2-Dimensional beamsteering using dispersive deflectors and wavelength tuning

Transcription

1 2-Diensional beasteering using dispersive deflectors and wavelength tuning Trevor Chan, 1,* Evgeny Myslivets, 1 and Joseph E. Ford, 1 1 Departent of Electrical and Coputer Engineering, University of California, San Diego, 9500 Gilan Dr. La Jolla, CA 92093, USA * Corresponding author: tkc@ucsd.edu Abstract: We introduce a 2D beascanner which is controlled by wavelength tuning. Two passive dispersive devices are aligned orthogonally to deflect the optical bea in two diensions. We provide a proof of principle deonstration by cobining an arrayed waveguide grating with a free space optical grating and using various input sources to characterize the beascanner. This achieved a discrete 10.3 by 11 output field of view with attainable angles existing on an 8 by 6 grid of directions. The entire range was reached by scanning over a 40 n wavelength range. We also analyze an iproved syste cobining a virtually iaged phased array with a diffraction grating. This device is uch ore copact and produces a continuous output scan in one direction while being discrete in the other Optical Society of Aerica OCIS codes: ( ) Free-space optical counications; ( ) Diffractive optics; ( ) Array waveguide devices. References and links 1. G. Nykola, G. Raybon, B. Mikkelsen, B. Brown, P. F. Szajowski, J. J. Auborn and H. M. Presby, A 160 Gb/s free space transission link in Proceedings of Conference on Lasers and Electro-optics, (Washington, D.C., 2000), pp M. Cole and K. Kiasaleh, Signal intensity estiators for free-space optical counications through turbulent atosphere. IEEE Photon. Technol. Lett. 16, , (2004). 3. L. Zhou, J. M. Kahn, K. S. J. Pister, Scanning icroirrors fabricated by an SOI/SOI wafer-bonding process, J. Microelectroech. Syst. 15, 24-32, (2006). 4. V. Nikulin, R. Khandekar, J. Sofka, Perforance of a laser counication syste with acousto-optic tracking: An experiental study, Proc. SPIE 6105, 61050C, (2006). 5. A. Yariv, P. Yeh, Optical Waves in Crystals, (Wiley, Hoboken, 2003), Chap B. Winker, M. Mahajan, M. Hunwardsen, Liquid crystal bea directors for airborne free-space optical counications, in 2004 IEEE Aerospace Conference Proceedings, (Big Sky. Montana, March 2004). 7. I. Filinski and T. Skettrup, Fast dispersive bea deflectors and odulators, IEEE J. Quantu Electron. 18, , (1982). 8. T. K. Chan, J. Karp, R. Jiang, N. Alic, S. Radik, C. F. Marki, J. E. Ford, 1092 channel 2-D array deultiplexer for ultralarge data bandwidth, J. Lightwave Technol. 25, , (2007). 9. J. E. Sisarian, A. Bhardwaj, J. Gripp, K. Sheran, Y. Su, C. Webb, L. Zhang, M. Zirngibl,, Fast switching characteristics of a widely tunable laser transitter, IEEE Photon. Technol. Lett. 15, , (2003). 10. M. Shirasaki, Virtually iaged phased array, Fujitsu Sci. Tech. J. 35, , (1999). 1. Introduction Free-space optical (FSO) counications provide a counication link by encoding inforation in a laser bea between the points of counication. A FSO counication link is soeties chosen over RF wireless counications because the nature of the laser allows secure, high-bandwidth wireless links. However, an additional coplexity in establishing a counication link is that the optical bea ust be aied to its targeted receiver. This is routinely accoplished between stationary nodes 1 where large angle, active alignent is not required. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14617

2 This ethod becoes ore coplex when the points of counication (either or both) becoe obile, as this requires real-tie alignent. In order to aintain the counication link, the laser bea ust be constantly pointed at the receiver, thereby requiring fast beasteering speeds for a large angular range. In addition to the relative otion of the receiver, atospheric effects such as scintillation ay also isdirect the bea, thereby calling for beasteering speeds uch less than the pointing errors caused by scintillation (on order of a few illiseconds 2 ). Several ethods have been used to steer a bea over a range of angles. An obvious solution involves siply reflecting the light off a tiltable irror, such as a galvanoetric scanner. Although these large irrors suffer in speed and power dissipation, tiltable irrors can be sall when they are ade as icroelectroechanical systes 3 (MEMS). These irrors operate considerably faster with ulti-khz sweeping speeds, but are liited to aperture sizes of a few illieters. This trade-off between size and speed is a concern because a saller aperture leads to a greater bea divergence which in turn decreases the range. Acousto-optic 4 deflectors are liited in their total angular range and electro-optic 5 crystals require kilovolt drive voltages to obtain an adequate angular range and aperture. The ideal beasteering device for FSO links would have a cobination of fast switching speeds, large aperture, pointing accuracy and low power. Usually, the desired perforance paraeters are intertwined such that when high speed (> khz) is achievable, it is reached at the expense of one of the other paraeters. An indirect way to increase aperture size without decreasing perforance is by segenting the aperture into an array of sub-apertures. Each sub-aperture functions as its own device and therefore, aintains the advantages of a sall aperture device. However, the cost of segregation is that each aperture has to be phase atched to one another. Alternatively, phase atching itself can be used as a eans of beasteering. This has been done with arrays of liquid crystals 6 where the phase delay fro each eleent in an array is given a phase delay such that eitting wavefronts atch in the desired bea propagation direction. However, liquid crystals can only be tuned at kilohertz speeds. In our work, we accoplish phased array beasteering with passive phased arrays instead of active ones. Passive phased arrays can be diffraction gratings, arrayed waveguide gratings (AWG) or virtually iaged phased arrays (VIPA). More conventional beasteering optics control the properties of the steering device to deflect an entire wavefront (or iage). Instead, orthogonal diffraction uses fixed-response optics and directs a single bea of onochroatic light by changing the wavelength of the bea. This effectively decouples the speed dependence fro the other paraeters by utilizing wavelength flexibility as another degree of freedo; one that can have a negligible effect on the FSO link D Wavelength bea-scanning concept A diffraction grating has the ability to disperse light in one diension. When a single wavelength is transitted through a grating, it is diffracting according to the grating equation, Eq. (1), where P is the grating period, the diffraction order, λ the wavelength, and θ and θi the transitted and incident angles respectively. P sin θ sinθ (1) ( ) λ i = This equation shows that the deflection by the grating changes as the wavelength is changed. Thus, when a laser is tuned across a wavelength range, the bea scans along a line 7. This alone is insufficient for FSO pointing and tracking since objects exist in a 2-diensional field of view. To reach the second diension, we can include another diffraction grating that is oriented orthogonally to the first. This arrangeent is shown in Fig. 1. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14618

3 Fig diffraction gratings oriented to achieve beascanning in 2 diensions. The difference in diffraction orders between the two diffractive structures allows for raster scanning as the wavelength is increased. There are three characteristics which these gratings ust have in order to function as a viable bea scanner that raster scans an entire 2-diensional range. Firstly, both gratings should be fabricated such that ost of the power is diffracted into one diffraction order. For exaple, in the case of a diffraction grating, a blaze angle would be required to restrict the power within an angular range that contains only one order. Second, the free spectral range of the lower order grating should be greater than the total wavelength scanning range. Thirdly, the total wavelength range should be greater than N ties the free spectral range of the higher order grating, where N is the nuber of lines that are scanned across in the raster scan. When the second and third conditions are cobined, a scan through the wavelength range will ake one pass across the angular range of one grating while also aking N passes through the range of the other. Since the gratings are oriented orthogonally, this creates a raster scan through a 2-D range of angles. 2.1 An AWG and a free space optical grating One such arrangeent that can eet the above criteria is an AWG followed by a free-space optical grating. The AWG has high dispersion because it operates in a high diffraction order. The diffraction order is deterined by the increent in length, L, between adjacent waveguides in the array, as defined by Eq. (2). Here, n awg is the index of refraction of the waveguides and λ is the central output wavelength within the th diffraction order. n awg ΔL = (2) λ The wavelength separation between diffraction orders is known as the free-spectral range (FSR). Equation (3) gives the FSR which is directly related to the final angular separation created by the diffraction grating. λ FSR = λ λ+ 1 = (3) + 1 To deterine all of the achievable directions, we ust first know the exact wavelengths that are transitted through our syste. When we assue no wavelength dependant losses fro reflections and transission through optical aterials, the transitted wavelengths fro our syste are entirely deterined by the diffractive properties of the AWG. To explore this, we start with the grating equation, and odify it to account for optical path delays in the AWG. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14619

4 This yields Eq. (4) which now includes the index of the waveguides, n awg, and the phase shift induced by the increental length, L, of successive waveguides. n awg P awg ( θ sinθ i ) + nawg ΔL = λ sin (4) In the AWG, the diffracted light out of the waveguide array is coupled into another array of output waveguides. Each of these waveguide channels collect the light at a certain angle, θn, coing out of the waveguide array. The wavelength at channel N and order is found by isolating λ in Eq. (4) to arrive at Eq. (5). λ n P awg awg ( ΔL Pawg sin θ i ) + sinθ n awg = (5) Here, we ake the assuption that the output waveguides are arranged with an angular period θo, and that the output waveguides are centered such that θ is zero at the iddle of the output array. When these are assued, the first ter on the right hand side of Eq. (5) is equal to the wavelength hitting the iddle of the array, λc. These are found using Eq. (3). With these considerations, we now arrive at Eq. (6) where N is the output channel nuber and N ax is the total nuber of channels. nawg Pawg N ax + 1 λ N, = λc + sin θ o N (6) 2 Eq. (6) can be rewritten as Eq. (7) to explicitly yield N. N = sin 1 ( λ λ ) n θ N, awg o P c awg N The wavelengths within a diffraction order are separated spatially into free-space by the V- groove array. The output array of spots yielded by the v-groove array (pitch P v ) is re-iaged by the free-space grating deultiplexer before it is directed by the output lens (with f o focal length). The output angle is given by Eq. (8). o ax N ax + 1 PV N = 2 tanθ 1 (8) f The output angle in the orthogonal direction is created by the dispersion of the diffraction orders by the free-space grating deultiplexer. This deultiplexer ight contain a 4-f iaging syste with a reflection grating of period P g, tilted by θg, at the Fourier plane and uses identical Fourier lenses with focal length f. The resulting output angle deflection caused by the diffraction grating is given by Eq. (9). f 1 λ tanθ 2 = tan sin + sinθ + g θ g (9) fo P g (7) (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14620

5 3. Experiental setup For our experiental deonstration of beascanning, we used the syste illustrated in Figs. 2 and 3. We used a JDSU 1x8 channel arrayed waveguide grating as the first diffractive eleent which operated in the 194 th diffraction order at 1550 n. Since the AWG gives discrete outputs at specific wavelengths, the achievable scanning angles exist in a nearly rectangular array rather than a set of lines. The AWG channels were transferred into free space through a 635 µ pitch v-groove array. This led into a free space optical grating deultiplexer, taken fro a Network Photonics WDM switch, consisting of a 100 focal length Fourier lens followed by a 3 rd order, 300 lp/ reflection grating (Newport catalog code 53-*-013R). The architecture is siilar to that used in our prior work in developing a large channel wavelength deultiplexer 8. Like the deultiplexer, the output fro the two deultiplexers is a grid of single ode Gaussian spots where each one corresponds to a different wavelength. This is converted into a bea scanner with the addition of a gold irror (to divert the light away fro the coponents) and a Nachet Vision 8x, 25 focal length icroscope objective to colliate and direct the bea. Fig. 2. Syste and coponents used to deonstrate beascanning. The deultiplexed channels fro the AWG are ejected into free space by a V-groove array which sits at the focal plane of a Fourier lens. This lens is part of a free space grating deultiplexer which separates the ultiple orders fro the AWG. The output is focused at the focal plane of a icroscope objective which colliates and directs the light. Fig diensional illustration of a onochroatic bea propagating through the syste. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14621

6 The AWG and free-space optical grating were both designed for the C-band ( n) of the telecounication spectru. These wavelengths are also popular for free-space optical counications; thus, our deonstrations are appropriately carried out using wavelengths near 1550 n. Figure 4 shows the predicted output angles of the directed bea for wavelengths fro 1546 to 1590 n using the equations shown previously and the paraeters illustrated in Fig. 2. Here, we used a waveguide index, n awg, of 1.5 and a L of 199 µ to have the AWG operate in the 194 th diffraction order at 1550 n. We also assued that there is no apodization fro any apertures which ight have liited the field. The figure shows 6 coluns and 8 rows of achievable directions. Each row corresponds to a single channel fro the AWG, and the horizontal separation between the rows is created by dispersion fro the diffraction grating. The 6 coluns are a result of fitting 5 FSRs (~ 8n for the AWG) in the total wavelength range. Fig. 4. Calculated output angles fro the bea scanner. Ideally, the range of a bea scanner should be continuous instead of a discrete array of spots. This functionality can be incorporated into our bea scanner with the addition of a continuous bea scanner to cover the range in between the spots. The requireents of this bea scanner are achievable with existing technology (such as an electro-optic scanner) since the required scanning range is now liited to the distance between our discrete directions. 4. Experiental results Figure 5 shows a photograph of the actual setup where the output bea is projected onto a white screen 20 inches distant. We used a Sensors Unliited 320M caera with visibility in the NIR wavelength range to iage the projected beas. This photograph also shows an Agilent tunable laser as the optical source. In our experients, we used a nuber of different optical sources to characterize different properties of the bea scanner. For fast beascanning, the ideal source would be a fast, wavelength tunable laser. To the author s knowledge, the fastest tunable lasers are grating-assisted codirectional coupler with rear sapled reflector (GCSR) lasers which can switch across its entire wavelength range in under 50 ns 9. The wavelengths fro these lasers are liited to a discrete set of wavelengths; however, this does not har our syste s scanning ability. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14622

7 Free-space reflection grating Tunable laser n sweep Microscope objective AWG V-Groove fiber array Fig. 5. Photograph of the bea scanner with a tunable laser acting as the optical source. We used a C- and L-band ASE source to siultaneously transit power into the entire output array of beas. Figure 6 shows the captured iage of all of these beas projected onto the screen. Fig. 6. An ASE optical source is used to send power towards every achievable direction of the bea scanner. The caera is saturated to show all spots. Figure 6 shows that the achievable range fro our bea scanner was 10.3 and 11.0 and exists as discrete directions in an 8 by 6 array. This is consistent with the predictions shown previously. The apparent power variation over the different directions is priarily due to the spectral profile of the ASE source. Previous work using these coponents as a wavelength deultiplexer instead of a beascanner 8 suggests that the insertion losses are unifor across the wavelength tuning range with approxiately 8 db loss. For the next part of our experients, we used an Agilent 81642A tunable laser to send power to a single direction. The output bea of a single wavelength is shown in Fig. 7(a). While this laser s slow tuning speed (2.8 seconds to scan through 100 n) akes it infeasible for FSO beascanning, the Agilent tunable laser was still useful in observing the output bea quality of the bea scanner. The bea profile is a result of a re-iaged fiber output (fro the v-groove array) directed and colliated by the icroscope objective. Figure 7(b) shows an iage of the single spot. This was captured using a bare InGaAs sensor placed iediately after the output of the bea scanner. We easured a 6 1/e 2 bea diaeter which (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14623

8 atches the expectations of a bare fiber output being colliated by the icroscope objective. We note that this bea waist can be changed by selecting a different colliating lens. Figure 7(b) also shows that the bea aintains a priarily Gaussian for which is indicative of low distortion through the syste and yielding a sooth wavefront. In previous work 8, we showed that in these 2D dispersive systes, the crosstalk inherent in the AWG and the linewidth of the optical source deterines the aount of crosstalk between neighboring wavelength channels. The prior results also show that there is virtually no (> 30 db) crosstalk between channels which are separated by one FSR. Using the devices in our setup as an exaple, the Agilent laser linewidth is 100 khz which is negligible copared to the 100 GHz AWG channel pitch. This eans that the crosstalk is deterined by the crosstalk between AWG channels which we easured as greater than 35 db. Figure 7(a) also shows no apparent leakage of the optical power into neighboring channels. Fig. 7. (a). The output of a single wavelength and b) its profile iediately exiting the beascanner. Since this syste is intended for use in FSO counications, the transitted bea ay carry a odulated signal and would inherit a finite bandwidth. Since our syste is dispersive, a significant bandwidth would stretch the output bea over a linear range of directions. The aound of stretching would depend on the degree of dispersion of the free space optical grating. Crosstalk ay also cause a liitation when the signal is odulated with a bandwidth greater than those of the AWG channels. An alternative to using a tunable laser as the source is to use an ASE source passed through a tunable filter. We deonstrated this option using the sae ASE source used above along with a CoreTek MEMS tunable etalon filter. Two separate detectors were placed at different points in the field of view. These points corresponded to n and n wavelengths. Figure 8 shows the transient response of the power at the siultaneously easured positions when the filter was switched between these two wavelengths. This shows a 183 µs switching tie which is purely deterined by the switching speed of the tunable filter. The CoreTek MEMS filter used here had a passband of 0.47 n which is larger than the channel pitch of our AWG. Thus, this filter produced approxiately 5 db of crosstalk in adjacent channels and would not be suitable as a secure FSO beascanner. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14624

9 Fig. 8. Transient response when switching between two directions. 5. Bea scanner with VIPA An alternative ipleentation of this bea scanner is one that uses a VIPA in place of the AWG. The VIPA, invented by M. Shirasaki 10, is a glass slab that acts uch like a highly dispersive diffraction grating. The VIPA operates on the principle of virtual iages interfering with one another, as illustrated in Fig. 9. These virtual iages are of a line source which is created by focusing a colliated bea with a cylindrical lens. This line source is focused at the entrance aperture of the glass slab which is tilted at a sall angle. The back surface of the slab is nearly 100% reflective while the front slab has a graded reflectivity to allow output light. In this arrangeent, the light will undergo ultiple reflections and create a series of staggered virtual line sources. The light leaks out of the front surface into free-space where the cylindrical waves created by the virtual line sources interfere with each other. The graded reflectivity equalizes the leaked power of successive reflections. Ultiately, this resebles a diffraction grating operating with a large tilt and very high order. Fig. 9. A VIPA with the virtual line sources illustrated. One beneficial advantage of the VIPA over the AWG is that the VIPA operates in a uch higher diffraction order. This allows for ore raster lines over a chosen wavelength range. This is described by the grating equation for the VIPA, Eq. (10), where v is the diffraction order, t v is the thickness of the VIPA, n v is its index of refraction, θv is the tilt angle of the VIPA and θo is the output angle. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14625

10 vλ 2t n v v π sin + θ o θ 2 = v The output angle, θo is shown explicitly in Eq. (11). π 1 vλ θ = + o θ v sin (11) 2 2tvnv The f-nuber of the cylindrical lens deterines the field of view of the virtual line sources. This subsequently restricts the range of possible output angles, θo and will therefore keep the optical power in diffraction orders that fall within this angular range. In a beascanner, especially one used for secure free-space optical counications, only one diffraction order should exist within the field of view for each wavelength. To ensure this, the optical path length of the VIPA, t v n v, inherits a axiu liit. When this optical path length is sall, the value of the last factor in Eq. (11) is significantly different at adjacent diffraction orders. This difference should be large enough such that the adjacent diffraction orders fall outside the possible range of θo and only one diffraction order exists for each wavelength. This requireent is shown in Eq. 12 where f # is the f-nuber of the cylindrical lens. tan 2 1 vλ < 1 ( v 1) sin sin f # 2tvnv 2tvnv 1 λ The VIPA is ade into a 2D beascanner by layering a diffraction grating on the VIPA s output surface. This is depicted in Fig. 10 where every wavelength, represented by different colors, is shown projecting out of the beascanner. Also, a planar waveguide creates the line source entering the VIPA instead of a cylindrical lens. This creates a ore sealess and copact integration of fiber optic input into the VIPA. The angular range of the cylindrical wave is now deterined by diffraction fro a slit. In other words, the angular range of the beasteerer is inversely proportional to the width of the slit and independent of the total aperture. To increase the aperture size, the length of the line source can be echanically increased and the transissivity through exit surface of the VIPA can be decreased to create ore internal reflections and distribute the power over a longer distance. Unlike the fiber coupled AWG, the VIPA gives a spatially continuous, free-space output eaning that the beascanner can scan continuously along one of the directions. (10) (12) Fig. 10. Bea scanner consisting of a VIPA and a free-space optical grating deultiplexer. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14626

11 Equation (10) can be used to find the wavelength range of each diffraction order in the VIPA s field of view as well as the output angle of each wavelength in one direction. The grating deterines the output angle in the orthogonal direction which is predicted by Eq. (1). We envisioned a VIPA beascanning syste with the following paraeters to deonstrate its potential perforance. The cylindrical wave feeding into the VIPA has a nuerical aperture of 0.05 and the diffraction grating has a 500 lp/ pitch operating in the 1 st order. VIPAs typically have a 1.5 index, 1 thickness. Figure 11(a) shows the output if we use a typical VIPA tilted at 2.5 where the spots at 1553 n are circled. This figure reveals that ultiple diffraction orders are created. As entioned earlier, this is because the optical path, t v n v of the VIPA is too large. In order to create a viable beascanning syste which transits only one bea, the VIPA would have 1.5 index of refraction and 0.1 thickness. The thinness of the VIPA is atypical and ay pose additional probles such as coupling the light into the slab, sturdiness or unifority of the VIPA. For the purposes of this investigation, we assued that such a device could exist and operate with the sae properties of a typical VIPA. The sae section of the output is shown in Fig. 11(b) with the thinner VIPA. Figure 12 shows the entire calculated output field where the optical wavelength range used is n. With these paraeters, Fig. 12 shows approxiately 5 x 9 total output range existing over 26 coluns of output angles. Fig. 11. A section of the output of a 2D deultiplexer which uses a) a 1 thick VIPA and b) a 0.1 thick VIPA. Spots created at 1553 n are circled and show degeneracy through ultiple diffraction orders when the VIPA is too thick. For secure FSO counications, only one diffraction order should exist. Fig. 12. Calculated output directions (in degrees) using a VIPA in the bea scanner. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14627

12 6. Conclusion and discussion We presented a beascanning ethod where the input wavelength is directly related to the output direction in a 2-diensional field of view. This is accoplished by cobining two wavelength dispersive devices orthogonally and in series. With the proper devices, the bea raster scans through the 2-diensional field as the wavelength fro an optical source is increased. We have experientally deonstrated beascanning by cobining an AWG with a freespace optical grating deultiplexer. Using this cobination of parts, we obtained 2- diensional beascanning on an 8x6 grid of discrete directions. This output was directed using a 25 focal length icroscope objective resulting in a total angular range of 10.3 by These results were consistent with calculations based on diffractive theory and geoetric optics. We achieved 183 µs scanning ties using an ASE source followed by a CoreTek tunable filter which switched the power fro n to n. We also discussed an iproved version of the bea scanner which uses a VIPA cobined with a free-space optical grating deultiplexer. Because the VIPA has a continuous output, one diension in the output field of view can be scanned continuously. In the other diension, the high dispersion of the VIPA allows us to point to a greater nuber of discrete directions which are spaced closer together. This was verified with theoretical calculations on a conceptual syste. Acknowledgents The authors would like to acknowledge DARPA Systes Microsystes Technology Office for their support in this work. (C) 2008 OSA 15 Septeber 2008 / Vol. 16, No. 19 / OPTICS EXPRESS 14628