1. Introduction

The Shannon capacity limit of single-mode fiber (SMF) systems in the presence of fiber nonlinearity has been extensively studied in the last decade [1–4]. This capacity has been quickly approached by the recent demonstration within some practical engineering margin [5]. It is impossible to enjoy the same dramatic capacity improvement in the future as in the past two decades if we continue to stay with the SMF platform. Therefore, the spatial-division multiplexing (SDM) has been explored recently to overcome the capacity barrier [6–13]. There are a few techniques available at the moment for SDM, including multi-core fiber (MCF) [6–8] and few-mode fiber (FMF) [9–13]. During the last few years, there has been impressive progress in the research of MCF and FMF transmission. However, there have also been some intriguing questions regarding comparison between MCF and FMF systems.

It has been shown that FMF has the advantage of power efficiency over MCF in terms of optical amplification [14]. So far, FMF compatible optical amplifiers based on Raman effect [15] and few-mode EDFA [16] have been reported. By using well-controlled pump modes, mode dependent gain variations can be minimized, opening the ways for long-haul FMF transmission. Similar to the trends in single mode fiber (SMF)-based optical networks, the capacity scaling in FMF systems will be achieved by wavelength-division multiplexing (WDM). In such systems, reconfigurable optical add-drop multiplexers (ROADM) that support all propagation modes will be a key element for realizing flexible networking. Although point-to-point WDM transmission over FMF has been demonstrated, the study of FMF-compatible ROADM, to the best knowledge of ours, has not been reported.

This paper touches upon an important perspective of different implementations in OADM for FMF- and MCF-based systems. In MCF-based systems, the optical signals in different cores are transmitted independently. Therefore, the system complexity, more specifically the number of OADMs, are multiplied by the number of cores. However, in FMF-based systems, there exists a great opportunity to simplify the system design. Figure 1 shows the future high-capacity few-mode fiber networks where the wavelength channels containing all the modes are routed as one entity which is subsequently optically added/dropped, and amplified without mode-demultiplexing/multiplexing during transmission. We call this wavelength channel containing few-mode tributaries the mode-division multiplexed (MDM) superchannel. An analogy can be drawn between such a FMF-based system and polarization-multiplexed (PM) system where the two polarizations are not separately processed during the transmission, which greatly simplifies the system design while doubling the system capacity. In this paper, we show for the first time the feasibility of FMF-compatible OADM, which is an important first step towards FMF-compatible ROADMs. We demonstrate successful reception of CO-OFDM signal passing through a dual-LP₁₁ mode FMF-based OADM. We found that the power penalties for add, drop and through ports are 2.6, 2.4, 0.7 dB, respectively for 3x318 Gb/s OFDM signals.

 

Fig.1 Illustration of few-mode compatible ROADM in few-mode fiber based optical networks. MDM: Mode division multiplexing/multiplexed, Mux: Multiplexer, DeMux: De-Multiplexer, FMF: Few-mode fiber.

Download Full Size | PDF

2. OADM architecture

The architecture of the FMF compatible OADM can be similar to their SMF counterparts. We consider that mode dependence of beam divergence angle in a FMF to be small for a practical implementation of OADM with free-space components. This argument is true since the beam divergence $\Theta$can be calculated as $\Theta =D_m/f$, where D_m is the mode field diameter (MFD) and f is the focal length of collimators. We use collimators of 11-mm focal length, and the MFDs are 11.0 µm and 11.3 µm for LP₀₁ mode and LP₁₁ mode respectively. The above parameters give less than 0.06 degree divergence, which makes the transmission characteristics insensitive to different modes. Such free-space optics based OADMs can be miniaturized and the number of input/output ports can be scaled up by using well-established technologies such as MEMS and LCOS [17]. In this work, we design and build a two-mode compatible free-space OADM using collimators and a pair of thin film filters (TFF). Figure 2 shows the OADM architecture. The TFFs act as a band-pass filter in transmission mode and as a notch-filter in the reflection mode. We use double reflection in order to sufficiently suppress the drop channel in output, which would otherwise serve as in-band crosstalk for the add channel. We tilt the TFFs by 5° in order to achieve lateral separation of the in/add and drop/through ports in the double reflection configuration. The 5° tilt launch also prevents any parasitic reflection from the drop or add ports due to accumuled multiple reflections. The free-space beam diameter is 2 mm.

 

Fig. 2 Schematic diagram of the two-mode OADM.

Download Full Size | PDF

In order to characterize the few-mode compatible OADM, we conduct measurement for the OADM using both LP₁₁ modes and LP₀₁ mode. First, we attach two-mode fiber (TMF) to the four ports (in, add, through, and drop) and use LP₁₁ mode to measure the transmission characteristics. The LP₁₁ mode (LP_11a or LP_11b) is generated by a conventional SMF transmitter, a long-period fiber grating (LPFG)-based mode converter (MC) [10,18], and a mode stripper (MS). The generated LP₁₁ mode is fed through a collimator lens, a prism, the other collimator lens, and then coupled to a TMF fiber which is connected to an OADM. At the through and the drop ports of the OADM, the TMF fiber is directly attached to a multimode OSA to measure the spectrum. The transmission characteristics of the two LP₁₁ modes measured by the OSA is shown in Fig. 3(a) . It can be seen from in-to-out transmission characteristics for both LP_11a and LP_11b that, the dropped channel (at the center of the spectrum) is 25 dB below the through channel over a bandwidth of 0.5 nm. This ensures that the added signal will not be affected by in-band crosstalk. The loss for add, drop, and through channel are about 2.2, 2.1, 2.5 dB under the best operation condition, respectively. The add/drop transmission characteristics resembles a band-pass filter but with some ripples, which is caused by the beating between LP₁₁ and residual LP₀₁ modes. The residual LP₀₁ can be attributed to many sources, such as imperfection of the mode converter, mode combiner, and even OADM. The transmission characteristics for LP₀₁ mode of the OADM is also measured by removing the MC and MS and replacing TMF with SMF at each port. The LP₀₁ mode is presented in Fig. 3(b), showing much smooth transmission characteristics due to immunity from LP₀₁ and LP₁₁ beating. This point to the difficulty of characterizing FMF based devices such as OADM. In other words, two-port transmission characteristics should be represented by a 4x4 input to output matrix, with both inputs and outputs could be a choice from any two polarizations of any two LP₁₁ modes. This poses a huge challenge to fully measure the FMF compatible devices in terms of measurement complexity and a need for high-precision mode multiplexing and de-multiplexing devices.

 

Fig. 3 (a) Measured transmission characteristics of the OADM based on TMF. (b) Measured transmission characteristics of the OADM based on SMF. Th: Through.

Download Full Size | PDF

To identify the polarization dependence, we insert SMF based polarization controller before MC and MS. By rotating the polarization state, we find there is no observable change on the OSA. From this we conclude that the polarization dependence of this OADM is insignificant (< 0.2 dB). The reason for such negligible polarization dependence is that we utilize thin-film filter with an angle of incident (AOI) as small as 5 degree.

3. Experimental setup

We use 3 channels to measure add, drop, and through functionalities of our custom-designed OADM, with the center channel to be dropped and added, and two outer channels to pass through. Each channel contains a multiband CO-OFDM signal consisting of 9 densely-spaced bands. Figure 4 shows the experimental setup used. The optical mode coupling from SMF to FMF and vice versa is achieved using two pairs of MS and MC. Free space mode combining and splitting is achieved with collimators and prism beam splitters (BS). There is an adjusting key assembled to each collimation port, to enable convenient adjustment of the orthogonality of the two LP₁₁ patterns. Each channel is created by combining 3 external cavity lasers (ECLs), each carrying 3 tones spaced at 6.5185 GHz. Thus there are 9 tones in total for each channel. The frequency guard band between different lasers is 500 MHz. The 27 tones from 3 channels are combined and fed into an optical IQ modulator, which is driven by an OFDM signal. Then polarization multiplexing is emulated by splitting the CO-OFDM signal from IQ modulator into two branches that are delayed with each other by 1 OFDM symbol (500 ns), and recombined with a polarization beam splitter (PBS). We then emulate the signals for two orthogonal LP₁₁ modes by splitting the signal and delaying one branch by two OFDM symbols (1 µs delay). These two signals are mode converted into LP₁₁ mode using LPFG-based MCs, combined and launched using free-space optics into a 4 meter TMF fiber. The signal is then inserted into the input port of the OADM module. All the amplification is achieved by SMF EDFAs (namely, the amplifiers are placed either before LP₀₁-to-LP₁₁ conversion or after LP₁₁-to-LP₀₁ conversion). Since the TMF is linear in the demonstration, TMF channel effect is insignificant in this work which is focused on the OADM impact.

 

Fig. 4 Experimental setup for TMF compatible OADM performance measurement. AWG: Arbitrary waveform generator, PBS: Polarization beam splitter, MS/C: Mode stripper/converter, BS: Beam Splitter, BPF: Band-pass filter, PD: Photodiode, ADC: Analog-to-digital converter, LD: Laser Diode.

Download Full Size | PDF

The OFDM parameters are: OFDM symbol length of 2560 points; cyclic prefix of 452 points; 4 OFDM training symbols are employed to represent alternate launch of 4 combinations of polarizations and modes, which is used for 4x4 channel matrix by means of intra-symbol averaging [19]; unique word length of 32 used at each end of OFDM symbol to seed phase estimation for data symbols [20]. The phase estimation of training symbols is done by using RF pilot subcarrier [20,21]. The data symbols use DFT-spread (DFT-S) signal, and the phase estimation is done using decision feedback seeded by using unique word [20]. The three CO-OFDM channels each occupied 59.67 GHz bandwidth, carrying a net data rate of 318 Gb/s. All the BERs are calculated using 998,400 bits. Signal processing uses 4x4 MIMO OFDM procedure [18].

4. Experimental results and discussion

Figure 5 illustrates the optical spectra of transmitted superchannels spaced at 100 GHz and the MDM superchannels after OADM. From the spectra of the dropped channel and the through channel without add, it can be seen that the center channel have been effectively eliminated. The neighbor channel isolation is more than 25 dB determined from the spectrum at the drop port.

 

Fig. 5 Measured spectra for transmitted signal, OADM through channels (w/ and w/o add channel), and the drop channel. (Curves have been shifted along the y-axis for comparison).

Download Full Size | PDF

Figure 6(a) shows the Q factors of all the functionalities of OADM. The OSNRs are measured for the entire channel at 318 Gb/s. The Q factor (Q² factor) is calculated from BER. It can be seen that the penalty at a 7% FEC limit (Q = 8.53 dB) for add, drop and through ports are 2.6, 2.4, 0.7 dB, respectively (through is averaged between left and right channels). Figure 6(b) presents the performance of all the bands for add, drop, through channels measured at an OSNR of 22.8 dB. It can be seen all the bands are below 7% FEC threshold of 8.53 dB. The inset shows the constellation for drop channel at an OSNR of 22.8 dB. We attribute the relatively large penalty to the beating between LP₀₁ and LP₁₁ modes.

 

Fig. 6 (a) Measured Q factor versus OSNR for add, drop and through channels at 318 Gb/s. Inset shows constellations for the add port signals at OSNR of 22.8 dB. (b) Q factor for all bands after OADM at an OSNR of 22.8 dB.

Download Full Size | PDF

Our system is limited to 4x4 MIMO and does not capture the LP₀₁ mode due to the following two reasons. First, the free-space based mode couplers (shown as BS in Fig. 4) seriously degrade the performance by inducing low modal extinction ratio and large mode-dependent loss. Second, due to the extreme large differential-mode-delay (DMD) between the LP₀₁ and LP₁₁ for our TMF, the transmission demonstration is not considered in this paper. However, in the 4x4 MIMO configuration, any coupling between LP₀₁ and LP₁₁ modes in the OADM will be detrimental to the performance. This explains relatively large penalty shown in OSNR sensitivity in Fig. 6(a). Nevertheless, low DMD FMF fiber can be used aided by 6x6 MIMO signal processing [12] to compensate for this effect. Therefore we expect the OADM performance for 6x6 MIMO will be robust to inter mode coupling between LP₀₁ and LP₁₁ mode.

Our work also reveals tremendous challenges ahead in characterizing FMF based components. There are a few critical elements needs to be developed to fully characterize the FMF based components. For instance, pure higher-order mode sources, high extinction-ratio mode-multiplexers and de-multiplexers are essential to identify the inter-modal interference of the few-mode compatible components under test. The first-order estimate is that the mode extinction ratio for those characterization devices should be better than 20 dB.

5. Conclusion

We have demonstrated for the first time an OADM that supports two orthogonal LP₁₁ modes of a FMF. We perform add, drop and through functionalities for 3x318 Gb/s OFDM signals, and find that the OSNR penalties for add, drop and through ports are 2.6, 2.4, 0.7 dB, respectively.