JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 8, AUGUST 2002 1585 Add Drop Multiplexing By Dispersion Inverted Interference Coupling Mattias Åslund, Leon Poladian, John Canning, and C. Martijn de Sterke Abstract We demonstrate experimentally a previously proposed add drop multiplexer in a two-moded structure with a symmetric Bragg grating. The strong dispersion in the forward direction outside the reflective region of the grating is used to create a phase change for frequencies between the bandgaps of the supermodes of a twin core fiber. The phase change in transmission inverts the interference pattern at the end of the device. Signals within a narrow wavelength range couple out to one core; all other wavelengths couple out to the opposite core. Preliminary experimental results show a crosstalk ratio of 15 db. Index Terms Couplers, gratings, multimode waveguides, wavelength-division multiplexing. I. INTRODUCTION ADD DROP multiplexer devices allow access to single wavelength channels and are crucial devices in wavelength-division-multiplexed (WDM) networks [1]. Ideally, such devices have a low component count and require no fiber pig-tailing, thus allowing for low-cost manufacturing. Multiplexers based on Bragg grating technology enable the most spectrally efficient filter responses because of their spectrally steep edges. The most commonly deployed methods use the narrow bandwidth reflective properties of gratings to separate wavelength channels [1]. However, in practice, these devices require additional optical components, such as circulators, to separate input and reflected signals. There is, therefore, an interest in novel structures that can reduce manufacturing complexity and component count. This is most easily achieved in all-fiber devices. Some all-fiber designs already exist [2] [4]. All-fiber devices based on Bragg gratings and couplers can be divided in two categories: the Mach Zender interferometer with Bragg gratings outside the coupling regions [2] and devices with Bragg gratings imprinted inside the coupling region. The former device is interferometric with two separate fibers as the optical paths. The paths require careful and complex balance, which must be maintained throughout the lifetime. This instability is avoided by using an integrated planar waveguide Mach Zender [5] at the manufacturing expense mentioned earlier. The second category of devices, grating imprinted couplers, can be divided into two subcategories: the grating frustrated Manuscript received November 26, 2001; revised April 1, 2002. This work was supported by the Australian Research Council and by Ericsson Australia Pty Ltd. M. Åslund, L. Poladian, and J. Canning are with the Optical Fiber Technology Centre, University of Sydney, Sydney, Australia (e-mail: m.aslund@ oftc.usyd.edu.au). C. M. de Sterke is with the School of Physics, University of Sydney, Sydney, Australia. Digital Object Identifier 10.1109/JLT.2002.800355 coupler [3] and the grating assisted coupler [4]. In this paper, we introduce a third subcategory, the dispersion inverted interference coupler. The device is based on the idea proposed in [6] in the context of a two-moded planar waveguide device. These devices are compared and discussed in Section II. In Section III, we show how we experimentally verified the proposed principle. In Section IV, we discuss the experimental results. We also demonstrate that the concept applies to any structure with two interfering modes. II. BRAGG GRATING IMPRINTED COUPLERS Devices with Bragg gratings imprinted in the coupling region can be divided into two types: symmetric or asymmetric. In the symmetric type, the grating is uniform across both cores, which are identical, and it does not matter for the functionality of the device into which core the signals are launched. In the asymmetric type, the grating is not uniform in both cores, and it is important for the functionality of the device into which core the signals are launched. Let us first consider the grating frustrated coupler [3], which is an asymmetric device. It consists of a matched directional coupler (two cored coupling region) with a Bragg grating written in the opposite core from where the input light signals are launched. At wavelengths unaffected by the grating dispersion, the light couples out fully to the opposite core from where the signal enter. In contrast, at wavelengths affected by the grating dispersion, the coupler becomes mismatched and the light remains in the input core. The grating dispersion effect is often enhanced by using dissimilar cores, which are only matched and couple over fully for a wavelength range relatively close to the Bragg wavelength [7]. If the signals are launched into the core with the grating, they are just reflected back into the core where they came from, as can be confirmed from the coupled mode equations given in [8]. The asymmetry of this device prevents it from being used as an add drop multiplexer; it has only a drop function and does not have an add function. It is not considered any further in this paper. Let us now consider symmetric grating imprinted couplers: the grating assisted coupler [4] and our new proposed device, the dispersion inverted interference coupler. They both consist of a coupling region with a grating written uniformly across both waveguides. Both devices are explained using the supermodes [9] of the twin-core structure since the formalism works well for symmetric devices. As the supermode method is not commonly used, we start our device descriptions with a brief review of the supermode method. 0733-8724/02$17.00 2002 IEEE
1586 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 8, AUGUST 2002 Fig. 1. Transverse symmetrical fields of the even and odd supermode (solid lines) of a twin-core structure (dashed circles). Fig. 3. Accumulated phase at the end of the grating structure for the two supermodes as a function of wavelength (solid line: fundamental supermode; dashed line: second-order supermode). Gray boxes mark the position of each supermode s rejection band. Fig. 2. Interference beating of symmetrical supermodes during propagation (matched directional coupling). A schematic of the transverse electric fields of an even and odd supermode of a symmetrical twin-core fiber is presented in Fig. 1. The dashed circles represent the two cores. The solid curves indicate the supermode profiles on the central axis connecting the cores. Light launched into one core excites both supermodes equally. During propagation, there is no power transfer between the supermodes. However, because the two supermodes have different propagation constants, they go in and out of phase during propagation and interfere with a period equal to their beat length, so that the optical power transfers between the cores; see Fig. 2 (matched directional coupling). A return to the original interference pattern occurs every time the accumulated difference in phase between the supermodes is a multiple of 2. The essential physical difference between the grating assisted coupler [4] and the dispersion inverted interference coupler is the length and strength of the grating; the essential functional difference is that one works in reflection and one in transmission. To understand why they work so differently, it is necessary to remember that each supermode reflects light at a Bragg wavelength ( ), which is proportional to the effective index ( ) of the supermode (, where is the grating period). Hence, there are two different rejection bands, one for each supermode. The width of each rejection band is proportional to the refractive index modulation of the grating. The two quantities and give an intuitive relation between the center separation and the widths of the rejection bands. In the grating assisted coupler,. Thus the rejection bands of the supermodes essentially coincide. In contrast, in the dispersion inverted coupler,, so the supermodes have well-separated rejection bands. The grating assisted coupler can now be understood as an add drop multiplexer working in reflection. The reflected light from wavelengths within both rejection bands (coinciding region) couples out to the opposite core from where they entered [4]. The dispersion inverted interference coupler works in transmission. To clarify the dispersion and interference effects taking place, we show first the accumulated phase for the two supermodes over the length of the grating structure used in the experiments in Fig. 3 as a function of wavelength. The diagram shows the calculated phase accrued over the length of the device for each supermode. The solid line represents the phase of the fundamental mode and the dashed line the phase of the second order mode. is a constant. The results are based on estimations of the grating strength and effective index of the two supermodes (see Section III for grating and waveguide estimations). It shows the deviations from a straight line leading up to and away from each mode s rejection band, marked with gray boxes. In the wavelength region between the two rejection bands, both supermodes experience strong dispersion, but of opposite sign. However, the grating-induced dispersion on the shorter and longer wavelength sides of the two bandgaps is of the same sign. This difference is very important for the functionality of the device, as the only important factor is the accumulated difference in phase between the modes. This is highlighted in Fig. 4, which shows the difference in phase between the two supermodes; the rejection bands are marked with gray boxes. Most wavelengths propagate unperturbed through the grating structure, and for those wavelengths, the supermodes accrue a constant difference in phase (indicated by the dashed line at ). The only features, in practice, that deviate from the constant difference in phase between the supermodes are grating induced. Consequently, the grating parameters are adjusted so that at the end of the grating, the accumulated
ÅSLUND et al.: ADD DROP MULTIPLEXING 1587 Fig. 4. Difference in phase at the end of the grating structure for two supermodes as a function of wavelength. Gray boxes mark the position of each supermode s rejection band. Fig. 6. A cross-section of the twin core fiber used in the experiment. Fig. 5. Grating-induced group delay of device as a function of wavelength. Gray boxes mark the position of each supermode s rejection band. grating-induced difference in phase in the wavelength region between the supermodes equals. This grating-induced phase shift will invert the interference pattern, and light will couple out to the opposite core. In other words, wavelengths between the two rejection bands experience one more intensity maximum over the length of the grating (see Fig. 2). In an earlier paper [6], we theoretically designed an add drop multiplexer within a two-moded planar waveguide based on this principle. In this paper, we verify experimentally the predicted effect, and also show that the supermodes of a twin core structure can be used in this fashion. In Fig. 5, the group delay due to the presence of the grating is shown. Gray boxes mark the position of the rejection band of each supermode. In the middle of the drop band, there is zero group delay, but as the grating is unapodized there is some ripple present. III. EXPERIMENT All experiments were carried out in a twin-core fiber instead of in directional couplers, as it was readily available (estimated data: core center separation 13.3 m, core diameter 6.8 m, cladding refractive index 1.443, and core refractive index 1.449). A photograph of the twin-core fiber cross-section using transmitted white light can be seen in Fig. 6. The gratings were directly written through a phase mask ( nm, mm) by translating an ultraviolet (UV) beam (frequency-doubled continuous-wave Ar 0.2 1.0 kj/cm, nm) along the fiber. The fiber was hydrogen loaded (170 atm, 80 C, 72 h) before grating writing proceeded. The measurement setup is shown in Fig. 7. It allows free choice of supermode excitation/collection, polarization, and relative phase difference between the modes at the end of the twin-core fiber. The choice of mode excitation is provided by butt coupling to and from the twin cored fiber with standard si1ngle-mode fiber mounted on three axis positioning stages (labeled A and E in the figure). The choice of polarization is provided by an in-line polarizer before the input butt-coupling takes place. The control of the relative phase difference between the modes is determined by cutting back a section of the twin-core fiber without a grating in it for rough adjustment and stretching the remaining section of fiber between stages B and C without a grating with a micrometer precision controlled stage (labeled B in the figure) for fine tuning. The near field of the output of the twin-core fiber can be monitored before grating writing commences, to ensure that the UV light illuminates the cores uniformly so that no core shields the other. The near field is then imaged onto a vidicon camera by replacing the fiber on stage A with a lens. Though great care was taken, it cannot be guaranteed that the fiber was free from slight twisting due to limitations with the fiber holders and the elliptical shape of the fiber. The transmission properties were measured using the amplified spontaneous emission from an erbium-doped fiber amplifier as source and collecting the transmitted power with an optical spectrum analyzer. A weak grating was first written, thus minimizing dispersion effects, to determine the separation of the rejection band. The wavelength separation was found to be 0.23 nm, which corresponds to a and a beat length in the fiber of mm. The sought dispersion effect is proportional to the product of the grating length and strength. This gives some freedom in the choice of parameters. For spectral clarity, we chose to write the gratings as long as possible ( mm) to allow the rejection bands to be as narrow as possible. A reliable measure of the actual rejection depth as a measure of the grating strength for each supermode could not be measured due to interference effects between the supermodes. After three attempts with varying writing speeds, we found the correct grating strength. The fiber was cleaved just before the start
1588 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 8, AUGUST 2002 Fig. 7. Measurement setup. A and E are xyz-positioning stages, B is a z-positioning stage, and C and D are fixed fiber mounts. Fig. 8. Spectral transmissivity for the drop (solid line) and the throughput port (dashed line). of the grating on one side, and 100 mm beyond the grating on the other side, to allow cutback and stretching for phase adjustment. The fiber was subsequently cleaved to a length approximately 1 mm short of an interference peak, which was found by monitoring the near field. This allowed the phase difference between the modes to be fully adjusted by stretching to give an interference maximum. IV. RESULTS AND DISCUSSION The typical transmission results for gratings with near optimum strength are displayed in Fig. 8. The dashed line represents the transmissivity of the throughput port, and the solid line represents the transmissivity of the drop/add port. The two gray boxes mark the locations for each of the supermode rejection bands. The maximum rejection of the throughput port is 15 db, and the maximum transmissivity of the drop/add port is 18 db. These results should be compared with the optimal theoretical results in Fig. 9, where the dashed line represents the transmissivity of the throughput port and the solid line represents the transmissivity of the drop/add port. Fig. 9 was constructed as follows. As discussed earlier, we took the experimentally measured Bragg wavelengths to find the effective indexes of the supermodes. Using the length of the grating, we then used our theoretical model [6] to identify the value of leading to the highest extinction ratio, which was found to be. The resulting spectral response is shown in Fig. 9. Note that the rejection ratio in this figure is 100%, which is higher than in [6]. This is possible as the exit modes are equally excited by the supermodes, which was not the case in [6]. Further, the previous statement for dispersion inverted interference devices is verified as. Fig. 9. Theoretical spectral transmissivity for the add/drop (solid line) and the throughput port (dashed line). The essential features of the experimental response of the device follow the theoretical predictions very closely, and the difference is attributed to poor fiber geometry. There was evidence for poor fiber uniformity, as interference effects with multiple periodicities were taking place over the whole spectrum. This suggested that the two cores had different propagation constants and/or possibly supported another set of supermodes. The central dip in the drop channel of Fig. 8 is probably from reflecting cross-coupling from one supermode to the other due to asymmetrical grating writing. It can be removed in principle by writing a transversally uniform grating. The response of the device is in principle fully symmetric, so the drop or add ports should be interchangeable, and indeed in practice we found this to be qualitatively true. Subsequent aging of the theoretical device will decrease the rejection ratio by reducing the phase difference. When the grating strength of the theoretical device has dropped by 2%, the rejection level of the throughput port has reduced to 20 db from 100%, and the rejection bandwidth at 10 db rejection level is reduced by 2.5%. To simplify device fabrication, the cores should be closer together, thus shortening the grating length restrictions. In the end, the twin-core region should be made by fusing two photosensitive fibers together to form a purpose-made twin-core directional coupler of suitable length. Fine-tuning of the relative phase between the modes could then be accomplished by UV-tuning of the short twin-core region leading up to or away from the grating section. The known effect of slight mismatch between cores of an directional coupler is not a serious impediment to applications in the telecom industry; and that effect on our device is equivalent and should therefore not be a serious constraint.
ÅSLUND et al.: ADD DROP MULTIPLEXING 1589 The main practical limitation of the device is that the spectral response is confined to the region between the two rejection bands, which is given by the difference in effective index of the supermodes. However, the performance of the device is not affected by the detailed shape of the spectral response of the grating. Therefore, the usual design tools can be brought to bear to improve the filtering characteristics of the device. V. CONCLUSIONS We have presented experimental results showing a new way to manufacture an all-fiber optical add drop multiplexer. The device is located inside the twin-core region of a fiber and relies on Bragg grating dispersion inverted interference in the forward direction for its functionality. Initial results show a 15-dB spectral efficiency. REFERENCES [1] G. E. Keiser, A review of WDM technology and applications, Opt. Fib. Technol., vol. 5, no. 1, pp. 3 39, Jan. 1999. [2] F. Bilodeau, D. C. Johnson, S. Theriault, B. Malo, J. Albert, and K. O. Hill, An all-fiber dense-wavelength-divison multiplexer demultiplexer using photoimprinted Bragg gratings, Photon. Technol. Lett., vol. 7, pp. 388 390, Apr. 1995. [3] J. L. Archambault, P. St. J. Russell, S. Barcelos, P. Hua, and L. Reekie, Grating-frustrated coupler: a novel channel-dropping filter in single-mode optical fiber, Opt. Lett., vol. 19, no. 3, pp. 180 182, Feb. 1994. [4] I. Baumann, J. Seifert, W. Novak, and M. Sauer, Compact all-fiber add drop multiplexer using fiber Bragg gratings, Photon. Technol. Lett., vol. 8, pp. 1331 1333, Oct. 1996. [5] R. Kashyap, G. D. Maxwell, and B. J. Ainslie, Laser-trimmed 4-port bandpass filter fabricated in single-mode photosensitive Ge-doped planar wave-guide, Photon. Technol. Lett., vol. 5, pp. 191 194, Feb. 1993. [6] M. Åslund, L. Poladian, J. Canning, and C. M. de Sterke, Add drop multiplexing by grating-induced dispersion in multimode interference device, Photon. Technol. Lett., vol. 13, pp. 969 971, Sept. 2001. [7] L. Dong, P. Hua, T. A. Birks, L. Reekie, and P. S. Russell, Novel add/drop filters for wavelength-division-multiplexing optical fiber systems using a Bragg grating assisted mismatched coupler, Photon. Technol. Lett., vol. 8, pp. 1656 1658, Dec. 1996. [8] A.-C. Jacob-Poulin, R. Valle, S. LaRochelle, D. Faucher, and G. R. Atkins, Channel-dropping filter based on a grating-frustrated two-core fiber, J. Lightwave Technol., vol. 18, pp. 715 720, May 2000. [9] A. W. Snyder and J. D. Love, Optical Waveguide Theory. London, U.K.: Chapman and Hall, 1983. Mattias Åslund, photograph and biography not available at the time of publication. Leon Poladian, photograph and biography not available at the time of publication. John Canning, photograph and biography not available at the time of publication. C. Martijn de Sterke, photograph and biography not available at the time of publication.