Performance monitoring of reconfigurable optical networks based on average Q factor

0 Introduction

With the development of multimedia services, the proportion of packet signals in optical backbone transmission networks is increasing. Due to the suddenness and unpredictability of packet data services, optical networks are required to be reconfigurable. Reconfigurable optical networks are optical networks that automatically and dynamically allocate bandwidth resources and are suitable for optical networks dominated by data services. Monitoring of reconfigurable optical networks, especially performance monitoring, is particularly important.

Great progress has been made in the research of reconfigurable optical networks and their performance management. Among them, the Q factor measurement method is an effective monitoring method and an effective means to determine the quality of the optical channel. By testing the Q factor to evaluate the BER performance, the minimum BER in the optimal design device can be obtained. It can use a dual decision circuit for online monitoring or a single decision circuit to work in a suspended service state. The eye diagram is the preferred tool for observing waveform distortion, noise level and signal quality. At present, there are many different sampling methods based on eye diagrams at home and abroad that can realize the monitoring of optical performance.

Among the existing methods, the standard deviation and average power value of the signal can be obtained through eye diagram sampling, thereby obtaining the Q value. This method requires clock timing extraction, and the implementation process is relatively complicated; while the asynchronous eye diagram sampling method is used to monitor the performance parameters of return-to-zero code and non-return-to-zero code in the optical channel (mainly the dispersion of the optical channel), the measured parameters are relatively simple and cannot fully reflect the channel quality; based on the corresponding relationship between Q value and bit error rate, a scheme for online detection of the Q value of the optical signal is proposed, and a detection module is designed based on digital signal processing chip technology, thereby realizing online Q value monitoring of the signal. The decision level of this method is not easy to determine, and when the noise amplitude is large, the error is also large; the method of using synchronous eye diagram to calculate the average Q value requires clock extraction for synchronization during sampling, which is complex, costly, and the bit rate of this method is not transparent; while the asynchronous eye diagram sampling method is used to sample the influence of pulse width on the test result, clock recovery is not required, and the bit rate is transparent and the cost is low.

Most of the above methods do not monitor the relationship between the test results and the actual system. To this end, this paper proposes a new monitoring method based on the summary of previous studies, that is, under the condition that clock timing extraction is not required, the average Q value is calculated and obtained by asynchronous eye diagram sampling method and computer simulation. By comparing the relationship between the average Q value and the initial Q value, the performance of the system can be quickly monitored. This method can be used in multi-wavelength systems to obtain the relationship between the sampling interval and the test results, thereby realizing the role of effectively monitoring multi-wavelength optical networks. This method is transparent in bit rate, simple in structure, and easy to implement.

1 Implementation Principle

There are many parameters that reflect the signal performance quality of optical communication systems. However, no matter which parameter is used, it can reflect the performance of the optical channel to varying degrees. Among these parameters, BER reflects the optical channel performance most accurately, but its measurement requires complex operations such as electro-optical conversion and clock recovery. Currently, most systems use OSNR to evaluate the performance of optical channels, but the accuracy is not enough. Therefore, a new parameter, the Q factor, is introduced.

The Q factor is an important parameter that reflects the signal-to-noise ratio (SNR) of an optical fiber communication system. It is defined as the ratio of the electrical signal power to the noise power at the receiver under the optimal decision threshold. It is applicable to digital signals of various signal formats and rates, and does not require the frame structure to be unraveled. Therefore, it is relatively simple and easy to perform system analysis. The establishment of the Q factor parameter and the realization of the measurement method are of great significance for the actual maintenance and testing of reconfigurable optical networks.

Since the Q factor is measured before the optical receiver makes a decision on the data, it is easy to establish a relationship with the BER:

In addition, since the noise power part of the Q factor calculation is caused by ASE, the relationship between Q factor and OSNR can be formally obtained as follows:

Where r is the extinction ratio of the optical signal at the transmitting end.

Figure 1 shows the asynchronous eye diagram of the average Q value and its amplitude histogram. The amplitude diagram shows the amplitude distribution of the signal pass and space, and the definition of the average Q value is:

Where μ1 is the mean level of signal 1, σ1 is the standard deviation (noise) of the signal 1 level, μ0 is the mean level of signal 0, and σ0 is the standard deviation (noise) of the signal 0 level.

[page]

This method can be used to monitor the SNR of the signal in the transmission fiber. The asynchronous eye diagram of the optical signal is measured by an optical receiver and an oscilloscope, and the average estimated Q value is obtained by sampling the eye diagram. Through a large number of computer simulations to select the appropriate sampling point interval, the author found that there is a linear relationship between the average estimated Q value and the initial Q value. This linear relationship can be used to achieve performance monitoring of optical networks. At the same time, after optical sampling, it can also be used for monitoring multi-wavelength channels by simply adding wavelength identification.

2 System Structure

Figure 2 shows a system flow chart of a sampling method, and the simulation is based on this system. The input optical signal is divided into two networks through a splitter, one part is output through the optical fiber, and the other part is output to the regulator. The sampling pulse and the input optical signal can enter an independent polarization modulator driven by a low-frequency pulse generator. Since this pulse modulation and data are asynchronous, sampling and averaging can be performed. When the signal enters the next system, the system will use a photodiode to detect the power of the modulated signal within a specific time. The digital signal processor can be used to collect the measurement data and obtain the corresponding probability density function. These probability density functions are then used to analyze the monitoring information. This method can be used to estimate the Q factor of different wavelength channels at the same time.

3 Simulation Implementation

FIG3 shows a simulation flow chart of the optical network performance monitoring system. A cosine signal plus random normal noise can be used to simulate the signal transmitted in the optical fiber (including the NRZ code signal of the pseudo-random binary sequence and the spontaneous radiation noise) to obtain the initial Q value. Then select the appropriate input signal sampling point, which can be in units of period, and the interval between two sampling points is T+n·△t, and then put the sampled values ​​into a one-dimensional array. Then, according to the range of the sampling value level, the sampled values ​​in the one-dimensional array are counted, and the amplitude Gaussian distribution diagram is drawn according to the statistical value. Finally, the average estimated Q value is obtained according to the Gaussian distribution diagram. In this way, by comparing the initial Q value and the average estimated Q value, the relationship diagram between the two can be obtained.

4 Simulation Results

By selecting different numbers of sampling points and sampling intervals, the average Q values ​​listed in Tables 1 and 2 can be obtained under different initial Q values.

[page]

It can be seen from Table 1 and Table 2 that in the sampling process, when the number of sampling points is small, the interval T+n·△t between two sampling points has a greater impact on the simulation results, and the test results often cannot reflect the actual Q value. Generally, when the sampling points are greater than 40,000, the difference between the average Q values ​​caused by different sampling positions is less than 0.01, and the simulation results are basically independent of the sampling position (T+n·△t). Since the system does not require clock timing extraction, the sampling point position is random, so the function value may be taken as 0, resulting in an average Q value that is much smaller than the initial Q value. However, the average Q value is stable and can still reflect the size of the initial Q value. The relationship between the initial Q value and the average Q value obtained by the author is shown in Figure 4.

From Table 1, Table 2 and Figure 4, we can see that when the initial Q value decreases from 6 to 5, the average Q value of the single wavelength channel decreases from 3.0086 to 2.9082, a decrease of 0.1004. When the sampling point is greater than 20,000, the fluctuation caused by the sampling position is less than 0.002. At this time, the degradation of the Q value can be monitored.

In the wavelength division multiplexing system (4 wavelength divisions), 20,000 points are also taken for simulation. When the initial Q value of one wavelength channel is reduced from 6 to 5, the Q values ​​of the other three channels are still 6. At this time, the average Q value of the four-wavelength system is reduced from 3.0085 to 2.9835 (reduced by 0.025), and the fluctuation error caused by the sampling position is also less than 0.002. The degradation of this Q value can also be monitored. Therefore, this method can be used in wavelength division multiplexing systems.

Further simulations also show that in a wavelength division multiplexing system with n channels, the average Q value of a wavelength system will decrease by 1/n due to the decrease in the Q value of a certain wavelength channel, so the fluctuation error caused by the sampling position must be smaller. Simulations show that the number of sampling points in a 16-channel wavelength division multiplexing system should be greater than 80,000.

5 Conclusion

This paper mainly discusses a reconfigurable optical network performance monitoring technology based on average Q factor. This method uses asynchronous eye diagram sampling. No clock synchronization is required. Through a large number of numerical simulations, the direct impact of the number of sampling points on the estimated Q value is obtained. The results show that when the number of sampling points is less than 5000, the actual results cannot be reflected. When the number of test points is greater than 40000, the actual Q value of the multi-wavelength system can be tested. This method can not only quickly monitor the performance of reconfigurable multi-wavelength optical networks, but also has transparent bit rate, simple structure and easy implementation. It is of great significance for the maintenance and testing of actual reconfigurable optical networks.