See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/267595834 A New Calibration Method and Device for Certified Flow Measurements With Laser Velocimetry Conference Paper · November 2012 DOI: 10.1115/IMECE2012-93100 CITATIONS READS 0 183 4 authors, including: Katsuaki Shirai Jürgen Czarske Shibaura Institute of Technology Technische Universität Dresden 78 PUBLICATIONS 183 CITATIONS 392 PUBLICATIONS 1,868 CITATIONS SEE PROFILE SEE PROFILE Carsten Kykal TSI Inc. 19 PUBLICATIONS 37 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Multi Instrument Manager (MIM) Software Solution View project Laserbasierte tomographische Messung der lokalen akustischen Impedanz von überströmten Linern (TOMLIM) View project All content following this page was uploaded by Carsten Kykal on 12 June 2017. The user has requested enhancement of the downloaded file. Proceedings of the ASME 2012 International Mechanical Engineering Congress & Exposition IMECE 2012 November 9-15, 2012, Houston, Texas, USA IMECE2012-93100 DRAFT: A NEW CALIBRATION METHOD AND DEVICE FOR CERTIFIED FLOW MEASUREMENTS WITH LASER VELOCIMETRY Katsuaki Shirai∗†, Lars Büttner, Jürgen Czarske Laboratory of Measurement and Testing Techniques Carsten Kykal TSI GmbH Faculty of Electrical and Computer Engineering Technische Universität Dresden Email: juergen.czarske@tu-dresden.de 52068 Aachen, Germany Email: carsten.kykal@tsi.com ABSTRACT We aim to establish traceability at calibration and hence to enable a certified flow measurement with a calibrated measurement system. A new calibration method is presented for laser velocimetry. We develop a simple, unique method which establishes traceability of its uncertainty. The device is transportable and calibratable by any users for their own instruments on-site. Our new method requires only a rotating disk and a precision linear stage providing positional information. In former calibration methods, the uncertainty of the orbit radius of a scattering object was dominant due to the diffficulty of accessing the true center of the rotation. The diffuculty was solved in our new method. The new method provides an accurate estimate of the orbit radius and hence the velocity of the calibration object through a linear regression. The calibration constant is obtained even without the need of direct access to the rotation radius. The uncertainty budget is examined throughout the calibration procedure. The traceability chain is established once the traceabilities are maintained to the translation stage and the motor used for rotating the calibration disk. The new method has been realized with three different calibration setups and their performances were investigated. We demonstrate that the new calbration method can achieve reasonable uncertainty values. NOMENCLATURE A B C0 d f frot M N r u ucal x x0 xj z α δ Λ λ ω σ σj ξ c ∗ Address all correspondence to this author. † Present affiliation: Laboratory of Energy-Conversion Technology, Kobe University, Rokkodai 1-1, Nada, 657-8501, Kobe, Japan, Email: shirai@mech.kobe-u.ac.jp cal j, k sc 1 slope of the linear fit [1/(m·s)] intercept of the linear fit [1/s] constant fringe spacings [m] Doppler frequency [Hz] rotation frequency [Hz] number of locations for local Gaussian fit [m] number of scattering objects radius or radial coordinate [m] velocity of tracer particle [m/s] calibration velocity [m/s] stage coordinate [m] offset of the stage coordinate [m] constanc in the Gaussian fit [m] axial coordinate (optical axis) [m] half-angle between the two laser beams of LDV [rad] parameter variation value defined in Eq.(21) [m2 ] laser wavelength [m] angular velocity of rotation [rad/s] uncertainty (standard deviation) constant in the Gaussian fit [m] SNR or amplitude of Doppler signals center values calibration values j, k-th values stage-coordinate values c 2012 by ASME Copyright ⃝ INTRODUCTION Quality control gains more attentions in production fields, which leads to the rapid increase of sensor developments and applications. Sensor technologies are expected to play a major role in the future and the sensor market will show futher growth in industry [1]. Measurement is the basis for production control together with tolerance. Tolerance values are managed through quantitative measurements using various sensors. Nowadays, the interexchange and combination of tolerance values are required between different components produced in different places even beyond borders and continents with the globalization of modern economy. Hence, establishment of traceability in measurements obtains higher attentions in production field. Flow measurement is not an exception. Many different techniques are available for knowing flow velocities. Modern laser diagnostics allow nonintrusive measurements without inserting a physical probe into a flow such as the location behind rotator blades in turbomachinery. Laser Doppler velocimetry (LDV) has been widely used as a well-established diagnostic tool in fluid mechanics [2] since it was invented by Yeh and Cummins [3]. LDVs are used for measurements requiring relatively high resolutions of time and space without disturbing the flow. They are capable of measuring multiple velocity components with exceptionally high dynamic ranges in unsteady flows. Developments of laser Doppler techniques is continued toward achieving high spatial resolutions into the range of micrometers [4]. LDV is also known to have small measurement uncertainty when it is properly used. Besides, LDVs are even set for transferring velocity standard of gas flowrate in some countries because of their wide ranges of linearity. In this paper, we propose a new calibration method toward establishing a traceability of uncertainty in flow measurements using laser velocimetry such as an LDV. The calibrated velocimetry is applied to flow measurements. Our challenge is to establish measurement traceability at the calibration. The new calibration method is based on the linear relationship between the Doppler frequency and the orbit radius of scattering object attached on a disk rotating at a constant angular velocity. Calibration uncertainty of the method is examined based on the Guide to the expression of Uncertainty in Measurement (GUM) [9]. We show the measurement uncertainty is traceable with this new method as long as the traceabilities are achieved for the two fundamental elements required in the system. Three different setups of the new calibration are realized and their experimental results are examined. Measurement Volume x time u d α α z 1/f Doppler Burst Signal Laser Beams (wavelength: λ) Tracer Particle FIGURE 1. Schematics of the measurement volume and Doppler signal of a laser Doppler velocimetry (LDV). particles are seeded into the flow in advance or naturally existing particles may function as tracer particles. The local flow velocity is measured by evaluating the Doppler frequency shifts of the particles. Fig. 1 depicts schematics of the measurement volume and a Doppler burst signal of a typical differential LDV. The measurement volume is formed at the intersection of a pair of coherence laser beams. The resulting interference fringes become nominally parallel in the direction of optical axis when the beams intersect at their focal points. The fringe spacings d of this nominally parallel fringes can be described as d= λ , 2 sin α (1) being λ and α are the wavelength and the half crossing angle, respectively. When a tracer particle passes through the fringes, it scatters the incident laser beams. The scattered light generates modulated signal called Doppler burst such as in the right side of Fig. 1. The beat frequency f of the Doppler signal is proportional to the physical velocity u perpendicular to the bisector plane of the fringes u f= . d (2) Hence, the velocity of the particle is obtained by measuring the Doppler frequency. As long as the tracers follow the fluid motions, the local flow velocities in the measurement volume are obtained. Calibration of LDV The calibration can be carried out using a scattering object which has a constant, well-defined local velocity. With a known velocity ucal of the scattering object, the nominal value of the fringe spacings d is precisely determined from the measured Doppler frequency as LDV AND ITS CALIBRATION Laser Doppler velocimetry (LDV) The principle of an LDV is based on the fact that the Doppler beat frequency of the scattered light by a small tracer particle is proportional to its velocity of the physical movement. The d = ucal / f, 2 (3) c 2012 by ASME Copyright ⃝ which is called as a calibration factor. In other words, the calibration of an LDV is equivalent to obtain the nominal value of the constant fringe spacings d in Eq.(2). Hence, velocity measurement is identical to the reproduction of the calibration velocity. The uncertainty originating from the calibration induces systematic contribution to the total uncertainty budget of the velocity measurements. Indeed, the calibration uncertainty is crucial for laser velocimetries, especially when the traceability is demanded in the measurement. Ultimately, the calibration eventuates the realization of a constant known velocity ucal of a scattering object. In the metrological field, LDV has been set as a primary velocity standard for gas flows in seveal countries such as the USA, Japan and Germany [5–7]. They are all based on a nominally constant velocity realized by a scattering object attached on a rotating system. The tangential velocity of the scattering object is expressed as ucal = rω = r2π frot . Scattering Object (e.g., Pinhole) True Center (not known in advance) r Disk ( ∂d σf x=0 FIGURE 2. Principle of the new calibration method using multiple scatterin objects (r: orbit radius, x: stage coordinate). (4) uncertainty as systematic one. The estimate uncertainty of the frequency can be described based on the theory of Crámer-Rao’s lower bound (CRLB) [10], depending on the sample rate, the record length of the signal and the signal-to-noise ratio (SNR). The fractional uncertainty of Doppler frequency σ f / f becomes usually around in the order of 10−4 ∼10−3 with a typical set of conditions of the parameters. The uncertainty of the calibration velocity σucal appearing in the second term is further decomposed into √ ( ∂u )2 )2 ( ∂u cal σ frot + σr ∂ frot ∂r √ ( σ )2 ( σ )2 frot r + . = ucal frot r σucal = ucal (6) (7) The uncertainty of the calibration velocity becomes from 0.6 % to 1.7 % as a sense of single standard deviation σucal in conventional calibration methods using a single scattering object [8]. This large uncertainty value originates from that of the orbit radius of the scattering object. The uncertainty comes from the difficulty to access the true rotation radius under the fabrication tolerances and the plays required between some elements such as the one between the disk and the rotating shaft. Compared to the radial uncertainty, the disk rotation and frequency estimate typically ranges around 0.05 %, which are one order of magnitude smaller the radial one. The motor rotation can be stablized by a phase-lock-loop and an encoder, while the estimate uncertainty of the frequency is limited by the bound mentioned in the above. Hence, the reduction of the radial uncertainty at the calibration is the most effective strategy in order to reduce the total uncertainty of the calibration. )2 ( ∂d )2 σucal + ∂ucal ∂f √ )2 ( ) 1 ( ucal σ f + σucal 2 . = f x xj Unknown Offset: x0 Calibration Uncertainty The calibration uncertainty, which is equivalent to the relative uncertainty of thre fringe spacings σd /d, is analyzed along the GUM. In the latest version of the GUM, the uncertainties are classified into type A in the form of a standard deviation, and type B such as a tolerance or a worst case estimate with a lower and an upper limit bound. Eq.(3) already shows that the calibration depends on two sources of uncertainties – the velocity of a scattering object and the frequency estimate of Doppler signals. The uncertainty of the fringe spacings can be expressed by following the law of uncertainty propagation being the uncertainty elements assumed to be independent, √ uj =rjω fj rj being r , ω, frot the orbit radius, angular velocity and rotation frequency, respectively. The calibration devices in the national metrology insitutes tend to become bulky and not designed to be transported for on-site calibrations. Outside of national metrology institutes, a limited number of sysmtems have been tested and used for calibrating laser velocimetries. Such methods with their performance evaluations are summarized by Shirai (2011) [8]. σd = d f (5) The first term in the square root of the right-hand side of the equation is the statistical uncertainty but it contribute to the final 3 c 2012 by ASME Copyright ⃝ NEW CALIBRATION METHOD We propose a new calibration method for accessing the inaccessible orbit radius of a scattering object. The new method provides the calibration constant even without the need of accessing the radius. Principle The new method is based on the linear relationship between the tangential velocity and the true orbit radius as schematically shown in Fig. 2. We use multiple, pointwise scattering objects (i.e., pinholes) attached on a single rotating disk. The angular velocity of the disk is assumed to be constant. Each of the multiple scattering objects can act as a single scattering object with the Doppler frequency depending on the orbit radius. A linear precision stage is utilized to mechanically shift the pinholes in the radial direction (see Fig. 3). The radial shift between the successive scattering positions is known with relatively small uncertainty from the displacement of the precision stage. As long as the disk rotates at a constant angular velocity, the tangential velocity of the pinholes linearly increases depending on the radial position from the rotation center. Hence, it provides the exact rotation radius of a scattering object with a simple linear extrapolation. This is possible because the unknown value is included commonly. Even without knowing the commonly included unknown value, the differential distance between respective orbits can be precisely recorded using a linear precision stage with either an encoder or a displacement sensor. The mathematical expression of the calibration method is described as following. We use N pieces of single scattering objects ( j=1,. . .,N) and the resulting Doppler frequency f j at j-th object is proportional to the orbit radius r j fj = 2π frot r j. d FIGURE 3. Schematic of the new calibration method with a combination of rotating disk and a linear precision stage. The offset x0 is an unknown but it is commonly included in the stage coordinate. The Dopppler frequencies of the scattering objects are acquired while their locations are recorded in the stage coordinate. The resulting position-frequency pairs are fitted with a linear regression using Eq.(9). The linear fit provides the unknown constants A and B, and hence the initially uknown offset x0 . Once the offset value is known, the actual radius r is known r j = x j − x0 , The fringe spacings d is obtained as d= 2π frot , A Uncertainty of New Method The uncertainty of the new method is examined in the following. The calibration velocity is written as (9) The constants A and B are ucal = rcal ω = (xcal + x0 )ω, 2π frot , d 2π frot B= x0 . d B where x0 = − . A A= (14) where the rotational frequency frot is known. Since the constant A is the slope of the linear regression, the fringe spacing is obtained even without the need of knowing the offset value. The stage coordinate can be known with a relatively small uncertainty by means of a position sensor or an encoder embedded in the linear stage. (8) In fact, the true orbit radius is inaccessible. Instead, we have the radial displacements from the linear precision stage. Since the linear stage is traversed in the radial direction, the relationship between the Doppler frequency and the stage coordinate x j becomes linear f j = Ax j + B. (13) (10) (15) with rcal and xcal being the radius and stage coordinate values at the calibtation. Hence, the radial uncertainty becomes (11) σrcal = (12) 4 √ ( σ xcal )2 ( )2 + σ x0 , (16) c 2012 by ASME Copyright ⃝ being σ x the uncetainty of the stage coordinate, equivalent to the positional uncertainty such as unidirectional/bidirectional repeatability. The stage coordinate and the offset are assumed to be statistically independent here. The offset uncertainty is further expressed as Mounted Pinholes Rotating Disk Nomial Center of Rotation √ ( σ x0 )2 ( )2 ∂x0 ∂x0 σA + σB = ∂A ∂B √ ( )2 )2 (B 1 σA + − σ B . = A A2 Pinhole (17) FIGURE 4. Multiple pinholes attached on the chopper blade of a commercial optical chopper in the first experiment. (18) The estimation uncertainty of the fit constants are provided as [11] √ σA = σ f σB = σ f √ N , Λ (19) Σx j 2 , Λ (20) where Λ = N N ∑ j=1 2 N ∑ x j 2 − x j . Optical Chopper Blade (21) j=1 Rotating Disk (Pinholes clamped inside) where the σ f is the uncertainty of the Doppler frequencies being in the order of 10−4 ∼10−3 at a typical condition. The remaining is the uncertainty σ x j of determining the respective scattering location. It depends on the realization described in the second experiment. The resulting uncertainty of the claibration velocity follows Eq.(7) in the same way as the single scattering object σucal = ucal ucal √ (σ rcal rcal )2 ( σ )2 frot + . frot FIGURE 5. The rotating disk in the second experiment. It contained five pieces of unmounted pinholes clamped inside. The white labels with numbers indicate the pinhole number 1–5 from inside to outside. whose rotation was regulated using a phase-lock-loop. A simple LDV was built using a laser diode as a laser source and a diffractive grating for splitting beams. The nominal value of the fringe spacings was 4.5 µm with the working distance of 50 mm. A non-automated manual linear stage was employed and a laser triangulation sensor (Micro-Epsilon: ILD1700-10) was used together. The position sensor monitored the displacement of the linear stage with a precision of 0.5 µm. The feature of this experiment was the determination of the scattering location using an oscilloscope with visual inspection of the signals by the experimentalist. This half manual method caused certain ambiguity, which would be improved in the second experiment. The resulting variation of the fringe spacings regarded as the uncertainty became 0.24 % [8]. (22) Finally, the calibration uncertainty is obtained through Eq.(5). Feasibility Experiments The new calibration method was realized with three different setups. Calibration experiments were carried out in order to examine their feasibilities and performances. First Experiment The first setup was built only for checking the fundamental feasibility of the method. Five pieces of mounted pinholes with an aperture diameter of either 2 µm or 5 µm (Edmund Optics: mounted precision pinholes) were attached to a specially fabricated disk such as shown in Fig. 4. The disk was attached to an optical chopper (Thorlabs: MC1000), Second Experiment The second experiment had two improvements compared to the first one – in terms of axial uncertainty of the pinholes and the scattering locations. In constrast to the mounted pinholes, unmounted pinholes of 2 µm aperture diameter (Edmund Optics: unmounted precision pinhole) were 5 c 2012 by ASME Copyright ⃝ Linear Actuator with an Integrated Encoder Rotating Disk attached with 5x Pinholes Self-Made LDV System Detection Optics (Forward Scattering) Measurement Volume FIGURE 6. The second calibration experiment using multiple pinholes and motorized lienar precision stage. FIGURE 7. Local Gaussian fit for determining the accurate, reproducible scattering point inside the measurement volume. clampted flush to the flat inner surface of the disk as shown in Fig. 5. This way of mounting pinholes eliminated the axial ambiguity of the scattering locations. The other improvement was the determination method of the scattering location in the stage coordinate. In order to determine the precise location of every scattering object, a stable referece point is required inside the measurement volume. We chose the center of the measurement volume as the refernece point, since the optical intensity becomes maximum and hence the resulting Doppler signals always exhibit the largest amplitude and the highest signal quality at the center. Taking into account of these features, we recorded both the SNR and the amplitude values of the Doppler singals in time domain. The radial traverse was done at small steps of the stage coordinate around each pinhole such as shown in Fig. 7. Either the recorded SNR or amplitude values were fitted with the stage coordinate values assuming the Gaussian distribution ( ξ = C0 exp x j − x jc σj δξ = (24) The first two terms can be neglected since the variation of the first two terms should be sufficiently small compared to that of the parameter x j in the third term. Hence, the variation of the x j becomes δx j = δξ ∂ξ ∂x j . (25) The variation is calculated for all the local points around j-th scattering object (k=1,. . .,M) and the resulting uncertainty for the x j is obtained as an average )2 , ∂ξ ∂ξ ∂ξ δC + δσ + δx . ∂C0 0 ∂σ j j ∂x j j √ (23) σx j = where ξ is either the SNR or the amplitude values of the Doppler signals. The parameters C0 and σ j are constants. Both the SNR and the signal amplitude should take maximum value at the center of the measurement volume. Hence, the scattering center of the j-th object x jc is the position where the Gaussian distribution becomes locally maximum. The maximum point is obtained through an interpolation either nonlinear Gaussian fit or simple three-point parabolic fit in natural logarithmic domain commonly used for Doppler frequency estimate of an LDV [2]. This Gaussian fit provides very precise locations of the scattering points in the final calibration and hence reduced the umbiguity of the procedure in the former visual method. The determination uncertainty of the scattering center is calculated through a linear approximation, neglecting the higher order terms ∑M k=1 (δ x j ) M 2 . (26) Once this uncertainty of the scattering position is determined, the uncertainty of the calibration radius is obtained through Eq.(16). Further details of the other potential uncertainty sources not mentioned in the present paper are discussed in Shirai (2010) [8]. Fig. 6 shows the setup of the second experiment. Four pieces of the pinholes out of the five functioned properly during the experiment. We used a motorized linear precision stage with an integrated encoder (Newport Corportation: CONEX TRA25CC). The positional uncertainty of the stage was σ sc =2µm according to the specification sheet of the manufacturer. The typical result of Gaussian fit for the SNR and the amplitude are shown in Fig. 8. The SNR was expected to be more robust than the amplitude of the signal in time domain. In contrast to this expectation, we 6 c 2012 by ASME Copyright ⃝ (a) SNR 18 original data Gaussian fit fitted center 17 16 AMP SNR [dB] TABLE 1. List of the relative uncertainty values for respective pinholes 2–5 (pinhole 1 did not function) based on SNR and amplitude (AMP) values of the signals. (b) Amplitude original data Gaussian fit fitted center 15 14 13 12 11.58 11.6 11.62 11.64 11.66 11.68 11.7 11.72 0.24 0.22 0.2 0.18 0.16 0.14 0.12 0.1 0.08 11.58 11.6 11.62 11.64 11.66 11.68 11.7 11.72 x [mm] σr /r [%] x [mm] FIGURE 8. Comparison of the Gaussian fit in the second experiment, (a) SNR, (B) amplitude. σd /d [%] pinhole♯ SNR AMP SNR AMP 2 0.767 0.190 0.961 0.190 3 2.501 0.183 0.910 0.183 4 0.256 0.172 0.912 0.172 5 0.980 0.144 0.843 0.144 1.35 f [MHz] 1.3 four pinholes initial linear fit iterative linear fit 1.25 1.2 sult. However, this way of the realization is the simplest and may be the most realistic without complication, because an off-theshelf LDV system with a backward-scatter detection mode can be directly applied without any modification to the system. A commercial LDV system with a fiber-optic probe (TSI: TR260) with an air cooled Ar laser for two velocity-component measurement was used for this experiment (see Fig. 10). A blackpainted disk was attached to a DC motor (Maxon Motor: ECmax 30). Only the blue-beam pair (λ=488 nm) was used in this experiment and the scattered light was detected with a photomultipleier unit. The Doppler frequencies were direclty caliculated by an equipped hardware processor (TSI: FSA3500). One of the typical results is shown in Fig. 11. Not all the measurement points are aligned on the perfect linear line, hence only the points on the line were used for the linear fit. The selected points seem to be on an almost perfect linear line. An iterative procedure included in the fitting algorithm has slightly improved the fit result but the improvement is hardly discernible in comparison with the simple linear fit. The calibration was done with the optical chopper operated at frot =20 Hz and the working distance of the probe at 261 mm. The resulting fringe spacing from this condition (Fig. 11) was d=2.52±0.06 µm, which indeed covers the specification value of d=2.55 µm for this probe. The relative uncertainty becomes 2 % in the sense of single standard deviation in this case. This relatively large uncertainty was attributed to the scattering which took place at multiple locations. All the illuminated locations inside the measurement volume scattered the light, so the scattering locaton was not fixed to a single point such as in the case of the second experiment in which pinholes were used as single scattering objects. Hence, the scattering location was determined with a large uncertainty around the half diameter of the laser beams at the measurement volume. In spite of the uncertainy, the realization of the third experiment was the simplest compared to the other ones and its feasibility was confirmed accordingly. 1.15 1.1 39 40 41 42 43 44 45 r [mm] FIGURE 9. Linear fit of the Doppler frequency to the stage coordinate of the four pinholes in the second experiment. The local scattering positions were determined by using the signal amplitudes in this figure. observed that the amplitude provided more reliable center in the present experiment (compare the two plots in Fig. 8). The resulting linear fit of the Doppler frequency to the stage coordinate of the four pinholes is shown in Fig. 9. Based on the determined orbit radii of the four pinholes, calibration was done for the respective pinholes based on the SNR and the amplitude, respectively. The radial uncertainty σr /r was determined and the fringe spacings were caliculated with their uncertainties σd /d. The results are summarized in Tab. 1. The resulting fringe spacings value were consistent each other between the ones determined based on the SNR and the other on the amplitude (AMP), although the ones based on the SNR has larger scatters of the measurement points in Fig. 8. The resulting uncetainty values of the fringe spacings have been reduced into 0.14 %–0.19 %. This improvement is achieved by the local Gaussian fit for determining the scattering locations with small uncertainty. Third Experiment The scattering object is not limited to pinholes. The third calibration setup was built in a more simple way. The pinholes were replaced with the surface roughtness of a disk. The scatteirng may occur everywhere on the illuminated area and hence the resulting Doppler signal may be the superposition of multiple scaterings. This would increase the ambiguity of the positional determination as we would see in the uncertainty re7 c 2012 by ASME Copyright ⃝ Rotating Disk (The surface was painted black color.) ACKNOWLEDGMENT This work was partly supported by the Deutsche Forschungsgemeinschaft (CZ55/20-1,2). The authors thank Mr. Janarthanan Gnanaprakasam for his work in the second experiment. The calibration method and the devices reported in the present document is patent pending in the world wide. Please contact with the Knowledge and Technology Transfer (GWT) at TU Dresden for your further interest on technology transfer. Optical Chopper (Thorlabs Inc.: MC1000) Measurement Volume DC Servo Actuator (Newport Corpotation: CONEX TRA25CC) Commercial LDV Probe (TSI Inc.: TR260) REFERENCES [1] Andersen, P.D., Jørgensen, B.H., Lading, L., Rasmussen, B., 2004. “Sensor foresighttechnology and market”. Technovation, Vol. 24, pp. 311–320. [2] Albrecht, H.E., Borys, M., Damaschke, N., Tropea, C., 2003. Laser Doppler and Phase Doppler Measurement Techniques. Springer, Berlin. [3] Yeh, Y., Cummins, H.Z., 1964. “Localized fluid flow measurements with an He-Ne laser spectrometer”. Appl. Phys. Lett., Vol. 4, pp. 176–178. [4] Czarske, J., 2006. “Laser Doppler velocimetry using powerful solid-state light sources”. Meas. Sci. Technol., Vol. 17, pp. R71–R91. [5] Yeh, T.T., Hall, J.M., 2008. “Airspeed calibration service”. NIST Special Publication 250–79 (34 pages). [6] Kurihara, N., Terao, Y., Nakao, S., Takamoto M., 2005. “An uncertainty analysis of laser Doppler velocimeter calibration for air speed standard (in Japanese)”. J. Trans. JSME. Series B, Vol. 71, pp. 136–143. [7] Müller, H., Kramer, R., Strunck, V., Mickan, B., Dopheide, D., 2001. “Laser-Doppler-Anemometer zur Darstellung und Weitergabe der Einheit, Strömungsgeschwindigkeit (in German)”, 9. Fachtagung, Lasermethoden in der Strömungsmesstechnik: Deutschen Gesellschaft für LaserAnemometrie (GALA), Winterthur, Switzerland, 18–20th September 2001, pp. 24.1–24.8. [8] Shirai, K., 2010. “Development and application of novel laser Doppler velocity profile sensors for high spatially resolved velocity measurements in turbulent shear flows”. Dissertation, Dresdner Berichte zur Messsystemtechnik, ed.: Czarske, J., Vol. 3, Shaker Verlag, Aachen, Germany. [9] BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML, 1995. “Guide to the expression of uncertainty in measurement”. 2nd edn., International Organization for Standardization (ISO). [10] Rife, D.C., Boorstyn, R.R., 1974. “Single-tone parameter estimation from discrete-time observations”, IEEE Trans. Inform. Theory, Vol. 20, pp. 591–598. [11] Taylor, J.R., 1997. An Introduction to Error Analysis (2nd ed.), University Science Books, New York. FIGURE 10. The third calibration experiment using a commercial LDV system and the surface scattering of a rotating disk. 1.6 f [MHz] 1.4 1.2 original points points used for analysis initial polyfit iteration 1 0.8 0.6 0.4 0.2 0 0 5 10 15 20 25 30 35 r [mm] FIGURE 11. Typical result of the lienar fit with the surface scattering in the third experiment. Conclusion We developed a new calibration method for laser velocimetry such as an LDV toward establishing the traceability of uncertainty in measurements. As the measurement is equivalent to reproduce the calibration velocity influenced by the accompanying systematic uncertainties, an eatablishment of the uncertainty at the calibration is required. The principle of the new method was described and its uncerainty was analyzed along the GUM. The analysis showed that the measurement uncerainty can be traced throughout the caliration. Hence, it is possible to realize a qualified measurement with this new method as long as the traceabilities are maintained to the translation stage and the motor. The new calibration system was tested in the three experiments with the differnet realizations. The experiments indicated that the calibration uncertainty of down to 0.14 % can be achieved with one of the setups. It was demonstrated that the new method was also compatible with an off-the-shelf commercial LDV without any modification. Development is continued toward the automation and the improvement of the system with a further smaller uncertainty. This new portable calibration method provides measurement traceability to the end-users of laser velocimetries for their own measurements which requires low effort but still establishes tracerable uncertainty. 8 View publication stats c 2012 by ASME Copyright ⃝