1. Introduction

To achieve higher network resource utilization, the architecture of optical networking is evolving gradually from ring to mesh. As the core subsystem in an optical network, the reconfigurable optical add/drop multiplexer (ROADM) has also developed from a simple two-dimensional structure to complex multi-dimensional (> 2) structures. The capacity of ROADMs has grown increasingly larger because of the growth in the number of wavelength channels in each transmission fiber. Meanwhile, to efficiently support the character of dynamic and time-varying internet business, the optical network layer is required to control the optical channels dynamically, which means that greater flexibility and controllability are necessary at the optical switching node. The architecture of next-generation optical switching nodes must also meet requirements of being ultra-high capacity, colorless, directionless, and contentionless [1]. To further improve optical spectrum efficiency and support the ultra-broadband optical channel, the optical switching node should satisfy gridless spectrum operation, which is regarded as another important feature. Roorda and Collings [2] summarized multiple types of architectures of colorless directionless optical nodes in 2008. J. Wagener et al. compared two kinds of optical nodes based on the large MEMS matrix optical switch and the technology of the wavelength selective switch (WSS), respectively [3]. The results indicated that the latter has advantages of higher reliability, lower price, and supporting multicast function. In 2010, based on analysis of colorless, directionless optical node architecture with obstruction, AT&T found that the business blocking rate was more serious in a dynamic optical network at a such node [4]. Polatis Inc. implemented a kind of colorless, directionless, and contentionless optical switch node by embedding a large M × N MEMS matrix optical switch in 2011 [5]. In addition, the researchers improved the performance of multidimensional colorless optical switch node by designing a 1 × 43 WSS [6]. A gridless spectrum operation mode was favored by the operator, Verizon, and the optical network equipment company, Ciena [7].

However, with the development of optical communication systems into higher dimensional net structures, the traditional ROADM technology based on 1 × N WSS encountered its bottleneck [8–12] and could not meet the needs of system development. Wavelength conflicts occur when business of the same wavelength from different directions, at the same time, cannot be switched up and down at a single ROADM node. Currently, manual intervention is needed, but poses an inconvenience to the system application and increases labor cost. To address this problem, the development of M × N WSS has attracted extensive attention from global researchers. M × N WSS can effectively solve the problem of wavelength conflict and implement in ROADM node as a core device. Different input and output ports correspond to different regions of a liquid crystal on silicon (LCOS) device in an M × N WSS, so we can use M × N WSS to route different channels of the same wavelength from different input ports to different output ports. However, in a 1 × N WSS, the input and output ports correspond to the same regions of LCOS, so the device cannot offer non-blocking switching. Some previous demonstrations of M × N WSS have been reported [13–21]. In [13–15], the researchers employ couplers, optical switches, and 1 × N WSSs to construct an M × N WSS. The discrete components led to both higher system complexity and worse performance stability. The idea of using spatial light paths to design an M × N WSS was proposed in [16–18], while these works represent the trend of future development, the high values of insertion loss (IL) are a serious restriction. In [16] the proposed 5 × 5 WSS had a best IL of 13 dB, and a worst of 20 dB; in [17], the achieved IL for a 2 × 3 WSS was 15 dB; in [18], the proposed 3 × 3 WSS could not switch the input same wavelength to random output port independently; in [19–21], the proposed 8 × 8 wavelength selective cross connect (WSXC) had a best IL of 12dB, and a worst of 20dB.

In this paper, we propose a LCOS-based 3 × 4 tunable bandwidth WSS (TBWSS) using compact spatial light paths. In order to obtain high coupling efficiency between the input and output ports, cemented cylindrical lenses are used to eliminate aberration of optical path. The designed functional prototype has a best IL of 8.4 dB, and a worst value of 12.5 dB, over the full C-band. The function of tunable attenuation was added and a tuning range up to 35 dB was achieved. The range of the bandwidth was set from 25 GHz to 5 THz, with the adjusting step less than 7 GHz. Moreover, when increasing the number of collimators on the basis of existing optical paths, the number of input and output ports can be extended to 10 and 20.

2. The 3x4 TBWSS design

Figure 1 shows the schematic diagram of optical path through the wavelength direction (y-axis) and the switching direction (x-axis), and the components of the designed 3 × 4 TBWSS are displayed in Table 1 (also see Fig. 2). In this paper, the positions and focal lengths of the used lenses are theoretically calculated based on Gauss beam coupling equation. Then they are further optimized by using the optical design software of Zemax. This operation ensures the Gaussian beams of the input optical fiber collimator couples with the output optical fiber collimator very well after passing through the optical path, which makes sure better IL performance. Above all, all the optical elements are solidified on optical substrate compactly, and no need of any mechanical adjustment is required. In Fig. 1(a), the input light is firstly passed through an input collimator array (Component 1) and separated by polarization conversion unit (Component 2) to achieve the optimal polarization for both diffraction gratings (Components 5, 19) and LCOS chips (Components 10, 14). Then the input light beam is expanded by using two cylindrical lenses along the y-axis, with focal lengths of 4 mm and 60 mm, respectively. A diffraction grating (Component 5) is employed to separate the different wavelengths of the expanded beam into different angles. The wide-spectrum input signal is focused onto the LCOS chips (Components 10, 14) along the y-axis using a cylindrical lens with focal length 120 mm (Component 8). The LCOS chips are located on the front and back focal planes of a 4-f system composed of two cylindrical lenses on the y-axis, with focal length of 100 mm (Components 11, 13). A cylindrical lens (Component 16) on the y-axis with focal length 120 mm is used to focus the light from one LCOS chip (Component 14) at different wavelengths and different pixels. A diffraction grating (Component 19) is adopted to transfer the different angles of incident light into the same emergent angle at the same position. Two cylindrical lenses (Components 20, 21), with focal lengths of 60 mm and 4 mm, respectively, are used to compress the light beam along the y-axis. After passing through the polarization conversion unit (Component 22) and output collimator array (Component 23), the output light beam is finally obtained across the y-axis.

 

Fig. 1 Schematic diagrams of optical path, (a) through the wavelength direction (y-axis), (b) through the switching direction (x-axis).

Download Full Size | PDF

Table 1. Components of the Designed 3 × 4 TBWSS

View Table

 

Fig. 2 Photograph of the inner structure of the proposed 3 × 4 TBWSS.

Download Full Size | PDF

The schematic diagram of optical path through the switching direction (x-axis) is displayed in Fig. 1(b). Because the optical path of the experimental setup is folded, the input parts (Components 1 to 9), the switching parts (Components 10 to 14) and the output parts (Components 15 to 23) should be drawn on a central axis according to rule of projection at the switching direction. But such drawing leads to a lot of optical elements overlapped, which could not explain the principle of Gauss transform method at the switching direction clearly. Thereby, the three parts drawing in Fig. 1(b) present a certain included angle. Component 1 is an input collimator array. The light beam from the three input ports along the x-axis direction is separated by polarization conversion unit (Component 2) to achieve the optimal polarization for both diffraction gratings (Components 5, 19) and LCOS chips (Components 10, 14). The two cylindrical lenses on the x-axis (Components 6, 9), with focal lengths of 55 mm and 110 mm, respectively, are used to focus the light onto the LCOS chip (Component 10) at three different regions. The LCOS chips are located on the front and back focal plane of the cylindrical lens on the x-axis, with focal length of 200 mm (Component 12). By controlling the phase distribution in the three areas of the LCOS chip (Component 10) on the x-axis, the light beams of the three input ports can be controlled independently and switched to four different areas of the LCOS chip (Component 14) on the x-axis. These four areas correspond to four output ports at the output collimator array (Component 23). The phase distribution of the four different areas are independently controlled by the LCOS chip (Component 14) on the x-axis. The light beams at the four areas are passed through two cylindrical lenses on the x-axis (Components 15, 18), with focal lengths of 85 mm, and finally reach the four output ports at the output collimator array (Component 23).

We note that the beam waist at the input collimator array (Component 1) was 0.22 mm. After passing through the two cylindrical lenses on the x-axis (Component 6, 9) with focal lengths of 55 mm and 110 mm, respectively, the beam waist on the LCOS chip (Component 10) along the x-axis was 0.22 × 110/55 = 0.44 mm. When going through the cylindrical lens on the x-axis with focal length of 200 mm (Component 12), the beam waist on LCOS chip (Component 14) was 1.55 × 10⁻³/($\pi \times 0.44$) × 200 = 0.22 mm. The size of the LCOS chips (Components 10, 14) along the x-axis is 1920 × 8 × 10⁻³ = 15.36 mm. The size of complete beam is three times the beam waist. The size of the complete beam for the input collimator array (Component 1) on the LCOS chip (Component 10) on the x-axis was 0.44 mm × 3 = 1.32 mm. When the complete beam interval on the LCOS chip (Component 10) on the x-axis is 1.5 mm, the complete beam interference between different areas can be completely abolished. Thus, the corresponding number of collimators for the input collimator array (Component 1) are 15.36/1.5 = 10.24 > 10. The size of the complete beam for the output collimator array (Component 23) on the LCOS chip (Component 14) on the x-axis is 0.22 mm × 3 = 0.66 mm. When the complete beam interval on the LCOS chip (Component 14) on the x-axis is 0.75 mm, the complete beam interference between different areas can be completely abolished. Thus, the corresponding number of collimators for the output collimator array (Component 23) are 15.36/0.75 = 20.48 > 20. Thereby, based on the proposed compact spatial light paths, it is convenient to expand the number of input and output ports to 10 and 20, respectively. The compact spatial light paths in this paper are defined that all optical elements are solidified on optical substrate compactly without any mechanical adjustment. Compared with the previous works in [16] and [17], the proposed method can achieve the optimal optical structure and reduce the size of light path using reflecting mirror. Due to the large size and irregular shape of mechanical adjusting bracket, enough space should be reserved, which will cause large distances and included angles between optical elements. Eventually, the mechanical adjustment method in [16] and [17] leads to the deterioration of the IL.

Figure 3 shows the schematic diagram of distribution for LCOS chips and switching lens. Three different areas on input LCOS chip (Component 10) and four different areas on output LCOS chip (Component 14) are located on the front focal and back focal of switching lens (Component 12) as shown in Fig. 3. The input optical signal of three different areas is deflected to four different output areas on LCOS chip (Component 14) using input LCOS chip (Component 10). The included angle between the optical signal on the four areas and output optical fiber collimator is adjusted to zero by using the output LCOS chip (Component 14). Thus, the input optical signal is switched to output port. The deflection angles of input LCOS chip (Component 10) and output LCOS chip (Component 14) are $\arctan (d_1/f)$ and $\arctan (d_2/f)$. Where $d_1$ is the distance between the target output port on LCOS area and center of switching lens (Component 12). $d_2$ is the distance between the target input port on LCOS area and center of switching lens (Component 12). $f$is the focal length of switching lens (Component 12). In our proposed 3 × 4 TBWSS, the input port2 and output port3 are located on the center of switching lens (Component 12). The maximum deflection angle is achieved when the optical light is switched from input port1 to output port1 or from input port3 to output port1. In this paper, the values of $d_1$, $d_2$and $f$ are 3mm, 3mm and 200mm, respectively. Thus we can obtain the maximum deflection angle is 0.86° according to the above analysis. When the designed 3 × 4 TBWSS is extended to 10 × 20 TBWSS, the maximum deflection angle is changed from 0.86° to 2°. Based on the measured IL on LCOS chip, extra maximum 0.7dB IL is introduced because of deflection angle variation. Thus, 1.4dB IL is introduced due to light deflection of input LCOS chip (Component 10) and output LCOS chip (Component 14). Meanwhile the larger light beam deflection angle will cause larger field angle, which will worsen the aberration and reduce the Gauss beam coupling efficiency. Due to the advantage of using cemented cylindrical lens and Gaussian transform light path in this paper, we solve the issue of larger aberration caused by larger field angle very well. The extra IL introduced by aberration is expected to be 0.5dB. Thus, the total extra increased IL for 10 × 20 TBWSS is 1.4 + 0.5 = 1.9dB compare with the designed 3 × 4 TBWSS.

 

Fig. 3 Schematic diagram of distribution for LCOS chips and switching lens.

Download Full Size | PDF

In this paper, we propose a new method based on cemented cylindrical lens and Gaussian transform light path to reduce IL. Compared with the previous works [19–21], the advantages are as follows:

3. Experimental results

3.1 Port isolation test

We first measured the values of IL as a function of wavelength along the entire C-band. We use an ASE source as a transmitter and an optical spectrum analyzer (OSA) AQ6370B as a receiver to measure the values of IL in this paper. The range of input optical signals was 1530–1570nm. When only port1 had the input optical signal, the measured results are shown in Fig. 4(a). It can be seen in Fig. 4(a) that the measured values of IL for output port1, port2, port3 and port4 were 8.4 dB at 1545.07 nm, 8.8 dB at 1549.4 nm, 10.4 dB at 1547.7 nm, and 10.6 dB at 1549.5 nm, respectively. The measured isolation between the signal output port and the other three output ports was > 30 dB.

 

Fig. 4 Measured values of IL as a function of wavelength from 1530 nm to 1570 nm.

Download Full Size | PDF

The same phenomena were observed when inputting optical signals into input port2 and port3, as shown in Fig. 4(b) and (c). In Fig. 4(b), the measured values of IL for output port1, port2, port3 and port4 were 8.7 dB at 1530.1 nm, 8.8 dB at 1546.6 nm, 10.3 dB at 1547.8 nm, and 10.6 dB at 1547.9 nm, respectively. While in Fig. 4(c), the measured values of IL for output port1, port2, port3 and port4 were 10.6 dB at 1530.3 nm, 9.2 dB at 1533.1nm, 10.8 dB at 1534.7 nm, and 12.5 dB at 1530.6 nm, respectively. Also, the measured isolation between the signal output port and the other three output ports were all > 30 dB, as illustrated in Fig. 4(b) and (c).

3.2 Function of tunable bandwidth and attenuation test

Figure 5(a) shows the measured spectrum of the minimum 3 dB bandwidth and the minimum value is 25 GHz. The tunable bandwidth step was 6.25 GHz and the tunable range was 25 GHz to 5 THz, as displayed in Fig. 5(b) and (c). We also added the function of tunable attenuation. The maximum value we achieved was 35 dB, and the measured spectra with different attenuation values are given in Fig. 5(d). According to the measured results in [19], the 0.5dB and 3dB bandwidths are 28GHz and 40.5GHz under the case of 50GHz channel spacing. The passband rolloff from 0.5dB to 10dB is 14.5GHz. Our measured results in Fig. 6 show that the 0.5dB and 3dB bandwidths are 22.5GHz and 42GHz under the case of 50GHz channel spacing. The passband rolloff from 0.5dB to 10dB is 16.3GHz, which is slightly worse than the results in [19].

 

Fig. 5 (a) The spectrum of minimum 3dB bandwidth; (b) Tunable bandwidth step; (c) Range of tunable bandwidth; (d) Function of tunable attenuation.

Download Full Size | PDF

 

Fig. 6 Switching function with wavelength range from 1530 nm to 1570 nm.

Download Full Size | PDF

3.3 Switching function of tunable bandwidth with one port input signal

The switching function of tunable bandwidth with only one port input signal was investigated. The wavelengths of the input signal occupy the bandwidth 1530–1570 nm, which can be distributed and switched to the four output ports. In Fig. 6(a), only port1 input optical signal, while port2 and port3 were idle. The central wavelength of 1532.87 nm with 3 dB bandwidth of 1.22 nm, central wavelength of 1546.11 nm with 3 dB bandwidth of 3.15 nm, and central wavelength of 1563.55 nm with 3 dB bandwidth of 4.48 nm were extracted and switched to output port1. The central wavelength of 1535.4 nm with 3 dB bandwidth of 1.28 nm and central wavelength of 1551.63 nm with 3 dB bandwidth of 4.62 nm were extracted and switched to output port2. The central wavelength of 1538.29 nm with 3 dB bandwidth of 2.2 nm and central wavelength of 1556.15 nm with 3 dB bandwidth of 1.3 nm were extracted and switched to output port3. The central wavelength of 1542.17 nm with 3 dB bandwidth of 2.25 nm and central wavelength of 1558.75 nm with 3 dB bandwidth of 1.8 nm were extracted and switched to output port4. The measured results are illustrated in Fig. 6(a) and significant performance is observed. When inputting optical signal into port2 and port3, similar performance was achieved, as shown in Fig. 6 (b) and (c).

We further added the function of output attenuation and bandwidth adjustment. The measured results are shown in Fig. 7. In Fig. 7(a), only port3 input optical signal, and port1 and port2 were idle. The central wavelength of 1532.97 nm with 3 dB bandwidth of 1.22 nm and 7.2 dB attenuation, central wavelength of 1546.11 nm with 3 dB bandwidth of 3.15 nm, and central wavelength of 1563.55 nm with 3 dB bandwidth of 4.48 nm were extracted and switched to output port1. The central wavelength of 1535.4 nm with 3 dB bandwidth of 1.28 nm and 7.1 dB attenuation, and central wavelength of 1551.63 nm with 3 dB bandwidth of 4.62 nm were extracted and switched to output port2. The central wavelength of 1538.29 nm with 3 dB bandwidth of 2.2 nm and 6 dB attenuation, and central wavelength of 1556.15 nm with 3 dB bandwidth of 1.3 nm were extracted and switched to output port3. The central wavelength of 1542.17 nm with 3 dB bandwidth of 2.25 nm and 5.9 dB attenuation, and central wavelength of 1558.75 nm with 3 dB bandwidth of 1.8 nm were extracted and switched to output port4.

 

Fig. 7 Switching function with wavelength range from 1530 nm to 1570 nm with output attenuation.

Download Full Size | PDF

The extracted wavelengths can be switched to any of the four output ports. In Fig. 7(b), the central wavelength of 1532.97 nm with 3 dB bandwidth of 1.22 nm was switched to output port1. The central wavelength of 1535.4 nm with 3 dB bandwidth of 1.28 nm, central wavelength of 1542.17 nm with 3 dB bandwidth of 2.25 nm, central wavelength of 1551.63 nm with 3 dB bandwidth of 4.62 nm were switched to output port2. The central wavelength of 1538.29 nm with 3 dB bandwidth of 2.2 nm, central wavelength of 1556.15 nm with 3 dB bandwidth of 1.3 nm, central wavelength of 1563.55 nm with 3 dB bandwidth of 4.48 nm were switched to output port3. The central wavelength of 1546.11 nm with 3 dB bandwidth of 3.15 nm and central wavelength of 1558.75 nm with 3 dB bandwidth of 1.8 nm were switched to output port4. It is noticed that the proposed 3 × 4 TBWSS with function of tunable bandwidth will be used in next generation colorless, directionless, contentionless and gridless ROADM. The output optical signal in Fig. 7 are switched from random input ports. The wavelength and bandwidth at each output port are also random, which verifies the gridless property.

3.4 Switching function of three input ports with three different wavelengths signal

Figure 8 shows the optical spectra of the three different wavelengths signal for the three input ports. The central wavelengths with 3 dB bandwidth of 0.65 nm for port1, port2 and port3 are 1561.43 nm, 1553.33 nm, and 1547.71 nm, respectively. We first switched the three input signals into one output port. The performance of three inputs with one output is illustrated in Fig. 9. We also investigated the performance of three input signals switching to three independent output ports. In Fig. 10(a), input port1, port2 and port3 were switched to the corresponding output port1, port2 and port3. In Fig. 10(b), the input port1, port2 and port3 signals were switched to output port4, port1 and port2, respectively. We further added the attenuation function; the measured results are displayed in Fig. 11. The input port1, port2 and port3 signal were switched to output port4 with 0.4 dB attenuation, port1 with 4.9 dB attenuation and port2 with 10 dB attenuation, respectively.

 

Fig. 8 Optical spectra of the three different wavelength signals for the three input ports.

Download Full Size | PDF

 

Fig. 9 Performance of three different wavelengths inputs, one output. Input port1, port2 and port3 signals were all switched to (a) output port1; (b) output port2; (c) output port3; and (d) output port4.

Download Full Size | PDF

 

Fig. 10 Performance of three different wavelengths inputs, three outputs. (a) Input port1, port2 and port3 signals were switched to the corresponding output port1, port2 and port3. (b) Input port1, port2 and port3 signals were switched to the corresponding output port4, port1 and port2.

Download Full Size | PDF

 

Fig. 11 Performance of three different wavelengths for input three output with attenuation.

Download Full Size | PDF

3.5 Switching function of three input ports with the same wavelength signal

Finally, we investigated the case that the signal for the three input ports are set to the same wavelength of 1553.33 nm with 3 dB bandwidth of 0.65 nm as shown in Fig. 12. The performance for the three input-three output with the same wavelength is displayed in Fig. 13. In Fig. 13(a), the input port1, port2 and port3 were switched to the corresponding output port1, port2 and port3. In Fig. 13(b), the input port1, port2 and port3 signal were switched to output port2, port4 and port1, respectively.

 

Fig. 12 Optical spectra of the same wavelength signal for the three input ports.

Download Full Size | PDF

 

Fig. 13 Performance of switching three input to three output ports with the same input wavelength. (a) Input port1, port2 and port3 signals were switched to the corresponding output port1, port2 and port3; (b) Input port1, port2 and port3 signals were switched to the corresponding output port2, port4 and port1.

Download Full Size | PDF

4. Conclusions

In this paper, we have proposed and demonstrated a LCOS-based 3 × 4 TBWSS using compact spatial light paths. The performance of IL, tunable bandwidth, tunable attenuation and switching function were experimentally investigated. The measured results indicate that the proposed and designed 3 × 4 TBWSS has significant potential to be implemented in future dynamic networks. Moreover, the proposed method can conveniently be expanded to a 10 × 20 TBWSS.