Time Synchronization in XG-PON Systems: An Error Analysis

Panagiotis Sarigiannidis

, Georgios Papadimitriou

, Petros Nicopolitidis

, Emmanouel Varvarigos

and Malamati Louta

Department of Informatics and Telecommunications Engineering, University of Western Macedonia, Kozani, Greece

Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, Greece

Computer Technology Institute and Press ”Diophantus” N. Kazantzaki, University of Patras, Campus, Rio, 26500, Greece

Keywords:

Analysis, Passive Optical Networks, Synchronization, XG-PON.

Abstract:

Synchronization in modern optical systems is of paramount importance. The sharing of a common time be-

tween different network components constitutes a crucial factor towards system stability. Latest advancements

in optical access networks, such as Passive Optical Networks (PONs), allow transmission rates of 10 Gbps.

Hence, a very accurate synchronization method is required in order to keep the network free of blocking and

collisions. In this work, we focus on 10-gigabit-capable passive optical network (XG-PON) systems, one of

the most popular standards for PONs. A rigorous analytic approach is devised so as to investigate the re-

quired precision of underlying synchronization method. Closed equations for computing the no blocking and

blocking probabilities are developed. The devised analytic framework is veriﬁed using simulation methods.

1 INTRODUCTION

Passive Optical Networks (PONs) architecture ef-

fectively apply a cost-effective solution of optical

technology to address the bandwidth mismatch be-

tween the backbone and the last mile (Lam, 2011).

They present numerous inﬂuential assets: a) a cost-

effective, optical-based solution, without needing re-

generators, ampliﬁers, or active converters, b) a ﬂex-

ible infrastructure allowing scaling up at low cost,

c) broad deployment which means that a PON could

cover over 60 Km distance between the central of-

ﬁce (CO) and the most distant user, and d) various,

concurrent service level agreements (SLAs) enclos-

ing heterogeneous quality of service (QoS) guaran-

tees among numerous users withing the network (Lee

et al., 2006),(Sarigiannidis et al., 2015). Indeed, there

is a growing interest on optimizing the performance

of many inner components and processes of the PON

(Sarigiannidis et al., 2010).

Next generation PONs (NG-PONs) have been re-

cently inaugurated in order to further empower the

network capabilities (Kani et al., 2009). By delivering

shared Internet access up to 10 Gbps, the latest stan-

dard of the telecommunication standardization sector

of the international telecommunication union (ITU-

T), called 10-gigabit-capable passive optical network

(XG-PON), is envisioned to provide even more ca-

pabilities to even more users (ITU-T). Nevertheless,

having in mind that the number of users potentially

connected to a PON is growing, the responsibility

of delivering data following multiple user trafﬁc re-

quirements becomes more pressing. Data delivery

should be carried out without violating QoS agree-

ments, while user requests should be met in time ac-

cording to their Service Level Agreements (SLAs).

Accordingly, users that share the same SLA should

experience the same level of services. This implies

that the network performance is efﬁcient enough to

provide good and fair data delivery to all users.

However, such a demanding network system re-

quires accurate synchronization mechanisms. The

Time of Day (ToD) distribution over the system

should be accurate enough in order to avoid collisions

between bandwidth allocations. Various techniques

are available to support time synchronization, such

as Global Positioning System (GPS), Network Time

Protocol (NTP), and Precision Time Protocol (PTP).

According to the standard speciﬁcation, the Optical

Line Termination (OLT), which is the main network

component in the XG-PON systems, is equipped with

an accurate real time clock; however, the means of

provisioning an accurate clock to the OLT is left be-

yond the scope of the standard recommendations.

The time synchronization in XG-PON systems is

the main focus of this paper. The ToD distribution

Sarigiannidis P., Papadimitriou G., Nicopolitidis P., Varvarigos E. and Louta M..

Time Synchronization in XG-PON Systems: An Error Analysis.

DOI: 10.5220/0005577600140021

In Proceedings of the 6th International Conference on Optical Communication Systems (OPTICS-2015), pages 14-21

ISBN: 978-989-758-116-8

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

method, according to the standard speciﬁcation, is

adopted. The Optical Network Units (ONUs) receive

the ToD through downstream frames that the OLT

send in a periodic manner. The ToD accuracy is cru-

cial since it dramatically affects the bandwidth alloca-

tion schedule integrity. For example, if two ONU re-

ceive different ToD, a bandwidth allocation collision

may be happened. This phenomenon may lead to net-

work performance degradation since the collided allo-

cations should be re-scheduled again in the forthcom-

ing allocation opportunities, resulting in considerable

delays. To this end, an analytic framework is devised

in this work in order to examine the blocking prob-

ability of a burst (upstream allocation that received

erroneous time from the OLT. Furthermore, the num-

ber of involved bursts is calculated under saturation

conditions. Analytic results are veriﬁed using a rig-

orous simulation environment. The obtained results

indicate the capability of various accuracy levels.

The remainder of the paper is organized as fol-

lows. Section 2 introduces several features of the XG-

PON system. In Section 3 existing research efforts

towards timing synchronization are outlined. A de-

tailed description of the proposed analytic approach

is provided in Section 4. Section 5 illustrates the ob-

tained results, followed by detailed reports. Finally,

conclusions are given in Section 6.

2 BACKGROUND

The XG-PON framework deﬁnes a point-to-

multipoint optical access infrastructure that provides

(nominal) 10 Gbps data rate in at least one direction.

According to the standard speciﬁcations (ITU-T),

the resource allocation process is designed to satisfy

speciﬁc downstream and upstream requirements,

while speciﬁc principles are met. Typically, the

XG-PON system has a physical tree topology with

the Central Ofﬁce (CO) located at the root and the

subscribers connected to the leaf nodes of the tree.

The main network entities of a PON are the ONUs,

providing to users connection to the network, and the

OLT, providing to ONUs access to the backbone. A

passive optical splitter/combiner connects the OLT

and ONUs, receiving a single optical ﬁber from

the OLT and distributing the incoming signal to

multiple single optical ﬁbers and vice versa. Figure 1

illustrates a typical XG-PON architecture.

The XG-PON transmission convergence (XGTC)

layer thoroughly describes the functional protocols

and procedures, including resource allocation and

QoS provisioning. Speciﬁcally, the XTGC layer is

responsible for receiving service data units (SDUs)

Central

Office

Passive

Splitter/

Combiner

VoIP

IPTV

Optical

Network

Unit

Optical Line

Terminal

Figure 1: A typical XG-PON architecture based on a tree

topology.

from the upper layers and providing an uninter-

rupted bitstream at the nominal interface, support-

ing a 9.95328 Gbps (nominal) data rate in the down-

stream direction and a 2.48832 Gbps (nominal) data

rate in the upstream direction. Downstream and up-

stream frames of the same duration (125 µsec) are de-

ﬁned; this corresponds to a downstream frame size of

155520 bytes and an upstream frame size of 38880

bytes. Both downstream and upstream frames con-

tain control information. Speciﬁcally, considering

the downstream frame, the physical synchronization

block ﬁeld (PSBd) includes a synchronization bit-

stream, the PON identiﬁcation number, counters and

other control information. Additionally, BWmap, an

important control ﬁeld comprised in XGTC header

that follows, is associated with the bandwidth allo-

cation process. Speciﬁcally, it is used by the OLT

to inform the ONUs about the granted transmission

opportunities; it deﬁnes the start time of the transmis-

sion opportunityand the grant size of each recipient of

the allocation within each ONU. In essence, the OLT

continuously broadcasts data to ONUs, including re-

quested data delivery, messages and bandwidth allo-

cation information. Considering the upstream frame,

the physical synchronization block ﬁeld (PSBu) con-

tains the preample and delimiter ﬁelds. Then the

XGTC burst follows that includes XGTC header and

trailer. Dynamic Bandwidth Report (DBRu) deter-

mines the adopted resource allocation process. Two

options are deﬁned by the standard, namely, a) the sta-

tus reporting (SR) method, according to which each

allocation encloses the DBRu header and reports to

the OLT its buffer status a d b) the trafﬁc monitoring

(TR) method, according to which the OLT monitors

the idle upstream frames to perceive the bandwidth

pattern of each recipient. According to the speciﬁ-

cations, the XG-PON OLT should support both tech-

niques in a separate way or even combined. An in-

TimeSynchronizationinXG-PONSystems:AnErrorAnalysis

terested reader could refer to (Effenberger, 2010) for

more information.

In order to provide an uninterrupted and effective

upstream schedule, the OLT should apply an accurate

time synchronization method to the connected ONUs

(Topliss et al., 1995). In essence, the OLT informs

the ONUs about the ToD using speciﬁc downstream

XGTC frames. The ONUs receive the ToD infor-

mation by reading the ONU Management and Con-

trol Interface (OMCI) channel, which allows OLT and

ONUs to exchanges control messages. The ToD dis-

tribution process implies a master-slave clock rela-

tionship, where the OLT maintains the master clock,

whereas the ONUs are synchronize their slave clocks

to the OLT’s master clock by monitoring certain

downstream XGTC frames which are used as timing

reference. In any case, a 64-bit guard time is consid-

ered between upstream bursts from different ONUs

to prevent allocation overlaps. Nonetheless, the mas-

ter clock of the OLT is considered accurate while the

clock implementation details are considered beyond

of the scope of the standard recommendations. In this

work, we try to shed light to the case that the master

clock is not fully accurate using analytic techniques.

In addition, the upstream allocation block probability

is calculated by considering multiple accurate clock

levels.

3 SYNCHRONIZATION

METHODS AND RELATED

WORK

Many time synchronization protocols could be used

in XG-PON systems. The NTP, deﬁned in IETF RFC

1305, is considered a clock synchronization proto-

col of mature technology. It has been used for many

years to ensure ToD sharing in distributed network de-

vices. The synchronization process employed times-

tamps which are piggybacked in data packets to ad-

vertise the relative time. The function of NTP al-

lows a time accuracy of milliseconds. For example,

it is used in synchronous optical networking to pro-

vide alarm, billing, control, and signaling messages.

However, the usage of NTP in high-speed optical ac-

cess networking seems problematic due to NTP’s low

accuracy levels.

On the other hand, PTP, also known as IEEE

1588 standard, allows higher accuracy time distribu-

tion than NTP. NTP version 1 has been employed to

provide time synchronization in Ethernet Local Area

Networks (LANs). It engages time-messaging con-

trol packets between master and slave clocks, where

the time information is carried using timestamps. It

also included several improvements such as higher

packer rate and hardware-based time-stamping. In

LAN environments, it supports one microsecond ac-

curacy. NTP version 2 has been used in Wireless Area

Networks (WANs). The concepts of boundary clock

and transparent clock are strong enhancements of this

version. The boundary clock uses slave ports for time

recovering from an upstream master; then it utilizes

this recovery time as a basis for a set of master ports

to synchronize downstream slave ports. The transpar-

ent clock is used as a measure device. It monitors the

time a message needs to be processed and updates the

clock. In PON environment, the usage of NTP seems

inadequate due to link delay asymmetry (Luo et al.,

2012).

The involvement of GPS systems in determining

the current time is not a new concept (Lewandowski

et al., 1993). Today, high precision GPS systems have

been used in distributed network systems for time

synchronization purposes. Recent efforts demon-

strated very effective GPS clock systems using short-

term frequency stability of crystal oscillator (Zhang

et al., 2013). In addition, advanced memory and pro-

cessing techniques are used in (Shan et al., 2014),

where an accurate GPS system for mobile communi-

cations was presented. Modern GPS systems support

about 100 nsec time precision; however the installa-

tion cost as well as the implementation is high.

Despite several research efforts towards time syn-

chronization in modern network systems, the syn-

chronizationissue has not been extensivelyaddressed,

especially in modern PONs. For example, there is

no evidence about the impact of a clock error in the

PON operation. To be more speciﬁc, it is important to

specify the impact of an clock error in the upstream

allocation process, i.e., how a timing error affects the

integrity of the allocation process in the upstream di-

rection in XG-PON systems. In this paper, we intend

to cover this gap by examining the synchronization

deﬁciencies in XG-PON systems.

4 ERROR ANALYSIS

4.1 Formulation

Let E(µsec) denotes the timing error of an ONU that

received erroneous time by the OLT. Assume that the

XG-PON under study consists of N ONUs. It is as-

sumed that the error distribution follows a normal dis-

tribution, having µ = 0 (Mar´oti et al., 2004), (Bregni

and Tavella, 1997):

E(µsec) → N(0, σ

) (1)

OPTICS2015-InternationalConferenceonOpticalCommunicationSystems

Due to the presence of the guard time between

upstream bursts, a minor timing error could be ab-

sorbed. Based on the ITU-T recommendations, the

upstream guard time is at least 64 bits. Given that

the upstream transmission rate is 2.48832 Gbps, this

guard time corresponds to GT = 0.257 µsec. Thus,

a timing error larger than −GT and lower than GT

can be absorbed. Given that Φ yields the Cumulative

Distribution Function (CDF) of the standard normal

distribution, the probability of error absorption, p

equal to:

= E(−GT ≤ X ≤ GT) =

Φ(

) − Φ(

−GT

) =

Φ(

) − Φ(−

) =

Φ(

) − (1− Φ(

)) =

Φ(

) − 1+ Φ(

) =

2Φ(

) − 1 =

2(1− Φ(

−GT

)) − 1 =

1− 2Φ(

−GT

)

(2)

In case that a timing error is larger than |GT|,

the error impact is inevitable. Hence, the probabil-

ity of violating the upstream transmission integrity is

1 − p

= 2Φ(

−GT

). Let UB

denotes the length of

Table 1: Notation Table.

E(µsec) ONU Timing Error

N Number of ONUs

GT Guard Time (64 bits)

Probability of Error Absorption

Length of the i

Upstream Burst

Single Blocking Discrete Probability

Distribution

Double Blocking Discrete Probability

Distribution

bursts

Average Blocking Bursts

No Blocking Probability

(Probability of Error Absorption)

Probability of a Single Blocking

Probability of a Double Blocking

bursts

Average Number of Discarded Bursts

the i

upstream burst in terms of Bytes. Also, as-

sume that the timing error affects the n

upstream

burst, within an upstream PHY frame of 125µsec.

The synchronization disorientation of this upstream

burst may harm one or two other upstream bursts.

For example, the n

upstream burst may collide with

the (n − 1)

burst with probability E(−GT ≤ X ≤

−2GT −UB

n−1

). In a similar way, the n

upstream

burst may collide with the (n+ 1)

burst with proba-

bility E(GT ≤ X ≤ 2GT +UB

n+1

). In particular, the

probability of the n

upstream burst to collide with

one of its neighbor bursts is as follows:

E(−2GT −UB

n−1

≤ X ≤ −GT)+

E(GT ≤ X ≤ 2GT +UB

n+1

) =

Φ(

−GT

) − Φ(

−2GT −UB

n−1

Φ(

2GT +UB

n+1

) − Φ(

) =

1− Φ(

) − (1− Φ(

2GT +UB

n−1

))+

Φ(

2GT +UB

n+1

) − Φ(

) =

Φ(

2GT +UB

n−1

) + Φ(

2GT +UB

n+1

)

−2Φ(

)

(3)

If the timing error of an erroneous burst is slightly

larger than the guard time plus the upstream length of

its neighbor burst, then the erroneous burst may col-

lide with two or more successive bursts. This depends

on the burst lengths. In order to simplify our analysis,

we consider identical burst lengths for all upstream

bursts:

UB = UB

= UB

, ∀i, j, 1 ≤ i, j ≤ N (4)

Under this assumption, the n

upstream burst

may collide with two other bursts at most. To be

more speciﬁc, the n

upstream burst collides with the

(n− 1)

and (n− 2)

upstream bursts when the tim-

ing error is in the range [−GT − 2UB, −2GT −UB].

Similarly, the n

upstream burst collides with the

(n + 1)

and (n + 2)

upstream bursts when the

timing error is in the range [2GT + UB, GT + 2UB].

At this end, the probability of the n

upstream burst

on colliding with its two consecutive neighbor bursts

is calculated as follows:

TimeSynchronizationinXG-PONSystems:AnErrorAnalysis

E(−GT − 2UB ≤ X ≤ −2GT −UB)+

E(2GT +UB ≤ X ≤ GT + 2UB) =

Φ(

−2GT −UB

) − Φ(

−GT − 2UB

Φ(

GT + 2UB

) − Φ(

2GT +UB

) =

1− Φ(

2GT +UB

) − (1− Φ(

GT + 2UB

))+

Φ(

GT + 2UB

) − Φ(

2GT +UB

) =

2(Φ(

GT + 2UB

) − Φ(

2GT +UB

))

(5)

In the same way, the probability of the n

up-

stream burst to collide with one of its neighbor bursts

(Eq. (3)) becomes:

E(−2GT −UB ≤ X ≤ −GT)+

E(GT ≤ X ≤ 2GT +UB) =

Φ(

−GT

) − Φ(

−2GT −UB

Φ(

2GT +UB

) − Φ(

) =

1− Φ(

) − (1− Φ(

2GT +UB

))+

Φ(

2GT +UB

) − Φ(

) =

2(Φ(

2GT +UB

) − Φ(

))

(6)

Obviously, the probability of the n

upstream burst to

collide with the (n+ 2)

burst will be:

E(GT + 2UB ≤ X ≤ 2GT + 3UB) (7)

Based on Eq. 6 and Eq. 7, it is easy to devise the recur-

sive probability distribution for the n

upstream burst

to collide with the (n± i)

, i ≥ 1:

i = 1, E(−2GT −UB ≤ X ≤ −GT)+

E(GT ≤ X ≤ 2GT +UB)+

i = 2, E(−2GT − 3UB ≤ X ≤ −GT − 2UB)+

E(GT + 2UB ≤ X ≤ 2GT + 3UB)+

i = 3, E(−2GT − 5UB ≤ X ≤ −GT − 4UB)+

E(GT + 4UB ≤ X ≤ 2GT + 5UB)+

. . . +

(8)

Based on Eq. 8, the discrete probability distribution

of occurring a single blocking, F

, when an upstream

burst received erroneous timer, is deﬁned:

(i) =

E(−2GT − (2(i− 1) + 1)UB ≤ X ≤

−GT − 2(i− 1)UB)+

E(GT + 2(i− 1)UB ≤ X ≤

2GT + (2(i− 1) + 1)UB) =

2(Φ(

2GT + (2(i− 1) + 1)UB

−Φ(

GT + 2(i− 1)UB

)), i = 1, 2, 3, ...

(9)

In accordance, the recursive probability for the n

upstream burst to collide with the (n ± i)

, i ≥ 1 and

(n± (i+ 1))

is given as:

i = 1, E(−GT − 2UB ≤ X ≤ −2GT −UB)+

E(2GT +UB ≤ X ≤ GT + 2UB)+

i = 2, E(−GT − 4UB ≤ X ≤ −2GT − 3UB)+

E(2GT + 3UB ≤ X ≤ GT + 4UB)+

i = 3, E(−GT − 6UB ≤ X ≤ −2GT − 5UB)+

E(2GT + 5UB ≤ X ≤ GT + 6UB)+

. . . +

(10)

The discrete probability distribution of occurring a

double blocking, F

, when an upstream burst re-

ceived erroneous timer, is given as follows:

(i) =

E(−GT − 2iUB ≤ X ≤

−2GT − (2(i− 1) + 1)UB)+

E(2GT + (2(i− 1 + 1))UB ≤ X ≤

GT + 2iUB) =

2(Φ(

GT + 2iUB

−Φ(

2GT(2(i− 1) + 1)UB

)),

i = 1, 2, 3, ...

(11)

4.1.1 Average Number of Blocking Bursts

The average blocking bursts, when a burst receives

erroneous time, is given as follows:

bursts

∑

i=0

(12)

The p

probability represents the probability of occur-

ring no blocking (i = 0), one blocking (i = 1), and two

blocking (i = 2). The probability of no blocking is di-

rectly obtained by Eq. 3:

OPTICS2015-InternationalConferenceonOpticalCommunicationSystems

= −2Φ(

) (13)

Accordingly, the probability of occurring a blocking

is given in accordance to Eq. 9:

∞

∑

i=1

(i) =

∞

∑

i=1

(2(Φ(

2GT + (2(i− 1) + 1)UB

)

−Φ(

GT + 2(i− 1)UB

)))

(14)

In the same way, the probability of an erroneous burst

to cause two burst blocking is given in accordance to

Eq. 11:

∞

∑

i=1

(i) =

∞

∑

i=1

(2(Φ(

GT + 2iUB

)−

Φ(

2GT + (2(i− 1) + 1)UB

)))

(15)

Hence, the average blocking bursts are:

bursts

= 0p

+ 1∗ p

+ 2∗ p

= p

+ 2p

(16)

Under the assumption that upstream bursts do not

utilize FEC mechanisms, we consider that a block-

ing occurrence totally disorientates the bursts being

affected. As a result all these bursts are totally dis-

carded. Hence, an erroneous burst may discard two

(single blocking) or three (two blocking) bursts in-

cluding the erroneous bursts itself. In the light of

the aforementioned remarks, Eq. 12 is transformed to

yield the average number of discarded bursts:

bursts

∑

i=1

(i+ 1)p

= 2p

+ 3p

(17)

5 PERFORMANCE EVALUATION

This section is devoted in verifying the accuracy of

the proposed analytic model.

5.1 Environment

A simulation environment was implemented in Mat-

lab in order to verify the incorporated analysis. In

particular, a full XG-PON operation has been imple-

mented. The downstream and the upstream transmis-

sion rates were set 9.95328 and 2.48832 Gbps re-

spectively. The network operation has been examined

under saturated conditions. This means that all the

available upstream bandwidth is utilized by the con-

nected ONUs. As a result, the available upstream

bandwidth, which corresponds to 38880 Bytes, is

equally shared to the ONUs. Thus, assuming a num-

ber of N = 30 ONUs, each ONU burst corresponds to

1296 Bytes including a guard time of 64 bits. Thus,

UB = 1296− 8 = 1288 Bytes.

For each conducted simulation experiment, an up-

stream burst is randomly selected as an erroneous

burst. The timing error depends on the synchroniza-

tion device between OLT and ONU. In order to study

different synchronization methods, ﬁve accuracy lev-

els were devised: a) Very High Accurate Synchro-

nization (VHAS) with time precision of ±10nsec,

b) High Accurate Synchronization (HAS) with time

precision of ±100nsec, c) Accurate Synchronization

(AS) with time precision of ±1µsec, d) Low Accu-

rate Synchronization (LAS) with time precision of

±10µsec, and d) Very Low Accurate Synchronization

(VLAS) with time precision of ±100µsec. For each

synchronization method the value of standard devia-

tion (σ) is conﬁgured as its precision divided by 4, in

terms of µsec, so as to maximize the cumulative per-

cent of normal distribution values. Hence, the values

of σ per synchronization method are: a) σ

VHAS

0.01

b) σ

HAS

0.1

, c) σ

, d) σ

LAS

, and e)

VLAS

100

5.2 Results and Discussion

First, the probability of no blocking is examined for

each synchronization accuracy level. Figure 2 depicts

the results from both analytic and simulation frame-

works. By observing this ﬁgure it is clear that VHAS

and HAS are deemed as adequate to be applied as syn-

chronization accuracy levels since they offer almost

100% time precision. On the contrary, AS, LAS, and

VLAS may cause burst blocking. AS offers about

69% absorption probability, which means that it is

VHAS HAS AS LAS VLAS

0.2

0.4

0.6

0.8

Synchronization accuracy

No blocking probability

Analysis

Simulation

Figure 2: No blocking probability of an erroneous upstream

burst.

TimeSynchronizationinXG-PONSystems:AnErrorAnalysis

VHAS HAS AS LAS VLAS

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Synchronization accuracy

Single blocking probability

Analysis

Simulation

Figure 3: Single blocking probability of an erroneous up-

stream burst.

31% possible a blocking occurrence, and therefore

burst losses. The worst performance is observed in

LAS and VLAS.

It is worth mentioning that applying a synchro-

nization device at VLAS levels, it is quite sure that

a timing error will induce burst losses.

Figure 3 illustrates the single blocking probabil-

ity of an erroneous upstream burst. As expected,

VHAS and HAS present zero blocking probability,

which means that an upstream burst that receives er-

roneous time from the OLT does not cause any block-

ing with any other burst. On the other hand, AS pol-

icy presents 30% single blocking probability. Hence,

it is 30% possible of an erroneous burst to cause a

single blocking, resulting in discarding two upstream

bursts. LAS obtains a single blocking probability of

83%, fact that implies that in case of timing failure,

the erroneous burst will probably collide with another

burst. Lastly, VLAS presents a lower single blocking

probability than LAS. This is attached to the fact that

VLAS presents high double blocking probability.

VHAS HAS AS LAS VLAS

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Synchronization accuracy

Double blocking probability

Analysis

Simulation

Figure 4: Double blocking probability of an erroneous up-

stream burst.

The double blocking probability is inspected in

Figure 4. As expected, only LAS and VLAS may

suffer from double blocking, which causes the dis-

orientation of three bursts, including the erroneous

one. VLAS presents the highest probability, fact

that indicates its inefﬁciency to handle time-sensitive

applications in XG-PON systems. Moreover, LAS

demonstrates noticeable double blocking probability

(> 10%), thus LAS is also deemed as inadequate to

guarantee sensitive services.

VHAS HAS AS LAS VLAS

0.5

1.5

2.5

Synchronization accuracy

Average discarded upsream bursts

Analysis

Simulation

Figure 5: Average discarded upstream bursts.

The average number of discarded bursts due to an

erroneous timing synchronization is depicted in Fig-

ure 5. This ﬁgure offers a clear picture of the timing

error impact. VHAS and HAS are deemed as ade-

quate synchronization accuracy levels since they min-

imize the impact of a timing error. On the other hand,

all other schemes cause data losses. AS keeps data

losses below 1 upstream burst, which means that 1288

Bytes at most are discarded in case of an error. LAS

and VLAS reach high values of data drops. VLAS

may cause the drop of almost 3 bursts, resulting in

the loss of 3864 Bytes under saturation conditions.

This loss is deemed high since sensitive services like

voice and real time video may be disoriented. In a

nutshell, VHAS and HAS are the synchronization ac-

curacy levels that are suggested to handle potential

timing errors in XG-PON systems.

6 CONCLUSIONS

A rigorous analytic approach for determining the im-

pact of a timing error in XG-PON systems was pro-

posed in this work. Several synchronization methods

on sharing common time between OLT and ONUs are

investigated in terms of burst blocking probability. In

addition, the number of affected upstream bursts was

calculated. Analytic results were veriﬁed by simula-

tion experiments. Both results indicate that at least

100nsec timing accuracy is required in order to fully

absorb potential burst blocking due to an erroneous

burst.

ACKNOWLEDGEMENTS

This work has been funded by the NSRF (2007-2013)

SynergasiaII/EPAN-II Program ”Asymmetric Passive

Optical Network for xDSL and FTTH Access,” Gen-

eral Secretariat for Research and Technology, Min-

istry of Education, Religious Affairs, Culture and

Sports (contract no. 09SYN-71-839).

OPTICS2015-InternationalConferenceonOpticalCommunicationSystems

REFERENCES

Bregni, S. and Tavella, P. (1997). Estimation of the per-

centile maximum time interval error of gaussian white

phase noise. In Communications, 1997. ICC ’97 Mon-

treal, Towards the Knowledge Millennium. 1997 IEEE

International Conference on, volume 3, pages 1597–

1601 vol.3.

Effenberger, F. (2010). Tutorial: Xg-pon. In Opti-

cal Fiber Communication (OFC), collocated National

Fiber Optic Engineers Conference, 2010 Conference

on (OFC/NFOEC), pages 1–37.

ITU-T. 10-gigabit-capable passive optical networks (xg-

pon): Transmission convergence (tc). Technical re-

port, Rec. G.987.3.

Kani, J.-i., Bourgart, F., Cui, A., Rafel, A., Campbell, M.,

Davey, R., and Rodrigues, S. (2009). Next-generation

pon-part i: Technology roadmap and general require-

ments. Communications Magazine, IEEE, 47(11):43–

49.

Lam, C. F. (2011). Passive optical networks: principles and

practice. Academic Press.

Lee, C.-H., Sorin, W., and Kim, B.-Y. (2006). Fiber to the

home using a pon infrastructure. Lightwave Technol-

ogy, Journal of, 24(12):4568–4583.

Lewandowski, W., Petit, G., and Thomas, C. (1993). Pre-

cision and accuracy of gps time transfer. Instru-

mentation and Measurement, IEEE Transactions on,

42(2):474–479.

Luo, Y., Effenberger, F., and Ansari, N. (2012). Time syn-

chronization over ethernet passive optical networks.

Communications Magazine, IEEE, 50(10):136–142.

Mar´oti, M., Kusy, B., Simon, G., and L´edeczi, A. (2004).

The ﬂooding time synchronization protocol. In Pro-

ceedings of the 2Nd International Conference on Em-

bedded Networked Sensor Systems, SenSys ’04, pages

39–49, New York, NY, USA. ACM.

Sarigiannidis, A., Iloridou, M., Nicopolitidis, P., Papadim-

itriou, G., Pavlidou, F.-N., Sarigiannidis, P., Louta,

M., and Vitsas, V. (2015). Architectures and band-

width allocation schemes for hybrid wireless-optical

networks. Communications Surveys Tutorials, IEEE,

17(1):427–468.

Sarigiannidis, P., Petridou, S., Papadimitriou, G. I., and

Obaidat, M. (2010). Igfs: A new mac protocol exploit-

ing heterogeneous propagation delays in the dynamic

bandwidth allocation on wdm-epon. Systems Journal,

IEEE, 4(1):49–56.

Shan, Q., Jun, Y., Le Floch, J.-M., Fan, Y., Ivanov, E., and

Tobar, M. (2014). Simulating gps radio signal to syn-

chronize network-a new technique for redundant tim-

ing. Ultrasonics, Ferroelectrics, and Frequency Con-

trol, IEEE Transactions on, 61(7):1075–1085.

Topliss, S., Beeler, D., and Altwegg, L. (1995). Synchro-

nization for passive optical networks. Lightwave Tech-

nology, Journal of, 13(5):947–953.

Zhang, Y., Xia, W., Li, C., and He, Z. (2013). Research and

realization of high-precision gps time transfer system.

In Computational Problem-solving (ICCP), 2013 In-

ternational Conference on, pages 334–337.

TimeSynchronizationinXG-PONSystems:AnErrorAnalysis