Open Access
Add to BibSonomyAbstract
In recent years, the diffusion of mobile terminals has brought about an explosive increase in communication traffic of mobile RANs. The number of radio base stations and optical fiber lines between them is becoming larger. For this reason, we studied effective optical network technologies for mobile RANs and propose the use of TWDM-PON as a means of enabling RANs to be operated flexibly and have wideband communication capability. We confirmed the feasibility of TWDM-PON for this application by numerical simulation. The results show that TWDM-PON can accommodate the bandwidth more than TDM-PON and completely eliminate unused bandwidth in TDM-PON.
© 2013 Optical Society of America
1. Introduction
Communication traffic in mobile networks has been sharply increasing in recent years due to the explosive diffusion of smart phone applications. A very important method for and trend in dealing with the sharp increase is making network cells smaller and dividing the large amount of traffic in one cell into smaller amounts of traffic in a number of smaller cells [1,2]. This has resulted in a huge increase in the number of wireless base stations and the number of optical fiber lines between them. This in turn has resulted in a serious congestion problem in radio access networks (RANs) that connect a wireless base station and a core network. An effective broadband RAN configuration is needed to combat this problem.
It has been found that the passive optical network (PON) technology used in fiber-to-the-home (FTTH) service is an effective solution to the RAN configuration problem [3]. Its use achieves not a point-to-point network configuration but a configuration in which a part of an optical fiber is shared by using a passive branch (optical splitter) in an optical fiber network. An example of this is the Ethernet-PON technology that has been standardized [4]. This technology makes it possible to achieve an effective equipment configuration through the effective sharing of optical fibers. It is also able to provide low-delay communication between base stations and improved flexibility in cell configuration. However, in time division multiplexed (TDM)-PON, communication speed is limited by physical data rates. With mobile network traffic still increasing markedly, the current TDM-PON bandwidths will become deficient in the near future. However, there is a wideband alternative to TDM-PON, namely wavelength division multiplexed (WDM)-PON [5].
WDM-PON uses a number of wavelengths assigned to the optical network units (ONUs) at user terminals at the far end of the access network. Because each ONU uses its assigned wavelength to communicate to an optical line terminal (OLT) in a base station, the communication becomes point-to-point and independent even though the physical construction of the fibers is shared between ONUs and OLT. However, in WDM-PON, the wavelength assignment of the arrayed waveguide grating (AWG) for multiplexing/demultiplexing wavelength channels is fixed, as is the network configuration, and this makes it hard to achieve cell configuration flexible enough to handle differences in radio conditions.
To address this issue, we studied the idea of applying dynamic time and wavelength division multiplexed (TWDM)-PON technology [6], which makes use of both TDM and WDM technology, to a mobile RAN [7]. The use of TWDM provides flexibility in network configuration and enables dynamic control of broadband WDM communication [8]. By using this flexibility and dynamic control, we can easily construct various cell structures and coordinate network activity between the wireless base stations.
Applying dynamic TWDM-PON to RAN, the upstream latency and packet jitter in the current TDM-based PON system are critical issue. However, wireless scheduling based PON scheduling we are developing in other study reduces PON equipment latency and jitter and make it possible to apply TDM to RAN.
In this paper, we confirm the effect of applying dynamic TWDM-PON technology to a mobile RAN. In section 2, we describe the technology’s basic functions for the RAN and some examples of network scenarios. In section 3, we demonstrate the feasibility of applying the technology to the RAN by two numerical simulations that compared a dynamic TWDM-PON configuration with a static PON configuration. In section IV, we summarize the main points of the paper.
2. Dynamic TWDM-PON
2.1 Basic functions
Figure 1 shows the schematic of a RAN using dynamic TWDM-PON. Baseband units (BBUs) are the signal processors and remote radio heads (RRHs) compose the radio frequency part of the RAN. First, we give an overview of the stream of signals in the PON. When downstream signals go from the OLT to the ONUs, the lasers in the line cards (LCs) inside the OLT output the optical signals. The signals are multiplexed in the n-by-m splitter or the n-by-m passive wavelength router (which is assumed to be a cyclic AWG) [9,10], where n is the number of LCs and m is the number of feeder fibers output from the OLT. In this section we call the splitter (or the wavelength router) “the first splitter.” It should be noted that the first splitter ensures that each ONU can access any BBU without the need to put any active switches between the BBUs and the LCs; this simplifies the overall system configuration. Themultiplexed signals in the feeder fibers are distributed in the second splitters, which are near to the ONUs. (The second splitters can be set outside of the base station where the OLT is set.) The second splitters can be located flexibly to meet the network configuration requirements. The signals are received in the ONUs through the second splitters, which distribute the signals. When upstream signals go from the ONUs to the OLT, they are multiplexed in the second splitters and reach the first splitter in the OLT. The signals are distributed according to their wavelengths at this point and thus reach their corresponding LCs.
Figure 2 shows an example of the structures of the LCs and the ONUs in the configuration enabled by the use of dynamic TWDM-PON when the splitter in the OLT is used as the first splitter. The part denoted as A is the circulator when the wavelength band of the upstream and downstream signals is the same. When it is different, A is the band filter whose output ports change according to the wavelength band of the input signal. The reason the LCs and ONUs have the same structure is because, in dynamic TWDM-PON, the LCs and ONUs work in the same way as the WDM devices. The lasers in the LCs, as well as the ONUs, are wavelength tunable and communicate with each other in corresponding wavelength sets. When using the splitter rather than the wavelength router as the first splitter, wavelength allocation becomes relatively simple as follows. The wavelength setting procedure for the ONUs is divided into groups in accordance with the number of LCs (for example, there are three groups when there are three LCs). The transmission wavelength of the wavelength tunable filter in each group’s ONU and that of the LCs corresponding to the ONUs are set to be the same to achieve communication.
If the wavelength router (which, as mentioned above, is assumed to be a cyclic AWG) instead of the splitter is used as the first splitter, the wavelength tunable filter in Fig. 2 can be omitted from the OLT. The loss of the cyclic AWG is typically 3 to 6 dB, which is much lower than that of the splitter, especially when the number of ports is large (e.g., 8). On the other hand, with the cyclic AWG the wavelength setting procedure becomes relatively complex; we must consider the cyclic characteristics of the AWG in allocating appropriate wavelengths to connect each LC to each feeder fiber [10]. For simplicity, we focus on the use of the splitter rather than the wavelength router in the following discussion.
In each wavelength, a group comprising one LC and a number of ONUs using the same wavelength communicates as one independent PON because the wavelengths of the other groups are different and the networks of the groups are independent from each other. Thissetup enables the group of LCs to constitute new PONs by wavelength division multiplexing. We call a PON constructed by this WDM a “virtual PON” and one physically constructed by the splitter a “fixed PON”.
Figure 3 shows the relation between wavelengths and time in virtual PONs. All the ONUs and LCs are divided into groups and construct virtual PONs; the number of PONs is the same as the number of wavelengths so that the communication bandwidth in each group is in the same range. The number of wavelengths, which is the number of virtual PONs, is decided so as to be able to use the maximum communication bandwidth in each virtual PON according to the number of LCs. First, the virtual PONs 1 ~n in the figure are constructed. In each virtual PON, the LCs and the ONUs communicate through TDM in the same way as in a conventional TDM-PON at wavelength λ1 ~λn. Since the wavelength differs for each group, the filter breaks communication for a different group and the communication is discarded automatically. When the bandwidths of the virtual PONs change because of a change in the mobile terminal number and the communication bandwidth, the combination of ONUs and LCs also changes so that, by changing the wavelength of the ONUs and LCs, the bandwidths of the new virtual PONs are in the same range. First, the virtual PONs 1 ~n continue for time tA, during which time the virtual PONs 1’ ~n’ are constructed from a combination of new ONUs and LCs; they also use the wavelength λ1 ~λn. The new virtual PONs are constructed so that they use the same bandwidths as the initial virtual PONs. This reconstruction of virtual PONs occurs in time tA + tB and also in tA + tB + tC when the bandwidth of each virtual PON changes and bandwidth bias occurs. Changes to the network’s communication bandwidths and wavelength control are conducted in the controller of the OLT. The controller outputs signals that specify which wavelengths should be used in the LCs and in the ONUs. The wavelength tunable filters of the LCs and ONUs use the signals to change the transparent wavelengths. This procedure enables the network configuration to be changed flexibly and dynamically, since physical considerations such as the splitter connection to the ONUs can be ignored.
Using TWDM-PON makes network very flexible. Additionally, using wavelength variable filter in ONUs and OLTs makes all branch point cheap passive power splitter and the construction of PON very flexible. For example, in traditional PON spec the branch point is only one in out of doors. However, dynamic TWDM-PON don’t confine network to one branch network and multistage branches such as Fig. 1 can be constructed. Especially, because the traffic fluctuation in RAN is very fast, the flexible network construction in wide area network by multistage branches is very useful. In this way, dynamic TWDM-PON can be used without limitation to the standard spec in RAN and it is also required by the character of mobile traffic.
Here, we explain the mobile RAN to which dynamic TWDM-PON was applied. In the mobile RAN of LTE, the network consists of two parts [11]. One is the mobile backhaul (MBH), which is the network between the wireless access base stations and the core network. The other is the mobile fronthaul (MFH), which is the network between the wireless access base stations and the user terminals. The S1 and X2 interfaces have been studied for use as the MBH network interface [12]. In the MFH, the CPRI (Common Public Radio Interface) [13] link is widely used as the network interface between the BBUs and the RRHs of a radio base station. In all three interfaces, the effects of assigning a wavelength dynamically are not very significant since the signal bandwidth is not proportional to the number of terminals or the amount of traffic. However, since a required bandwidth is proportional to the number of terminals or the amount of traffic if the data compression technique [14] is used, it is possible to achieve a flexible bandwidth setup based on traffic. Considering future traffic increasing of RAN, the bandwidth constant spec such as the three interfaces will not be able to respond to the increasing. Therefore, the bandwidth variable network by other spec or by using data compression technique [14] will be used in RAN future.
It is very difficult applying TWDM-PON with ordinary DBA in current TDM-based PON system to MFH because of upstream latency and packet jitter requirement of CPRI. In the conventional DBA scheme, the upstream latency in the PON system is about 1 ms and the packet jitter is about 500 μs, and it is too large for MFH requirement (lower than several hundred μs). To resolve these issue, we are developing wireless scheduling based PON scheduling that can drastically reduce PON equipment latency (< 50 μs) and jitter (several tens μs), and the packet jitter can be absorbed by a proper buffer [15]. We recognize that the latency of PON system is the latency of optical transmission line, and this latency is not limited by CPRI equipment latency of 2.5 μs. In concerning to the clock jitter, PON based synchronization technique can provide highly accurate clock from ONU. Therefore, TWDM-PON is quite applicable to MFH in RAN.
Next, we describe the functions of dynamic TWDM-PON in applying it to the mobile RAN. Since a wavelength tunable filter is used, a wavelength and a combination of ONUs can be flexibly combined with each virtual PON. For example, when the communication bandwidth usage of a certain PON has been tight, by changing the communication wavelength of an ONU that uses a large amount of bandwidth in the PON, the communication can be switched to another virtual PON that has a margin bandwidth, thus enabling an efficient bandwidth setup. When there is a margin in the communication bandwidth of all the PONs, it is also possible to switch the LC of an OLT that does not need to be used into sleep state, thus saving power. When only one ONU’s traffic is sharply increasing, it is possible to assign only that ONU to a virtual PON and to make it communicate with one LC in the OLT not as a PON but as an SS (Single Star) configuration. In this case, it is distinguished from other PONs as simple WDM rather than TDM. If two or more LCs are installed in the ONU at a time when the bandwidth of other LCs in OLTs is small, it is also possible to assign that ONU one more wavelength, that of a surplus LC in the OLT, thus achieving twice as wide a bandwidth.
2.2 Example network scenarios
Here we explain examples of typical network scenarios in TWDM-PON. Figure 4 shows three network scenarios.
Example 1; In Fig. 4, LC1 and part A show a network scenario in which the total communication bandwidth is small, about 10 Gbps or less, with uniformity and one wavelength per LC in the OLT. In this scenario, conventional TDMA control can be conducted. This example shows control being conducted over only one wavelength, for the components whose blocks are filled in with dots. Unused LCs can be put in sleep mode to save electricity.
Example 2; In Fig. 4, LC1 and LC2 and part B show a network scenario in which the total communication bandwidth is large, about 10 or more Gbps, with uniformity, and the ONUs are divided into parts and each part communicates over each wavelength. This example shows communication over two wavelengths, for the components whose blocks are filled in with dots and slashes. Again, the unused LC can be put in sleep mode to save electricity. When the bandwidth becomes larger, more LCs are needed and so the unused LC is taken out of sleep mode.
Example 3; In Fig. 4, LC3 and part C show a network scenario in which there is large non-uniformity bias in the RRH communication bandwidths. When one RRH has very large bandwidth, one LC is assigned to that RRH and the others are controlled by other LCs in the OLT. In this example the RRH filled in with dots has large communication bandwidth and the LC filled in with dots is assigned to it. The other RRHs are controlled by the LC filled in with slashes. The number of LCs assigned to one RRH is set by the service menu of the user in the RRH.
3. Numerical simulation
To confirm the validity of our proposed use of dynamic TWDM-PON, we numerically simulated the available bandwidth for the case where TWDM-PON is configured dynamically and for the case when the configuration is fixed (static).
3.1 Simulation for varying number of mobile terminals
As an example, three LCs in an OLT-based PON configuration were simulated for the case where the bandwidth of one LC and ONU is 10 Gbps. Since there are three LCs, the number of physical PONs is set to three. The user number ratio for the three PONs is set to 7:2:1 (in order of urban, residential, and rural area users) by assuming a varying number of mobile terminals. We simulated the bandwidth assigned to each PON and the bandwidth used for the case when the number of users increased in accordance with this ratio. The use bandwidth for a single user is set to 100 Mbps. These network parameters are summarized in Table 1.
Table 1. Network parameters in numerical simulations.
In static configuration, the bandwidth assigned to one PON easily becomes 10 Gbps by using an LC in the OLT. Therefore, when the total number of users becomes about 143 (143x7/10~100), the number of urban area users exceeds 100 and the required bandwidth exceeds 10 Gbps.
In dynamic configuration, two LCs can constitute one virtual PON for an increase in urban area users. Conversely, rural and residential areas are set to one PON. In this case, when the total number of users becomes about 286 (286x7/10~200), the use of two LCs causes the required bandwidth to exceed the assigned bandwidth by 20 Gbps. Figure 5 compares the bandwidth demand vs. supply rate for a static PON and a virtual PON (each with two LCs) for urban area users when the user number randomly increases at the abovementioned ratio (7:2:1). As the figure shows, the rate starts decreasing when the number of users reaches 140 for the static PON and 281 for the virtual PON. This indicates that a dynamic PON is better able to deal with changes to the larger user numbers than a static PON.
Next, we simulated a case for a different user number ratio for an increase to three PONs, assuming variation in the number of mobile terminals. We set the assumed ratio at x:2:1. Whatever x is, the user number at the rate decrease points shown in Fig. 5 is twice that of the fixed PON case because there are two LCs for the virtual PON and one for the fixed PON.
Table 2 shows the number of users at the rate decrease point for several increasing ratios x. When x is large, the number becomes small because the number of users in residential and rural areas is very small when x is very large. In every case, the ratio was about 2:1 for the fixed and virtual PONs. We also confirmed that the larger the user number is, the better the virtual PON using the dynamic TWDM-PON can deal with the increase.
Table 2. Number of users at the rate decrease point for several increasing ratios x.
3.2 Simulation for bandwidth fluctuation
Next, we calculated the bandwidth fluctuation when the number of mobile terminals fluctuates in both the fixed PON and virtual PON cases.
The assumed network parameters are identical to those for the simulation described in 3-1 (Table 1). Three LCs are used and the bandwidth of one LC and ONU is 10 Gbps. Thenumber of physical PONs is set to three. The user number ratio for the three PONs is set to 7:2:1 (in order of urban, residential, and rural area users) by assuming a varying number of mobile terminals. The used bandwidth for a single user is set to 100 Mbps. In this simulation, the mean numbers of the mobile terminals in the three areas are fixed at 280, 80, and 40. These are the maximum numbers in a virtual PON producing the effects found in the 3-1 calculations and shown in Fig. 5. Incidentally, the number 280 means that the one ONU covers 35 mobile terminals when one physical PON consists of eight ONUs.
In this simulation, we calculated the bandwidth fluctuation when the mobile terminal numbers in the three areas fluctuate. As the random number distribution of the user mobile terminal number fluctuation, we used the Poisson distribution, which is the distribution used in telecommunications call models. Figure 6 shows the random fluctuation by Poisson distribution of the mobile terminal numbers in the three areas. The horizontal axis shows the number of events where the terminal number fluctuation occurs. The mean numbers for the three areas are assumed to be 280, 80, and 40. Figure 7 shows the mobile terminal number distribution in the three areas. Because we used random numbers from the Poisson distribution, the average number and variance were almost the same which is the feature of the Poisson distribution. Figure 8 shows the bandwidth used by one mobile terminal. Becausein the fixed PON case one LC is fixed to one physical PON, the 10 Gbps is shared by the terminals in each area. Therefore, in this case the bandwidths are calculated by dividing 10 Gbps by the number of terminals in the area. In the virtual PON case, the terminal number variation is uniformized by the WDM function. Each LC can communicate with the same number of terminals by assigning the wavelength for the ONUs in three areas to three other parts not part of the physical PON construction so that the new three new parts have the same number of terminals. Therefore, because the terminal number variation is uniformized, to calculate the bandwidth we only have to use the total number of terminals in three areas and the total 30 Gbps bandwidth of three LCs. Figure 8 shows the bandwidth calculated this way for one terminal in the virtual PON case.
As the figure shows, in the fixed PON case the bandwidth is 100 Mbps for one terminal in a residential area ONU and a rural area ONU. Of course, this is because the total number of terminals in the areas is less than 100 (10 Gbps / 100MHz), which is the maximum number where one terminal can use 100 Mbps. In the fixed PON case the mean bandwidth of one terminal in an urban area is about 36 Mbps, i.e., 10 Gbps / 280. On the other hand, in the virtual PON case the mean bandwidth of one terminal is about 75 Mbps, i.e., 30 Gbps / 400. This means that the variation of one terminal bandwidth is uniformized by using TWDM-PON. Thus, urban area users can use a wider bandwidth than they can in the fixed PON case. In the calculations shown in Fig. 8, even if the bandwidth provided to one terminal islarger than 100 Mbps, one terminal can use only 100 Mbps. This is because there is a limit to the bandwidth a mobile terminal can use, which in this simulation is 100 Mbps. In this case, the unused bandwidth goes to waste. Figure 9 shows the unused bandwidth in a fixed PON and in a virtual PON. In the fixed PON case, the unused bandwidth is due to residential and rural area ONUs. In these two areas, the terminal numbers are 80 and 40, respectively. The mean unused bandwidth was 8 Gbps (2 × 10Gbps – (80 + 40) × 100 Mbps). On the other hand, in the virtual PON case, the unused bandwidth is eliminated completely by using TWDM-PON to uniformize the unused bandwidth for all users in the three areas. Thus, we find that the utilization rate of three physical PONs is substantially increased by using TWDM-PON.
4. Conclusion
We proposed the use of dynamic time and wavelength division multiplexed - passive optical network (TWDM-PON) technology to achieve flexibility in mobile radio access networks (RANs). Since the technology enables a virtual PON to be configured with a wavelength, it can be effectively applied to a RAN in which the communication situation changes in real time and enable effective use of bandwidth.
We confirmed the feasibility of applying dynamic TWDM-PON to a mobile RAN by numerical simulation for three areas. It was found that when the user number ratio of the three areas was 7:2:1, the dynamic TWDM-PON can accommodate twice as many user terminals as the fixed PON. Additionally, even when the bias rate was very large, the dynamic TWDM-PON can deal with the traffic bias in the area. Moreover, even when there was fluctuation of the communication traffic in the areas that were unique to a mobile RAN, the dynamic TWDM-PON uniformizes the bias of the traffic, completely eliminates the unused bandwidth in the areas, and greatly increases the usage efficiency of the bandwidth.
Since dynamic TWDM can accommodate changes in network composition, it has promise for helping to achieve more efficient network construction and operation.
References and links
1. 3GPP RAN Workshop on Release12, June 2012, http://www.3gpp.org/Future-Radio-in-3GPP-300-attend.
2. K. Etemad and M. Baker, “Evolution of 3GPP LTE in release 11 and beyond,” IEEE Commun. Mag. 51(2), 73 (2013). [CrossRef]
3. C. Choi, Q. Wei, T. Biermann, and L. Scalia, “Mobile WDM Backhaul Access Networks with Physical Inter–Base-Station Links for Coordinated Multipoint Transmission/Reception Systems,” in Proceedings of GLOBECOM. 2011, pp.1–5.
4. IEEE 802.3ah, http://www.ieee802.org/3/ah.
5. S. Narikawa, H. Sanjoh, N. Sakurai, K. Kumozaki, and T. Imai, “Coherent WDM-PON using directly modulated local laser for simple heterodyne transceiver,” in Proceeding of ECOC. 2005, pp. 449–450. [CrossRef]
6. ITU-T SG15 standard G.989.1 (general requirement) and G.989.2 (PMD specifications).
7. D. Iida, S. Kuwano, J. Kani, and J. Terada, “A proposal of dynamic TWDM-PON for mobile radio access network,” APMP2013, TuA-3.
8. S. Kimura, “WDM/TDM-PON technologies for future flexible optical access networks,” in Proceedings of OECC. 23GPP010, pp.14–15.
9. J. Kani, “Enabling technologies for future scalable and flexible WDM-PON and WDM/TDM-PON systems,” IEEE J. Sel. Top. Quantum Electron. 16(5), 1290–1297 (2010). [CrossRef]
10. J. Kani, “Power saving techniques and mechanisms for optical access networks systems,” J. Lightwave Technol. 31(4), 563–570 (2013). [CrossRef]
11. 3GPP TS 36.300 v.11.3.0, Sep. 2012.
12. 3GPP TS 36.420 v10.2.0, Sep. 2011.
13. CPRI, http://www.cpri.info/spec.html.
14. N. Shibata, S. Kuwano, J. Terada, and N. Yoshimoto, “Data bandwidth reduction based on wireless resource allocation for digitized radio over TDM-PON system,” OFC2013, OTh4A.6. [CrossRef]
15. G. Yazawa, T. Tashiro, S. Kuwano, J. Terada, and N. Yoshimoto, “Low-latency transmission technique for mobile fronthaul based on TDM-PON,” IEICE Society Conf. 2013, B-8–38 (in Japanese).
