Length metrology apparatus and methods for suppressing phase noise-induced distance measurement errors

Abstract

Length metrology apparatuses and methods are disclosed for measuring both specular and non-specular surfaces with high accuracy and precision, and with suppressed phase induced distance errors. In one embodiment, a system includes a laser source exhibiting a first and second laser outputs with optical frequencies that are modulated linearly over large frequency ranges. The system further includes calibration and signal processing portions configured to determine a calibrated distance to at least one sample.

Claims

1. A method comprising: producing a first laser chirp and a second laser chirp, wherein at least one of the first laser chirp and the second laser chirp is an optical frequency sideband; directing the first and the second laser chirp towards an object; receiving a first interference signal based, at least in part, on the interaction of the first laser chirp with the object; receiving a second interference signal based, at least in part, on the interaction of the second laser chirp with the object; and processing the first interference signal and the second interference signal to determine a range to the object.

2. The method of claim 1, wherein the directed first and the second laser chirps are scanned laterally across at least a portion of the object.

3. The method of claim 1, wherein there is relative lateral motion between the directed first and second laser chirps, and the object.

4. The method of claim 1, wherein a surface roughness of the object is greater than an optical wavelength of the first laser chirp and an optical wavelength of the second laser chirps.

5. The method of claim 1, wherein the laser output simultaneously includes an up-chirped sideband and a down-chirped sideband.

6. The method of claim 5, wherein processing the first interference signal and the second interference signal to determine the range to the object includes compensating for speckle noise based, at least in part, on a first portion of the received light associated with the up-chirped sideband and a second portion of the received light associated with the down-chirped sideband.

7. The method of claim 1, wherein processing the first interference signal and the second interference signal comprises: calculating a first signal phase based on the first interference signal; calculating a second signal phase based on the second interference signal; determining at least one corrected phase signal based on the first signal phase and the second signal phase; and determining a distance to at least a portion of the object based on the at least one corrected signal phase.

8. The method of claim 7, wherein using the first signal phase and the second signal phase comprises determining a sum or a difference of the first signal phase and the second signal phase.

9. The method of claim 7, wherein calculating the first signal phase or the second signal phase comprises performing a Hilbert transform.

10. The method of claim 7, further comprising performing corrections to at least one of the first or second signal phases based on a wavelength or a chirp rate of the laser output.

11. A method comprising: producing a laser output beam including at least one chirped optical frequency sideband; directing the laser output beam including the at least one chirped optical frequency sideband towards an object; receiving interference signals using at least one optical detector, wherein each interference signal represents interference between a local oscillator signal and a return signal from the object, the local oscillator signal and the return signal being chirped signals including multiple chirps each having a chirp duration, and wherein the return signal is based on an interaction of the laser output and the object; and determining a distance to at least a portion of the object, at least in part by calculating a respective phase of the interference signals as a function of time over a respective chirp duration.

12. The method of claim 11, further comprising shifting an optical frequency of at least a portion of the laser output beam.

13. The method of claim 11, wherein the multiple chirps includes a first laser chirp having a first chirp rate and a second laser chirp having a second chirp rate different from the first chirp rate.

14. The method of claim 13, further comprising: receiving a first interference signal and a second interference signal; calculating a first phase signal based on the first interference signal and a second phase signal based on the second interference signal; determining at least one corrected phase signal based on the first and the second interference signal; and determining the distance based on the corrected phase signal.

15. A system comprising: a laser source configured to produce a laser output including at least one chirped sideband; an optical system configured to direct the laser output towards an object and receive light from the object; a beam scanner configured to scan the laser output across at least a portion of the object; a detector configured to produce a signal based on the received light; and a processor configured to determine a distance to the object based on the signal.

16. The system of claim 15, further comprising an optical frequency shifter configured to shift an optical frequency of the laser output.

17. The system of claim 16, wherein the frequency shifter is configured to shift the optical frequency of the laser output by an offset frequency.

18. The system of claim 17, wherein the processor is configured to use frequencies below the offset frequency to detect a down-chirped sideband of the laser output and frequencies above the offset frequency to detect an up-chirped sideband of the laser output.

19. The system of claim 15, further comprising a modulator configured to produce at least one chirped sideband from the laser output.

20. The system of claim 19, wherein the modulator is external to the laser source.

21. The system of claim 15, wherein the laser output includes an up-chirped sideband and a down-chirped sideband.

22. The system of claim 21, wherein the processor is configured to correct a speckle noise in the signal based, at least in part, on a first portion of the received light associated with the up-chirped sideband and a second portion of the received light associated with the down-chirped sideband.

23. The system of claim 15, wherein the optical system comprises a beam splitter configured to split the laser output into a transmitted (TX) portion and a local oscillator (LO) portion, wherein the optical system is configured to direct the TX portion towards the object and receive a received (RX) portion, and wherein the optical system is further configured to combine the LO portion and the RX portion to produce an interferometric signal at the detector.

24. The system of claim 23, further comprising an optical frequency shifter configured to shift a frequency of at least one of the TX portion and the LO portion.

Description

DESCRIPTION OF THE DRAWINGS

(1) FIG. 1A is a diagram 100 showing some principle elements of a swept frequency metrology system according to disclosed embodiments;

(2) FIG. 1B is a plot showing a linearly swept LO and time-delayed Rx optical frequencies as functions of time according to disclosed embodiments;

(3) FIG. 2A is a plot of range profile data of a brushed alloy sample showing the peaks from two measurements (an up chirp and a down chirp) from the same target surface according to disclosed embodiments;

(4) FIG. 2B is a plot of range measurement data of a brushed alloy sample according to disclosed embodiments;

(5) FIG. 3 is a plot of a discretized sample plane with Gaussian measurement beam according to disclosed embodiments;

(6) FIG. 4A is a plot showing frequency versus time for first and second lasers according to disclosed embodiments;

(7) FIG. 4B is a plot showing phase versus time for a stationary sample. In this case the phase with and without speckle noise are nearly identical according to disclosed embodiments;

(8) FIG. 4C is a plot showing phase versus time of the measured signals for a sample with lateral motion. 1,0 and 2,0 represent the phase without speckle-induced phase noise, while 1 and 2 represent the measured phase with speckle-induced noise according to disclosed embodiments;

(9) FIG. 5A is a plot showing an example of simultaneous measurement of up and down chirps without frequency offset showing that f.sub.beat is the same for the up and down chirps according to disclosed embodiments;

(10) FIG. 5B is a plot showing an example of simultaneous measurement of up and down chirps with frequency offset showing that f.sub.beat is the different for the up and down chirps according to disclosed embodiments;

(11) FIG. 5C is a simplified block diagram showing components in a chirped sideband modulation setup. An AOM is used to shift the LO (or Tx) off DC and thereby isolate speckle information from range information according to disclosed embodiments;

(12) FIG. 6A is a plot showing filtered (black) and unfiltered (gray) FMCW range errors for two different surfaces undergoing lateral motion of 50 mm/s with respect to the measurement beam. 4901 measurements from a 2 m surface roughness piece of ground glass according to disclosed embodiments;

(13) FIG. 6B is a plot showing filtered (black) and unfiltered (gray) FMCW range errors for two different surfaces undergoing lateral motion of 50 mm/s with respect to the measurement beam. 2000 measurements from a business card with 25 m surface roughness according to disclosed embodiments;

(14) FIG. 7 is a plot showing an example of threshold settings for the phase RMSE and peak SNR filter according to disclosed embodiments; and

(15) FIG. 8 is a block diagram showing components used in one embodiment to compensate for Doppler and speckle phase noise.

DETAILED DESCRIPTION

(16) The invention described herein teaches how multiple optical phase-sensitive measurements can be made of a surface and used to significantly suppress phase noise-induced distance measurement errors during lateral motion, such as those due to speckle. In each embodiment described, a difference in phase-sensitivity to the sample surface distance between the multiple measurements is used to suppress the phase noise-induced errors.

(17) A measurement may be defined as the time-varying phase of the interference between light received from a reference surface and a sample surface for a single laser. It is understood that additional surfaces and lasers may also be considered. A depiction of a system that may perform two simultaneous measurements using two separate lasers is shown in FIG. 1A. Below is a mathematical description of the signal for a single measurement that may follow the formalism provided in Z. W. Barber, et al.; Accuracy of active chirp linearization for broadband frequency modulated continuous wave ladar, Appl. Opt., 49, 213 (2010). The light received from the reference surface will be referred to as the local oscillator (LO). The time-varying optical frequency for LO electric field can be represented in the form
E.sub.LO(t,z=0)=E.sub.0e.sup.i(.sup.0.sup.t+t.sup.2.sup.),(1)

(18) where .sub.0 is the angular optical frequency at the beginning of the chirp, and is the angular chirp rate. Propagation of the LO field to the sample surface can be treated by performing a Fourier transform to the frequency domain.

(19) $\begin{array}{cc}E\left(,z=0\right)=\frac{E_0}{\sqrt{2}}{}_{-}^{}{e}^{-i\left({}_0\mathrm{.Math.}+\frac{1}{2}{\mathrm{a.Math.}}^2\right)}{e}^{it}\mathrm{dt}=E_0\frac{1-i}{\sqrt{2}}{e}^{\frac{{i\left(-{}_0\right)}^2}{2}}& \left(2\right)\end{array}$

(20) The LO field is then propagated to the sample surface and back to the reference surface, where it interferes with the LO field, by applying a Taylor expanded form of the propagator e.sup.iz.
E.sub.Rx(,z=2R)=E(,z=0)e.sup.i2.sup.0.sup.Re.sup.i2.sup.1.sup.(.sup.0.sup.)Re.sup.i2.sup.2.sup.(.sup.0.sup.).sup.2.sup.R(3)

(21) Here R is the distance to the sample surface, and

(22) ${}_0=\frac{{}_{0}n}{c},{}_1=\frac{}{}{.Math.}_{={}_0}=\frac{1}{\mathrm{vg}},\mathrm{and}{}_2=\frac{{}^2}{{}^2}{.Math.}_{={}_0}.$
Also, c is the speed of light, n is the refractive index of the medium between the reference and sample surfaces, and .sub.g is the group velocity in the medium. The time-domain description of the field reflected from sample surface, back to the LO surface, is given by

(23) $\begin{array}{cc}{E}_{\mathrm{Rx}}\left(t,z=2R\right)=E_0\sqrt{\frac{{}^{}}{}}{e}^{-i\left({}_0\left(t-2\frac{n}{c}R\right)+\frac{1}{2}{{}^{}\left(t-2{}_{1}R\right)}^2\right)},& \left(4\right)\end{array}$

(24) where

(25) ${}^{}=\frac{}{1+2R{}_2}.$
The interference between the fields E.sub.LO and E.sub.Rx comprises a single distance measurement, and takes the form

(26) $\begin{array}{cc}S\left(t\right)~E_{\mathrm{LO}}\left(t,z=0\right)E_{\mathrm{Rx}}\left(t,z=2R\right)~e^{-i\left(2{}_0\frac{n}{c}R-\frac{1}{2}{{}^{}\left(2{}_{1}R\right)}^2+2R{}^{}{}_{1}t+\frac{1}{2}\left(-a^{}\right)t^2\right)}.& \left(5\right)\end{array}$

(27) For many cases, terms involving .sub.2 and .sub.1.sup.2 can be neglected, and the signal can be adequately approximated by

(28) $\begin{array}{cc}S\left(t\right)~e^{-i\left(2{}_0\frac{n}{c}R+2R{}_{1}t\right)}.& \left(6\right)\end{array}$

(29) However, we have included terms to second order in equation (5) to aid the discussion in later sections of this document.

(30) FMCW Carrier Measurements for Compensating Phase Noise-Induced Errors

(31) FIG. 1A is a diagram showing some principle elements of a swept frequency metrology system 100 according to disclosed embodiments. In FIG. 1A, black arrows indicate optical paths.

(32) As shown in FIG. 1A, the swept frequency metrology system 100 includes a first frequency-chirped laser 110, a second frequency-chirped laser 115, a beam combiner 120, a circulator 130, a reference surface 135, a first sample surface 140, and a detection and processing unit 150.

(33) The first frequency-chirped laser 110 and the second frequency-chirped laser 115 each output light of an optical frequency that changes substantially linearly (chirps) in time over a given chirp duration.

(34) The beam combiner 120 is configured to receive and combine at least part of the first and second laser outputs into a combined laser output. In some embodiments, a single laser may produce an output with both frequency-chirped components, in which case the beam combination occurs internal to the laser.

(35) The combined laser output from the beam combiner 120 is then directed through the circulator 130 and a plurality of optical paths configured to direct at least part of the combined beam onto an optical detector to produce an interference signal.

(36) In FIG. 1A, an optical path may include a transmitted portion denoted Tx. An optical path may include reflection from the reference surface 135, the reflected portion from the reference surface 135 being denoted LO. An optical path may include reflection from the first sample surface 140, the reflected portion from the first sample surface 140 being denoted Rx.

(37) A sum of LO and Rx is directed to the detection and processing circuit to determine the distance measurement, as noted below.

(38) FIG. 1B is a plot 150 showing a linearly swept LO and time-delayed Rx optical frequencies as functions of time according to disclosed embodiments.

(39) In some embodiments of the invention, the optical phase-sensitive measurements may be performed using the FMCW ladar technique, and where the optical carrier may be linearly swept, or chirped, in time. Performance Analysis of Next-Generation LADAR for Manufacturing, Construction, and Mobility, (cited above) describes the FMCW chirped ladar technique and is incorporated herein by reference in its entirety. A simplified block diagram showing a setup that may be used to compensate speckle noise is shown in FIG. 1A, the laser radiation from two frequency-chirped lasers (one for each optical phase-sensitive measurement) may be combined, directed through an optical circulator, and transmitted (Tx) toward a sample surface. A portion of the combined light (LO) may be reflected from a reference surface, while a second portion (Rx) may be reflected from a sample surface. Interference may occur between light reflected from the reference and sample surfaces. The distance between the reference and sample surfaces may be determined by measuring the frequency of the heterodyne beat generated by the interferometric combination of the Rx and LO resulting from either of the two frequency-swept lasers. As shown in FIG. 1B for a single laser, the beat frequency is given by the equation f.sub.beat=, where =/2=B/.sub.chirp is the chirp rate (B is the chirp bandwidth and .sub.chirp is the chirp temporal duration), and =2Rn.sub.g/c is the time delay between the Rx and LO chirp waveforms where n.sub.g is the group index of the measurement path. Solving for R as a function of heat allows for determination of the range by measuring the heterodyne beat frequency.

(40) However, when the sample surface is rough, using the measurement of the distance from just one of the frequency-swept lasers may result in distance errors due to speckle. These errors may increase dramatically when the sample surface is translated perpendicularly to the beam propagation direction (e.g. scanning the beam across the surface or vice versa). FIG. 2A shows an FMCW range profile resulting from one surface and from two lasers with different chirp rates. FIG. 2B shows the determination of range from such a range profile for 100 consecutive range measurements of a brushed alloy surface (with sub-resolved surface roughness) using an FMCW chirped ladar system with resolution (full-width at half max) of approximately 1.5 mm. The plus and circle symbols in FIG. 2B represent the distance measurements when only one optical phase-sensitive measurement (data from one laser) is used. In FIG. 2B, the sample remained stationary for the first 28 measurements, and then was laterally translated for the remaining measurements. The nearly two order of magnitude increase in distance measurement errors observed during lateral motion confirm the observations made by the NIST researchers.

(41) FIG. 2A is a plot 200 of range profile data of a brushed alloy sample showing the peaks from two measurements (an up chirp and a down chirp) from the same target surface.

(42) FIG. 2B is a plot 250 of range measurement data of a brushed alloy sample. Sample lateral motion begins after the 28th data point. Circles and pluses represent range measurements using either down or up chirps. Square data points represent range measurements that combine data from the up and down chirps resulting in speckle-compensated measurements.

(43) Measurement Method

(44) To solve the problem noted above, the disclosed embodiments teach how the use of two simultaneous distance measurements with different phase sensitivities on the surface distance can mitigate the speckle and other phase noise effects. In some embodiments of the invention, the different phase sensitivities are achieved by chirping the two lasers at different chirp rates. Even though both lasers are used to measure a single distance, for one chirp rate, the phase of the received signal evolves at one rate in time, while for the second chirp, the phase of the received signal evolves at a different rate in time. Conversely, it is important to note that the phase noise caused by speckle is common-mode for the two measurements, and can therefore be removed while maintaining the distance information. The measurement setup shown in FIG. 1A allows a single detector and single digitizer to acquire the sample surface distance information from both lasers simultaneously because , and therefore f.sub.beat, is different for the two lasers. The different range peak frequencies for the two measurements, shown in FIG. 2A, highlight how the two measurements exhibit different optical phase-sensitivity to the sample surface distance. The + points in FIG. 2B represent the range measurement results from the first chirp rate, the o points represent the same for the second chirp rate, and the black square points are the speckle-compensated results from combining the phase information from the first and second chirps, as described below. It is clear from closer inspection of the data that the + and o data are anti-correlated. The algorithm described below capitalizes on this anti-correlation to compensate the speckle-induced phase noise.

(45) The following is a mathematical description for speckle phase error compensation of FMCW laser radar measurements from a diffuse target with surface roughness .sub.z, where .sub.z R, and R=c/2B is the distance measurement resolution. The mathematical model relies on discretizing the sample plane into a uniform grid of j cells, and assigning a random height z.sub.j to each grid cell as shown in FIG. 3. For simplicity, we will let n=n.sub.g=1 in following description.

(46) FIG. 3 is a plot 300 of discretized sample plane with Gaussian measurement beam according to disclosed embodiments.

(47) The measured FMCW distance signal is a sum of the returns from each grid cell in the sample plane,

(48) $\begin{array}{cc}S\left(t\right)=\underset{j}{.Math.}{e}^{\left[-\left({\left(x_j-x_0\right)}^2+{\left(y_j+y_0\right)}^2\right)/2w_0^2\right]}{e}^{\left[\frac{4i}{c}{z}_j\left(v_0+t\right)\right]},& \left(7\right)\end{array}$

(49) where is the laser chirp rate, .sub.o is the laser start frequency, and z.sub.j is the distance to the j.sup.th grid cell. One can express equation (7) in polar form as,

(50) $S\left(t\right)=A\left(t\right)e^{\left[\frac{4i}{c}{z}_0\left(v_0+t\right)+\left(t\right)\right]},$

(51) where

(52) $Z_0=\frac{1}{N}\underset{j}{.Math.}{z}_j,$
and (t) and A(t) are defined by equation (7). (See, P. Pavlicek, et. al. Theoretical measurement uncertainty of white-light interferometry on rough surfaces, Appl. Opt. 42, 1809-1813 (2003).) Due to the small surface roughness, and in the limit that the measurement bandwidth is small compared to the laser frequency .sub.0, the phase and amplitude functions can be approximated by first order Taylor expansions: (t).sub.0+.sub.1t, and A(t)A.sub.0+A.sub.1t. In this regime, the range errors due to speckle take the form, z=c/2.sub.1z.sub.0. Equations (7) and (8) were developed to describe the complicated behavior of the amplitude and phase of coherent distance measurements from diffuse surfaces. References Baumann and Pavlicek both describe the degradation of their respective measurements due to speckle from diffuse surfaces, but offer no solutions for compensating the measured phase errors. In the following paragraphs we will describe how to use two FMCW laser radar measurements with different sensitivities of the phase to the sample distance to compensate speckle-induced phase errors.

(53) FIG. 4A is a plot 400 showing frequency versus time for first and second lasers according to disclosed embodiments.

(54) FIG. 4B is a plot 430 showing phase versus time for a stationary sample. In this case the phase with and without speckle noise are nearly identical according to disclosed embodiments.

(55) FIG. 4C is a plot 470 showing phase versus time of the measured signals for a sample with lateral motion. .sub.1,0 and .sub.2,0 represent the phase without speckle-induced phase noise, while .sub.1 and .sub.2 represent the measured phase with speckle-induced noise according to disclosed embodiments.

(56) FIGS. 4A-C illustrate conceptually the behavior of speckle-induced phase errors for a two-laser FMCW distance measurement in scenarios with stationary and laterally moving samples. FIG. 4A shows the frequency chirp for both lasers. FIG. 4B shows the phase versus time for each FMCW range measurement for a stationary sample. In this case the time evolution of the speckle phase (t) is entirely due to the time-rate-of-change of the laser frequency. For typical values of the chirp bandwidth (B>100 GHz) and surface roughness (<30 m) the measured speckle-induced range noise due to the time-rate-of change of the laser frequency is small, on the order of the surface roughness. FIG. 4C shows the phase versus time for each FMCW range measurement when the sample is moving laterally. If the sample surface were specular (i.e. without speckle noise) the measured phase would evolve linearly with a slope that is proportional to the sample range as represented by .sub.1,0 and .sub.2,0, and there would be no error in the distance measurement. When the sample surface is diffuse the speckle phase (.sub..1 and .sub..2) evolves rapidly due to the changing position of the surface features relative to the measurement beam leading to large changes in the interference between returns from the individual grid cells shown in the left side of FIG. 3.

(57) Without compensation, these phase excursions can result in range errors that are on the order of the distance measurement resolution R=c/2B.

(58) This case can be described by equation (8), with first order Taylor expansions of the phase and amplitude functions, for time intervals where the phase fluctuations are approximately linear. The entire measurement duration is then modeled by combining many sequential regions defined by first-order Taylor expansions in and A. If the two lasers are sufficiently close in wavelength, and the surface roughness is sufficiently small, the speckle phases for the two measurements are approximately equal at every point in time [.sub.1(t).sub.2()]. The condition for measuring correlated speckle phase with lasers 1 and 2 is provided in equation (9).

(59) 0$\begin{array}{cc}\frac{2.Math.{}_2-{}_1.Math.}{{\left(\frac{{}_2+{}_1}{2}\right)}^2}{}_{z}1.& \left(9\right)\end{array}$

(60) Finally, the phases from the two separate measurements may be combined to form a compensated phase whose range may depend only on the average distance to the sample z.sub.0, and the two laser chirp rates. A linear fit of the compensated phase may then be performed to extract the phase slope (e.g. the angular frequency) of the compensated range peak from which the distance measurement may be calculated. These steps are shown mathematically in equations (10) and (11).

(61) $\begin{array}{cc}{}_{\mathrm{comp}}={}_2\left(t\right)-{}_1\left(t\right)=\frac{4z_0}{c}\left(v_2+{}_{2}t\right)+{}_2\left(t\right)-\frac{4z_0}{c}\left(v_1+{}_{1}t\right)-{}_1\left(t\right).& \left(10\right)\\ z_m=\frac{\left(\mathrm{phase}\mathrm{slope}\right)c}{4}\left(\frac{1}{{}_2-{}_1}\right).& \left(11\right)\end{array}$
Chirped Sideband FMCW

(62) FIG. 5A is a plot 500 showing an example of simultaneous measurement of up and down chirps without frequency offset showing that f.sub.beat is the same for the up and down chirps according to disclosed embodiments.

(63) FIG. 5B is a plot 530 showing an example of simultaneous measurement of up and down chirps with frequency offset showing that f.sub.beat is the different for the up and down chirps according to disclosed embodiments.

(64) FIG. 5C is a simplified block diagram showing components in a chirped sideband modulation setup 570. An AOM is used to shift the LO (or Tx) off DC and thereby isolate speckle information from range information according to disclosed embodiments.

(65) As shown in FIG. 5C, the chirped sideband modulation set up 570 includes a laser 572, a waveform generator 574, an electro-optic modulator 576, a splitter 578, a circulator 580, an acousto-optic modulator 582, an acousto-optical modulator signal generator 584, a target 586, a combiner 588, and a detection and processing unit 590.

(66) In some embodiments of the invention, the optical phase-sensitive measurements may be performed using the FMCW ladar technique, and where the waveforms used may include frequency-chirped sideband modulation (i.e. homodyne rather than heterodyne) following U.S. Pat. No. 7,742,152 Coherent Detection Scheme for FM Chirped Laser Radar. U.S. Pat. No. 7,742,152 is incorporated herein by reference in its entirety. In U.S. Pat. No. 7,742,152, the authors describe a signal fading problem that hinders the measurements during motion. We have determined that this signal fading is a result of the fact that, while two phase-sensitive measurements are present, the two measurements cannot be easily distinguished. This is because the two phase-sensitive measurements utilize sideband chirps with opposite signs (i.e. one is increasing in frequency and one is decreasing in frequency), but the same chirp rate magnitude. In this case, the up and down frequency chirps are measured at common or similar RF frequencies because the measurements are performed symmetrically about DC. As shown in FIG. 5A, f.sub.beat is the same for both the up and down frequencies. When using the same detector, the signals received from the up and down chirps can therefore interfere, resulting in signal fading. As a result, the demonstrated measurements can suffer from phase-noise errors during surface (or beam) movement.

(67) To separate and utilize the two, phase-sensitive measurements, the disclosed embodiment shows that by shifting the measurement off of DC, the up and down chirps can be made to not share similar RF frequencies and the measurements can be made without signal fading because they don't interfere with one another.

(68) FIG. 5A discloses an embodiment to accomplish the shifting of the measurement off of DC. Light from a laser 572 is directed through an electro-optic modulator 576 that produces chirped sidebands on the optical carrier frequency of the laser. The electro-optic modulator 576 is driven by a waveform generator 574. The modulated light is received by a splitter 578, which outputs a local oscillator (LO) and a transmitted (Tx) portion. The Tx portion is directed through an optical circulator 580 to a target 586. Light returning (Rx) from the target is redirected and received by a beam combiner 588. The LO portion is directed through an acousto-optic modulator 582 to allow separation of the contributions from the up-chirped and down-chirped sidebands. The acousto-optic modulator is driven by an acousto-optic modulator signal generator. The frequency-shifted LO is also received by the beam combiner. Output from the beam combiner is directed to a detection and processing unit 590.

(69) In FIG. 5B, we show the result of shifting the Tx beam with an optical modulator. In this case, f.sub.beat is shown to be different for the up and down chirps. A simplified example chirped sideband modulation embodiment is shown in FIG. 5C that has been used to solve the signal fading problem and suppress phase noise-induced distance measurement errors. The figure shows the use of an acousto-optic modulator (AOM) in the LO path to enable the separation of the up and down chirp information. In the example shown one may use the RF frequencies below f.sub.offset to detect the down chirp signals and above f.sub.offset to detect the up chirp signals. Once the up and down chirps are separated, the suppression of phase noise-induced distance measurement errors follows analogously as for the carrier chirp case described in the previous section. Again, the different chirp rates (one positive and one negative in this case) provide the different optical phase sensitivities for suppression of the phase noise-induced distance measurement errors.

(70) Measurement Filtering

(71) The non-specular reflectivity of diffuse surfaces introduces the possibility for multipath interference in FMCW measurements of rough surfaces. Multipath interference refers to secondary reflections or scattering of the measurement beam between two or more surface features that may ultimately scatter back into the receiver. Multipath interference may cause time-varying phase shifts that result in errors in FMCW range measurements. These errors may become more pronounced when the sample undergoes lateral motion due to the rapid phase evolution of the interfering reflections. Specifically, large range errors may be observed in cases where the separation between the contributing surface features is sufficiently large that the inequality expressed in equation (9) is no longer valid. In such cases, the speckle-induced phase may not be well compensated by the measurement approach described in the previous section, and the resulting FMCW range measurement can exhibit errors on the order of the FMCW range resolution. FMCW measurements made on several types of rough surfaces indicate that the locations where multipath interferences occur, the spatial frequency of these effects, and the magnitude of the measurement errors may have the following properties: Their locations and magnitudes may be repeatable; the spatial frequency and magnitude of the errors are dependent on the material type; and the statistics of resulting range errors may not be Gaussian. Measurements of Lambertian scattering materials may exhibit more frequent and larger magnitude range errors while measurements of pseudo-diffuse materials, those that appear diffuse at low observance angles but reflective at high observance angles, yield less frequent and smaller magnitude range errors. Examples of range errors due to multipath interference are shown in FIG. 6A for ground glass and FIG. 6B for a paper business card.

(72) FIG. 6A is a plot 600 showing filtered (black) and unfiltered (gray) FMCW range errors for two different surfaces undergoing lateral motion of 50 mm/s with respect to the measurement beam according to disclosed embodiments. 4901 measurements from a 2 m surface roughness piece of ground glass. The filter threshold is set to reject 25% of the measurements. The filtered data set (black) contains 9 points with range errors exceeding 25 m while the unfiltered set (gray) contains 121 points with range errors exceeding 25 m.

(73) FIG. 6B is a plot 650 showing filtered (black) and unfiltered (gray) FMCW range errors for two different surfaces undergoing lateral motion of 50 mm/s with respect to the measurement beam according to disclosed embodiments. 2000 measurements from a business card with 25 m surface roughness. The filtered data set (black) contains 2 points with range errors exceeding 250 m while the unfiltered set (gray) contains 71 points with range errors exceeding 250 m. The difference in the frequency and magnitude of the range errors between the two materials is due to the fact that ground glass is pseudo-diffuse while the business card is a Lambertian scatterer.

(74) Fortunately, measurements that exhibit large range errors due to multipath interference contain signatures that may allow for detection of the errors. Once detected the errors may either be weighted or removed from the data set.

(75) The disclosed embodiments teach two filtering methods to detect measurements containing large range errors. Both methods rely on the idea that peaks containing interference from multiple unresolved surface features may often be deformed as a result of the multi-surface interference, compared to an ideal single specular reflection. One embodiment uses peak shape analysis to detect misshapen peaks. In this embodiment, the FMCW range peak may first be fit with a Gaussian or other appropriate function. Next, the root-mean-squared error (RMSE) between the measured peak and the fit function may be computed. Finally, the RMSE is compared against a threshold value to identify peaks containing large range errors. In situations where multiple range measurements are averaged, the threshold value may be computed based on the statistics of the RMSE values for the set of points being averaged. In single-point measurement scenarios the threshold may be computed in the same way as for averaged measurements using the assumption that the average peak SNR changes slowly compared to the measurement rate.

(76) FIG. 7 is a plot 700 showing an example of threshold settings for the phase RMSE and peak SNR filter according to disclosed embodiments.

(77) The second embodiment for filtering compares the range peak SNR and the RMSE of the signal phase to detect misshapen range peaks. This embodiment was used to filter both data sets shown in FIG. 6, and its implementation is illustrated in FIG. 7. For this embodiment, the SNR of the range peak may first be computed. Next, the instantaneous phase of the signal may be computed as a function of time. If the measured signal is real-valued, this step may be carried out using a Hilbert Transform to derive the complex-valued representation. The instantaneous phase may then be computed from the complex-valued signal using the tangent function. The phase is unwrapped yielding a phase versus time curve that is approximately linear. The phase curve may then be fit with a low-order polynomial, and the RMSE between the fit line and the measured values may be computed. Finally, peaks containing large range errors may be identified by comparing the measured peak SNR and phase RMSE against threshold values, as shown in FIG. 7.

(78) The data sets in FIG. 6 demonstrate the benefits of this type of filtering. In each case the number of outlier measurements is reduced by more than a factor of 10, and the standard deviation of each data set is reduced by nearly a factor of 2. During initial testing both filtering methods achieved similar performance, and the preferred approach (or combination of approaches) may depend on the specific measurement scenario and the details of the signal processing work flow.

(79) Dispersion Compensation of FMCW Measurements

(80) In measurement scenarios where the target, the beam delivery optics, or the optical medium between the reference surface and sample surface has dispersion the ()t.sup.2 term in equation (5) may become significant, and the dispersion may require compensation to produce accurate distance measurements. Fortunately, compensation of such measurements can be achieved by averaging an up-chirp and down-chirp measurement that cover roughly the same spectral region, and have approximately the same, but opposite sign, chirp rate. In practice this can be accomplished by averaging temporally sequential up and down chirps from the same chirped laser source. This technique will be illustrated by considering a measurement where the reference surface and the sample surface are separated by 0.5 m of SMF-28 fiber. The dispersion coefficient for SMF-28 is .sub.2=0.022 ps.sup.2/m, and the group velocity is .sub.g=c/1.4682 at 1550 nm. For this example the measurement duration will be 200 s, and the chirp rate will be 600 MHz/s. At the end of the measurement (t.sub.c=200 s) the up-chirp (e.g. .sub.up>0) will have accumulated an FMCW phase of .sub.u=.sub.u.sub.1zt.sub.c+(.sub.u.sub.u)t.sub.c.sup.2. The accumulated phase for each term in .sub.u is given by .sub.u.sub.1zt.sub.c=3669.4 rad, and (.sub.u.sub.u)t.sub.c.sup.2=0.0063 rad. That (.sub.u.sub.u)t.sub.c.sup.2>0 reflects the fact that .sub.u>.sub.u for the up-chirp. For the down-chirp the accumulated FMCW phase at the end of the measurement follows the same relation, .sub.d=.sub.d.sub.1zt.sub.c+(.sub.d.sub.d)t.sub.c.sup.2. However, for the down-chirp .sub.d.sub.1zt.sub.c=3669.4 rad, whereas (.sub.d.sub.d)t.sub.c.sup.2=0.0063 rad. We can now use equation (11) to calculate the distance between the reference and sample surfaces. Due to the dispersion phase term the up-chirp measurement appears too long by 2 ppm, and the down-chirp measurement appears too short by 2 ppm. However, the averaged distance measurement,

(81) $z=\frac{\left({}_u-{}_d\right)v_g}{2t_c\left({}_u-{}_d\right)}=0.5m,$
provides the correct answer.
Using Phase Reconstruction to Compensate Phase Noise-Induced Distance Measurement Errors

(82) As described in previous sections, compensation of phase-noise-induced distance measurement errors due to speckle for coherent ladar measurements may be important for obtaining accurate and precise measurements of dynamic and diffuse targets. In this section, we demonstrate an effective method for accomplishing the processing involved with this compensation using phase reconstruction, as shown in FIG. 8. For one embodiment of this invention, we may simultaneously acquire heterodyne beat signals resulting from a sample surface using two different chirp rates. As an exemplary case, an up-chirp and a down-chirp, as represented by the functions f.sub.up(t) and f.sub.dn(t), are assumed for the following discussion. We may transform these heterodyne beat signals to recover/reconstruct the signal phases as functions of time during the measurement, as represented by .sub.up(t) and .sub.dn(t). This may be accomplished using methods such as a Hilbert Transform. If the center wavelength or chirp rate is different for the up and down chirps, then one may apply scaling and offset corrections to account for the differences, as represented by .sub.up(t) and .sub.dn(t). One may then average the corrected up-chirp and down-chirp phases as functions of time to suppress the phase noise, as represented by .sub.avg(t). If desired, the noise-suppressed time domain oscillating heterodyne beat function may be reconstructed from knowledge of the noise-suppressed phase as a function of time, as represented by f.sub.cor(t).

(83) FIG. 8 is a block diagram showing components used in a processing chain 800 to compensate for Doppler and speckle phase noise according to disclosed embodiments.

(84) As shown in FIG. 8, the processing chain 800 includes an up-chirp time-domain beat determination element 810, an up-chirp phase reconstruction element 820, an element for correction of the up-chirp for and differences 830, a down-chirp time-domain beat determination element 840, a down-chirp phase reconstruction element 850, an element for correction of the down-chirp for and differences 860, an averaging element 870 for averaging up and down phases, and a construction element 880 for constructing a corrected signal. These various elements 810-880 could be implemented as electrical circuits or digital processing.

(85) The up-chirp time-domain beat determination element 810 is configured to determine an up-chirp time-domain beat f.sub.up(t).

(86) The up-chirp phase reconstruction element 820 is configured to reconstruct an up-chirp phase .sub.up(t) based on the up-chirp time-domain beat f.sub.up(t).

(87) The element for correction of the up-chirp for and differences 830 is configured to correct the up-chirp phase reconstruction .sub.up(t) based on and differences to generate a corrected up-chirp phase reconstruction .sub.up(t).

(88) The down-chirp time-domain beat determination element 840 is configured to determine a down-chirp time-domain beat f.sub.down(t)

(89) The down-chirp phase reconstruction element 850 is configured to reconstruct a down-chirp phase .sub.down(t) based on the down-chirp time-domain beat f.sub.down(t).

(90) The element for correction of the down-chirp for and differences 860 is configured to correct the down-chirp phase reconstruction .sub.down(t) based on 0 and differences to generate a corrected down-chirp phase reconstruction .sub.down(t).

(91) The averaging element 870 is configured to average the corrected up-chirp phase reconstruction .sub.up(t) and the corrected down-chirp phase reconstruction .sub.down(t) to generate an average phase reconstruction .sub.avg(t).

(92) The construction element 880 is configured to construct a corrected signal f.sub.corr(t) based on the average phase reconstruction .sub.avg(t).