illumination
of the specimen. Having large emission intensities in the blue-violet region of the
visible spectrum, good signal-to-noise ratios can be achieved over the whole visible
spectrum. The detector head consists of a lens and a very small integrating sphere which
is coupled to a spectrometer/CCD detection system with a quartz fibre. The diameter of
the lens is 20mm. The distance from the specimen to the detector head is 750 mm. The
use of an automated filterwheel carrying three neutral density filters extends the
dynamic range to 6 decades. While measuring the dark current, the incident light beam
is directed to a Si photodiode. The response of this detector allows us to compensate for
the fluctuations in the light source output. As we measure the incident power on the
sample with the same detector head, absolute BSDF-values qe,λ can be calculated from
both dark current corrected CCD readings (counts) according to Eq. (2). More details of
the instrument are described in Ref. [6].
Fig. 1. Picture of the BSDF goniospectroradiometer
Standard grey fabric
A standard grey fabric which is used as a starting material for several textile products
was studied. All the nylon piles have the same direction, which is slightly tilted from the
substrate normal. To investigate the influence of this directionality on the measured
gloss value, the spectral BRDF of the sample was measured at four different
illumination/viewing geometries.
Reference axes are fixed to be relative to the sample. The z-axis is coincident with the
sample normal. Four half-planes normal to the sample can be defined and are called ϕ 0 ,
ϕ 90 , ϕ180 , ϕ 270 . The top of the pile is directed towards the ϕ180 half-plane, as illustrated
in Fig. 2.
Fig. 2. Reference axis and four half-planes normal to the sample. The top of the pile is
directed towards the ϕ180 half-plane. In this figure, the illumination is from the ϕ 0 halfplane.
The spectral BRDF of the material was measured with four directions of illumination,
each lying in one of the half-planes. The anormal angle of incidence θ i is taken to be
equal to 60° in accordance with the standard angle of incidence for specular gloss
measurements [7]. Anormal viewing angles θ s ranged from -55° to 85° in the plane of
incidence, with an increment of 5°. Negative values refer to anormal angles in the halfplane of incidence, while positive values refer to anormal angles measured in the halfplane of the specular reflection.
Measurement results of the spectral BRDF at 500 nm as a function of the viewing angle
at four illumination geometries are shown in Fig. 3.
Fig. 3. BSD-functions at 500 nm of a standard grey fabric illuminated in four different halfplanes with an anormal angle of incidence equal to 60°.
In the specular reflection direction (60°), no reflection peak can be observed for any
illumination geometry. Furthermore, a reflectance minimum is found at an angle of
viewing that equals approximately 0°. In fact, the minima are located a few degrees offzero and are slightly dependent on the illumination geometry. Probably this small shift
is caused by the slightly off-normal direction of the pile. It is not clear why the fabric
appears darker when viewing near the substrate normal. Probably, the light will be
trapped in the cavities formed by the closely spaced pile.
BRDF-values increase when the anormal viewing angle increases in both the
illumination half-plane and the reflection half-plane. However, the increase of the
scattered reflection is not symmetrical and is dependent on the illumination direction.
This will result in a visual appearance which is strongly directional. The directionality
of the lightness of the sample seems to be much more important when the sample is
observed from the illumination half-plane; in the reflection half-plane, directionality
becomes only significant at anormal viewing angles beyond 60°. These conclusions are
in agreement with visual observations.
The increase of the reflectance with viewing angle, although illuminated from the
sample normal, was also reported in Ref. [8]. Textile samples were measured by a
Murakami Model GCMS-3B Gonio-Spectrophotometric color measurement system at
anormal receiving angles 15°, 45° and 60°, each combined with azimuthal angles
ranging from 0° to 345° at 15° interval.
As can be seen from Fig. 3, the curves corresponding to an illumination from the ϕ 90
and ϕ 270 half-plane are quite similar. Both illumination conditions are indeed
equivalent with respect to the pile direction of the fabric. Both curves are inbetween the
curves obtained with illumination from the ϕ 0 and the ϕ180 half-plane. At grazing
viewing angles, the fabric looks darker when illuminated from the ϕ180 half-plane
compared to illumination from the ϕ 0 half-plane. In the former situation, the direction
of the incident light beam is opposite to the pile direction. This probably causes
increased light-trapping in the pile and a higher absorption. At 85° viewing angle, the
ratio of the reflectance for both illumination conditions reaches nearly two.
As a comparison, test measurements with a Byk-Gardner micro-TRI-Gloss gloss meter
[9] were executed. In accordance to the ASTM Standard Method D523 [7], this
instrument measures specular gloss using the 20°, 60° and 85° geometry. However, at
each geometry, a value of zero specular gloss units (SGU) was measured and no
directionality could be detected, although differences in appearance are clearly
observed. A limited dynamic range is probably the reason why commercial gloss
measuring instruments are not very useful for characterizing matt samples.
In Fig. 4, the spectral BRDF of the standard grey fabric illuminated in the four halfplanes and viewed from an anormal angle of -45° are represented. As expected, the
spectral BRDF does not change very much with wavelength. The absolute values are
classified in the same order as shown in Fig. 3.
spectral BRDF at -45° (1/sr.nm)
wavelength (nm)
Fig. 4. Spectral BRDF at an angle of viewing of -45° and illuminated in four half-planes at
an anormal angle of 60°.
Coloured printed fabrics
Two coloured purple-pink textile samples were manufactered by printing on the
standard grey fabrics. The colour of sample A, taken from the first batch, was visually
more saturated than the colour of sample B, which belonged to a batch fabricated at the
end of the production run. The directionality of the measured BRDF was analogous to
Fig. 3. As expected, the reflectance in the specular direction is even lower than for the
neutral grey samples. With the industrial gloss meter, a value of 0 SGU in the 60°
geometry as well as in the 85° geometry was measured.
The spectral BRDF of the samples illuminated in the ϕ90 half-plane plane at 60° angle
of incidence and viewed at an anormal angle of 45° and 85° respectively is shown in
Fig. 5. The colour difference is clearly due to the difference in spectral reflectance in the
wavelength interval from 560 nm to 660 nm. Sample A will appear more reddish and
yellowish. At 85° viewing angle, the lightness of both samples has increased and the
difference in chromaticity is more pronounced, in agreement with the visual assessment.
From the BRDF-data, CIELAB values were calculated, using CIE illuminant D65 and
the CIE 2° Standard Colorimetric Observer. The results are shown in Table I.
Spectral BRDF (1/sr.nm)
sample A (45°)
sample B (45°)
sample A (85°)
sample B (85°)
380 400 420 440 460 480 500 520 540 560 580 600 620 640 660 680 700 720 740 760 780
wavelength (nm)
Fig. 5. Spectral characteristic of two coloured fabrics illuminated at 60° in the
ϕ 90
plane and viewed at an anormal angle of 45° and 85°.
Table I. CIELAB values of printed samples A and B at two different viewing angles
Viewing angle 45°
Sample A
Sample B
Viewing angle 85°
Sample A
Sample B
From the spectral BRDF values, ΔL*, Δa*, Δb *and ΔE * were calculated at all angles of
viewing, with sample A taken as reference. Results are shown in Fig. 6. As can be seen,
the variation of ΔL* with viewing angle is more pronounced than the variation in Δa*
and Δb*. The difference in appearance between the two samples increases with viewing
-50 -45 -40 -35 -30 -25 -20 -15 -10
Angle of viewing (°)
Fig. 6. CIELAB colour differences of two coloured textile fabrics at different angles of viewing.
Both samples were also characterized using an industrial colorimeter (Minolta CR-300).
CIELAB values were generated using illuminant D65 and the 2° Standard Observer. In
contrast to the BRDF measurements, d:8 geometry was used in the instrument.
Unfortunately, measurements were not quite reproducible as can be seen from Table II.
CIELAB values of five consecutive measurements show a large spreading for both
samples, especially for the lightness. An explanation can be found in the fact that the
colorimeter needs to make a close contact with the sample to measure the colour. When
exerting pressure on the fabric, the pile direction is modified. This will not change the
chromaticity very much, but it will affect the lightness substantially.
Table II. CIELAB values of five consecutive measurements on two coloured textile fabrics, obtained
with a Minolta CR-300 colorimeter.
Sample A
Sample B
Due to a specific pile direction, neutral grey fabrics can exhibit strong differences in
appearance when illuminated and viewed from different angles. Such type of fabrics can
be classified as achromatic gonio-apparent materials, manifesting an angle-dependant
lightness. When viewing the fabric from the sample normal, a rather dark appearance is
observed. At grazing angles, the lightness of the sample increases. Furthermore, the
orientation of the illumination with respect to the pile direction of the fabric will result
in a different light trapping in the pile, resulting in a strong directionality of the
These kinds of fabric are difficult to characterize with commercial colour and gloss
meters. Because of the total absence of any specular reflection, gloss meters do not give
any relevant information because of their limited dynamic range. Standard colorimeters
need to make close contact with the sample, which modifies the pile direction and
results in a large spread in the readings, particularly of the lightness. For the
characterization of those materials, non-contact spectrogoniometers are required.
CIE 15:2004, “Colorimetry”, 3rd edition, 2004.
ASTM E 179-96, ‘Standard Guide for Selection of Geometric Conditions for Measurement of
Reflection and Transmission Properties of Materials”, 1996.
Leloup, F., Hanselaer, P., Pointer, M., Versluys, J. “Characterization of gonio-apparent colours”. In
Proceedings of 10th Congress of the International Colour Association, Granada (Granada 2005), pp
M.P. Gashti, A. Peyravi. “Colorimetric properties of direct dyed cotton fabrics after treatment with
softeners”. In Proceedings of 10th Congress of the International Colour Association, Granada
(Granada 2005), pp. 721-724.
ASTM E 1392-96, “Standard Practice for Angle Resolved Optical Scatter Measurements on Specular
or Diffuse Surfaces”, 1996.
Leloup, F., De Waele, T., Versluys, J., Hanselaer, P., Pointer, M. “Full 3D BSDF
Spectroradiometer”. In Proceedings of the ISCC/CIE Expert Symposium: 75 Years of the CIE
Standard Colorimetric Observer, National Research Council of Canada, Ottawa, Ontario, 2006.
ASTM D523, “Standard Test Method for Specular Gloss”, 1999.
K. Suzuki, G. Baba, “Prediction of Color and Appearance of Textiles”. In Proceedings of 10th
Congress of the International Colour Association, Granada (Granada 2005), pp. 817-820.
Provided by Analis, official supplier Byk-Gardner Belgium.
Measurement of Bidirectional Reflectance in the Visible and Infrared
Peter Raven
QINETIQ, Malvern Technology Centre, St Andrews Road, Malvern, Worcestershire,
WR14 3PS
The appearance of a surface in a particular environment depends on many factors,
including the intrinsic optical properties of the constituent materials and the illumination
and observation geometry. The surroundings can affect the appearance of uniform ‘solid
colour’ materials, and gonio-apparent coatings place even greater demands on the
measurement fidelity required. In addition to dependence on wavelength,
comprehensive assessment of surface appearance must characterise the variation of
reflectance with the directions of both illumination and observation, introducing the
concept of bidirectional reflectance.
Bidirectional Reflectance Distribution Function
The dependence of the optical properties of a material on the illumination and
observation directions can be described by the Bidirectional Reflectance Distribution
Function (BRDF). The BRDF is a uniquely defined quantity, with no reference to the
instrument configuration. Fig. 1 illustrates the definition of BRDF. Mathematically the
BRDF is defined as a function per unit solid angle using:
dL (θ , φ , λ )
BRDF (θ i , φi , θ r , φ r , λ ) = r r r
dM i (θ i , φi , λ )
where dMi is the incident irradiance and dLr is the reflected radiance. A perfectly
reflective and perfectly diffuse (Lambertian) surface possesses a constant BRDF of 1/π
sr-1. For a perfect mirror the BRDF is a delta function, with an infinite value at a
specular reflectance condition, and zero elsewhere. The definition of BRDF is
convenient for incorporation into radiometric ray tracing codes used for scene
simulation or optical design applications. It allows calculation of reflectance factors
appropriate to the situation under consideration by integration over the desired solid
angle of either reflection or illumination.
Reflected beam
Incident beam
Fig. 1. Bidirectional Reflectance Geometry
BRDF Measurement
BRDF is measured at QinetiQ using an SOC-200 bidirectional reflectometer (Fig. 2).
The instrument is configured to measure the far-field BRDF. The angular resolution is
0.9°. Only reflecting optics are used, to allow the measurement waveband to be selected
across the visible and infrared (IR) by changing the radiation source and detector. A
tungsten-halogen lamp is used for visible and near IR measurements, and a 1600°C
black body source for mid and far IR measurements. The waveband is selected with a
bandpass filter placed in the detector arm.
Source radiation
Fig. 2. SOC-200 Bidirectional Reflectometer
The BRDF measurements are taken relative to a reference standard. Spectralon® is used
for wavebands between 0.3μm and 2.0μm, and a diffuse gold surface, produced by gold
coating bead blasted aluminium, for wavebands between 2.0μm and 14μm. The
reference value of the BRDF of the standard is specified for 0° incident polar angle and
45° reflectance polar angle. In order to calibrate the BRDF measurements for a general
polar incident angle (θi) a three stage process is used:
1. Measurement of reference standard at (0°,45°)
2. Measurement of reference standard at (0°,θi)
3. Measurement of reference standard at (θi,0°)
By the Helmholtz reciprocity theorem the BRDF of measurements (2) and (3) are
identical, so the measurement sequence allows the irradiance at θi to be related to the
irradiance at 0°. The throughput of detector beam path is assumed to be independent of
reflectance polar angle.
The measurement calibration process relies on a value having been established for the
BRDF of the reference standard. At QinetiQ Malvern this is achieved by taking a
directional hemispherical reflectance (DHR) measurement. The DHR can be calculated
from the BRDF by integration over the hemisphere of all possible reflectance angles:
DHR (θ i , φi ) = ∫∫ BRDF (θ i , φi , θ r , φr ) ⋅ cosθ r ⋅ sin θ r ⋅ dθ r ⋅ dφr
Ω r = 2π
The DHR can be measured directly, and reference standards are readily available. At
QinetiQ Malvern a Cary 5G and Labsphere integrating sphere is used to measure DHR
in the UV/Vis/NIR, and in the IR an SOC-100 is used.
1. Measure DHR directly at 8° incident angle
2. Measure BRDF with 0° incident polar angle, with the reflectance polar angle
varied from 0° to 90° at fixed reflectance azimuth angles of 0°,90°,180°,270°.
This measurement includes four readings to provide the initial value of the
BRDF for 0° incident polar angle and 45° reflectance polar angle,
3. Integrate BRDF data to obtain value for DHR
4. Calculate correction factor C =
5. Obtain correct BRDF at incident polar 0°, reflectance polar 45° using:
BRDFnew (0,45) =
⋅ BRDFold (0,45)
This calibration process assumes that the DHR for 0° and for 8° incident angles are
identical. The accuracy of the integration algorithm relies on the isotropy of the surface.
The surface of the standard does need to be diffuse, but it is not essential for the surface
to exhibit a perfect Lambertian BRDF.
The system throughput and detectors are typically configured to achieve a noise
equivalent BRDF in the range 10-4 – 10-3 sr-1. For specularly reflecting surfaces the
dynamic range can exceed that available from the detector, so neutral density filters are
used to extend the dynamic range of the system.
An example BRDF dataset is shown for Spectralon in Fig. 3. Each trace shows the
BRDF as a function of reflectance polar angle for a particular incident angle. The
azimuth reflectance is fixed in the plane of incidence, and any specular reflectance
would appear at positive reflectance polar angles. At near normal (8°) the BRDF is
relatively constant, as expected for a near Lambertian material. For this sample, at
incident angles of 60° and higher the forward reflectance increases.
BRDF from Spectralon at 550nm
BRDF (sr )
Polar reflectance angle (°)
Fig. 3. BRDF scans taken from Spectralon at eight incident polar angles, and the
wavelength of 550nm, with the reflectance polar angle scanned in the plane of incidence.
4 Polarised BRDF
Materials can exhibit different reflectivity depending on the polarisation state of the
illuminating radiation and the polarisation state used for observation. A common
example is reflectance from smooth surfaces, where incident unpolarised radiation can
become linearly polarised due to the greater reflectance for radiation polarised
perpendicular to the plane of incidence. Materials exhibiting a high degree of multiple
scattering, e.g. paints, can depolarise, removing preferential polarisation present in the
incident radiation. Measurement of polarised BRDF can therefore provide extra
information regarding the composition and morphology of the material, allowing
analysis of the reflectance processes responsible for the appearance of the material.
In order to account for possible changes of polarisation state when measuring polarised
BRDF, the polarisation states of the incident and reflected radiation need to be defined
separately. The convention we use is to define the polarisation state relative to the plane
formed by surface normal and direction of propagation, i.e. the incident ray for incident
polarisation state and reflected ray for reflected polarisation state. Basic polarisation
analysis, considering only linear polarised radiation with perpendicular (S) and parallel
(P) orientations, requires four measurements, the co-polarised terms (SS, PP), and the
cross-polarised terms (SP, PS). Comprehensive polarimetric analysis, including circular
polarisation and linear polarisation at ±45° is usually considered by taking
measurements to calculate the 16 elements of the Mueller matrix.
Reflected beam
Incident beam
Fig. 4. Polarised BRDF geometry
The calibration of the S and P polarised BRDF is carried out using the average BRDF.
In order to correctly account for the division of radiation between polarisation states by
reflection, the average BRDF is calculated using:
Ave = ½(SS+SP+PS+PP)
The calibration factors can be determined using the average BRDF, following the same
steps used for the polarisation independent BRDF discussed in section 3. This procedure
does not require the reference standard to be perfectly depolarizing. However the
procedure does make the assumption that the instrumental throughput is independent of
polariser orientation.
The orientation of polarisers is determined using the specular reflection from a smooth
surface, e.g. black glass. The extinction ratio of polarisers is typically of order 103.
Although for many applications this is adequate, smooth surfaces can exhibit low
depolarisation, where the SP and PS values can be more than three orders of magnitude
smaller than the SS and PP values.
An example S and P polarised BRDF dataset is shown for Spectralon in figure 5, with
the incident angle of 60°. The reflectance of Spectralon is dominated by volume
scattering from within the material, so the reflectance of all four polarisation states is
very similar for most reflectance angles. However, at the incident polar angle of 60°,
Spectralon starts to exhibit increased reflectance from the surface of the material. This
causes the co-polarised SS and PP BRDF to increase towards +90°, which is not
observed in the cross-polarised SP and PS BRDF.
Polarised BRDF from Spectralon at 633nm and
60° incident angle
BRDF (sr )
Polar reflectance angle (°)
Fig. 5. Polarised BRDF scans taken from Spectralon at 60° incident polar angles, with a 633 nm laser
The Bidirectional Reflectance Distribution Function (BRDF) can be used to describe
macroscopic appearance of materials and surfaces. It is convenient for taking data into
radiometrically accurate graphical and scene simulation models.
At QinetiQ Malvern Technology Centre, BRDF measurements are taken using an SOC200 reflectometer. The instrument takes measurements relative to a reference standard
with a known BRDF value at the incident polar angle of 0°, and reflectance polar angle
of 45°. The BRDF of reference standard is determined via comparison with integrating
sphere measurement of the directional hemispherical reflectance.
Polarised BRDF can provide greater understanding of the reflectance processes
responsible for the appearance of a material. Different scattering processes can create
different degrees of preferential polarisation and depolarisation.
Part of this work was carried out as part of the Weapons and Platform Effectors Domain
of the MOD Research Programme.
The Imaging Sphere – the First Appearance Meter?
Robert Yeo 1, Ron Rykowski, Doug Kreysar & Kevin Chittim 2
Pro-Lite Technology LLP, Innovation Centre, University Way,
Cranfield, MK43 0BT, UK
Radiant Imaging Inc., 15321 Main Street NE, Duvall, WA, 98019, USA
The Imaging Sphere™ is a unique tool that can be used to measure the angular luminous
intensity and colour variation of light sources, the view angle performance of displays, and
the light scatter from surfaces (bi-directional reflectance distribution function or BRDF).
Combining a CCD imaging photometer with a hemispherical reflecting chamber, the
Imaging Sphere is based on novel technology jointly developed by Radiant Imaging and
Royal Philips Electronics. For light source and display characterisation, the Imaging
Sphere functions as a far-field goniophotometer, except there are no moving parts, the
measurements take just a few seconds and the source is characterised in all directions in a
single measurement. For surface appearance testing, the Imaging Sphere functions as a
BRDF instrument, capturing the full 2π hemispherical reflected intensity distribution for a
given angle of illumination in a matter of seconds. In addition, the Imaging Sphere costs a
fraction of the price of traditional goniometer and BRDF systems, allowing its use in
routine production testing as well as R&D.
1. Introduction
The newly developed Imaging Sphere™ from Radiant Imaging, Inc. is a novel instrument
which allows for the quantification of the magnitude and colour of the light output emitted
or reflected from a device under test and how they vary with viewing angle (or with
viewing and illumination angles for non-emitting samples). By providing data on the
variation of colour, intensity and reflectance as a function of viewing and illumination
angles, the Imaging Sphere provides much more information to better describe the
appearance of a light source, display or material.
This paper will review the need for angular light, colour and reflectance measurements on
light sources, displays and materials. Traditional measurement devices will be compared
with the Imaging Sphere and results from initial application studies conducted using the
Imaging Sphere will be presented.
2. The Need for Angular Light Intensity and Colour Measurements
The “appearance” of an object does not depend solely upon its spectral reflectance but also
upon the directionality of the light upon reflection from the material. Few materials exhibit
the ideals of specular (i.e. mirror-like) or diffuse reflectance; in reality, most samples
exhibit some combination of diffuse and specular reflectance. Moreover, the spectral
distribution of the light upon reflection may vary with both viewing and illumination
angles. Examples of difficult to measure coatings and finishes include metallic effect
finishes, so-called “colour flip” (or gonio-apparent) special-effect decorative paints and
holographic foils. Colour flip pigments are multi-layer paints which cause the object to
change colour dramatically when viewed (or illuminated) from different angles leading to a
very complex analysis if one were to try to quantify the appearance of the material.
The luminous intensity (photometric power per unit solid angle) from a light source such
as a light emitting diode (LED) will vary with angle. In addition, the colour of light emitted
from certain types of white LED also varies with angle (for those LEDs that produce white
light as a result of propagating the output from an ultraviolet or blue LED through a
phosphorescent coating). The photometric modelling of LED-based light sources or
luminaires generally (and incorrectly) assumes that the LED performs as an isotropic, point
source. Thus, computer-aided design models of LED-based lighting rarely give an accurate
prediction of the illumination performance of the final product due to the light from the
LED being more concentrated in particular directions. Standards such as CIE 127 attempt
to apply a universal frame of reference to LED measurements by specifying that the
luminous intensity of an LED should be reported in the direction defined by the LED
package’s mechanical axis, but that simply serves to avoid the issue that the direction of
peak intensity will probably be aligned along a different axis altogether. LED
manufacturers also generally specify the view angle for their device, but this is a single
number which defines the range of angles (in a single plane) over which the intensity from
the LED falls to half that at its peak. Such scant information is wholly insufficient for LED
system designers to know how the LED will perform in the intended application.
The visibility of LCD displays varies greatly with the direction from which one views the
flat panel. For display manufacturers, three of the most important parameters to test are the
luminance (often called the brightness or photometric intensity per unit area), colour and
contrast ratio as a function of angle. In addition, the readability of a display will be greatly
impacted by the ambient light to which it is exposed. The reader will be familiar with the
problems of glare caused by specular and near-specular reflections from the surface of a
display which can almost completely mask the image being viewed. Fully characterising a
flat panel display requires not only a spatial (i.e. on-axis) measurement of luminance and
colour uniformity but also how the display performs at off-axis angles.
3. Traditional Measurement Solutions
Traditionally, the direction-dependent variation of light intensity and colour emitted by a
source, or reflected from a material have been determined using a goniometer (sometimes
referred to as a goniophotometer – Fig. 1). A goniometer typically comprises a moving
platform which rotates a sample about one or two axes (azimuth and inclination) and a
photodetector which views the sample in a fixed direction. Measurements are performed
incrementally, one angle at a time over the complete hemisphere (2π steradians) or in a
single plane. The platform is often motorized so that the measurements can be automated;
depending upon the angular range and resolution required, goniometric measurements can
take a long time to perform, sometimes several hours. The photodetector will be
photometrically filtered for luminance or luminous intensity measurements, or comprise a
tristimulus response photodetector for colorimetric measurements. Alternatively, a
spectroradiometer can be used for spectral analysis, from which the photometric and
colorimetric parameters are calculated. The photodetector will be positioned at a defined
distance away from the light source in order that the correct geometric conditions for
luminance or luminous intensity measurements are met.
Fig. 1. Goniometer for LED Angular Luminous Intensity Measurements
For intensity, measurements are performed in the “far-field”, whereas luminance
measurements are made in the “near-field”. The far-field is defined as the region within
which the source behaves as if it were a point source and as such, the inverse squared law
applies. A star is a large object, but when viewed from Earth, has no physical extent and
behaves as a point source. For many traditional lamps, the far-field is defined loosely as
being between 5 and 10 times the source radius away from the lamp. For LEDs, the
analysis is more complex and readers are referred to CIE 127 for the definition of how to
perform average intensity measurements, which are – in effect – near-field intensity
measurements performed under defined geometric conditions. Conversely, luminance
measurements are made in the near-field, with the source and receiver physically close. In
the near-field, the source has a physical extent and measurements are made of source
intensity per unit area, which is the luminance.
3.1 LED Measurements
In the production of encapsulated LEDs or LED-based optical assemblies, an important
step is the alignment of the optics to the LED die. Typically, this is performed by means of
a goniometric measurement. A dual-axis goniometer provides all the data necessary for
this task, but the measurement speed is quite slow. A 2D goniometric analyser only
samples the intensity distribution in one plane. This makes the quality of the results highly
dependent upon the placement of the LED in the device, which cannot be accurately
controlled. Furthermore, most commercially available 2D goniometric analysers are not
configured to handle larger sources, such as multi-die chip-on-boards, or systems with
large optical components. Motorized goniometer systems tend to be slow, complex and
relatively expensive devices.
These limitations severely limit the amount and quality of data which LED manufacturers
can realistically make available to integrators of their devices. As a rule, LEDs are supplied
with a view angle specification and a typical plot of the intensity versus angle for a single
plane. Such scant performance information makes the job of designing optical systems
around an LED much more complicated and very much a “hit and miss” affair. Moreover,
it would be impractical for an LED manufacturer to sample and bin devices according to
their directionality (as is common practice in the industry for LED intensity and colour)
due to the time involved in individually testing devices as they come off the production
line. It should be noted however that the colorimetric accuracy of goniometric
measurements can be very high provided that a spectroradiometer is used as the detector.
Conversely, the colorimetric accuracy of filter photometers or colorimeters can be quite
poor unless the detector is calibrated against a source possessing a similar spectral power
distribution to the samples it is intended to measure.
3.2 Flat Panel LCD Display Measurements
The method applied to testing LEDs is also used to measure the angular output of flat panel
displays and LED arrays and clusters. A goniometric stage is used to position and rotate
the sample with respect to a fixed detector which views a defined area on the display (Fig.
Fig. 2. Goniometer for Flat Panel Display Measurements
The luminance variation from a display is measured in the near-field, hence the detector
will be positioned relatively close to the source. Goniometers for display testing will, by
necessity, be much larger than for discrete LED emitters, and their slow speed and high
cost limits their use to off-line QA sampling and developmental use in the display industry.
A further drawback of classical goniometers for display view angle testing is the
measurement spot size varies as the angle of the display is changed, leading to an
inconsistent sampling position. On the other hand, if used with a spectroradiometer
detector, the colorimetric accuracy of a goniometer can be very high and less susceptible to
errors arising from stray light.
An alternative method for testing display angular luminance and colour exists in the form
of the Conoscope (Fig. 3). This is an instrument which combines a high resolution CCD
camera with Fourier optics which maps an emitting spot on the display onto the CCD
detector. Each pixel on the CCD corresponds to a specific emission angle from the
measured spot on the display. The Conoscope is a well established technique but its utility
is restricted by its high cost, susceptibility to stray light and limited measurement spot size
(which makes it sensitive to the display’s pixel size). The Conoscope is however very fast,
gives a consistent measurement area for all angles and offers good angular resolution.
Fig. 3. Conoscopic Method for Display Testing
A more recent refinement on the goniometer theme has been to replace the spot detector
with a CCD-based imaging photometer, as implemented in the Radiant Imaging FPMS Flat
Panel Measurement System. The CCD imaging photometer employs a 2D detector array
with a spectral response scaled to match the CIE V(λ) or tristimulus functions to image the
whole area of the display in one measurement. Conceptually, the imaging goniometer
provides measurements of the display luminance and colour for all positions, one angle at a
time. From this data, the uniformity as a function of view angle can be easily computed, as
can the view angle performance for any location on the display. Further advantages include
a selectable measurement spot size, a measurement area that remains constant for all view
angles and the ability to provide measurement data from an infinity perspective (all
measurement locations measured at normal incidence).
3.3 Measurement of Surface “Appearance”
Colorimeters are instruments which are commonly used to measure the colour or spectral
reflectance of materials. A variety of standardised measurement geometries are employed
in such instruments, including directional illumination (e.g. 8°) with diffuse (hemispheric)
collection using an integrating sphere or directional illumination (e.g. 0°) with directional
collection (e.g. axial or annular 45°). Whilst such instruments provide measurements which
conform with internationally accepted conventions, they suffer from severe limitations
when used to measure the colour of materials whose spectral reflectance varies greatly
with viewing or illumination angles, such as certain metallic finishes and special effect
paints. To fully quantify the appearance of such materials, it would be necessary to
measure the reflected colour at all angles of illumination and viewing. Certain
manufacturers have attempted to address this problem with so-called “multi-angle”
colorimeters. These measure the reflected colour in several directions (e.g. 15, 25, 45, 75 &
110°) for a single illumination direction (e.g. 15°). Such instruments can only be a partial
solution to the complex problem of measuring exotic paints and finishes however.
Single or multi-angle colorimeters are simple, inexpensive and fast tools but can only
report colour for a fixed angle of illumination and a single or limited number of viewing
directions in one plane. As such, they cannot provide a meaningful measurement of
appearance for complex materials.
A much more sophisticated (and expensive) device is the scatterometer. This is a generic
name for the type of instrument which measures angular reflectance (Fig. 4). A
scatterometer is normally configured to measure the BRDF (bi-directional reflectance
distribution function) of a surface, the ratio of incident irradiance to reflected radiance for
defined angles of illumination and viewing. The light source is usually a laser, and BRDF
data is given at specific wavelengths. A scatterometer is a very powerful device, providing
high resolution and accuracy. However, measurements are very slow (reflected radiance is
measured sequentially at one angle of elevation and at one azimuth angle for each angle of
illumination), and scatterometers are very costly and complicated to use.
Fig. 4. Scatterometer for BRDF Measurements
4. New Technology – the Imaging Sphere
There has not been an instrument for luminous intensity, colour, view angle or BRDF
measurements that offers speed, angular resolution and low cost – until now. The Imaging
Sphere™ from Radiant Imaging provides a unique and innovative solution to the problem
of obtaining meaningful angular light distribution data for both research and development
and on-line production QA testing.
4.1 The Philips Parousiameter
The genesis for the Imaging Sphere began in 1996. Philips was in the process of
developing its “Matchline” series of television sets and companion video recorders. These
featured plastic cabinets with metallic-effect finishes. The cabinets for the TV, VCR and
stand all came from different factories. Colorimetrically, all the parts met the colour
matching specification, and the parts also matched in terms of their gloss levels, however
upon visual inspection, the parts clearly had different finishes. Existing colorimetric and
gloss measurements were simply unable to detect these differences, so Philips decided to
develop a new method for evaluating the visual “appearance” of the plastic parts. Philips
Applied Research in Eindhoven, The Netherlands was entrusted with this project and the
result was the Parousiameter, a name chosen after the Greek word for appearance. The
original Philips Parousiameter is shown below (Figure 5) together with its inventor, Sipke
Fig. 5. The Original Philips Parousiameter with its Inventor, Sipke Wadman
The Parousiameter showed that the differences in appearance of the Matchline parts was
due to the directionality of the anisotropy of the reflections when viewing in the direction
of the short axis on one part and the long axis on the adjacent part. This caused a
brightness difference in one direction of view, that could flip over in another viewing
With data obtained from its new instrument, Philips was better able to control the
production of the metallic effect finishes applied to its Matchline products such that all the
parts looked the same under a variety of lighting conditions. Philips realised that it had
developed a powerful new technique for assessing the colour and appearance of materials
and applied for and was granted protection for its intellectual property. Since then, Philips
sought a partner company to help commercialise its invention and in February 2005, signed
a cross-licence agreement with the Seattle-based light measurement specialists, Radiant
Imaging, Inc. Radiant Imaging was chosen as Philips’ partner based on its experience and
leadership in CCD-based light and colour measurement instrumentation.
4.2 Commercialising the Parousiameter – the Radiant Imaging Imaging Sphere™
The main optical elements of the Imaging Sphere are a diffuse, low reflective hemisphere,
a curved secondary mirror and a CCD-based photometer or colorimeter (Fig. 6). The
camera software automatically applies corrections for image offset, stray light and flatfield distortion so that the displayed image is a true representation of the light source’s
angular intensity and colour distribution.
Fig. 6. Basic Elements of the Imaging Sphere
The hemisphere is attached to a flat, non-reflecting baseplate containing a small aperture at
its centre. Light enters the light-tight device through the aperture in the baseplate (Fig. 7),
strikes the inner surface of the coated hemisphere, and is then reflected by a convex mirror
(Fig. 8) as an image of this illumination pattern onto the CCD camera (a Radiant Imaging
ProMetric™ Imaging Photometer or Colorimeter). The convex mirror enables the camera
to image almost the entire inner surface of the hemisphere at once. The image thus contains
all the information necessary to reconstruct the entire angular intensity profile of the
illumination at a resolution determined by the camera’s image sensor. Almost 2π
steradians of data can be recorded in a single measurement. For automated display testing,
the imaging sphere or the device under test is mounted for rapid relative motion (X,Y
stepped translation) so that the sphere or device under test can be repositioned to sample
the angular profile from multiple test points on the display, according to the specific test
protocols required.
Fig. 7. Light Enters the Imaging Sphere and Creates an Illumination Pattern on the Inside of the Dome
Fig. 8. The CCD Camera Images the Illumination Pattern from the Convex Mirror
Those readers familiar with the operation of integrating spheres may question the Imaging
Sphere’s ability to retain image quality and hence angular data integrity due to the multiple,
diffuse reflections that occur within the dome. An integrating sphere is a hollow sphere
which serves to spatially integrate (i.e. average) the spatial light distribution from a light
source propagated into it. Integrating spheres are typically coated with a high reflectance
(>90%) diffuse coating which causes multiple reflections to occur within the sphere. The
multiple reflections lead to a uniform radiance distribution within the sphere, which can
then be sampled by a photodetector mounted on the sphere wall. Thus, an integrating
sphere can measure the beam flux from all manner of extended area and divergent optical
One of the keys to the successful operation of the Imaging Sphere is a novel coating on the
inner surface of the hemisphere. This is a grey (18-20%) diffuse reflectance coating
designed to deliver only scattered reflections. The baseplate is coated black to approach
zero reflectivity. As a result, the CCD camera sees an intense image of the first-order
illumination pattern with a very weak, uniform background created by second strikes and
beyond. This background level depends upon the total light flux and is typically below 1%
of the peak image intensity. The Imaging Sphere samples the background light level and
this is then subtracted out from the measured source image in the camera software.
The angular resolution of the Imaging Sphere is defined by two parameters. The ProMetric
CCD cameras used on the Imaging Sphere are available with a range of pixel resolutions,
but the camera chosen for most applications has a full frame CCD sensor with 512 x 512
pixels. Over the complete hemisphere, each pixel sees a view of 0.35°. However, the size
of the sample must also be considered. Light from two locations on the same source can
impinge upon the same spot on the wall of the dome (Fig. 9). With a 7mm sample aperture,
the angular uncertainty θ is calculated from arc tan (4/254) = ± 0.79°.
Fig. 7. Limiting Resolution of the Imaging Sphere
4.3 The Three Imaging Sphere Configurations
The configuration outlined above (referred to as the model IS-LI Imaging Sphere) applies
to the measurement of angular luminous intensity and colour from point light sources such
as LEDs and small, planar LED clusters or arrays. Compared with traditional goniometric
testing, the Imaging Sphere is much faster (measurements of the full hemispherical
intensity and colour distribution take a few seconds compared with many minutes if not
hours for goniometric testing) and provides data in all planes, not just one. The imaging
sphere has no moving parts to go wrong, has a high angular resolution and is significantly
less expensive than traditional instrumentation. Combined, these attributes allow the
Imaging Sphere to be employed not just in R&D, but also on the production line for quality
assurance and device selection and binning. LED manufacturers who adopt on-line testing
of the angular variation and direction of peak intensity of their devices will be rewarded
with a significant competitive edge in the market place.
Two further versions of the Imaging Sphere have been developed for the testing of displays
(model IS-VA) and for scatterometry and surface appearance testing (model IS-SA). In
display testing, the Imaging Sphere measures the luminance and colour variation at one
point on the display at a time. The measurement spot is selectable up to 40mm diameter.
To test the luminance, colour and contrast variation over the whole display, the device
under test must be translated and the required locations on the display measured
sequentially (Fig. 10). It is instructive to compare the approach taken with the Radiant
Imaging Imaging Goniometer (all locations on the display, one angle at a time) with that of
the Imaging Sphere (all angles, one location at a time). The Imaging Sphere can be used to
test a variety of display technologies, including OLED, LCD, PDP, LED and LED
backlight. Compared with goniometric and Conoscopic techniques for display testing, the
Imaging Sphere is significantly faster and less expensive, possesses a high angular
resolution, is insensitive to pixel dimensions and stray light and measures with a constant
spot size at all angles.
Fig. 8. Imaging Sphere Configuration for Display Testing
As a scatterometer for surface appearance testing, the IS-SA version of the Imaging Sphere
can be equipped with a probe beam to illuminate the sample at either a variety of fixed
angles of illumination, or at any angle if equipped with an adjustable illuminator (Fig. 11).
Both illumination configurations have been evaluated for the Imaging Sphere, but that
chosen for the production version uses a single, variable angle, white light illuminator (Fig.
12) over that which uses fixed illumination directions (Fig. 13).
Fig. 11. Imaging Sphere Configuration for Appearance and BRDF Testing
The reflected radiation pattern imaged onto the CCD camera from the dome will be
characteristic of the sample under test. If the sample is glossy, there will be a distinct
specular reflection whereas matte materials will create a uniform light distribution within
the dome. The Imaging Sphere software will analyse the light distribution and compute the
BRDF of the material.
Fig. 12. The Imaging Sphere IS-SA for Surface Scatter Testing with Adjustable Illumination Direction
Fig. 13. The Imaging Sphere IS-SA with Prototype Fixed Illumination Positions
Geometrically, the Imaging Sphere holds a major advantage over conventional goniometric
scatterometers in that the reflected radiation distribution is measured simultaneously at all
azimuth and elevation angles for each direction of illumination (Fig. 14). Traditional
goniometric scatterometers are limited to measuring the reflected light in one direction at a
time for each angle of illumination. Characterising the complete hemispheric scatter
distribution for every angle of illumination with a traditional instrument would clearly be a
prohibitively time consuming task.
Fig. 14. Imaging Sphere Performs Out of Plane Scatter Analysis in Real Time
The speed, ease of use, relatively low cost and ability to perform measurements in ambient
light conditions are further advantages of the IS-SA over conventional technologies. All
three versions of the Imaging Sphere employ the same basic hardware which enables the
user to configure a single instrument for luminous intensity, view angle and scatter
measurements if he or she so wishes.
Validation testing of the Imaging Sphere has been performed by comparing the variation of
colour and luminance from a flat panel LCD display on the IS-VA with a traditional
goniometer. These results are shown below (Fig. 15).
Fig. 15. Inter-Comparison of IS-VA and Traditional Goniometer
The Imaging Sphere software provides for instrument calibration and recalibration, data
acquisition, and a diverse range of data analysis functions. Measurement results are
presented graphically and numerically. The parameters reported include luminous intensity,
luminance, CIE colour coordinates (xy, u’v’ & L*a*b*), tristimulus values (XYZ),
correlated colour temperature (CCT), view angle, BRDF and cosine-corrected BRDF
5. Application Overview
To illustrate the utility of the imaging sphere in each of its three areas of application,
results are presented below from an LED, a flat panel display and from a gonio-apparent,
special-effect paint sample. A single LED emitter was placed at the input port on the
Imaging Sphere (model IS-LI) and the complete hemispheric luminous intensity and colour
variation recorded. A screen grab of the measurement results in shown below (Fig. 16).
The isometric plot shown on the left displays regions of constant intensity in the same
colour, whilst the cross sectional Cartesian plot shown to the right is the luminous intensity
sampled through a user-defined axis in the isometric plot. Of course, the value of the
Imaging Sphere is that it records the spatial light distribution in all axes at the same time,
and this data can be displayed in a 3D plot which greatly simplifies the visualisation of the
output of the LED (Fig. 17). It should be noted that this measurement took about 5 seconds
to perform. By comparison, a traditional goniometric measurement might take about 30
seconds to perform, but this measurement is limited to one plane only; to sample the LED
at every 5° of azimuth would take the operator over an hour.
Fig. 16. Angular Intensity Profile of an LED
Fig. 17. 3D Plot of the Angular Luminous Intensity of an LED
An LCD flat panel display was tested with the Imaging Sphere (model IS-VA). A single
location on the display was selected for simplicity. The variation of luminance with angle
(view angle performance) is shown with the LCD set to display uniform white (Fig. 18).
The next figure shows the luminance of the FPD when it is set to display black. This plot
demonstrates the sensitivity of the IS-VA system, and also underscores the value of its
ability to determine an entire output distribution at one time. Specifically, this single
measurement, which took only seconds to make, clearly enables visualization of the
leakage pattern from the LCD in both qualitative and quantitative terms.
Fig. 18. View Angle Plot of LCD monitor set to display uniform white
Fig. 19. View Angle Plot of LCD monitor set to display uniform black
A sample of Helicone paint (a gonio-apparent, colour flip finish) was tested using the
Imaging Sphere (model IS-SA). The probe beam was brought into the dome at an azimuth
angle of 45° and at a number of elevation angles. The variation of the CIE u’ and v’
chromaticity coordinates at a viewing azimuth of 45° is shown below (Fig. 20). With the
illumination at normal incidence, the paint sample exhibits a consistent yellow colour
(mean chromaticity of u’ 0.20 and v’ 0.49). When the illumination angle is increased to 45°
(Fig. 21), the colour of the paint flips over to a distinct green hue at high view angles.
From the data obtained in a matter of a minute or so using the Imaging Sphere, it is
possible to determine the colour shift for the gonio-apparent paint for any viewing
direction and for the particular illumination conditions recorded.
Fig. 20. Angular Colour Variation of Helicone Special-Effect Paint Illuminated at Normal Incidence
Fig. 21. Dramatic Colour Shift of Helicone Paint Illuminated at 45 Degrees is Revealed by the Imaging
6. Summary
The Imaging Sphere from Radiant Imaging represents a powerful, enabling technology that
provides access to angular light and colour distribution data from all manner of light
sources (in particular LEDs and small arrays), displays and from materials and coatings. Its
reduced measurement times, lower capital cost, simple operation and robust construction
make the Imaging Sphere a viable proposition for both laboratory research and
development as well as in production line testing, an application that until now has not
been possible due to the slow speed of traditional goniometric techniques. Statistical
process control and the selection and binning of LEDs and other sources and materials by
view angle and direction of peak intensity is now made possible using the Imaging Sphere.
7. References
R. Rykowski, D. Kreysar & S. Wadman. “The Use of an Imaging Sphere for High-Throughput
Measurements of Display Performance – Technical Challenges and Mathematical Solutions”. SID
06 Digest, pp. 101-104.
ii. R. Rykowski, K. Chittim & S. Wadman. “Imaging Sphere. Photonics Spectra, September 2005, pp.
iii. S. Wadman & S. Baumer. “Characterisation of Appearance by a Parousiameter”. Annual
Proceedings SPIE, Vol. 48, August 2003.
iv. CIE 127-1997 (International Commission on Illumination Publication), “Measurement of LEDs”.
The Latest Revision of European Pharmacopoeia to 5.2, and its Effect on the
Qualification of UV-Visible Spectrophotometers
John Hammond
Optiglass Ltd, 52-54 Fowler Road, Hainault
Ilford, Essex IG6 3UT
United Kingdom
UV spectrophotometry is fundamentally both accurate and precise. However, it is
essential to initially establish compliance to specification, and thereafter check
instrument performance on a regular basis to ensure that it is within satisfactory
parameters i.e. 'under control'; and to allow corrective action when found to be outside
these limits.
In the European Pharmacopoeia, this requirement is fully documented in section 2.2.25
Absorption Spectrophotometry, Ultraviolet and Visible, and in the latest revision there
are some specific and important changes.
In the first instance, an additional requirement incorporates the use of a 600 mg/l
potassium dichromate solution to establish the ‘Control of absorbance’ at 430 nm.
Secondly, the use of potassium chloride to estimate the ‘Limit of stray light’ has been
revised; and lastly the use of ‘suitable certified reference materials’ has now been
approved in all appropriate procedures.
This paper describes the practical work undertaken to establish the stability of the 600
mg/l potassium dichromate solutions, and the results obtained. It also discusses all the
revisions detailed above, and the impact these will have on a day-to-day basis when
qualifying UV-Visible instrument systems.
Finally, the overall conclusions and future trends will be detailed as discussion points.
Infrared Spectrophotometry; Some of the Pitfalls
E. Theocharous
Optical Radiation Team, Quality of Life Division,
National Physical Laboratory (NPL),
Hampton Road, Teddington, TW11 0LW, UK.
Spectrophotometry is a relatively straightforward technique compared to, for example,
infrared spectroradiometry because the former does not require the definition of a
radiometric zero, nor does it require the use of a standard. Two of the most important
requirements of spectrophotometry are for the detection system to be stable and to have
a linear response over the range of incident radiant powers used for a particular
measurement. However, recent work by the author has shown that the identification of
an infrared detector with a responsivity which does not drift with time and has a linear
response is far from straightforward. The detector linearity requirement is particularly
difficult to satisfy in the case of spectrophotometers based on Fourier Transform (FT)
spectrometers because at the vicinity of the “zero path difference” point, the
photodetector receives the full spectral output of the source and is therefore likely to run
into a non-linear regime. Infrared radiometrists should also be aware of a number of
phenomena which give rise to changes/drifts in the responsivity of infrared detectors
and which can result in serious errors in spectrophotometric measurements. This paper
emphasizes the very limited range over which the responsivity of some infrared
detectors is linear and summarises a number of phenomena which give rise to large
drifts in the responsivity of infrared detectors. The paper also highlights how important
it is to have an infrared detector with a wide linearity range when developing
spectrophotometers utilising FT spectrometers.
The quality of a measurement is governed by the combined uncertainty associated with
that particular measurement. This assumes that a rigorous uncertainty budget has been
developed which includes all components that contribute to the total uncertainty of the
measurement. Recently NPL has reported, in a series of publications, a number of
phenomena which give rise to drifts in the responsivity of infrared detectors [1-5]. NPL
has also reported the linearity of response characteristics of a number of widely used
infrared detectors and showed that the range over which their responsivity is linear is
very limited [6-8]. This means that a number of new uncertainty contributions must be
introduced in the uncertainty budget which will cause the combined uncertainty of
infrared spectrophotometric measurements to increase.
The purpose of this paper is to summarise the non-linear response and drifts in the
responsivity of infrared detectors and highlight how they can give rise to serious
uncertainty contributions in infrared spectrophotometric measurements. It is important
to stress that there are a host of other phenomena such as atmospheric absorption [1],
the spatial uniformity of response of infrared detectors [9] etc which can also contribute
to the combined uncertainty of infrared spectrophotometric measurements. However,
these will not be considered in this paper.
Uncertainties due to the non-linear response of infrared detectors
A photodetection system is linear if its output is directly proportional to the level of
incident radiation. The linearity of the response of infrared detectors is a seriously
neglected phenomenon and the linearity of photodetection systems is often assumed
rather than demonstrated.
Until recently, there were no serious scientific studies/publications dealing with the
linearity characteristics of infrared detectors, despite their widespread use.
Radiometrists have adopted the “linearity factor” as a means of quantifying the linearity
characteristics of photodetectors [6]. The linearity factor L(VA+B) of a photodetector for
an output of (VA+VB)/2 is given by:
L(VA+ B ) =
VA + B
(VA + VB )
where VA, VB and VA+B represent the zero-corrected signals from the photodetector
when apertures A, B and A+B were open, respectively, with VA set to be approximately
equal to VB [6]. For a linear detector the value of L(VA+B) will be equal to unity. Where
L(VA+B) deviates from unity, its value can be used to estimate correction factors to apply
for different detector output signals to transfer the absolute responsivity calibration from
one level to another [10]. Another useful parameter in defining the non-linearity of a
detector is the deviation from linearity or non-linearity which is given by 1- L(VA+B).
Spectrophotometry involves the measurement of the radiant power of a beam of optical
radiation with and without an artifact inserted into the beam. The transmittance of the
artifact is defined as the ratio of these two measurements. The deviation from linearity
in the response of the photodetection system is, arguably, the parameter contributing the
largest uncertainty component in infrared spectrophotometric measurements.
The author was the first to report absolute linearity measurements of the response of
some infrared detectors. So far the non-linearity characteristics of Photo Conductive
(PC) and Photo Voltaic (PV) HgCdTe detectors [6], PbS detectors [7] and PbSe
detectors [8] has been reported. PC HgCdTe, PbS and PbSe detectors are intrinsic
photoconductors [11]. This means that they offer high D* values [11] at relatively high
operating temperatures which, in the case of PbS and PbSe as well as in some HgCdTe
applications, can be attained by thermoelectric cooling, hence their widespread use. PC
and PV HgCdTe detectors are widely used in infrared spectrophotometry in the 5 μm to
26 μm wavelength region. PbS was until recently the detector of choice for most
detection applications in the 1.6 μm to 3 μm wavelength region while PbSe detectors
are widely used in photodetection applications in the 2 μm to 5 μm wavelength range
Recent measurements by the author have shown that the non-linearity of HgCdTe, PbS
and PbSe detectors cannot be quantified in terms of the voltage output or indeed in
terms of the radiant power incident on the active areas of these detectors [6]. Instead, it
was shown that in the case of PC HgCdTe and PC PbSe detectors, it is the photon
irradiance which defines the linearity factor of these detectors [6, 8]. In the case of PC
PbS detectors the parameter of interest was shown to be the spectral irradiance [7]
which suggests that another effect such as the heating effect of the incident radiation
may have a significant role in the non-linear behaviour of PbS detectors. What is
particularly worrying to users of these detectors is that their non-linearity is significant
even for modest levels of incident irradiance. In the case of the PbS detector, for
example, a deviation from linearity of 1% was measured for an incident radiant power
of 1 nW concentrated in an area of 1 mm2 [7]. The author expects that some PbS
detectors currently being used in commercially available equipment are illuminated with
higher levels of spectral irradiance and are therefore likely to be operating in a nonlinear regime.
Fig. 1 shows a plot of the linearity factor of a PbS detector versus output voltage under
different illumination conditions (plots corresponding to different beam diameters and
wavelength of the incident beam). It is obvious that the non-linearity cannot possibly be
correlated to the output voltage without additional information. Fig. 2 shows the
corresponding data for a PbSe detector [8]. Again, the deviation from linearity cannot
be directly correlated to the radiant power incident on the detector. It is therefore clear
that attempts to correct for the effect of the non-linear response of these detectors based
on voltage output or even incident radiant power are bound to fail. The wavelength of
the incident radiation as well as the area of the detector being illuminated has to be
taken into consideration.
Linearity factor
2.2 μm, 1.8 mm spot
1.3 μm, 1.8 mm spot
2.2 μm, 1 mm spot
1.3 μm, 1 mm spot
2.2 μm, 0.76 mm spot
1.3 μm, 0.76 mm spot
Voltage output (V)
Fig. 1. Linearity factor of a Hamamatsu PbS detector as a function of the output voltage at 1.3 µm and 2.2
µm for 0.76 mm, 1 mm and 1.8 mm diameter spots illuminating the active area of the detector [7].
Linearity factor
2.2 μm, 1.78 mm
2.2 μm, 1.00 mm
2.2 μm, 0.78 mm
2.2 μm, 0.38 mm
4.7μm, 1.78mm
4.7 μm, 1.00mm
4.7μm, 0.78mm
Radiant power (μW)
Fig. 2. Linearity factor of the Judson PbSe detector as a function of the incident radiant power at 2.2 µm
and 4.7 µm wavelength, for different spot diameters [8].
In the case of the PC HgCdTe detectors, the non-linear behaviour can be explained on
the basis of the electron-hole recombination (Auger recombination) which is the
dominant loss mechanism of photo-generated charge carriers in this type of
photodetector. The density of the photo-generated charge carriers is proportional to the
photon irradiance. However, at higher carrier densities, there is a higher probability of
electron-hole recombination (Auger recombination) and therefore, as the photon
irradiance increases, the carrier lifetime decreases and therefore the non-linearity
Implications of the detector non-linearity in FT spectophotometry
Infrared spectrophotometry is now mostly being done using FT spectrometers because
spectrophotometers [12]. An FT spectrometer is basically a Michelson inteferometer
where the optical path difference between the passive arms can be varied. The radiant
power reaching the detector changes with path difference, resulting in the acquisition of
the interference function or “interferogram”. The voltage output of the photodetector
and hence the radiant power incident on it can be recovered by performing a Fourier
Transform on the interferogram. However, if the interferogram is not a true
representation of the radiant power incident on the detector due to (for example) the
response of the photodetector being non-linear, then the spectrum generated by the
Fourier Transform process of the interferogram will not be a true representation of the
spectral power reaching the photodetector.
How can the interferogram fail to be a true representation of the radiant power incident
on the photodetector? Fig. 3 shows a typical interferogram obtained from an infrared FT
spectrometer. It is characterised by small variations in the output except for some very
large variations near the zero path difference between the two interferometer arms. At
this position, the photodetector “sees” the full radiant output of the source captured
within the entendue/throughput [13] of the instrument. This is in contrast to a
monochromator-based spectrophotometer where the output at different wavelengths is
observed sequentially. The wavelength bandwidth of a monochromator-based system is
typically 1% of the working wavelength so the strength of the optical radiation incident
on the photodetector is (typically) two orders of magnitude lower than the signal
corresponding to the peak of an FT interferogram. The author has observed significant
non-linearities in radiometric measurements completed using monochromator-based
systems [7]. Since the photodetectors in FT-based spectrophotometers are illuminated
with values of radiant power which are two orders of magnitude higher than in a
monochromator-based system, it is safe to assume that spectrophotometric
measurements based on FT spectrometers can suffer from the effects of detector nonlinearity, unless precautions have been taken to prevent this from happening.
Photodetector output/AU
Mirror position/AU
3. FT interferogram demonstrating the higher output signal observed in the vicinity of the “zero path
length difference” between the two interferometer arms. (courtesy of C. J. Chunnilall, NPL)
Errors due to detector non-linearity in radiant power measurements using FT
spectrometers can be in excess of 30% [14, 15]. FT instrument manufacturers are aware
of these potential sources of error and market methods of correcting the errors which
can arise due to the non-linear response of infrared detectors. One method relies on
special preamplifier electronics to overcome the effects of detector non-linearities [16].
Others market software which promise to achieve the same goal [17]. However, a
correction method based on the electrical gain of a pre-amplifier assumes that the
detector non-linearity can be characterised in terms of the voltage output, something
which is contradicted by data publish recently [6 - 8]. A correction method based on
software algorithms also assumes that the conditions with which the photodetector is
illuminated are known, which, again, is not necessarily the case.
Fig. 4. The absorbance of a particular absorption band does not vary linearly with the concentration of tbutyl ether in CCl4. Deviation from Beer’s law is caused by the detector non-linearity (figure reproduced
from reference 18).
Fig. 4, reproduced from reference 18, shows how the absorbance of the 3.38 μm t-butyl
ether absorption band changes as the concentration of t-butyl ether in CCl4 increases.
These measurements were performed using an FT spectrophotometer utilising a
HgCdTe and a DLATGS pyroelectric detector. If Beer’s law was obeyed, then all the
plots in Fig. 4 should have been straight lines. Figure 4 shows that the plots exhibit
strong deviations from linearity, suggesting that Beer’s law is not obeyed. In fact
Griffiths and co-workers showed [18] that these deviations from linearity are caused by
the non-linear response of the photodetectors used. Plots using HgCdTe detectors
exhibit much more severe excursions from Beer’s law compared to the corresponding
plots acquired using DLATGS detectors. This is to be expected because the linearity
characteristics of HgCdTe detectors are considerably poorer compared to those of
DLATGS pyroelectric detectors. Fig. 4 shows that the concentration of t-butyl ether in
CCl4 will be estimated incorrectly even for concentrations as low as 2% and under
certain conditions, the measured absorbance will never exceed unity irrespective of the
concentration of t-butyl ether in CCl4. This demonstrates the importance of the detector
linearity in FT spectrometer based spectrophotometers. An extensive discussion on the
practical implications of the photodetector non-linearity on the accuracy of photometric
measurements acquired using FT spectrometers can be found elsewhere [18, 19].
Uncertainties due to drifts in the responsivity of infrared detectors
Some researches will argue that problems due to detector non-linearities can be
eliminated from FT-based spectrophotometric measurements by choosing detectors
such as InSb detectors which are believed to have a linear response. Unfortunately the
response of InSb detectors is limited to wavelengths shorter than 5.3 μm and HgCdTe
detectors remain (by far) the most widely used photon detectors capable of detecting
wavelengths longer than 5 μm at relatively high operating temperatures (77 K).
Furthermore, the range over which the response of InSb detectors is linear has also been
shown to be restricted [20]. Finally the responsivity of InSb detectors, like all cooled
detectors, has been shown to drift with time and this effect will also introduce
uncertainties in infrared spectrophotometric measurements. A brief summary of this
behaviour is presented below.
The performance of infrared photodetectors benefits greatly from cooling the
photodetectors to cryogenic temperatures [11]. For this reason, the vast majority of
infrared photodetectors are sold mounted in evacuated Dewars where they can be cooled
to temperatures below 80 K using cryo-coolants such as liquid Nitrogen or liquid
Helium. Work at NPL has shown that moisture can also go past the O-rings sealing the
Dewar window and evacuation port [21]. Once the Dewar is cooled, the Dewar cold
finger (which includes the infrared detector) acts like a cryo-pump so that the moisture
deposits as a thin film of ice on all cold parts of the Dewar, including the surface of the
infrared detector. Ice is known to have a number of strong absorption bands in the
infrared, so the film of ice deposited on the detector preferentially absorbs wavelengths
coinciding with the ice absorption bands [1, 4]. This causes the responsivity of the
infrared detector at wavelengths coinciding with the ice absorption bands to decrease
with time, causing drifts of many percent per minute at wavelengths around the 3.1 μm
absorption band [1], as can be seen from Fig. 5. Significant drifts have also been
observed around 12 μm where another strong ice absorption band is present [1, 3, 4].
Fig. 5 shows that the responsivity of this detector decreases by 30% over the period of
12 minutes and provides an indication of the magnitude of the expected drift. It should
also provide a warning to users of cooled infrared detectors and demand that infrared
spectrophotometric measurements utilising such detectors must include corrections as
well as uncertainty contributions to take into account the consequences of this
Normalised response
Time (h)
Fig. 5. Normalised response of an InSb detector monitoring radiation at 3.1 μm, immediately after
cooling to 77 K.
However, drifts are not restricted to wavelengths coinciding with the absorption bands
of ice. Ice has strong dielectric properties and when it is formed on the active area of
photodetectors, it interacts with the multi-layer dielectric anti-reflection coatings
deposited on the surface of the photodetectors. This causes the transmission of the antireflection coatings to change and thus the detector responsivities to drift with time [2,
3]. This effect was observed in InSb detectors [2], HgCdTe detectors [3] as well as in
silicon detectors (in the visible!) cooled to –40 oC. This implies that the responsivity of
a cryogenically cooled detector can be expected to drift for all wavelengths where the
detector has a finite responsivity.
The introduction of cold filters restricting the range of wavelengths reaching
cryogenically cooled infrared photon detectors and blocking some of the thermal
background radiation, results in a reduction in the total noise power at the output of the
detector [11]. This provides a well-established technique for improving the detector
Noise Equivalent Power (NEP) and specific detectivity (D*) of infrared detectors over
specific wavelength ranges [11]. The cold bandpass filters are mounted on the cold
shield of the detector so any moisture inside the Dewar vacuum deposits on the
bandpass filter rather than the detector which it shields. The bandpass filter is nothing
more than a stack of dielectric layers designed to have a specified transmission profile.
The deposition of a film of ice a few nanometres thick, interacts with the dielectric
structure of the filter and can alter its transmission characteristics, causing the response
of the filter/detector combination to change. The response of the filter/detector
combination was observed to decrease at some wavelengths and to increase at others
due to the deposition of ice. Numerical modelling and a dedicated experiment have
confirmed that the deposition of a thin film of ice on the dielectric bandpass filter is the
cause of the observed drifts [4].
In the absence of alternative good quality detectors, InSb detectors with low-pass filters
mounted on their cold shields are used extensively in the 1.6 µm to 2.6 µm wavelength
range. The filters are nothing more than glass plates which transmit wavelengths below
about 3 μm but block radiation of longer wavelength. Because they are cold (they are
mounted on the cold shield of the detectors) they block the thermal background
radiation of wavelength longer than 3 μm. When moisture is present in the Dewar
vacuum, it deposits on the cold glass plate and it behaves like a thin film of oil on the
surface of water. This means that as the thickness of the film of ice grows, the
transmission goes through maxima and minima [6]. The response of the filter/detector
combination exhibits an identical behaviour as can be seen in Fig. 6. The “oscillations”
are “fast” at first as there is a lot of moisture in the Dewar vacuum. As the moisture is
deposited as ice, it is gradually depleted, so the deposition rate decreases, causing the
frequency of the “oscillations” to reduce, as seen in Fig. 6.
Normalised response
Fig. 6. Normalised response of an InSb filtered detector monitoring radiation at 1.55 μm, immediately
after cooling to 77 K
Fig. 6 shows that the responsivity of this detector “oscillates” with a peak to peak
amplitude of approximately 14% and confirms that it can be an important uncertainty
contribution in infrared spectrophotometric measurements.
The drifts in the responsivity of cryogenically cooled detector are temporarily
eliminated after evacuation and baking of the Dewar at around 60 oC. It must be stressed
that the elimination of these drifts is only temporary. Drifts slowly reappear and they
grow in magnitude as moisture enters the Dewar vacuum past the O-ring seals.
Spectrophotometric measurements are based on the measurement of the radiant power
of a beam with and without the artefact inserted in the beam. These two measurements
can only be done sequentially and the period between these two measurements can vary
from a few seconds to many hours. In some field measurements, a measurement of the
reference (artefact out of the beam) is recorded at the beginning and end of the day
whereas all measurements during the day include the artefact in the beam. The work
described above indicates that such measurements can be seriously flawed due to the
drifts which can occur in the responsivity of infrared detectors. The extent of the
problem depends on the wavelength range of the measurements, the type of detector
used and on the type of Dewar in which the detector is housed [21].
Uncertainties due to drifts in the thermal background
Yet another effect which can give rise to significant drifts in the responsivity of infrared
detectors such as HgCdTe detectors has recently been reported [22]. The author has
shown that the responsivity of these detectors depends strongly on the unmodulated
(background) infrared spectral irradiance and therefore on the temperature of objects in
the FOV of these detectors. This effect was largest in PC HgCdTe detectors [22] and
was shown to arise due to the non-linearity of response characteristics of these
detectors. The incident radiation generates free charge carriers. Auger recombination is
a significant loss mechanism and this loss mechanism is known to rise with the
increasing density of charge carriers. This leads to the observed high non-linearity in the
response of HgCdTe detectors. Note that both the modulated radiation being detected as
well as the thermal background radiation can produce free charge carriers which
contribute to the non-linearity. The contribution due to the thermal background is
rejected by the phase sensitive detection utilised, but its contribution to the non-linearity
of response of the detector cannot be avoided. The irradiance due to the thermal
background was estimated to be over an order of magnitude greater than the irradiance
due to the modulated radiation being detected [22]. Thus the thermal background is the
dominant source of free charge carriers and is, therefore, the main component of the
observed non-linearity in the response of infrared HgCdTe detectors. This phenomenon
is expected to provide a practical limit to the accuracy of radiometric measurements
using not only HgCdTe detectors, but also other detectors whose linearity is a function
of the thermal background [22]. Fig. 7 shows the normalised response of a PC HgCdTe
detector at 4.7 μm versus time as the temperature of the FOV was raised in steps of
approximately 3 oC. Also plotted in the same figure are the values of the temperature of
the field of view. The responsivity of HgCdTe detectors was shown to decrease by
approximately 0.7% per oC increase in the background (ambient) temperature, but the
exact value of this change is governed by many factors such as the background
temperature, the FOV of the photodetector and detector spectral responsivity [22].
Observed drifts in the calibration of infrared radiometers can easily be assigned to this
phenomenon. This means that the accuracy of infrared spectrophotometric
measurements utilising cooled photon detectors such as HgCdTe detectors will be
affected by this phenomenon. The uncertainty contribution due to this effect must,
therefore, be estimated and included in the uncertainty budget of the relevant
Normalised response
Normalised response
Temperature ( C)
Time (h)
Fig. 7. Normalised response of the CMT9 detector at 4.7 μm versus time as the temperature of the FOV
was raised in steps of approximately 3 oC. Also plotted in the same figure are the values of the
temperature (reproduced from reference 22).
The main requirement of spectrophotometry is that the detection system has to have a
linear response over the range of incident radiant powers used for a particular
measurement. A photodetection system whose spectral responsivity does not drift is
also highly desirable. This paper shows that the identification of an infrared detector
with a linear response is not straightforward. Investigation of the linearity characteristics
of three types of infrared detectors has shown their responsivity to be highly non-linear
even for relatively low levels of radiant power incident on these detectors. In the case of
spectrophotometers based on Fourier Transform (FT) spectrometers the problems are
likely to be significantly worse because in the vicinity of the “zero path difference”
point, the detector receives the full spectral output of the source captured within the
instrument etendue, thus increasing the likelihood that the detector operates in its nonlinear regime. The responsivity of cooled infrared detectors was also shown to drift
significantly with time, thus requiring this drift to be quantified and included as a
correction and/or as an extra uncertainty component. Finally another phenomenon is
highlighted which can also generate apparent drifts in the responsivity of infrared
detectors. The origin of this source of drift is changes in the thermal background signal
incident on infrared detectors.
E. Theocharous and N.P. Fox, “Reversible and apparent ageing effects in infrared detectors”,
Metrologia, 40, S136-S140, 2003.
E. Theocharous, “On the stability of the spectral response of cryogenically cooled InSb infrared
detectors”, Applied Optics, 44, 6087-6092, 2005.
E. Theocharous, “On the stability of the spectral responsivity of cryogenically-cooled HgCdTe
infrared detectors”, Infrared Physics and Technology, 48, 175-180, 2006.
E. Theocharous, G. Hawkins and N.P. Fox “Reversible ageing effects in cryogenically-cooled
infrared filter radiometers”, Infrared Physics and Technology, 46, 339-349, 2005.
E. Theocharous, “On the drifts exhibited by cryogenically cooled InSb infrared filtered detectors
and their importance to the ATSR-2 and Landsat-5 earth observation missions”, Applied Optics,
44, 4181-4185, 2005.
E. Theocharous, J. Ishii and N. P. Fox, “Absolute linearity measurements on HgCdTe detectors in
the infrared”, Applied Optics, 43, 4182-4188, 2004.
E. Theocharous, “Absolute linearity measurements on PbS detectors in the infrared”, Applied
Optics, 45, 2381-2386, 2006.
E. Theocharous, “Absolute linearity measurements on a PbSe detector in the infrared”, accepted
for publication in Infrared Physics and Technology, May 2006.
E. Theocharous, N. P. Fox and T. R. Prior, ‘A Comparison of the performance of infrared
detectors for radiometric applications’, Optical Radiation Measurements III, Proc. SPIE, 2815, 5669, 1996.
C. L. Sanders, "Accurate Measurements of and Corrections for Nonlinearities in Radiometers", J.
Res. Nat. Bur. Stand, 76A, 437-453 (1972).
E Theocharous and J. Birch, “Detectors for Mid and Far Infrared Spectroscopy: Selection and
Use” was published in The Handbook of Vibrational Spectroscopy, J.M Chalmers and P.R.
Griffiths (editors), Volume 1, published by John Wiley and Sons, pages 349-367, 2002.
J. Chamberlain, “The principles of interferometric spectroscopy”, J Wiley & sons, Chichester,
page 15, 1979.
W.L. Wolfe and G.J. Zissis, (editors), “The Infrared Handbook”, Published by the Office of Naval
Research, Washington, 3rd edition 1989.
K. Rahmelow, "Electronic influences on an infrared detector signal: nonlinearity and amplification
," Applied Optics, 36, 2123-2132, 1997.
M. C. Abrams, G. C. Toon and R. A Schindler, “Practical example of the correction of Fourier
transform spectra for detector nonlinearity”, Applied Optics, 33, 6307-6314, 1994.
R. Curbelo, “Techniques for correcting non-linearity in a photodetector using predefined
calibration information”, U. S. Patent 5262635 16th November 1993.
A Keens and A Simon, Correlation of non-linearities in detectors in Fourier Transform
spectroscopy”, U. S. Patent 4927267, 22nd May 1990.
R. L. Richardson, H. Yang and P. R Griffiths, Effects of detector non-linearity on spectra
measured on three commercial FT-IR spectrometers”, Applied Spectroscopy, 52, 572-578, 1998
R. L. Richardson, H. Yang and P. R Griffiths, “Evaluation of a correction for photometric errors in
FT-IR spectrophotometry introduced by a nonlinear detector response”, Applied Spectroscopy, 52,
565-571, 1998.
E. Theocharous, “Absolute linearity measurements on InSb detectors in the infrared”, to be
E. Theocharous, “Drift in cryogenically cooled photodetectors: causes and cures”, April issue of
Photonics Spectra, 70-72, 2006.
E. Theocharous and O. J. Theocharous “On a practical limit of the accuracy of radiometric
measurements using HgCdTe detectors” to be published in 20th October issue of Applied Optics
Vol. 45, 7753-7759, 2006.
© Crown copyright 2007. Reproduced by permission of the Controller of HMSO and Queen’s Printer for
This work was supported by the National Measurement System Policy Unit of the Department of Trade
and Industry.
Integrated Sensors for Point of Care Detection
John de Mello
Dept. Chemistry, Imperial College London, SW7 2AY, United Kingdom
Microfluidic devices have shown themselves to be highly effective for laboratory-based
research, where their superior analytical performance has established them as efficient
tools for genetic sequencing, proteomics, and drug discovery. However to date they
have not been well suited to point-of-care applications, where cost and portability are of
primary concern. Although the chips themselves are cheap and small, they must
generally be used in conjunction with bulky optical detectors, which are needed to
identify or quantify the analytes or reagents present. Here we report the use of miniature
on-chip light sources and photodetectors based on light-emitting polymers (LEPs). LEP
devices have simple multilayered structures and may be fabricated directly on the
microfluidic chips by sequential deposition of appropriate polymers or electrodes. The
LEPs add minimal size and weight to the microfluidic chips, allowing for the creation of
low cost, quantitative, integrated diagnostic devices.
The Multi-Mode Optical Spectrometer : Towards the Simultaneous Measurement
of Absorption, Transmission, Turbidity, Linear Dichroism, Optical Activity,
Fluorescence (Normal, Linearly and Circularly Polarised) and Light Scattering.
Dr Alex F Drake
King's College London, Strand, London, WC2R 2LS, UK
In the 1980s, Drake reviewed [1,2,3] the various applications of polarisation modulation
in chemical spectroscopy with particular reference to the photoelastic modulator. In a
more recent paper [4], the basis of a multi-mode polarisation modulation spectrometer
was described.
This paper reviews our work to date and will report more recent results of
measurements in the fluorescence and light scattering modes.
The observations reported in this paper are being realised using the Chirascan
Spectrometer available from Applied Photophysics Ltd. Applications include protein
and nucleic acid conformation, binding studies and in-line monitoring.
A.F. Drake. “Polarisation modulation: linear and circular dichroism”. J. Phys. E. Sci. Instrument, 19,
170, 1986
A.F. Drake. “Chiroptical Spectroscopy”. Physical Methods of Chemistry - Volume 3B,
Determination of Chemical Composition and Molecular Structure, pp. 1-44, John Wiley & Son Inc.,
C.J. Barnett, A.F. Drake, S.F. Mason. J. Chromatogr. 202, 239, 1980
T. Arvinte, T.T.T. Bui, A.A. Dahab, B. Demeule, A.F. Drake, D. Elhag and P. King. “The multimode polarization modulation spectrometer: part 1: simultaneous detection of absorption, turbidity,
and optical activity”. Analytical Biochemistry, 332, 46–57, 2004
Posters Presented
Characterisation and Performance Validation of Spectral Scanning Fluorescence Microscopes with
Easy-to-Use Fluorescence Standards. U. Resch-Genger1, K. Hoffmann1, and R. Nitschke2.
Federal Institute for Materials Research and Testing (BAM), Berlin, Germany
Albert-Ludwigs-Universität, Life Imaging Centre, Institute of Biology I, Freiburg, Germany
SRMs 2241-2244: Relative Intensity Correction Standards for Raman Spectroscopy.
S. Choquette.
National Institute of Standards and Technology, Gaithersburg, USA
SRMs 2035, 2065 and 2036: UV-VIS-NIR Wavelength/Wavenumber Transmission and Diffuse
Reflectance Standards Utilizing a Rare Earth Oxide Glass. S. Choquette.
National Institute of Standards and Technology, Gaithersburg, USA
A Review of Integrating Sphere Measurements of the Optical Properties of Materials.
R. Yeo.
Pro-Lite Technology, UK
The Maintenance and Continuation of NPL's Mid-Infrared Standard Measurement Services. B.
Optical Radiation Measurement Team, National Physical Laboratory, UK
A Thorough Characterization of the Absolute Radiation Thermometer for the Determination of
Thermodynamic Temperatures. R. Winkler, E. Woolliams, B. Hartree, M. White, S. Salim, N. Fox.
Optical Radiation Measurement Team, National Physical Laboratory, UK
Optical Radiation Measurement Team. T. Burnitt.
Optical Radiation Measurement Team, National Physical Laboratory, UK
Photodiode Amplifier Developments and Best Practice. J. Mountford.
Optical Radiation Measurement Team, National Physical Laboratory, UK
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF