This action might not be possible to undo. Are you sure you want to continue?
J. Opt. Soc. Am. A / Vol. 12, No. 9 / September 1995
Yura et al.
Laser Doppler velocimetry: analytical solution to the optical system including the effects of partial coherence of the target
H. T. Yura
Electronics Technology Center, The Aerospace Corporation, P.O. Box 92957, Los Angeles, California 90009
S. G. Hanson and L. Lading
Risø National Laboratory, Post Box 49, DK-4000 Roskilde, Denmark Received November 8, 1994; revised manuscript received April 11, 1995; accepted April 28, 1995 Within the framework of ABCD matrix theory, analytical expressions are derived for the time-lagged covariance of a classical laser Doppler velocimetry system as a function of the laser spot size, the limiting aperture, and the measurement aperture size. Both partial and fully developed speckle as well as planar and rotating targets, are considered. Further, error estimates are presented that indicate how well one can determine in practice the velocity of both planar and rotating targets, and a comparison with time-of-flight velocimetry is given.
Laser velocimeters have found widespread use for localized measurements of fluid velocity. The most common system for this application is the so-called laser Doppler anemometer.1 We will consider such a system, not for fluid flow measurements but for measuring the velocity of solid surfaces or objects that are large compared with any scale defined by the instrument. In cases of solid objects we use the term laser Doppler velocimetry ( LDV ) rather than anemometry. Light scattering in these cases is often dominated by speckle phenomena. The correlation function and the corresponding power spectrum are evaluated for different types of surface statistics and for different parameters defining the operational mode2 of the instrument. The present analysis is performed by use of the method of generalized ABCD matrices.3 The case of fully developed speckle and a large collection aperture will give a power spectrum of the same type as is encountered in laser Doppler anemometry with a large number of particles.4 The shape of the spectrum is then exclusively given by the geometry of the measuring system, the optical wavelength, and the velocity of the scattering object. This is in contrast to the situation of partially developed speckle or to cases of speckle decorrelation. In such cases the spectral shape and the modulation depth of the signal will depend on both the spatial scales of the surface roughness and out-ofplane motions. The uncertainty of the estimated velocity has been investigated by Lading and Edwards5 in cases that correspond to fully developed speckle. Here we will expand the results so that partially developed speckle and speckle decorrelation also are incorporated into the analysis. We assume that photon and electron noise is negligible. In Section 2 we derive general expressions for the autocovariance of the photodetector current in LDV systems
and present closed-form analytical results for planar targets that have rough surfaces such that the reflected optical phase exhibits partial spatial coherence. This property is important for measurements on smoother surfaces that do not give rise to fully developed speckle. In Section 4 we obtain corresponding results for the case of fully developed speckle from curved surfaces, which is important for determining the angular velocity of rotating cylindrical shafts. Error estimates are derived in Section 5 that indicate how well one can estimate the velocity of both planar and rotating targets in practice. Finally, in Section 6 a comparison with time-of-flight velocimetry on similar targets is given.
2. TIME-LAGGED COVARIANCE FOR LDV SYSTEMS
We consider the classical LDV system depicted in Fig. 1. Two intersecting laser beams produce on a surface an interference fringe pattern that is imaged onto an intensity detector. It is assumed here for definiteness, as indicated in Fig. 1, that the x axis of the object-plane coordinates is perpendicular to the interference fringes and that the z axis is parallel to the optic axis. It is also assumed that the spacing between fringes is small compared with the overall laser spot diameter (i.e., the number of observable fringes is much greater than unity). Furthermore, the diffusing target is assumed to move with a constant velocity v that is nearly perpendicular to the z axis. We seek to determine the time-lagged (auto) covariance of the resulting photocurrent from the detector centered at the corresponding geometrical-image point of the center of the target spot. This quantity is given by Ci t itit1t 2 it it1t , (2.1)
where i t is the photocurrent obtained from the detector
© 1995 Optical Society of America
t G r. 9 / September 1995 / J.10) where v x 2f1 ks . The mean (diffuse) reflected interference fringe intensity is given by I0 r jUi r j 2 ! √ 4P0 2r 2 .2) where p is a vector in the image plane. (2. assumed here to be stationary. t . and G r. Schematic of a LDV system. p is the Green’s function for the circularly symmetric system.. given (to within an unimportant phase factor) by " # ik ik 2 2 exp 2 Dp 2 2r ? p 1 Ar G r. (2. (3. 2pB 2B (2.7) (2. 12. it is good to assume that the detector aperture is larger than the imaged spot. 2. Specifically. t . . t is the corresponding image-plane intensity distribution.8) (2. 2ik 2pB exp 2 x 1 r 2 v2 .Yura et al. ks 2 21 m . we model the reflected optical field for a planar target as U0 r. and a is a conversion factor (i.12) at time t (t is the time lag) and angle brackets denote the ensemble average over the realizations of the statistics of the reflected light. h is the detector quantum efficiency. (2. 3. 2 cos kx rx 2 exp 2 2 pss 2 ss (3. Opt. W p is the receiver-aperture weighting function.6) yields that G r. p m. t Ui r c r.2) Fig. U0 r. The instantaneous intensity function can be written as É É2 Z I p. 2 2 For definiteness we now assume that the reflected laser interference fringe pattern. (2. Am. which is fixed in space. (2. (2. The instantaneous photocurrent is given by 6 Z it a dpW p I p. Here the transmitter and the receiver are shown as separate units. (2. We assume that the amplitude of the reflection coefficient is constant. and D are the ray-matrix components of the system. and m is the geometrical magnification.9) into Eq.3) Equation (2. k is the optical wave number.e.4) where sa is the 1 e2 radius of the receiver-aperture weighting function. n is the optical frequency. In practice. 2. power to current) given by a qh hn . p Fig.6) where Ui r is the incident field. Here we model the (real and positive) receiveraperture weighting function by a circularly symmetric Gaussian-shaped function of the form W p exp 22p sa . results from reflection off a partially coherent diffuse planar reflector moving with uniform velocity parallel to the rx axis. No. Substituting Eqs. Optical diagram of a LDV system. t drU0 r. B. PLANAR TARGETS where q is the electronic charge. (2. Vol. while the phase exhibits partial spatial coherence. p . I p. 2 these quantities are given by A B D 2f2 f1 2 2if1 f2 . f1 and f2 are indicated in Fig. A 2041 where A. 2m .7)–(2.5) where we assume that the detector integration time is long compared with the coherence time of the incident laser light but short compared with the characteristic speckle fluctuation time. p 2 . The limiting aperture of 1 e2 radius s is positioned in the Fourier plane. Often they are combined and have a common optical axis. 1. and h is Planck’s constant.7 For the imaging system depicted in Fig. Soc. t is the reflected optical field in the object plane (assumed here to be in the rx – ry plane).10) expresses the Green’s function in terms of object-space variables v and x and is convenient to employ in the ensuing calculations. and hence we set W p 1 in this section.1) (2.9) 2f1 f2 where s is the 1 e2 radius of the imaging system’s limiting (Gaussian) aperture.11) (2.
12. 9 / September 1995 Yura et al. 1 and hence the contribution of the . (3. where P0 is the reflected power. (2. r1 . t U0 r1 0 .1) – (3.5) it is easily seen that the first term on the right-hand side of Eq. (2. Soc. for fully developed speckle. kx ss 2pss L . (3. r2t . p2 . t c r2 .2042 J. Hence the reflected field obeys circular complex Gaussian statistics. t 1 t U0 r2 . (3. (3. t U0 r2 0 . t U0 r2 0 . It can be shown that the direct terms yield. t .. and r2 are integration variables that result when the substitutions alluded to above are carried Substituting Eqs.12) Ci t exp 2 2 N 2 ss 1 rc 2 . which when expanded produce additive terms of the form exp 6ikx r1x 1 r2x 1 vx t (the direct terms) and exp 6ikx vx t (the cross terms). . Opt. out.11) That is. r2 . it can be shown that because the cross terms are independent of object-space coordinates they yield an overall multiplicative factor to the covariance. t U0 r2 .3) From the discussion proceeding Eq.8 In Appendix B we present general expressions for the covariance when the magnitude of the reflection coefficient varies spatially. many fringes are contained within the reflected spot). t 1 t U0 r2 . A / Vol. 4p < 2 (3. where d r is the Dirac delta function. (3.1). t j r. see Appendix A). p2 . rc ! 0). and (3. t 1 t U0 r1 0 .8) yields an integral whose integrand contains multiplicative factors of the form cos kx r1x cos kx r2tx . (2. t . (2. We further assume that jc r. we obtain that Ci t where K p1 . In order to perform the statistical average over the four indicated reflected fields.6) where r1 . (3. t U0 r2 . t is modeled by a Gaussian function given by !8 #9 √ " 2 r1 2 r2 2 = . Am. ss is the 1 e2 Gaussian spot radius. Here. satisfies the condition that L . t 1 t Fig. the correlation function Bc c r1 .e. On the other hand. (3. and kx 2p L . t 1 t . the reflected optical phase function is completely random and delta correlated.1) a term of the form U0 r1 . (3. Because we assume that there are many fringes contained within the reflected spot. 3. The resulting integrations were performed with the MATHEMATICA computer program.8) where L.1).5) 3 G r1 . No. Here we assume that r is constant and set it equal to unity in the following. after the integrations are performed over object space. t 1 t . p2 G r2 0 . t U0 r2 0 . where r is the magnitude of the reflection coefficient. we invoke the circular complex Gaussian statistics of the underlying speckle fields to write the average implied by expression (3. 1..e.9 with the final result that # " # " vt 2 i 2 1 1 cos kx vx t . p2 . we have that Bc ! 4p k2 d r1 2 r2 . t 1 t . t 0 U0 r2 0 .e. p1 G r1 0 .4) exp 2 Bc r1 . k2 : prc 2 rc 2 where rc is a measure of the phase correlation length of the target’s surface. 0 0 (3. t U0 r1 0 . multiplicative factors of the form exp 2c kx ss 2 . we assume diffuse reflection only. direct terms to the covariance are negligible. t 1 t 3 U0 r2 ..10) (3. p2 Bc r1 . r2 . a Taylor’s hypothesis). As an illustrative example. t 1 t c r 2 v t.7) will yield an expression that is identically canceled by the second term on the right-hand side of Eq. As a result.2).9) 3 where r2t r2 1 v t . (3. (3. On substituting Eqs. t 1 t U0 r2 . As discussed in Appendix A. which equals cos2 kx vx t 2 1 1 cos kx vx t 2. p1 K p 1 . where c is the dimensionless constant of the order unity.4). we assume that there is no specular component in the reflected field (this is equivalent to assuming that the reflected optical phase variance sf 2 . p2 3 U0 r1 . the reflected phase at any position in a coordinate system moving along with the diffuser does not change in time (i.5) we obtain that É Z dr1 Ui r1 G r1 . t 1 t U0 r1 .5) into Eq. In the limit of complete spatial incoherence (i. the fringe period. t Z dr1 Z dr2 Z dr1 0 Z dr2 0 Z dp1 W p1 Z dp2 W p2 K p1 . (3. t U0 r1 0 .2) and (2. 3 we plot the reflected intensity distribution for the case where L ss 5. ss (i.6) as U0 r1 . (2.7) 1 U0 r1 .10) into Eq. t U0 r1 . in Fig.10). p1 G r2 . we assume that the time evolution of the reflected phase is given by c r. (3. (3. Reflected intensity distribution for L ss 5. t Z É2 dr2 Ui r2 G r2 . In particular. we obtain from the first term on the right-hand side of Eq. By employing Eqs. That is. For a diffuser moving with a constant velocity v.
e.14) Fig. 12. 1 1 rc 2 v 2 1 rc 2 ss 2 (3. (3. In this case it can be shown by methods similar to the method used in obtaining Eq. the number of independent modes that passes through the optical system to the measurement aperture. 4. the number of observable oscillations in Ci t increases with increasing rc . it can be shown that. for complete coherent reflection rc ! ` examination of Eq. In all cases of interest the maximum time lag t and the rotational velocity are sufficiently small that the angular separation between a point on the surface at time t and the corresponding point at time t 1 t is much less than unity.e. interference effects due to the presence of the curved target cause a reduction in the temporal width of the covariance. (3. for a sufficiently large signal-tonoise ratio. Thus for specular reflection the LDV covariance of photocurrent is identically zero. as discussed in Section 1. hence increasing the sensitivity of the system. A 2043 where the mean current. i 2 1 1 cos kx v0 Rt exp 2 (4. the 1 e2 radius of the image of the reflected spot. In the limit rc ! 0 (i. (4. as indicated in Fig. vs . ry . Equation (3. as expected physically. respectively. and the detector aperture radius referred to object-space coordinates. (2.. As a result. in contrast to the planar case.1) becomes U0 r. (4. p G r2 .14) are identical to the result obtained by Lading and Edwards. (3. i it i t 1 t .. Measurement geometry for a rotating cylindrical shaft. No.13) and N. For completeness.3 Note that. Opt. (4. z 0) at the coordinate rx . the corresponding number of modes captured by the optical system. Eq. ROTATING TARGETS (4. in the presence of a rotating shaft and a finite detector measurement aperture sa . This is equivalent to the condition that the reflected spot size is small compared with the radius of the shaft (i. ss R . rc ! 0).1) That is.. 0 contains the multiplicative factor exp 2ikbrx 2 . Vol.e.12) that the covariance of the photocurrent. is given by " # " # v0 Rt 2 . whose weighting function is given by Eq. Nc vs m 2 1 s0 2 √ !2 √ !2 ss ksss . 1). this is the first analytical model that includes target partial coherence for LDV systems. p √ !2 1 s .4). 1 2 1 2 2 ss v 1 s0 ss 2 1 v 2 √ !2 1 s .2) sa . Nc . For detector apertures that are large compared with the size of the imaged spot.7) Nc vs 2 s0 4.5 On the other hand. is given by 6 N 1 1 ss 2 v 2 1 1 rc 2 ss 2 ..9) m The quantities N. we consider a finite receiver aperture.. a condition that is always met in practice. and s0 are.12) expresses the LDV covariance of the photocurrent as a function of the phase correlation scale rc for a planar diffuse reflecting target.4) (4. r2 Ui r1 Ui r2 3 G r1 .Yura et al.10 where b 1 2R .10) 1 1 v 2 ss 2 . 9 / September 1995 / J. t Ui r exp 2ikbrx 2 c r. (3.13) yields that i ! 0. . (4. 3 11 (4. the results expressed in Eqs. the number of independent (optical) modes that pass through the measurement aperture to the detector plane.3) becomes 8 √ " # !2 < v0 Rt i 2 1 1 cos kx v0 Rt Ci t exp 2 : Nc 2 ss #9 " kss v R 2 = sa ! ` . and hence a shorter corresponding transit time is obtained.8) In this section we consider. This is because for complete coherent reflection (i. which reduces the effective imaged spot size. 11 11 v 2f1 !2 √ 2f2 . mss 2 1 ks (4.12) reveals that the effect of partial coherence of the target’s surface is to increase the effective size of the measurement (target) area. fully developed speckle). Soc. specular reflection) there is no diffuse component of the photodetector current.3) Ci t N 2 D2 where v0 is the magnitude of the angular velocity of the rotating shaft. (3. Assuming that the laser beams are normally incident on the target. aP0 2f1 1 1 rc 2 v 2 1 rc 2 ss 2 (3.6) (4. Physically this arises from destructive interference.e.e. t . For definiteness we assume that the angular velocity of the shaft is parallel to the ry axis. Examination of Eq. 1 D2 i N 1 1 kss v R 2 . the reflected optical field Ui r in the initial plane (i. aP0 2f1 1 1 vs 2 sa 2 √ !1/2 v 2 1 s0 2 . . 4. To the best of our knowledge. Am. Eq. a rotating shaft of radius R that has a surface that reflects light with complete incoherence (i.5) (4.12) – (3. is given by Z Z Z i dpW p dr1 dr2 Bc r1 .
1 1 v 2 ss 2 (4.e. ERROR ESTIMATES # S f0 .6) is given Of primary concern in LDV measurements is the determination of the velocity of a target.12) and (4.11) yields . Note in particular that if one knows the precise location of the peak f0 . and variance sf0 2 . we expand s f near f f0 . . 1) Eq. As indicated above.3). (3. one can obtain a confidence interval for the unknown true value of f0 in terms of system parameters. (4.5) are based on the assumption that the filtered data behave like bandwidth-limited Gaussian white noise. (i. and f denotes temporal frequency. We consider the results derived in both papers that relate to measurements on targets that give rise to fully developed speckle fields. 11).. The central-limit theorem applies. 2 (5. (5. in contrast to the case of planar targets.5) For example. (4.2) On the basis of the normal distribution and Eq. Substituting either Eq. N . rotating targets (5. (5.13) (4.8) yields that sf0 2 p Df 2 . respectively. N . Opt. where Rmin is determined from the condition that kvss Rmin 2 . Therefore consider the autospectral density function of the covariance of photodetector current. For simplicity consider the case of large detector apertures.7). Df (5.12) or Eq.1) yields a spectrum in the region of positive frequencies.1) Assume next that these values of f are such that f follows a normal distribution with f f0 . velocity information is usually not determined directly from the covariance. In order to obtain an estimate for the error in determining the precise location of the peak of the autospectral density function.5) (with sa!` ) and (4.3). a comparison between the two systems can be made. The results given in Eq..e. the corresponding power spectrum is a peaked function. 8. Then examination of Eq. Nf . where i and Nc are given by Eqs. Thus. the effect of curved targets results in an increase in the spectral width Df [see Eq. Soc.2) into Eq. var S f by ˆ var S f f 2 f0 Df 4 4 S f0 2 . (3. near the peak. that Rmin kvss . rotating targets planar targets .e. from which it follows that the variance of the estimate of the location of the peak in the autospectral density function is given by sf0 2 p ˆ 1 3 Df 2 var S f0 1/2 where Ci t is given by Eqs. 1) we obtain that (4. where T is the finite record length and Be is the resolution bandwidth centered at frequency f (see Sec. in terms of estimates of f near f0 .4) 6. We next present expressions that indicate how well one can estimate the exact location of the peak of the power spectrum in practice. Because the current covariance is an oscillating function of the time lag t. This is an excellent assumption in practice when the filter resolution bandwidth Be is sufficiently small. i.12) whereas for v . which yields the estimate " ˆ S f 12 f 2 f0 Df 2 2 5. DISCUSSION AND CONCLUSION In general. (4.7) Ci t exp 22pif t dt .8) Substituting Eq. (5. based on any single estimate fˆ. 1. 2` (5. (5.11) For f ﬁ 0 the variance of the estimate of the autospecˆ tral density S f is given by11 ˆ var S f ˆ ˆ S f 2 S f S 2 f Be T . (5. equal to the . However. for v . the presence of the curved target causes a reduction in the temporal width of the covariance.. No. Rmin . (i. Then the fourth moment in Eq. (4.4)]. (5.3) into Eq. and the fact that Be is small means that the output spectrum must be essentially constant. the 95% confidence interval for determining the location of the peak of the spectrum is f f0 6 2sf0 . the corresponding velocity information follows directly from Eq. for example.9).2044 J.5 of Ref. which applies for almost all systems of practical concern. (4. Am. Based on the results obtained in this paper for LDV systems and the corresponding results obtained in a previous article2 treating the laser-time-of-flight velocimeter ( LTV). (5. To compare the two systems. 9 / September 1995 Yura et al. 12.9) # .7) is equal to 3sf0 4 . (5. from which velocity information can be extracted directly through knowledge of the location of the peak. Rmin kss 2 .10) reveals that the effects of rotation become significant for R . (5. which make the number of fringes in the LDV system. the pedestal term) S f where f0 vx L v0 R L 8 < v p ps 2 1 r 2 s c : v0 R pD ( planar targets . 3Be T (5. (5. defined as S f Z ` ˆ Thus. ss . ss . indicating that the filtered data should be more Gaussian than the input data.3) i2 f 2 f0 p exp 2 2 p DfN Df 2 " 2 S f0 . we assume that the transmitters for the two systems have the same numerical apertures. (5. A / Vol. of the form (excluding the low-frequency region.
This method is based on the assumption that all apertures are of Gaussian shape. (6.3) represents fundamental limits. Finite spatial coherence of the reflecting target has been considered for the LDV system only. Identical conditions for the detector and transmitter setups. for both systems. 2 gives the variance of the estimated value of the location of the peak of the cross covariance. so too . 9 / September 1995 / J. When we compare Eqs.e. Comparing the functional dependence of the normalized variances. Am. and a slight decrease in the magnitude of the detected signal. (5.3). as obtained from Eq.Yura et al. the number of collected modes.5 but even with the detailed signal statistics taken into consideration the conclusion appears to be valid. The autocovariance and the power spectrum have been obtained in closed analytical form. showing an apparent increase [see Eq. while the temporal widths of the respective covariances increases correspondingly with decreasing values of the spot size.2) for the case of large detector apertures.3). ss Time-of-Flight Velocimetry 2kdv 2kdss Laser Doppler Velocimetry kvss kss 2 relative distance between the focused spots divided by the spot radius in the LTV system. (6. Eq. Identical performance is obtained if the distance between the spots in the LTV system equals the spot size in the LDV system. A decrease in resolution is expected for variations from optimum alignment for the LDV system. Note that the bandwidth Be is different for the two systems.162) and Fig. this is accompanied by a decrease in the number of optical modes N. yield normalized covariance functions that are inversely proportional to the number of modes collected by the receiver. (6.1) nor Eq. The LTV system has a bandwidth that is given by the transit time through one of the focused spots. APPENDIX A In this appendix we present a physical model that relates the complex reflection coefficient to the surface-height fluctuations.1) and (6.23 of Ref. (6. Equation (57) of Ref. No. Further. it follows that the variance of the estimated velocity is therefore a factor Nf 21/2 smaller for the LTV system than for the LDV system. Soc. which is often used in analysis and reasonably accurate if the surface slopes are small. which can be easily converted to give the normalized variance for estimating the velocity in a LTV system: sv 2 1 . the numerical apertures of the detector systems are assumed to be identical. for the case of large detector apertures. Both LDV and LTV systems will suffer a decrease in the magnitude of the cross covariance and autocovariance as the target diameter decreases. (6. yielding a decrease in the variance of the estimated velocity. The comparison is made with respect to the error estimates.8 we adopt an elementary relation that leads to tractable results.2). In these respects the LTV system is superior to the LDV system. In particular. whereas the LDV system has a corresponding bandwidth determined by the transit time (which is a factor Nf larger) through the entire fringe pattern. Furthermore. is determined by the spot size for the LTV system and by the size of the fringe pattern for the LDV system. 12. Opt. 2. wide-sense-stationary random process. ss v . and for simplicity in notation we have suppressed the explicit dependence on time. is sv 2 1 p ~ 2 B T v2 Nf e √ vx v !2 . therefore. Following Goodman. and the effect of velocity misalignment. Because it is assumed that h r is a zero-mean. 8] c r where fr k 1 1 cos b h r . i. and therefore an increase in the normalized autocovariance. A 2045 Table 1. b is the angle between the direction of propagation of the laser beams and the normal to the surface. Further. The method of ABCD matrices has been applied to the analysis of LDV systems in conjunction with targets having correlated surface structures.9). System performance has been analyzed in that the variance of the estimated velocity has been obtained in terms relating to the optical system and the target surface reflectance characteristics. (2.1) and (6. Further. which provides a tool for parametric optimization of the optical system. (6. The resulting Huygens – Fresnel integrals can then be solved analytically. Vol. (A1) h r is the local surface height about the reference level. whereas the basic LTV system with circular spots in the measuring volume will lose signal in this case. is also given by Eq. The minimum radius of curvature for each of the two systems is given in Table 1. Targets giving rise to fully developed speckle will. result in a higher normalized time-lagged autocovariance for the LTV system. reveals that we arrive at identical expressions if the velocity in the LDV system is perpendicular to the fringes. whereby later mathematical approximations can be avoided.7). Be v 1 ss 2 1 1 v 2 1/2 . We note that neither Eq.. measurement restrictions for cylindrical rotating targets have been obtained and the results compared with previous results for time-of-flight velocimetry. p ~ v2 Nf 2 B e T (6. (4.3) where the spectral bandwidth. The corresponding normalized variance for the LDV system.12)] of the measuring volume. the minimum shaft radius. (6. we express the reflection coefficient as [see Eq.1) where the bandwidth for the LTV signal is determined by the effective transit time through one of the illuminated spots. (3. Minimum Radius of Curvature for Measurement on Cylindrical Surfaces in the Two Limiting Cases for LDV and Time-of-Flight Velocimetry Minimum Radius of Curvature v . Eqs. (A2) jc r jexp if r .
Grum. for fully developed speckle we require ad r1 2 r2 . r2 . T.10) becomes É K p1 . P. No. P. 3.e. September – November 1994. Then. p1 Ui r2 G r2 . where I0 is the incident intensity. A / Vol. (A7) we use a correlation function for partial coherent diffuse reflection given by √ Bc r1 . 1994). Buchhave. Hence it can be shown that a 4p k2 . p1 G rt . Lumley. (A9) Note that in the limit rc ! 0.2046 J. Opt. 312 – 362 (1973). That is. r2 c r1 c r2 can be expressed as B c r1 . L. t j p r r. J. Phys. H. . Angus. where r jr1 2 r2 j and rh is the lateral coherence length of the surface-height fluctuations. the mathematically convenient Gaussian form for the correlation function given by Eq. t . 4. 42. we have from Eq. pp. k2 : prc 2 rc 2 REFERENCES AND NOTES 1. T. Yura was performed while he was a guest scientist at the Risø National Laboratory. and J.. (3. 316 – 323 (1993). 85 – 125. it is straightforward to show that Eq. Bc r1 . (3. The research of H. Denmark. George and J. 12. Am. p2 (B2) 3 . and hence Bc can be expressed as Bc r where rc rh sf rh k 1 1 cos b sh (A8) exp 22 r rc 2 3 G r. Chichester. C. Assuming that jc r j 1 and that the surface-height fluctuations are a Gaussian random process. eds.” in Optical Diagnostics for Flow Processes. Fluid Mech. (B2) becomes É K p1 . K. Edwards.10). The Laser Doppler Technique ( Wiley. (B1) where bh is the normalized correlation function of the surface-height fluctuations.” J. Soc. Roskilde. For mathematical convenience we want to use a correlation function that is of Gaussian form and becomes a Dirac delta function in the limit of complete incoherent reflection. (A6) For almost all cases of practical concern the phase variance sf 2 is greater than unity. and A is the area of the illuminated spot (i. Wigley. G. where a is a constant that is that Bc r1 . A 10. r 2 exp i f r1 2 f r2 exp 2sf 2 1 2 bh r1 .e. 837 – 850 (1971). 60. r2t .” J. 1980). and G. 9 / September 1995 Yura et al. t p 4p Z dr r r r rt Ui r Ui rt k2 É2 . W. “The laser Doppler velocimeter and its application to the measurement of turbulence. Dunning. following similar arguments that led to Eq. t is the magnitude of the reflection coefficient. If sh 2 represents the variance of h. That is. L. Soc. (A7) (B4) ACKNOWLEDGMENTS This work was partly supported by the Danish Council for Technological Development and by DANTEC Measurement Technology A/S. bh r exp 22 r rh 2 where r r. r is the reflection coefficient. Lading. The quantity a can be determined from the requirement that the mean reflected radiance of fully developed speckle be given by I0 rA p. ( Plenum. (A9) . New York. L. W. Drain. Specifically. Hence the correlation function Bc becomes Bc r exp 2sf 2 1 2 exp 22 r rh 2 For the special case of fully developed speckle [i. we assume that jc r. is the phase lateral coherence length. (A4) APPENDIX B In this appendix we now assume that the magnitude of the reflection coefficient exhibits deterministic spatial variations over dimensions of the reflected laser spot. 4p < 2 exp 2 . “Spectral analysis of the signal from a laser Doppler flowmeter: time independent systems. S.. r2 !8 #9 " 2 r1 2 r2 2 = . p2 . t Z Z dr1 r r1 1/2 Ui r1 G r1 . is the phase angle f r . Equation (A4) provides us with a specific relationship between the correlation properties of the reflection coefficient and the correlation properties of the reflecting surface. Lading. On the basis of these considerations and Eq. (A3) that Bc ! 4p k2 d r1 2 r2 . R. r2t 4p k2 d r1 2 r2 2 vt . Eq. (A5) dr2 r r2 1/2 É2 3 Bc r1 . (B3) . r2 independent of the optical system. To proceed further. V. L. then the variance of f is given by sf 2 k 1 1 cos b sh 2 . p2 . 2. Opt. and T.” J. Am. we assume that the statistics of the reflecting surface are stationary and that the normalized correlation function of the surface heights is also of Gaussian form. p2 where rt r 1 vt . Appl. Hanson. (A9) contains the effect of partial coherence of the target and yields delta correlation for the special case of fully developed speckle. Lambert’s law). Yura. the correlation function Bc r1 . “Principles of laser anemometry. E. “Speckle: statistics and interferometric decorrelation effects in complex ABCD optical systems.
New York. 6. 1986). 1986). 9 / September 1995 / J. Vol. 20. Laser Speckle and Related Phenomena. 12. Am.” J. Yura and S. W. T. J. 1984). E. Piersol. Opt.. Berlin. MATHEMATICA: A System for Doing Mathematics by Computer (Addison-Wesley. Lasers ( University Science. Wolfram. Chap. Opt. 3855 – 3866 (1993). 1918 – 1924 (1993).” Appl. Dainty. A 2047 9. Opt. the factor of 2 in the exponent is replaced by 1 1 cos b . . S.Yura et al. 5. ed. Chap. “Laser-time-of-flight velocimetry: analytical solution to the optical system based on ABCD matrices. H. 1991). where b is the angle between the direction of propagation and the normal to the surface. 8. 32. “Laser velocimeters: lower limits to uncertainty. Goodman. G. V. Siegman. Bendat and A. Hanson. For nonnormal angles of incidence. C. Mill Valley. Redwood City. Chap. Soc. G. Am. Calif. 10. Edwards. 2. 8. A 10. Soc. 11. Calif. Random Data: Analyses and Measurement Procedures ( Wiley. S. Lading and R. (Springer-Verlag. A.. J. No. L. 7. J.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.