1. Introduction

5G mobile systems are strongly required for handling the mobile data traffic expansion caused by the spread of small devices [1,2]. To address this requirement, mobile telecom operators have introduced centralized radio access network (C-RAN) architecture that comprises two kinds of equipment: a centralized unit (CU) and a distributed unit (DU). These equipment units are connected via optical fibers as a mobile fronthaul (MFH). The MFH typically uses the common public radio interface (CPRI) [3] between CUs and DUs. However, its optical bandwidth in the MFH link reaches more than 10 times that of the original wireless data rate and thus low latency is required. In addition, the number of optical fibers will increase in the future MFH systems.

To address these issues, wavelength division multiplexing passive optical networks (WDM-PONs) are an attractive candidate [4]. The WDM-PONs can reduce the number of optical fibers, expand the capacity of each wavelength easily, and eliminate the additional latency by multiplexing signals through means such as a time division multiplexing (TDM)-PON. To achieve dense and effective deployment of small cells, optical network units (ONUs) have a tunable transceiver to enable the inventory to be easily controlled.

Controlling the wavelength of upstream signals in WDM-PONs is very important for achieving stable communication to support mobile systems. In WDM-PONs, the wavelength control is divided into two cases: the wavelength setting error at the initial connection between the optical line terminal (OLT) and the ONU, and the wavelength drift due to the aging degradation of the laser diode (LD). The conventional wavelength control method solves these cases, but needs an external filter with high accuracy specifications such as an etalon filter [5] and many photo detectors (PDs) for monitoring the optical power for each upstream wavelength [6–8].

To achieve these wavelength controls simply, the ITU-T recommendation G.989 series [9] specifies the auxiliary management and control channel (AMCC) scheme. The AMCC is a channel superimposed onto the client signals with a different physical path, such as a low-frequency pilot tone. This enables the OLT to relate the management information to the ONU independent of the client signal. Several studies [10–12] proposed methods to generate an AMCC scheme and investigated their feasibility. A procedure for activating a WDM-PON using an AMCC scheme was proposed and demonstrated by Sone et al. [13]. However, even if the AMCC messages in G.989.3 are used, the correct wavelength control is difficult, because the discrimination of the wavelength drifts from the other transmitter accidents such as the optical power degradation at the transmitter is difficult.

In our previous study [14], we proposed a wavelength adjustment method using an AMCC frame against the wavelength drift in WDM-PON and demonstrated it with an AMCC evaluation platform [15]. However, this study assumed that the initial emitted wavelength was completely set onto the accurate wavelength. Moreover, the system margin for the wavelength adjustment is not considered in order to keep the normal communication. In this paper, we expand our method to consider the initial tuning error and discuss the adjustment size of wavelengths based on the system margin of the loss budget. Our method makes it possible to control the wavelength setting error at the initial connection between the OLT and the ONU and the wavelength drift due to the aging degradation of the LD. We also clarify the control conditions needed to minimize the control time.

2. Upstream wavelength adjustment method using AMCC with power monitoring

2.1 WDM-PON architecture

Figure 1 shows a WDM-PON using our method with AMCC. The CUs and DUs are respectively connected to OLTs and ONUs. OLTs and ONUs are connected via an optical distribution network with an optical power splitter at the ONU side. We assume that an arrayed waveguide grating (AWG) that has a Gaussian shape of the transmittance characteristics is used as a wavelength multiplexing/demultiplexing filter at the OLT side. We make this assumption because a lot of ports can be provided at a cost-effective level better than that obtained with thin-film filters to accommodate many small cells. The OLT has an AMCC controller (master) and each ONU has an AMCC processing unit (slave). The OLT and ONUs send AMCC messages by superimposing a low-frequency modulated pilot tone on the client signals of the downlink and uplink. In the NG-PON2 specifications define the 500 kHz and 128 kb/s pilot tone as an AMCC implementation [9]. The G.989 series does not specify the required BER performance of AMCC signals but we confirmed that the estimated BER performance achieved error-free transmission (< 10⁻¹⁰) within 0.5 dB optical power penalty for the client signals [16,17]. The receiver in the OLT monitors the optical received power of the filtered upstream signals for each wavelength. In the ONU, the optical output power of the distributed feedback (DFB) laser is measured by the monitor PD. The output wavelength is changed in accordance with the temperature of the DFB laser diode controlled by a thermoelectric-cooler (TEC). The AMCC processor communicates with the AMCC processing unit by using the AMCC signals superimposed on the downstream or upstream client signals. Its frame format complies with ITU-T G.989.3.

 

Fig. 1. WDM-PON configuration for 5G MFH.

Download Full Size | PDF

In this system, we should adjust the wavelength of upstream signals for two cases: the initial setup and the laser degradation for the aging. In the initial setup state, the laser outputs a unique wavelength allocated by the OLT in the activation procedure. The wavelength of the upstream signal must not be set within the maximum tuning error (MTE) to meet the NG-PON2 specifications [9]. MTE is calculated from the maximum spectral excursion (MSE) and the minimum single-side spectral width at −15 dB (e.g. ±10 GHz for a 10 Gbps WDM-PON system). The second case is caused by the laser degradation due to many years of aging. In general, the output wavelength changes gradually (on the 10³ hour order) because the injected threshold carrier density increases [18]. These cases should be mitigated to provide a reliable WDM-PON. Accordingly, we propose the wavelength control method described in the following subsection.

3.2 Control algorithm for the upstream wavelength

Our wavelength control method and its parameters are respectively shown in Figs. 2 and 3. In the initial wavelength setting mode, the OLT and a newly installed ONU negotiate wavelengths for the uplink and downlink. This mode has two roles: adjustment of the initial tuning error and registration of the initial output/received power as references. After the activation, the OLT monitors and memorizes the optical received power (P_{rec, i}) at the monitoring point in the receiver and sets i = 0. The OLT sends a wavelength adjustment message (Adjust_Tx_Wavelength) defined in ITU-T G.989.3 with a negative adjustment frequency and i increments. The ONU shortens its output wavelength in accordance with the message. When the initial output wavelength exceeds the AWG’s center wavelength, the optical received power at the OLT increases after manipulating the wavelength. Then, the OLT sends the same message repeatedly until the optical received power decreases. When the OLT detects that no decrease or change in the optical received power has occurred, it checks the number i. When i = 1, it means the direction of the wavelength adjustment is on the opposite side of the center. Then, the OLT repeats the loop using the wavelength adjustment message with a positive number. If i ≠ 1, it implies that the wavelength is adjusted excessively. The OLT then sends a wavelength adjustment message that sets the wavelength of i so that it maximizes the optical received power and sets the initial optical received power (P_{rec_init}) in the OLT and the initial optical output power (P_{out_init}) from the ONU. These values are used as references and can mitigate the effect caused by the individual differences of devices or emitted wavelength power changed by the temperature.

 

Fig. 2. Flow chart of our wavelength control method.

Download Full Size | PDF

 

Fig. 3. Parameters used for wavelength adjustment and three-step adjustment process in wavelength fixed mode.

Download Full Size | PDF

After the initial wavelength setting mode is finished, the state changes into the normal mode and the OLT monitors the optical received power (P_rec) periodically. Considering that the time scale of the degradation is on the order of 10³ hours, we believe that it is sufficient to measure the power once a day to detect the wavelength drifts.

Wavelength drifts occur after many years of use. After a wavelength drift occurs, the output wavelength of the DFB laser in the ONU transmitter changes from the center wavelength of the AWG. Then, the P_rec degrades because of the Gaussian characteristic of the AWG transmittance. When the OLT detects the P_rec degradation from the periodic monitoring results, the wavelength drift detection mode begins. This mode aims to distinguish the wavelength drift from the optical received power fluctuation caused by other reasons (e.g. laser current degradation). The OLT requests the optical output power (P_out) to the target ONU be provided by AMCC. Here, ΔP_th is used for the threshold to detect the optical fluctuation transmitted from the ONU and received power degradation at the OLT. If the fluctuation of the received optical power is smaller than ΔP_th, the OLT regards it as negligible and keeps the normal mode. In another case, the OLT sends a request message (Request_Tx_Power) and the ONU reports the monitored P_out (Report_Tx_Power) to the OLT. To ask the output power level, the Change_Power_Level message is defined in G.989.3. However, ONU can only tell its attenuation level in 3dB to the OLT by this message, so the OLT cannot know whether the output degrade occur or not. Therefore, we defined above two messages. If the fluctuation of the optical output power (ΔP_out) becomes more than the ΔP_th, the OLT alerts the network operator that a problem has occurred with the ONU transmitter. If the ΔP_out is less than the threshold, it implies that the output power does not change and the received power is degraded. Therefore, the OLT judges that the wavelength drift has occurred in the ONU and the state changes into the wavelength fixed mode.

The OLT sends an Adjust_Tx_Wavelength message to request the upstream wavelength be shortened. When the upstream wavelength drifts to the longer wavelength, the ΔP_rec gets smaller as the wavelength adjusts according to the message. The OLT monitors the received optical power and calculates the ΔP_rec after waiting to finish the wavelength adjustment (e.g., 1 minute). If the ΔP_rec is more than ΔP_th, the OLT sends a further Adjust_Tx_Wavelength message. Some steps after that, when the ΔP_rec gets smaller than ΔP_th, the wavelength drift is mitigated and the state moves back to the normal operation mode. When the wavelength drifts to the shorter wavelength, the ΔP_rec becomes larger, so the OLT orders the ONU to adjust the wavelength to the longer wavelength.

Our proposed system enables the wavelength adjustment without any additional component such as etalon filter by using the optical power monitor in the TRx and exchanging the messages between the OLT and the ONU. Furthermore, the OLT can detect the ONU power degradation by comparing the optical output power with its initial value. In the next section, we describe the frequency adjustment size related to the setting time for each wavelength control mode.

3. Frequency adjustment size calculation

To calculate the maximum frequency adjustment size, we set three assumptions. First, that the frequency adjustment size was not changed during the initial wavelength setting mode and the wavelength fixed mode for the simple operation. Second, that the OLT can detect an optical power degradation ΔP_th as soon as possible. This assumption is related to the once a day monitoring period, which is short enough to detect the laser degradation. Third, that in the initial wavelength setting mode the optical received power is not permitted to go below the system margin for wavelength drifts. We define the system margin as the acceptable power degradation caused by wavelength drifting from the optical received power at the center wavelength of the AWG (P_center). This assumption considers that the wavelength might become distant from the center frequency if the first instruction of the wavelength adjustment tells the wrong direction.

Let the power degradation at wavelength λ, the system margin, and the adjustment frequency be respectively D(λ), M, and Δλ_adj. The maximum adjustment frequency Δλ_{adj_max} is obtained as

We calculated the minimum number of steps n_{step_min} when the uplink wavelength starts from MTE and the OLT instructs the wrong direction to the ONU in the first step. We assumed that the Δλ_{adj_max} is used for the initial wavelength setting because the common adjustment size makes this system simple. Then, we get nstep_min as

4. Results and discussion

Figs. 4(a) and 4(b) show the calculation results based on a 100-GHz-grid AWG with the Gaussian shaped transmittance characteristic that is experimentally measured. To calculate, we made an assumption that the upstream wavelength was set at the MTE defined in NG-PON2 after the ONU activation. For example, we set 1.0 dB as the system margin and 0.5 dB as ΔP_th. When the adjustment frequency is set to 4 GHz, the maximum wavelength drift is limited to ±2 GHz after the initial wavelength setting mode. The wavelength after the initial wavelength setting mode is selected from a set of the wavelengths during the loop of the initial wavelength setting mode. This wavelength set must include one wavelength fitting in the range from −2 GHz to 2 GHz (half of tuning frequency). If the 0.5 dB degradation from the initial wavelength setting mode occurs and the OLT gives the wrong direction to the ONU, the degradation goes down to 0.91 dB, but it still meets the system margin of 1.0 dB. However, for the 5 GHz case, the degradation might drop to below 1.0 dB and not meet the system margin. Therefore, 4 GHz is the maximum tuning frequency for this case. Then, from Fig. 4(b), the system requires adjustment to be made six times to finish the initial wavelength setting.

 

Fig. 4. (a) Maximum adjustment frequency vs. power threshold (b) Step size against power threshold in initial wavelength setting mode.

Download Full Size | PDF

We can estimate the control time of our wavelength control method by calculating this step size by using the tuning time of each tunable LD as a thermal control DFB of a distributed Bragg reflector (DBR)-LD. If the telecom operator chooses a smaller adjustment frequency than Δλ_{adj_max}, it increases the number of the steps at the initial wavelength setting and leads to additional time being needed to finish the initial wavelength setting. The initial wavelength setting mode finishes in minutes order for the thermal control DFB laser. This time is acceptable for operators, because the normal mode in which the data is transmitted starts after the initial wavelength setting mode. In the wavelength fixed mode, the client singles are not affected by the wavelength adjustment, because AMCC messages are sent via the different physical channel from the client signal and the wavelength adjustment is conducted within the system margin. Therefore, the OLT do not need to mind the adjustment time. Finally, when we use different AWGs (e.g., 200 GHz channel spacing), we can apply this design concept for adjusting the wavelength in the same way.

5. Conclusion

We described a method we proposed for controlling the wavelength of upstream signals using an AMCC scheme in a WDM-PON. Our method eliminates the need for external devices for the power monitor by using the power monitor function in an OLT receiver and makes it possible to control the wavelength setting error at the initial connection between the OLT and the ONU. It also makes it possible to control the wavelength drift due to the aging degradation of a laser diode. We clarified the conditions required for the amount of wavelength adjustment and the step size of the adjustment in order to minimize the control time within the target system margin of the loss budget.