1. INTRODUCTION

Typical LADAR systems are relatively bulky, expensive, and power hungry [1 –6]. Our goal is to develop a compact LADAR system that can be used for applications such as unmanned air and ground vehicle navigation. Using a single transmit/receive (TX/RX) aperture instead of two apertures allows the reduction of the system size and has other benefits, as described below. Previously, in [5], a shared aperture was used for the TX and RX beam paths so that the device could use a single small-polygon scanner to build an image. Another system [6] also used a single aperture to allow scanning with a micro-electromechanical system digital mirror deflector. A commercially available system also uses a shared aperture and scans with a rotating mirror [7]. Typical LADARs [8] are bi-static, with separate TX and RX optics that require precise alignment relative to each other.

LADAR implemented with all-fiber optic components provides a compact, lightweight, and low-cost system. Fusion splicing fiber components eliminates time-consuming and costly optical alignment procedures. As will be shown, the all-fiber approach also makes it possible to implement a compact beam scanner that uses a lead zirconium titanate (PZT) cantilever to vibrate a fiber tip, resulting in its large lateral movement. When the fiber tip is placed in the focal plane of a collimating lens, this lateral translation causes the angle scanning of the collimated TX beam. A position sensor placed close to the fiber tip is then used to determine the instantaneous tip position and resulting beam angle [9].

2. BLOCK DIAGRAM OF MONOSTATIC ALL-FIBER LADAR SYSTEM

The monostatic, all-fiber LADAR system is shown in Fig. 1. It consists of: (1) a fiber multiplexer (MUX) that uses a double-cladding fiber (DCF) with a near-single-mode core that transmits the signal from a (2) 1.5 μm pulsed fiber laser, and an inner multimode (MM) cladding that captures the RX light, which is coupled into an MM fiber that transmits the RX light to (3) an avalanche photodiode (APD), (4) a compound resonant cantilever with a fiber cantilever attached to its end, (5) a silicon two-dimensional (2D) position sensing detector (PSD), (6) a scanning/receiving lens, (7,8) and two analog/digital (A/D) systems on computer peripheral component interconnect (PCI) cards used to collect the data.

 

Fig. 1. All-fiber scanning LADAR design.

Download Full Size | PDF

The pulsed fiber laser consists of a distributed feedback (DFB), 1.5 μm seed-laser diode, a 2-stage erbium-ytterbium doped fiber amplifier (EYFA), and electronics to control the EYFA pump power and the triggering of the DFB laser. The EYFA generates 2.9 μJ, 1 ns full width at half-maximum (FWHM) pulses at a pulse repetition frequency of 50 kHz. Approximately 10% of the time-averaged output power was in the form of amplified spontaneous emission (ASE). The output fiber of the laser is fused to the DCF fiber of the MUX.

A. Fiber Scanner

A key component of the system is the 2D vibrating biaxial fiber cantilever. The beam-scanning fiber has a flattened cladding shape, resulting in two different mechanical resonant frequencies for its thick and thin axes. The double-cladding fiber, shown in Fig. 2, consists of a 12 μm diameter, $\mathrm{NA}=0.12$, near-single-mode core, a 90 μm, $\mathrm{NA}=0.22$ glass inner cladding, a $121\mathrm{\mu m}\times 100\mathrm{\mu m}$ glass outer cladding, and an acrylate jacket (not shown). This beam-scanning fiber is fusion spliced to a matched-core DCF fiber, with a circular 90 μm inner cladding, which is used to fabricate the fiber MUX. A similar scanning method has been proposed previously [10,11]. The perpendicularly cleaved scanning fiber tip has an anti-reflection (AR) coating to minimize the back reflection of the TX light into the fiber inner cladding and to the APD. This back reflection can cause damage to the APD or signal saturation which, because of slow detector recovery, results in the reduction of the minimum detection range.

 

Fig. 2. Image of the cleaved scanning fiber face.

Download Full Size | PDF

PZTs have been used extensively in scanning probe microscopy and in endoscopes [12 –16]. However, these systems typically used 4 PZTs arranged in a tubular shape to achieve an $X–Y$ scan. Our fiber is epoxied to a single bi-laminar PZT stripe actuator at a 45° angle, as can be seen in Fig. 3, where the cantilever PZT length is 3 cm, and the fiber cantilever length is 1.73 cm. This method is simpler in construction and achieves similar fiber deflection with a drive voltage that is an order of magnitude lower when compared to a recent tubular PZT implementation [12].

 

Fig. 3. Fiber scanning assembly and picture of scanning fiber tip positioned close to the position PSD.

Download Full Size | PDF

The PZT response is characterized by using a function generator and a microscope with a digital camera. Typical PZT resonant frequencies are near 300 Hz. The FWHM of the PZT response was measured to be 40 Hz. The fiber frequencies were set near the PZT resonance (Fig. 4) by choosing the fiber cantilever length. The fiber tip motion is amplified by the double resonances of the PZT and fiber bending modes. Micrometers of PZT movement are resonantly amplified to millimeters of fiber tip movement; typically less than $1{\mathrm{V}}_{pp}$ of PZT drive was required.

 

Fig. 4. Biaxial fiber oscillation model.

Download Full Size | PDF

For this vibrating fiber PZT arrangement, an illustrative fiber tip displacement of 2 mm is shown in Fig. 3. Significantly larger amplitudes were easily achievable. The low-frequency mode (vibration parallel to the short transverse fiber axis) is calculated to have a resonant frequency of 290.2 Hz based on an axis length of 100 μm; the high-frequency mode (vibration parallel to the long transverse fiber axis) is calculated to have a resonance at 336.5 Hz based on an axis length of 121 μm. The resonant frequencies agree well with measured resonant frequencies of 286 and 333 Hz, respectively. The measured fiber FWHM resonance widths are $\sim 8\mathrm{Hz}$, and the damping constant in Fig. 4 has been adjusted to match this. Sources of error for the calculated max frequency include fiber uniformity, length of the epoxied fiber, and details of the epoxy mounting to the PZT.

A single-drive voltage consisting of the sum of two sinusoids at the fiber’s resonant frequencies is applied to the PZT. This excites both orthogonal vibrations at their respective frequencies, creating a Lissajous scan pattern. The frequencies are fine tuned to give a low-pattern repeating rate and dense scan pattern in 0.4 s, as seen in Fig. 5. The Lissajous scan appears sparser in the center and denser near the edges because of sinusoidal dwell time factors. The edges of the scan pattern in Fig. 5 are not orthogonal. We hypothesize that this results from the coupling of the vibrational modes to a higher-frequency torsional mode of the fiber. Such coupling could be induced by the PZT vibration direction misaligned from the center of the fiber, or by asymmetry in the epoxied fiber attachment. A generalized coupled-mode model with the fiber vibrational modes coupled by a small perturbation gives results consistent with those in Fig. 5. Some mounted fibers have shown more orthogonal scan edges than the images shown in Fig. 5. Note that the outer shape of the scan pattern does not affect the shape of the scanned objects: square objects will appear square (see Fig. 10 below).

 

Fig. 5. $5°\times 7°$ Lissajous scan pattern for (a) 400 ms camera exposure and (b) a normalized computer display of the PSD signals after 400 ms.

Download Full Size | PDF

Equation (1) shows the relation between the scan angle, $\theta$, the fiber lateral tip displacement, $\mathrm{\Delta }x$, and the focal length, $f$, of the lens. This is also illustrated in Fig. 1.

Our TX beam collimating lens (LightPath GRADIUM) has a focal length of 40 mm and a clear aperture of 27 mm. The lens is sufficiently large so that with a $\sim 14°$ divergence of light cone emerging from the fiber core, and when $\mathrm{\Delta }x=2\mathrm{mm}$, the transmitted light is not clipped by the lens aperture. With these parameters, the maximum full scan angle is 6°. A larger scan angle can be achieved with a lower NA fiber core and a lower $F$-number lens. The collimated TX beam divergence was 0.3 mR.

B. Silicon Position Sensing Detector

A PSD is needed to instantaneously sense the fiber tip position since changes to the PZT, fiber, and environmental factors, such as temperature and external vibrations, would affect a one-time phase calibration that attempted to predict the fiber tip location. Small, mechanically induced position artifacts were observed in a scanning fiber system [13] that did not use a PSD. The silicon PSD [17], which is custom modified to operate in a transmissive configuration for the 1.5 μm light, is shown in Fig. 6. Since silicon has a bandgap larger than the 1.5 μm photon energy, it is highly transparent for linear one-photon absorption in the TX and RX beams. However, the sensor detects the high peak power 1.5 μm TX optical pulses due to two-photon absorption (TPA) in silicon.

 

Fig. 6. 2D position sensing silicon detector for 1.5 μm pulses.

Download Full Size | PDF

Although weak, the TPA during the nanosecond-wide, multi-kilowatt peak-power transmitted pulses causes sufficient electron-hole generation in silicon to produce a readily measured pulsed sensor electrical signal for the “$x$” and “$y$” positions. The PSD output pulse is bandwidth limited to 400 kHz, and its shape changes with location. Typically, PSDs are used in a continuous mode rather than a pulsed mode, as is done in this LADAR system. We obtain a low-noise signal, linear with displacement, by numerically integrating the PSD response during the first half of the 20 μs interval between pulses. This process is completed for each of the 4 PSD signals: $X_1-X_2$, $X_1+X_2$, $Y_1-Y_2$, and $Y_1+Y_2$. The $x$ and $y$ coordinates are then computed using Eq. (2). The position measured for each pulse is used to create the lateral display coordinates, as shown in Fig. 5(B).

The PSD had AR coatings on all optical surfaces; this resulted in negligible transmission loss for the TX light and minimized back reflection of the TX light into the receiver.

C. Fiber Multiplexer

The TX/RX MUX fiber coupler contains a DCF with a near-single-mode 12 μm, $\mathrm{NA}=0.12$ core imbedded in a 90 μm, $\mathrm{NA}=0.47$ multimode glass inner cladding, and a low index polymer outer cladding (Fig. 7). To fabricate the fiber MUX, the 90 μm fiber is laterally fused, over a length of several centimeters, to a polymer-clad, multimode 105 μm diameter, $\mathrm{NA}=0.47$ fiber. A fraction of the LADAR returned light propagating in the 90 μm fiber transfers to the 105 μm fiber and is then coupled into the receiver APD. The MUX performance is modeled and optimized by using RSoft finite difference physical optics simulation, which modeled the effects of fiber geometry, fused fiber overlap, and coupling length.

 

Fig. 7. Monostatic system multiplexer and transmitter/receiver.

Download Full Size | PDF

To calculate the power transfer ratio from the 90 μm fiber into the 105 μm fiber, a simulated randomized speckle field consistent with an expected return with an NA of 0.22 is launched into the core of the fiber. Since the core’s NA of 0.12 does not support this light, most of it is transferred to the cladding within 500 μm of the propagation distance in the fiber, as can be seen in the lower left insert of Fig. 8. The fibers are fused over a length of 1 cm, where the circular fiber areas overlap by 5 μm. The model shows that 53% of the power is transferred from the 90 μm fiber to the 105 μm fiber, 40% of the power remains in the 90 μm fiber, and 7% scatters out of the fibers. In the absence of scatter loss, a transfer ratio of 58% can be expected, based on the ratio of the 105 μm fiber area to the total area of the two fibers. An increase of the transfer ratio could be achieved with a larger multimode fiber; since the APD detector diameter is 200 μm, multimode fibers up to that size could be used in the MUX. The fabrication of fused fiber couplers was completed using a ${\mathrm{CO}}_2$ laser, beam scanning mirrors, and a pair of computer-controlled fiber-holding chucks; this fabrication is similar to the work of Dimmick et al. [18]. An MUX fiber transfer ratio of $\sim 50\%$ was typically measured, and is in reasonable agreement with the model. The light is coupled into the detector using a free-space matched-pair lens system to collect light from the 105 μm MUX fiber and focus it on the 200 μm diameter APD. In future systems, we plan to incorporate a fiber-coupled APD.

 

Fig. 8. Model of the MUX power transfer.

Download Full Size | PDF

An important issue in monostatic LADAR systems is the residual reflection of the transmitted pulse back into the receiver. In the all-fiber system shown in Fig. 1, an excessive reflection from the AR-coated end of the transmitted fiber can easily damage the APD or cause a large saturation of the APD, making it difficult to have a short minimum range for the LADAR system. We explored optical and electronic methods to mitigate these effects.

The core of the near-single-mode scanning fiber primarily supports the fundamental ${\mathrm{LP}}_{01}$ optical mode, with a weakly bound ${\mathrm{LP}}_{11}$ mode also allowed. If the cleave angle is not perfectly perpendicular, then light is also reflected back into the cladding modes of the fiber, as shown in Fig. 9.

 

Fig. 9. Normalized effect of cleave angle versus reflected light.

Download Full Size | PDF

The graph is normalized to unity reflection at 0°, so the received coupling fraction should be multiplied by the reflectance of the AR coating on the face of the fiber tip. Only one higher-order ${\mathrm{LP}}_{11}$ mode is important, since the laser has only one polarization and the cleave is considered to be in a single plane. As the cleave angle increases, more light is reflected into the ${\mathrm{LP}}_{11}$ and cladding modes. Only the cladding modes collected by the APD receiver are relevant to the measured cleave-reflection ($T_0$) effects. For an angle deviation of 0.1°, achievable by fiber-cleaving or polishing methods, the coupling fraction is predicted to be $4\times {10}^{-5}$. The fiber results presented here had a fiber tip cleave tilt guaranteed by the vendor to be less than 1°. For a fiber-face AR-coating reflectivity of 0.5%, this would result in $3.6\times {10}^{-4}$ of the incident ${\mathrm{LP}}_{01}$ power coupling into the DCF cladding; therefore, for a 50% DCF-to-MM fiber transfer ratio, $1.8\times {10}^{-4}$ of the laser power would couple into the APD. In our experiment, we measure $1.1\times {10}^{-4}$ of the TX (non-ASE) power at the APD. Given the uncertainty of the cleave angle, the measurement is reasonably consistent with the projections. Other contributors to $T_0$ are TX light that leaks out from the core into the cladding at the fiber fusion splices and in the MUX and is then is back reflected by the AR coated fiber tip, and the reflection of TX light from the PSD. It should be possible to achieve a significant reduction of the reflected power by further minimization of these effects and a reduction in the fiber tip cleave tilt angle.

3. RESULTS

The Acqiris AP240 pulse analyzer performs an analog-to-digital conversion of the RX pulse signal in order to calculate the range from the pulse time-of-flight. It samples at 2 giga-samples-per-second, operating in a peak finding mode that uses interpolated data every 31.25 ps (0.5 ns/16), resulting in a range precision of 4.7 mm.

As shown in Fig. 10, the system can resolve a clear image of a pedestal at an 8 m range with a 0.4 s frame using first returns. The PSD data provides the lateral coordinates, and the measured pulse delay gives the longitudinal ($z$) coordinates. The data points are also false-color coded by the delay. The pedestal consists of five square foam slices, ranging from 2.5 cm ($1^{\prime \prime }$) to 12.7 cm ($5^{\prime \prime }$) wide, each 1 cm thick. For these data, a PZT drive amplitude of 0.13 V gave a $\mathrm{\Delta }x$ of 0.32 mm, and a half-scan angle of $\theta =0.5°$.

 

Fig. 10. 3D scanning (monostatic) LADAR point-cloud output at 8 m.

Download Full Size | PDF

Data for targets at a 26 m range are shown in Fig. 11. The left figure is a user-rotatable planar view of the three-dimensional (3D) $(x,y,z)$ point cloud for each frame; the right part uses 2D PSD data for each frame (with fixed orientation), false-color coded by the range for each point. At longer target ranges and LADAR scan angles, added delay is apparent in the corners of the image compared to the center of the scan in Fig. 11. This is attributed to changes in the interpolated pulse peak location from variations in the sloping baseline and signal strength. The former is caused by the angular variation of the $T_0$ reflection from the PSD; the latter is caused by the larger lens aberration with an increasing scan angle. The resulting range error in Fig. 11 is less than 5 cm. A constant-fraction threshold circuit, reducing the $T_0$, or improving the anti-reflection coatings on the PSD, would alleviate this problem.

 

Fig. 11. 3D scanning (monostatic) LADAR output at 26 m. The left pedestal is the same used in the previous figure, and the right pedestal consists of four square foam slices, ranging from 7.6 cm ($3^{\prime \prime }$) to 30.5 cm (${12}^{\prime \prime }$) wide, each 1 cm thick.

Download Full Size | PDF

These results were obtained using an APD with a 100 MHz transimpedance amplifier (TIA). It uses a limiter capacitor between the APD and TIA that aids in $T_0$ recovery. In addition, there is a fast transistor that switches to the ground during the initial $T_0$ pulse.

4. CONCLUSION

This work describes the first implementation of a new type of compact, monostatic, all-fiber scanned LADAR system. It uses a combination of several unique components and techniques: an all-fiber MUX for separating the TX/RX signals, a monostatic all-fiber optical transmitter/receiver, and instantaneous pulse direction sensing. The system is compact, lightweight, low power, and low cost with robust TX/RX alignment.

We are continuing to investigate time-gating higher bandwidth detectors for higher resolution and faster $T_0$ recovery times. We are also making component improvements that will reduce the $T_0$ signal and allow for a shorter minimum range. The 3D images could be even clearer under operational conditions due to multiple look angles. This would allow further image manipulation using resampling and gridding [19 –21].