Timing measurement and simulation of activation process in GPON networks

Optica Applicata, Vol. XLV, No. 4, 2015 DOI: 10.5277/oa150403 Timing measurement and simulation of the activation process in gigabit passive optical networks TOMAS HORVATH1, PETR MUNSTER1, MICHAL JURCIK2, LUKAS KOCI1, MILOSLAV FILKA1 1Brno University of Technology, Faculty of Elecrical Engineering and Communication, Department of Telecommunications, Technicka 12, 616 00, Brno, Czech Republic 2Brno University of Technology, Faculty of Elecrical Engineering and Communication, Department of Physics, Technicka 8, 616 00, Brno, Czech Republic *Corresponding author: [email protected] The passive optical networks are widely used nowadays. Service providers have many customers in their distribution networks. The most important thing for communication between the end unit and the control unit is an establishment contact. Design and measurement of the activation process between optical network unit and optical line termination is presented. In general, the service providers have a big split ratio (up to 1:128) due to the connection eminent value of the customers per optical line termination port in chassis. We present the simulation of the connection process for 16, 32, 64, and 128 optical network units and the measurement for single optical network unit (the GPON Xpert is able to read a single connection process). We compare our results in simulation discussion. Keywords: gigabit passive optical network (GPON), GPON transmission convergence (TC) layer measurement, transmission convergence (TC) layer simulation, Matlab, GPON timing. 1. Introduction The bandwidth requirements are increasing every year from 20% to 30% due to the implementation of new services [1]. Nowadays, passive optical networks are widely developed. That is the reason why Internet services provider can increase the bandwidth for each customer. The connecting type according to fiber to the home/building (FTTH/B) dominated in Lithuania (from all European countries) in 2014 [2]. The situation is the same as foregoing in 2015. In general, FTTH Council does not take into consideration the technology for the customers. In Europe, gigabit passive optical network (GPON) is dominating, Ethernet passive optical network (EPON) is dominating in Asian access networks [3]. As was mentioned before, GPON is dominating for Europe networks thus we deal only with this technology. The description of GPON technology, timing relationship between optical line termination (OLT) and optical 460 T. HORVATH et al. network units (ONUs), related works, simulation models, and measurement results are discussed in the following sections. 2. Related works In recent years, many works related to GPON technology have been published. Works published up to date focus on a physical layer of GPON technology, especially line coding, dynamic bandwidth allocation (DBA) algorithm, end-to-end quality of services (QoS) and framing in a higher layer. SMITH et al. [4] presented the packet delay variance and bandwidth allocation algorithms for extended-reach GPON. They proposed two algorithms for an extended distance optical distribution network (ODN). In general, the paper does not contain the timing scheduler for ONUs. KYEONG-HWAN DOO et al. [5] dealt with a design of the retimed process for the long reach GPONs by an extender. The extender has been made with a field programmable gate array (FPGA). This solution probably could not be widely used due to a high price of the FPGA. On the other hand, the authors were able to use a split ratio up to 512 for a length 60 km of the distribution network. SALIOU et al. [6] analysed an optical transport network based on time division multiplex (TDM) to aggregate multiple protocols (Ethernet, business services, node B, and GPON traffic). They needed to solve an issue with various traffic formats, especially for the GPONs where the traffic is transmitted via bursts. In general, the data sources were combined with the TDM technique into 10 Gbit/s stream. Further, STEPNIAK et al. [7] introduced the bandwidth analysis of multimode fiber based PONs. They compared three architectures of the multimode PON numerically and experimentally. In general, the most important part of the PON is a splitter which may improve or decrease the bandwidth. LEE et al. [8] dealt in with the extended reach GPONs with distributed Raman amplifiers. They were able to reach a distance of the distribution network up to 60 km by remote pumping. On the other hand, they had this distance limit because the distribution network length is limited by ITU. The main disadvantage of this solution is that the distribution network contains the active elements. In our paper [9] we provided the simulation of the transmission convergence layer in the next generation PONs. We used the Matlab software for the simulation of 10 Gbit network numerical model. We simulated an equalization delay and influence of the refractive index on the timing. MERCIAN et al. [10] dealt with report message scheduling in 1G/10G EPONs and GPONs, especially in upstream direction. Downstream direction was transferred by the broadcast method, therefore it was not necessary to solve this direction. The upstream direction in EPON and GPON is attended in the different mode. In general, it is not possible to compare these networks, because GPONs do not have carrier sense multiple access (CSMA) access method. Timing measurement and simulation of the activation process... 461 Further, ALSHAER and ALYAFEI [11] introduced an end-to-end QoS scheme for GPON. They used the dynamic weighted fair queuing for prioritizing of selected traffic. The authors had a solution for end-to-end QoS with Report messages. YUANQIU LUO et al. [12] dealt with framing in GPON and NG-PON1. The authors compared the results for both standards. In general, they calculated only for framing but did not calculate all parameters in GPON and NG-PON1 networks. In both standards we need to count with round trip time (RTT), OLT response time, and relationship between OLT and ONU in a registering state. We analyse the GPON technology and timing relationship between OLT and ONU in the next section. 3. Gigabit passive optical networks (GPONs) GPON is a standard that offered sufficient bandwidth in the past. This standard was approved by ITU-T and has one important difference in comparison with older standards because GPON is able to use asynchronous transfer mode (ATM) and/or Ethernet protocols for data transmission. The general requirements of GPON are described in ITU-T G.984.1. In this document we can find the following information: specification of a split ratio, power budget, etc. It can be seen from Fig. 1 that the times for a propagation signal from OLT to ONUs are different. This is the main point of our research. We focused on timing relation for control and end units for different split ratios. In other words, the initialization connection between these units is most important for the first establishing connection. Section 4 deals with connection establishment between OLT and first ONUs connected to the network. OLT Last mile ONU1 t1 Splitter 1:64 ODN t2 tj ONU2 ONU3 ONUj t3 Fig. 1. The basic topology for GPON. OLT – optical line termination, ONU – optical network unit, ODN – optical distribution network, t – various times of propagation signal. 462 T. HORVATH et al. In general, the most important physical layer requirements for our design and measurements are the maximum length of the distribution network of 20 km, split ratio at least 1:64 or higher, and attenuation classes. For the GPON three attenuation plans were proposed: 5–20 dB, 10–25 dB, and 15–30 dB [13]. The limited length of the distribution network can be defined by the following equation: [km] (1) where Ac is the selected attenuation class [dB], Fl is the total loss of fiber in ODN [dB], Sloss is the splitter loss [dB], n represented the number of connectors, and Fa is the attenuation of used fiber [dB]. 4. Communication between ONU and OLT units The communication between ONU and OLT units is at first initialized by connecting ONU into the network. The OLT unit consecutively sends the downstream frames, which are exactly 125 μs long and contain physical synchronization block downstream (PSBd) header with Psync field of exact value used for ONU timing synchronization. The ONU, which is connected for the first time, needs to pass over the five states. The first state is the initial state (O1). In other words, the ONU needs to receive at least two frames with the same Psync value. In general, the initial state should be divided into three steps. At first, the ONU is in the hunt state, which means that the ONU is passive and receives the frames from the OLT. The parameters loss of signal/loss of frame (LOS/LOF) are set to 1. The ONU reads the incoming frames bit-by-bit and searches for the Psync field value. When this value is found, the ONU passes to the Pre-Sync state. At the same time the ONU sets the counter N1 for correct incoming frames to 1 and waits for 125 μs, where the subsequent frame with the same Psync field value should occur. When another frame with correct Psync field is received, the counter N1 is incremented and compared with the value M1 which sets the number of subsequent correct frames to the transition of ONU into the synchronized state known as standby state (O2). The parameters LOS/LOF are cleared. The ONU is now synchronized with the OLT in downstream direction but the subscriber still cannot transmit the data due to missing upstream synchronization. To obtain the upstream synchronization, the OLT unit periodically transfers the broadcast PLOAM message Upstream Overhead (PLOAM – physical layer operations, administrations and maintenance) which let the ONU configure the burst overhead fields for the subsequent upstream transmissions. The message is transferred three times and after the reception of at least one Upstream Overhead PLOAM message, the ONU sets the network parameters and move to serial number state (O3). The OLT unit then waits for 750 μs for ONU to process the message. lAcFl –Sloss –0.5n– Fa -------------------------------------------------------- = Timing measurement and simulation of the activation process... 463 In the O3 state, the ONU starts TO1 (timer 1) with default value of 10 s where ONU waits for Assign ONU-ID message from the OLT. Meanwhile the OLT constructs a quiet window of 250 μs length by sending a broadcast frame with empty BWmap (in this configuration without the pre-assigned delay), which stops upstream traffic for all ONUs in the operation state to avoid collisions. The next frame from the OLT contains a Serial Number Request message addressed to Alloc-ID 0xFE (which is used for activation) with 13 byte grant (to send only a PLOAM message Serial Number ONU) and start time of 77 μs. After message reception on the ONU side, it waits for a locally random delay (0–48 μs) and when the start time is reached it sends a Serial Number ONU message. This message also contains the generated random delay. Note that prior to sending of the Serial Number ONU message, the ONU has to construct and send the PLOu (physical layer overhead upstream) header immediately before the Serial Number ONU message. While the OLT receives the Serial Number ONU messages from ONUs, it normally operates in the downstream direction for the next 2 frames. After that the OLT sends consecutively three Assign ONU-ID messages with a serial number obtained from the Serial Number ONU message from one ONU (the ONU is selected by OLT from the first incoming Serial Number ONU message). The serial number is a unique parameter for each ONU which identifies ONUs on the provider side. The Assign ONU-ID message sets the ONU-ID for direct addressing in the network and has to be received before the timer TO1 expires. Normally the timer TO1 expires only when more than 10 ONUs are trying to connect the line at the same time because the Assign ONU-ID message is applicable just for one ONU. If this message is sent periodically every 1 second, it allows 10 ONUs to be connected in 10 s and all the others will move back to standby state (O2). After targeted reception of at least one Assign ONU-ID message through a serial number, the ONU moves to ranging state (O4). The ranging state (O4) is almost the same as serial number state (O3). The difference is in Range Request message which is sent directly to the specified ONU and in the absence of random delay because no collision can occur when only one ONU uses the line. The state O4 is the most important due to the setup of the equalization delay. The equalization delay synchronizes the communication of all ONUs in upstream direction. In other words, each ONU is in the unique distance from the OLT. With the equalization delay, the OLT keeps the synchronization of the time slots for the ONUs. The control unit can calculate the time equalization delay (Teqd) by the following equation [14]: (2) where RspTimei is the response time (μs), EqDi is the estimation of the equalization delay for the fiber distance from the previous formula, n1310 represents the group velocity refractive index for 1310 nm in the ODN, n1490 represents the group velocity Teqd T1490 i,RspTimeiEqDiT1310 i, +++ T1490 i, n1310 n1490 + n1490 ------------------------------------ RspTimeiEqDi ++ == = 464 T. HORVATH et al. refractive index for 1490 nm in the ODN. The fraction with group velocities can be called the index correction factor. It can be expressed as [14]: (3) The following equation defines the receiving time such as sum of the sending time and group velocity refractive index [14]: (4) When we institute Eq. (2) and Eq. (3) into Eq. (4), the representation of the receiving time of the actual time instance can be obtained when GTC frame N is delivered to ONUi [14]: (5) The index factor should be written out with the following equation [14]: (6) For the partial derivatives with the variations of n and Δn, we have the following equations [14]: (7a) (7b) The above mentioned Eqs. (7) can be simplified as n is about 3 orders larger than Δn [14]. The previous expression is much smaller than the second one, and can be neglected. The second expression states that small changes in Δn will be translated into small changes of the index factor with proportion 1/4n. T1490 i,Teqd RspTimei –EqD i –() n1490 n1310 n+1490 ----------------------------------- = TrecvNi,TsendNi,T1490 i, += TrecvNi,TsendNi,Teqd n1490 n1310 n+1490 ----------------------------------- OLT EqDiRspTimei +() n1490 n1310 n+1490 ----------------------------------- ONU – ++= n1490 n1310 n+1490 ----------------------------------- n1490 2n1490 n1310 n1490 –()+ -------------------------------------------------------------- n 2nnΔ+ ------------------------ == =2n2nnΔ– 4n2 ------------------------------ ≈1 2 -------- nΔ 4n ------------ –= ∂ ∂n -----------1 2 -------Δn 4n ----------- –   +Δn 4n2 -------------- = ∂y ∂Δn ---------------1 2 -------Δn 4n ----------- –   1 4n ----------- –= Timing measurement and simulation of the activation process... 465 After receiving the Serial Number ONU message, the OLT side sends a Ranging Time message which contains the equalization delay for targeted ONU. This message is sent 3 times. Once the equalization delay has been negotiated, the ONU moves into the operation state (O5). The operation state is the final state for the ONU to transmit the data from the user to the network. The review of the mentioned states is shown in Fig. 2. 5. Simulation results The variable nONU, which indicates a number of ONUs connecting to the network, is set to 16 from a possible range 1 to 128. Constant fd means the frame duration which is exactly 125 μs. The timer TO1 is set to 10 seconds. Variable curTime is used as the real time counter and is set to 0 and the variable delay adds a time when the counter curTime overflows timer TO1. Variable ONUs holds an array of the ONU classes which contains Distance and Eqd attributes. The frame flow shown in Fig. 2 repeats every second. This setup also expects no pre-assigned delay used in an Upstream Overhead message from OLT and that the ONU will not send any additional data after aSerial Number ONU message. The value of M1 in hunt state (O1) is set to 2. This means that only two consecutive frames with the correct Psync field value are needed for ONU to enter standby state (O2). We selected the attenuation class B (10–25 dB). The simulation generates a random distance from OLT for all ONUs from the range 0 to 20 km. The next step calculates the propagation delay and adds the random delay from the range 0 to 48 μs for each ONU. The OLT holds the Serial Number ONU messages from all ONUs but due to the OLT 125 125 750 125 125 125 750 0 Psync Psync OLT normal 125 125 125 111 236 250 14 750...125 125 125 ... 750 236 202 62 125 125 125 750 125 125 125 63 State O4 t [μs] State O1 State O2 State O3 operation state Upstream overhead Turn ON ONU time OLT normal operation state BWmap empty QW Assigned ONU-ID Range req PLOAM + SN req + SN ONU OLT normal operation state QW Ranging time OLT normal operation state + SN ONU Rnd = 0 PLOAM BWmap empty Fig. 2. Details of the ONU connecting into the passive network. 466 T. HORVATH et al. configuration, it first responses by Assign ONU-ID message to ONU where the first Serial Number ONU message was received from. All other ONUs have to wait until the next Assign ONU-ID message with their serial number occurs. As shown in Fig. 2, this will happen right after 1 second so the variable curTime will be incremented by 1. The variable curTime is in each step compared with the value of the timer TO1 and if curTime overflows TO1 all unhandled ONUs will move back to standby state (O2) and the variable delay will be incremented by 1 because the next Assign ONU-ID message will not have a recipient. The equation for calculating activation time is: (8) where ceil is the rounding function in Matlab, MRTD is the maximum round trip delay, QW is the quiet window length, fd is the frame duration, delay is the counter of TO1 expirations and i is the cycle counter. The simulations have been done by Mathworks Matlab tools and measurement with GPON Xpert professional device, which is designed for an analysis of the messages, frames, etc., at the transmission convergence layer and it has been developed by Tracespan. In our research we have two scenarios for the simulations and one for the measurement. The simulation scenarios use Eq. (8) and the process depicted in Fig. 2 for the calculation of the total times of each new connecting ONU to the PONs. Each ONU starts in the initial state (O1) and passes into the operational state (O5). The timer TO1 has the default value of 10 s and the distance of each ONU is generated from 0 to 20 km. The final results for 16 and 32 ONUs are depicted in Fig. 3 and for 64/128 ONUs – in Fig. 4. The maximum value of connecting activation time is around 32 s. The values seem random but they are not in order because the ONUs in the real ta35 2 ceil MRTD QW+ fd ---------------------------------------   +delayi1–()++= 16 ONUs 32 ONUs 35 30 25 20 15 10 5 0 0 5 10 15 20 25 30 35 ONU Time [s] Fig. 3. Activation timing for 16/32 ONUs in the gigabit PON. Timing measurement and simulation of the activation process... 467 networks are in various distances from the OLT. When the timing is cogitated in the blackout, then the farthest ONU will be connected after 32 s. On the other hand, the real network has around thousands ONUs connected to the OLT. Currently, a higher number of CPU (central processor unit) is used for parallel calculation to reduce the activation timing. 6. Measurement results The measurement topology is depicted in Fig. 5. The main difference is in the total insertion loss of the network caused by a splitter 1:4. For laboratory tests 64 ONUs 64 ONUs 128 ONUs 140 120 100 80 60 40 20 0 0 20 40 60 80 100 120 140 ONU Time [s] Fig. 4. Activation timing for 64/128 ONUs in the gigabit PON. Fig. 5. The laboratory topology with GPON Xpert. OLT – optical line termination, ONU – optical network unit, ODN – optical distribution network, t – various times of propagation signal. OLT Last mile ONU1 t1 Splitter 1:4 ODN t2ONU2 ONU3 t3 GPON Xpert