1. Introduction

Multi-core fiber (MCF) is widely utilized in space division multiplexing (SDM) technology to overcome the capacity crisis of current single-core fiber-based communication system [1]. However, inter-core crosstalk (IC-XT) is the essential difference between MCF and single-core fiber, which severely distorts signal and limits linear capacity increase from the use of each core as a complete individual spatial channel. Since all cores in MCF share the same cladding, random light coupling from other cores occurs at several random discrete phase matching points (PMPs) along the fiber, leading to the IC-XT. In order to maintain a higher core density for larger communication capacity and lower bending loss, reduced core pitch of MCF amplifies the crosstalk (XT) even in the weakly-coupled MCFs [2,3]. Some approaches in the fiber design, such as the trench-assisted cores [4], are proposed to deal with the IC-XT problem, but never get it to zero. Ultimately, IC-XT presents a fundamental adverse effect in the MCF-based transmission system and severely limits the signal-to-noise ratio (SNR) when its cumulative amount cannot be neglected. Moreover, the random variation of bit error ratio (BER) caused by IC-XT can result in the burst link outage, so online monitoring of IC-XT along the MCF is indispensable to realize a reliable transmission system [5,6].

Traditional measurement of XT mainly bases on the power meter (PM), related to the short-term average XT (STAXT) during a period [7]. Although it has the advantage in rapidity and accuracy for measurement, it cannot provide an XT online monitoring for an in-service communication link. This XT online measurement belongs to the area of optical performance monitoring (OPM), which need to continuously monitor various network performance parameters, such as optical power, optical signal to noise ratio (OSNR), chromatic dispersion (CD), polarization mode dispersion (PMD) and fiber nonlinearity [8,9]. However, since SDM is a relatively new approach for communication, there is no standard schemes on the OPM of IC-XT. Researchers have proposed the methods using pilot tones [10,11] or monitoring channels [12], which are all far away from the useful data in frequency domain, to monitor the IC-XT, but the monitoring parts cannot strictly reflect the accurate XT information due to the random characters of IC-XT in frequency domain [13]. A previous work in [6] has provided a good example for XT online monitoring using the pre-coded discrete multitone system, which employs some subcarriers as XT probes for dynamic instantaneous average XT measurement. However, it is a little complex for using precoding technology and unsuitable for time-domain modulation formats, such as on-off keying (OOK), making this method unable to be widely applied in current data-center interconnects or long-haul transmission systems. Other related works are based on neural networks [14] or deep learning [15], which needs extreme amount resource for storage and computation. Therefore, it is necessary to propose an efficient XT online monitoring method with relatively simple digital signal processing (DSP) and being appropriate for time-domain modulation formats at the same time.

In this paper, we propose an auxiliary management and control channel (AMCC) aided XT online monitoring of MCF. The technology of AMCC is from the area of wavelength division multiplexing passive optical networks (WDM-PON). It has been defined in the ITU-T G.989 series standards that WDM-PON requires AMCC for efficient network deployment, management and control [16–18]. Concretely, the technology uses a low-frequency or base-band signal, typically from 100 kbps to several Mbps, as the AMCC signal for management or control, while superposing it to the high-speed PON signal with a certain modulation depth at the same time. Due to the frequency-range disparity between AMCC signal and PON signal, they have the limited influence on each other. Previous studies related to AMCC have primarily focused on different kinds of AMCC superposition, feasibility and improvement of AMCC transmission performance [19–24]. Here, we mainly exploit the characters of AMCC to online monitor IC-XT in MCF, in which AMCC signal is specially used for IC-XT measurement with controllable influence on the in-service high-speed signal, that we call it PON signal in the following part but don’t restrict this scheme only for PON. The AMCC signals in transmitting terminal (TX) for different cores are selected as orthogonal sequences, while IC-XT is evaluated in the receiving terminal (RX) via cross-correlation (Xcorr). In order to detect the weak IC-XT, we collect a long-term signal to extract XT from noise via Xcorr. Since XT distributes on the frequency domain almost randomly when it is beyond the decorrelation bandwidth of IC-XT (typically 0.1-1 GHz) [13,25], the real-time IC-XT feature of AMCC signal has perfect accordance to the optical carrier. In the experimental verification, IC-XT of different wavelength from the useful signal, i.e. the heterodyne XT, is estimated and the results fit well with the measurement from PM, showing its effectiveness on the average short-term IC-XT measurement within the whole PON signal bandwidth. Then, we conduct the experimental tests for the IC-XT with the same wavelength as the useful signal in both C-band and L-band cases, i.e. the homodyne XT, more reasonably in the practical SDM transmission, as a result that the estimated XT perfectly matches with a single-frequency pilot tone close to the optical carrier. Overall, the estimation error is mostly less than 0.5 dB for all the cases mentioned above, which demonstrates our proposed scheme a promising prospect in XT online monitoring for MCF transmission.

2. AMCC aided XT online monitoring scheme

Figure 1 shows the schematic diagram of AMCC aided XT online monitoring scheme. The high-speed radio-frequency (RF) PON signal with the format of OOK is generated from the pulse pattern generator (PPG), while the low-speed AMCC signal, also in OOK, is from another PPG. They are added together with a certain modulation depth by a RF power combiner (PC). Then, the superposed signal modulates the continuous-wave (CW) light via Mach-Zehnder modulator (MZM). The modulated optical signals are launched into the corresponding cores, respectively, through the fan in device. In this work, we utilize core 1 for signal and core 2 for XT. Specially, the AMCC signals are selected as orthogonal sequences, so as to extract either of them from the noised received signal via Xcorr. The signal in core 1 is transmitted via MCF with the interference of XT from core 2. At the end of MCF, the signal enters the single-core fiber via the fan out device. Then, at the receiver side, the optical signal is divided into two parts via a coupler. One enters a photodetector (PD) for photoelectric conversion and then the received RF PON signal is evaluated by a bit error ratio tester (BERT) for communication performance judgement. The other chooses a low-bandwidth (low-BW) PD for photoelectric conversion, mainly for AMCC signal, and the received signal is sampled by a digital storage oscilloscope (DSO), from which we can calculate the XT by Xcorr using the offline data accurately.

 

Fig. 1. Schematic diagram of auxiliary management and control channel (AMCC) aided crosstalk (XT) online monitoring scheme. PPG, pulse pattern generator; MZM, Mach-Zehnder modulator; PC, power combiner; MCF, multi-core fiber; DSO, digital storage oscilloscope; PD, photodetector; BERT, bit error ratio tester; PON, passive optical network; RF, radio-frequency; BW, bandwidth; Xcorr, cross-correlation.

Download Full Size | PDF

The principle of AMCC has been described in previous works [23]. It mainly employs the superposition of high-speed signal and low-speed signal for data transmission, as well as auxiliary management and control at the same time. Thus, the eye diagram of the combined signal has two upper eyelids and two lower eyelids. In this paper, low-speed AMCC signal is mainly employed to measure the XT via Xcorr in the receiver side. The principle is shown as follows.

The complex electric field of the transmitted optical signal with two polarizations in core 1 can be simply described as:

The direct-current (DC) component of the received signal has been blocked in the offline data from DSO. Thus, the final received signal r = I – mean(I). The standard orthogonal sequences of the AMCC signals for the two cores are denoted as s₁ and s₂, with the distribution of ±1. We assume the extinction ratio (ER) and the AMCC modulation depth of the two cores are the same. Due to the principle of intensity modulation and direct detection (IM/DD), it is concluded that

Based on the scheme proposed above, experimental verifications of the online monitoring for both heterodyne XT and homodyne XT are conducted. Moreover, this method can be easily further extended to the case of more than two-core signal transmission simultaneously, as long as the AMCC signal sequences in all cores are orthogonal to each other.

3. Experimental verification of heterodyne IC-XT online monitoring

We first conduct the experiment to evaluate the online monitoring for heterodyne IC-XT with our proposed AMCC aided scheme. The experimental setup is shown in Fig. 2. In the transmitter, 10 Gb/s pseudo-random bit sequence (PRBS) with a periodical length of 2³¹−1, which is known as PRBS 31, and 1 Mbps PRBS 11 are generated as PON signal and AMCC signal, respectively, by different PPGs. They are superposed to form a combined signal with the 20% modulation depth of AMCC, which provides a proper performance tradeoff between the two signals in the experiment. The eye diagrams of PON signal and the combined signal are shown in Fig. 2(a) and Fig. 2(b). The combined signal is utilized to modulate the CW light centered at λ₁, equal to 1550.12 nm, which is the center of channel C34 in dense wavelength division multiplexing (DWDM), via an MZM biased at the quadrature point. The generated signal is first through a certain length of dispersion compensation fiber (DCF) to pre-eliminate the chromatic dispersion (CD) in the following 41 km 7-core fiber and then a variable optical attenuator (VOA) to control the launching power of about -20 dBm for core 1 in the 7-core fan-in. In order to monitor the launching power, a 1:99 coupler with a PM is used here. The information in core 1 is regarded as the useful signal, so core 1 is the mainly tested channel in the receiver. Meanwhile, the generated optical signal for core 2 is utilized as the XT, as a result that the CW light is centered at λ₂, equal to 1551.72 nm, as the center of channel C32 in DWDM, and the 1 Mbps AMCC signal is selected as PRBS 9, which is orthogonal with PRBS 11. In this experiment, the IC-XT for an adjacent core in 41 km 7-core fiber is basically below -40 dB, which is hard to be detected via Xcorr based on the current offline-data length due to the limitation in the storage of DSO. Thus, the signal after DCF is boosted by an erbium-doped fiber amplifier (EDFA) to guarantee the launching power up to about 10 dBm in the transmitter of core 2. Other settings for the transmitter for core 2 are almost the same as core 1. The signal-to-XT power ratio can be controlled by the two VOAs, so the IC-XT levels can be set freely. Then, the optical signal of the two channels is launched into the 7-core weakly coupled MCF, with the cross section shown in Fig. 2(c), that the cladding diameter is 150 µm and the core pitch is 42 µm. In the receiver side, we only detect the signal from core 1 and use the DSP technology to evaluate the IC-XT level. The received optical signal from the fan-out is firstly split into two parts by a 50:50 coupler, one of which is for the DSP calculation and the other is for reference (Ref.) XT measurement by WDM demultiplexing (DEMUX) and PMs. The WDM DEMUX extracts the signals with the wavelength centers of λ₁ and λ₂, corresponding to the useful signal and the XT in this heterodyne XT case. The results measured by PMs can be considered as accurate average IC-XT of the whole bandwidth. In the other part, the signal is amplified by an EDFA and through a waveshaper (WS) to filter the amplified spontaneous emission (ASE) noise as much as possible except for the two 0.16-nm passbands centered at λ₁ and λ₂, respectively. Then, a VOA attenuates the signal to a proper level to enter the low-BW PD, with the 3-dB bandwidth of 100 MHz, or the optical spectrum analyzer (OSA) for spectral measurement. The received RF signal is sampled by the DSO with a sampling rate of 4 MSample/s. Each offline data sequence has 2 M points, reaching the storage limit of the DSO. For the DSP, we firstly use a low-pass filter (LPF), with the 3-dB bandwidth of 700 kHz, to eliminate the noises from PD, EDFA, DSO and the residual PON signal, and then the received AMCC signal can be used to conduct the Xcorr with standard PRBS 11 and PRBS 9 for XT estimation by our scheme.

 

Fig. 2. Schematic diagram of experimental setup for the heterodyne IC-XT online monitoring. (a) Eye diagram of the standard PON signal; (b) eye diagram of the combined signal with 20% modulation depth of AMCC; (c) the cross section of 7-core fiber. Tx, transmitter; Rx, receiver; PRBS, pseudo-random bit sequence; EDFA, erbium-doped fiber amplifier; DCF, dispersion compensation fiber; VOA, variable optical attenuator; PM, power meter; DEMUX, demultiplexing; Ref., reference; WS, waveshaper; OSA, optical spectrum analyzer; LPF, low-pass filter.

Download Full Size | PDF

Figure 3 shows the results of heterodyne XT online monitoring, including the real-time measurements by Xcorr and PM during 20 minutes presented in Fig. 3(a) and 3(b), as well as an instant optical spectrum in the receiver in Fig. 3(c). We set the launching power of core 1 as -20 dBm for all the cases. In Fig. 3(a), the launching power of core 2 is 7 dBm and the distribution of XT is from about -18 dB to -15 dB. The mean absolute error (MAE) of our proposed Xcorr scheme is 0.41 dB in 20 minutes, compared with the standard PM method. Since the signal-to-XT power ratio is set with a 27-dB bias here, the real XT between the two cores is from about -45 dB to -42 dB during 20 minutes, that is reasonable to our knowledge. In Fig. 3(b), the launching power of core 2 is 4 dBm, providing a 24-dB bias for the signal-to-XT power ratio, and the distribution of XT is from about -20.5 dB to -15.5 dB with the MAE of about 0.53 dB in 20 minutes. In this case, the real IC-XT is about from -44.5 dB to -39.5 dB, which is also reasonable. It is clear that the curves of Xcorr in both Fig. 3(a) and 3(b) match PM well with limited errors. Since the scheme of PM gives the average XT of the signal in the whole bandwidth and Xcorr scheme only indicates the XT characters within the bandwidth of AMCC signal, the XT fluctuation range measured by Xcorr is slightly larger than the other. However, we regard this error acceptable to utilize AMCC for online monitoring the average XT of the whole bandwidth during a period. Figure 3(c) gives an instant optical spectrum of the received signal after the WS when the launching power of core 2 is 7 dBm, in which the crosstalk and the signal are centered at the wavelength of 1551.72 nm and 1550.12 nm, respectively, and the instant IC-XT is about -15.5 dB.

 

Fig. 3. The results of heterodyne XT online monitoring, including: the real-time measurements by Xcorr and PM in 20 minutes under the condition that the launching power of core 2 is (a) 7 dBm and (b) 4 dBm, respectively, as well as (c) an instant optical spectrum in the receiver with the XT of -15.5 dB.

Download Full Size | PDF

The results above denote that our proposed scheme works well for the IC-XT online monitoring. However, the IC-XTs in reality are mostly with the same wavelength as the signals, so that the traditional schemes based on PM cannot provide any support. In the following part, we experimentally verify the performance of our proposed scheme for the homodyne IC-XT.

4. Experimental verification of homodyne IC-XT online monitoring and detailed analysis for our proposed scheme

In this part, we experimentally verify the performance of our proposed scheme in online monitoring for homodyne IC-XT in both C band and L band, which present different levels of IC-XT. Meanwhile, in C band, the transmission results for the PON signal is also evaluated based on the AMCC superposition and detailed analysis for the scheme, such as the probability density function (PDF), Xcorr waveform and frequency spectrum of the received signals, are conducted to provide a comprehensive understanding for it.

Figure 4 shows the experimental setup for the AMCC modulated PON signal transmission and homodyne IC-XT online monitoring. Most parts and parameter setting in Fig. 4 related to IC-XT measurement are similar with Fig. 2. However, for a homodyne IC-XT case, the optical signals with the same wavelength enter the two channels and it is impossible to separate the signal and XT by WDM DEMUX at the receiver side. The wavelength of the laser is chosen as 1550.12 nm and 1610.06 nm, which are the centers of channel C34 and L62 in DWDM, respectively. In order to get a Ref. XT to evaluate the accuracy of our proposed scheme, a single-frequency pilot tone of 1 MHz is added in the transmitter of channel 2. Since AMCC signal has no frequency component at 1 MHz, the pilot tone has no influence on the Xcorr calculation after the LPF, and it can be easily detected by a fast Fourier transform (FFT). Moreover, in the receiver, the optical signal from core 1 enters an EDFA for amplification and an optical bandpass filter (OBPF) centered at certain wavelengths with the passband of about 0.2 nm to eliminate the ASE noise. Then, a VOA is to control the optical power for detection and a 10:90 coupler is for power splitter, in which 90% part for the PON signal Rx and 10% part for AMCC signal Rx. Here, we only test the performance of PON signal in C band, and the RF signal after PD enters a direct current block (DCB) with the passband of 0.01∼18 GHz to suppress the AMCC signal as much as possible. Then, the signal through a clock data recovery (CDR) is received by the bit error rate tester (BERT) for real-time performance evaluation. It is worth mentioning that the IC-XT of the L-band signal for a certain MCF is obviously larger than the C-band case due to the bigger mode-field sizes of light in longer wavelengths, resulting in easier inter-core optical mode coupling, which means the IC-XT in L band is larger and easier to be detected. Therefore, we add an EDFA only for L-band signal in channel 1 to decrease the gap of launching powers between the two cores and increase the OSNR before the PD detection.

 

Fig. 4. Schematic diagram of experimental setup for the AMCC modulated PON signal transmission and homodyne IC-XT online monitoring. OBPF, optical bandpass filter; FFT, fast Fourier transform; DCB, direct current block; CDR, clock data recovery; BERT, bit error rate tester.

Download Full Size | PDF

Firstly, we deal with the C-band transmission. Figure 5 shows the average BER together with the BER fluctuation bounds of received 10-Gbps-OOK PON signal in 30 minutes as a function of received optical power (ROP). The blue line is the case without AMCC signal, while the red line presents the condition of AMCC superposition with the 20% modulation depth. In this case, the launching power of core 1 and core 2 is about -10 dBm and 10 dBm respectively, corresponding to an approximate average IC-XT of -20 to -25 dB. The average BER decreases with the increase of ROP. The BER at each ROP suffers from fluctuation of 1∼4 orders of magnitude, which comes from the IC-XT. Obviously, the BER performance of the PON signal deteriorates slightly with the superposition of 20% AMCC signal due to the residual AMCC signal after DCB. This is acceptable when the communication requirements are not extremely demanding. Moreover, the BER fluctuation becomes gentler for the 20% AMCC case since the certain amount of residual AMCC distortion could reduce the variance of total fluctuation, which is formed by IC-XT, residual AMCC and the thermal noise from PD.

 

Fig. 5. Average BER together with the BER fluctuation bounds of received 10-Gbps-OOK PON signal in 30 minutes as a function of received optical power.

Download Full Size | PDF

 

Fig. 6. The probability density distribution (PDF) of some extracted AMCC signals in different levels of XTs: (a) -10.0 dB; (b) -13.2 dB; (c) -16.1 dB; (d) -19.2 dB; (e) -22.5 dB.

Download Full Size | PDF

Then, we conduct more detailed analysis for the received AMCC signals. Here, we maintain the launching power of core 1 to be about -20 dBm while setting the launching power of core 2 to be about 10 dBm, 7 dBm, 4 dBm, 1 dBm and -2 dBm, respectively. The AMCC signal is extracted by the lowpass filtering of the received signal. Figure 6(a)-(e) provides the PDF of a set of extracted AMCC signals with different levels of XTs, i.e., -10.0, -13.2, -16.1, -19.2 and -22.5 dB, respectively. The estimated SNRs of extracted AMCC signals are 5.37, 5.96, 5.93, 5.96 and 5.98 dB. Thus, the IC-XT is not the dominate noise anymore when it is less than -13.2 dB, while the major distortion comes from the residual PON signal indeed. Under this condition, the quality of the received AMCC signal is too poor to guarantee error-free detection. Nevertheless, the Xcorr operation of this signal with the standard PRBS sequences can still calculate the XT correctly.

Figure 7(a)-(e) presents the partial waveforms of the Xcorr between the extracted AMCC signals in different XTs and standard PRBS 9 sequence. The length of standard PRBS 9 sequence is larger than 2 M to guarantee the appearance of maximum PXC. Since the repetition period of PRBS 9 is 511 and the DSO samples each symbol with 4 samples, the PXC are repeated by 2044 samples. Figure 7(f) shows partial waveform of the Xcorr between the extracted AMCC signals and standard PRBS 11 sequence, with the gap of 8188 between two PXC. Since the launching powers of core 1 in all cases are the same, the Xcorr of extracted AMCC signal with PRBS 11 sequence is stable. The PXC are much higher than the noise in Fig. 7(f). It is obvious that the PXC value are lower when the XTs are smaller in Fig. 7(a)-(e). However, it is known from Fig. 7(e) that the PXC will be overwhelmed by noise when the peaks are lower than about 300, corresponding to the IC-XT of about -25 dB, which is the IC-XT detection limit of this scheme in current experimental setup and 2 M offline-data points each time. If there is more continuously sampled offline data or the SNR of received AMCC signal is larger, i.e. improving the AMCC modulation depth, the IC-XT detection limit will be improved further.

 

Fig. 7. The partial waveforms of Xcorr between the extracted AMCC signals in different XTs and standard PRBS sequences. (a) XT = -10.0 dB, PRBS 9; (b) XT = -13.2 dB, PRBS 9; (c) XT = -16.1 dB, PRBS 9; (d) XT = -19.2 dB, PRBS 9; (e) XT = -22.5 dB, PRBS 7; (f) PRBS 11

Download Full Size | PDF

Figure 8(a)-(c) demonstrate the frequency spectrums of the received AMCC signals with the 1-MHz pilot tones in different XTs, i.e. -19.2 dB, -21.9 dB and -24.2 dB, respectively. These signals have not been low-pass filtered. It is clear that the envelope of the received AMCC signal is a Sinc function with the zero point in 1 MHz and there is an obvious noise floor, mainly from the residual PON signal. We find it easier to get the level of pilot tones from the received signals by FFT, to realize a Ref. XT. Meanwhile, from Fig. 8(c), the pilot tone would be overwhelmed by noise when the XT is decreased by 3-4 dB, indicating the detection limit of the pilot tone in current case is about -28 dB, which is enough for our proposed scheme.

 

Fig. 8. The frequency spectrums of the received AMCC signals with the 1-MHz pilot tones in different XTs: (a) -19.2 dB; (b) -21.9 dB; (c) -24.2 dB.

Download Full Size | PDF

Figure 9 gives the results of homodyne XT online monitoring. Figure 9(a)-(e) presents the real-time measurements by Xcorr and pilot tone in 20 minutes. The launching power of core 2 is 10 dBm, 7 dBm, 4 dBm, 1 dBm and -2 dBm, respectively, and the corresponding MAEs of our proposed method are 0.17, 0.11, 0.42, 0.46 and 0.77 dB, compared with the detected pilot tone. Specially, the result in Fig. 9(c) shows unusual when the time is beyond about 14 minutes, that the XT is partially lower than the detection limits of Xcorr and pilot tone, which have been analyzed above. The results of detection limits match well with our analysis and the calculation of MAE in Fig. 9(c) doesn’t contain the data beyond 14 minutes. The larger errors are mainly distributed in some XTs close to the detection limits, in which the value of Xcorr peaks are close to the noise floor. The average real IC-XTs of the two cores are calculated as around -43 dB, -40 dB, -43 dB, -41 dB and -41 dB, respectively, for the above five cases, which are all reasonable for the utilized 41-km 7-core fiber. Figure 9(f) shows some instant optical spectrums in the receiver with some certain XTs when the launching powers of core 2 are set as above. We cannot separate the optical spectrums of signal and XTs at the same time because of the sameness of their wavelengths. However, the optical spectrum of signal is not time-varying and we can obtain the instant spectrums by turning off the signal channel or XT channel in measurement. The detected instant XTs in Fig. 9(f) are -10.7, -13.5, -16.5, 19.6 and -22.4 dB, respectively, which are all in the reasonable range. In summary, for C-band signals, our proposed scheme works very well in the homodyne XT online monitoring with low MAEs, mostly lower than 0.5 dB.

 

Fig. 9. The results of homodyne XT online monitoring in C band, including: the real-time measurements by Xcorr and pilot tone in 20 minutes under the condition that the launching power of core 2 is (a) 10 dBm, (b) 7 dBm, (c) 4 dBm, (d) 1 dBm and (e) -2 dBm, respectively, as well as (f) the instant optical spectrums in the receiver with some certain XTs.

Download Full Size | PDF

Finally, we will apply our proposed scheme to the IC-XT online monitoring in L-band transmission. Figure 10 presents the homodyne IC-XT online monitoring of L band. Figure 10(a)-(c) show the real-time measurements by Xcorr and pilot tone in 20 minutes under different launching powers, in which the signal-to-XT power ratios are set with 30-dB, 24-dB and 15-dB biases here. The detected XTs distribute from -8 dB to -3 dB in Fig. 10(a), from -22 dB to -8 dB in Fig. 10(b) and from -23 dB to -17 dB in Fig. 10(c), which denotes the real IC-XTs between the two cores are -38 dB to -33 dB, -46 dB to -32 dB and -38 dB to -32 dB, respectively. The corresponding average real IC-XTs are calculated as -35 dB, -38.5 dB and -35 dB, obviously larger than the levels in C band. The MAEs of our proposed scheme are 0.12, 0.34 and 0.45 dB, respectively, compared with the detected pilot tone. Also, the detected XT with larger errors distributes lower than -20 dB, close to the detected limit. Figure 10(d)-(f) present some instant optical spectrums in the receiver with some certain XTs of -5.8 dB, -11.2 dB and -18.9 dB, in which the settings of launching powers correspond to Fig. 10(a)-(c) and all the spectrums are in reasonable ranges. In summary, the IC-XT online monitoring for L band also demonstrates its effectiveness with MAE lower than 0.5 dB and the evaluated real IC-XTs in L band for 41-km 7-core MCF are about 5-dB larger than the C-band cases.

 

Fig. 10. The results of homodyne XT online monitoring in L band. The real-time measurements by Xcorr and pilot tone in 20 minutes under the condition of different launching powers: (a) core 1: -20 dBm, core 2: 10 dBm; (b) core 1: -20 dBm, core 2: 4 dBm; (c) core 1: -5 dBm, core 2: 10 dBm. The instant optical spectrums in the receiver with some certain XTs: (d) -5.8 dB; (e) -11.2 dB; (f) -18.9 dB.

Download Full Size | PDF

5. Conclusion

In this paper, we have proposed an IC-XT online monitoring scheme based on AMCC, in which orthogonal sequences are selected as the AMCC signals for the signal channel and XT channel. In the receiver, the IC-XT is accurately estimated by Xcorr. Comprehensive verifications of the IC-XT online monitoring in a 41-km 7-core MCF have been experimentally performed for both heterodyne XT and homodyne XT in C or L band. The results show that our proposed scheme has strong effectiveness with only slight error (mostly < 0.5 dB), compared with the traditional PM measurement or the detection of single-frequency pilot tone, which indicates this scheme is potential to be utilized in SDM links in the future for IC-XT online monitoring.