Ref.

Ares(2014)2247513 - 07/07/2014

D4.4
Optical layer supervision tools and
processes for long reach optical access

Dissemination Level: PU

• Dissemination level:
PU = Public,
RE = Restricted to a group specified by the consortium (including the
Commission Services),
PP = Restricted to other programme participants (including the Commission
Services),
CO = Confidential, only for members of the consortium (including the
Commission Services)
 Abstract:
This document includes an analysis of advanced technologies for optical
supervision and fiber management in Long Reach Passive Optical Networks
(LR-PONs).
A generic model for estimating the operational savings of optical supervision
in PONs is reported in Section 3.
Based on the primary reference architectures of the access network
considered in DISCUS, various approaches for integrating advanced
technologies for optical supervision in LR-PONs are reported in Section 4.
Optical Time Domain Reflectometry (OTDR) is studied both using external
devices and including OTDR embedded inside the PON transceivers. A
feasible framework for PON diagnostic using alarms of active elements in
DISCUS LR-PONs is also described.
In Section 5, an analysis of ultra-fast link restoration approaches in PONs
considering 1+1 and N:1 dual homing protection for application in DISCUS
access architecture is reported.
Finally, section 6 describes systems for fiber management based on Radio-
Frequency Identification technology and remotely powered (using light)
intelligent devices in the PON infrastructure, with the potential to provide
relevant operational advantages to DISCUS fiber access network.
 COPYRIGHT

© Copyright by the DISCUS Consortium.

The DISCUS Consortium consists of:

Participant Participant organization name Participant org. Country
Number short name
Coordinator
1 Trinity College Dublin TCD Ireland
Other Beneficiaries
2 Alcatel-Lucent Deutschland AG ALUD Germany
3 Nokia Siemens Networks GMBH & CO. KG NSN Germany
4 Telefonica Investigacion Y Desarrollo SA TID Spain
5 Telecom Italia S.p.A TI Italy
6 Aston Universtity ASTON United
Kingdom
7 Interuniversitair Micro-Electronica Centrum IMEC Belgium
VZW
8 III V Lab GIE III-V France
9 University College Cork, National University of Tyndall & UCC Ireland
Ireland, Cork
10 Polatis Ltd POLATIS United
Kingdom
11 atesio GMBH ATESIO Germany
12 Kungliga Tekniska Hoegskolan KTH Sweden

This document may not be copied, reproduced, or modified in whole or in part for
any purpose without written permission from the DISCUS Consortium. In addition
to such written permission to copy, reproduce, or modify this document in whole or
part, an acknowledgement of the authors of the document and all applicable
portions of the copyright notice must be clearly referenced.
All rights reserved.
Authors:
Name Affiliation

Julio Montalvo TID

Joerg Hehmann ALUD

Jiajia Chen KTH

Marco Ruffini TCD

Alan Hill TCD

Internal reviewers:
Name Affiliation

Nick Parsons Polatis

Rich Jensen Polatis

Marco Schiano TI

Andrea Di Giglio TI

Due date: (M18) 30 April 2014

TABLE OF CONTENTS
1 INTRODUCTION ....................................................................................................................... 6
2 REFERENCE SCENARIOS........................................................................................................ 7
3 OPERATIONAL EXPENDITURE MODEL FOR OPTICAL SUPERVISION IN PONS .. 9
3.1 OPEX SAVING BY USING AUTOMATIC FAULT MONITORING ....................................................................... 9
3.2 OPEX ANALYSIS FOR THE SUPERVISION SCHEMES COMPATIBLE TO THE RESILIENT PON
ARCHITECTURE .............................................................................................................................................................. 10

4 ADVANCED TECHNOLOGIES, MECHANISMS AND STANDARDS FOR PHYSICAL LAYER

SUPERVISION IN PONS ..................................................................................................................12
4.1 OPTICAL TIME DOMAIN REFLECTOMETRY (OTDR) .............................................................................. 12
4.1.1 External OTDR approach .................................................................................................................... 14
4.1.2 Transceiver-embedded OTDR approach...................................................................................... 19
4.2 OPTICAL LINE SUPERVISION AND DIAGNOSTIC IN PONS....................................................................... 24
4.3 BIDIRECTIONAL PERMANENT FIBRE MONITORING SYSTEM PROPOSAL ............................................. 26
4.4 ACTIVE DEMARCATION POINTS .................................................................................................................. 29
5 CONTROL PLANE-BASED LINK RESTORATION IN PONS .........................................32
5.1 1+1 DUAL-HOMING PROTECTION ............................................................................................................. 32
5.1.1 Failure Detection .................................................................................................................................... 33
5.1.2 Reactivation Methods ........................................................................................................................... 33
5.2 POSSIBILITIES FOR DUAL-HOMED N:1 PROTECTION ............................................................................. 34
5.2.1 Failure Detection .................................................................................................................................... 34
5.2.2 Reactivation Methods ........................................................................................................................... 35
5.2.3 Summary of Reactivation Requirements for Feeder Fibre Failure ................................. 38
5.3 ADDITIONAL DRIFT DUE TO WAVELENGTH CHANGES AFTER FAILURE ............................................. 39
6 SYSTEMS FOR FIBRE MANAGEMENT IN OPTICAL NETWORKS .............................40
7 SUMMARY AND CONCLUSIONS .........................................................................................43
8 REFERENCES ...........................................................................................................................44
ABBREVIATIONS ...........................................................................................................................46
1 Introduction
A simple schematic of the DISCUS architecture [1] is shown in Figure 1. It shows a dual
homed Long Reach Passive Optical Network (LR-PON) bypassing existing local
exchanges and terminating on the metro-core nodes. The metro-core (MC) nodes are
interconnected with an optical circuit switched wavelength layer.
The overall architecture consists of three major parts; the LR-PON Access
network, the metro-core node and the flat optical core-network.

Fig. 1. Schematic of the DISCUS architecture. DP: Distribution Point.

In the DISCUS architecture, the traditional metro network is shifted to and merged
with the flat optical network. The DISCUS access segment consists of a Long-Reach PON
with up to 100 km reach and large split ratio up to 1:512, consisting of a Wavelength
Division Multiplexing (WDM) backhaul fibre and cascaded amplified optical power
splitters in the outside plant.
In this Deliverable, we analyse the main monitoring challenges for the LR-PON fibre
infrastructure and study advanced optical supervision tools and processes suitable for
monitoring and managing the LR-PON fibre network.
2 Reference scenarios
In the following figures, we show the basic DISCUS LR-PON topologies. Fibres between
MC nodes and Local Exchanges/Central Offices (LE/CO) are named backhaul fibres;
fibres between the LE/CO and the 1st splitting stage are named feeder fibres; fibres
between the 1st and 2nd power splitting stages are named distribution fibres; and finally,
fibres connecting the 2nd splitting remote node (RN) and users are named drop fibres.
Fig. 2 shows the standard architecture with basic dual parenting and protection in
dense urban scenarios, and Fig. 3 shows a more cost-effective scenario for rural areas
with an increased length of the Optical Distribution Network (ODN) and a shrunken
backhaul.

Fig. 2. DISCUS LR-PON standard architecture with basic dual parenting and protection in dense urban
scenarios
 Fig. 3. DISCUS LR-PON design option with increased ODN length for rural areas

Fig. 4 shows an additional scenario where connected customers demand special

security requirements and thus they prefer a point to point physical communication.
Using wavelength filtering elements in the remote node, such as an Arrayed Waveguide
Grating (AWG), each customer only receives a dedicated optical channel, and the
channels corresponding to other customers are blocked in this port at the remote node.

Fig. 4. Filtered WDM-PON architecture for users requiring dedicated optical channel.

Both scenarios in Fig. 2 and Fig. 3 consider bi-directional amplification at least in the
first LE connected to the Metro-Core node, as well as 1+1 protection of backhaul fibre
and OLT, so that if a fault occurs either at the working OLT or backhaul fibre, the service
provided to ONT/ONUs can be restored using the protection backhaul fibre and a
backup OLT in a stand-by Metro-Core Node.
By including optical supervision tools and processes to manage the optical layer of the
fibre infrastructure, relevant operational savings can be achieved by the fibre network
operator.
In the following sections, we provide the analysis of advanced approaches for physical
layer supervision and management in PONs, specifically focusing on the main DISCUS
LR-PON reference scenarios shown in the former figures.

3 Operational Expenditure model for optical supervision in

PONs
The operational expenditures (OPEX) model for monitoring schemes here covered two
cases: 1) analysis of OPEX saving by using remote and automatic fault monitoring; and
2) analysis of OPEX saving by employing specific supervision schemes in different
deployment scenarios, e.g. urban and rural service areas, schemes with and without
protection, etc.

3.1 OPEX saving by using automatic fault monitoring

In this section, we calculate OPEX reductions achieved through remote fault
localization [2] with regards to an in-field approach requiring technician dispatches
and hand-held terminals.
C
Let one assume man-hour cost, m , to be 4 times higher in well-developed countries
with respect to developing countries (high versus low man hour cost, see Fig. 5), and K
to be the probability that drop-line fault occurred per year per Optical Line Terminal
(OLT) chassis supporting a total number of Optical Network Terminals (ONTs) Q=3072,
i.e. 12 line-cards with 8 PON ports each having a 1:32 split-ratio. The potential savings,
depicted in Fig. 5, start already at 0.065% and 0.25% of drop lines faults per OLT per
year, which correspond to less than 2 and less than 8 faulty lines in well-developed and
C
developing countries, respectively. The cost-savings per year, S , are calculated
according to the formula:

 C 
CS = 1 − remote  ⋅ 100%
 C 
 in − field 
(1)
where

Chandheld − h
Cin − field = K ⋅ Cm ⋅ Tin − field ⋅ Q +
Td (2)
 Cremote − h
Cremote =
Td (3)
are the costs resulting from in-field and remote fibre-fault localization per year,
T
respectively. The time needed for the in-field test, in − field is estimated as 3 hours which
includes transportation to and from the field, arrangements and tests, while remote test
does not need any labour costs. Besides, the above expressions embrace also the costs
C
of the necessary hardware for monitoring on a per-OLT-chassis basis. handheld−h is the
C
cost of handheld portable equipment and remote− h is the cost of hardware enabling
T
remote operation, both distributed over the depreciation time, d , of five years. We
C C
assume remote− h to be 30% more expensive than handheld − h due to extensive capabilities
and higher performance requirements.
Taking an example of German network where the average drop fibre length is 1.5km
for dense urban and 3.5 km for rural scenario, the percentage of drop line faults per OLT
per year could reach up to 0.876% and 2.044% considering a typical mean time
between failures of 1500000 hours for distribution cables. In Fig. 5, it can be seen
around 74% (in dense urban) and 87% (in rural) savings could be obtained for German
network by remote fault localization compared to in-field solution.

Figure 5: Potential cost savings achieved by in-field and remote fibre testing solutions

3.2 OPEX analysis for the supervision schemes compatible to the resilient
PON architecture
In this case, we analyze the failure related OPEX for fault supervision approaches and
take the supervision schemes presented in [2] using OTDR (optical time domain
reflectometry) as example to show the cost results. The optical supervision technology
is described later in more detail in section 4.1.1.
Our considered failure related OPEX consist of both the failure reparation and penalty
introduced by the service disruption. Penalty cost typically can be expressed as the
product of the penalty rate and the connection interruption time. In this study, the
operational time of an access network is assumed to be 20 years. We consider two
deployment scenarios, namely urban (collective) and rural (dispersive), which
correspond to dense and sparse populated areas, respectively.
In our proposed scheme [2] , if time T are needed to get one stable OTDR trace, for a
certain drop line the average time to automatically detect and localize its fault is
T*N*L/K/2. Therefore, compared to the single-thread case (i.e. K=1), the multi-thread
approach (i.e. K>1) can save T*N*L*(1-1/K)/2 in average for fault monitoring, which
indicates the mean time to repair (MTTR) for the distribution fibres (DFs) and optical
network terminal (ONTs) can be reduced correspondingly. According to the commercial
fibre plant manager product, typically 3 minutes are required in order to have an
acceptable signal-noise-ratio on the OTDR trace. The achieved operational saving by
using our scheme due to the reduction of the MTTR for the DFs and ONTs is presented
in Fig. 6, where the penalty and reparation costs of different considered scenarios are
normalized to the total OPEX of the unprotected PON with single-thread fault
monitoring in dense populated area (i.e. K=1 in collective case). The normalized added
capital expenditures (CAPEX) by employing the proposed supervision scheme (i.e. the
introduction of OTDR based system at central office, and the additional components at
remote node) is approximately 0.08%. Here, only trivial difference on the additional
CAPEX is caused between the various considered scenarios, because the OTDR module
is the main factor for the increase of deployment cost. It can be seen with a minor
CAPEX increase the proposed multi-thread based monitoring scheme can significantly
decrease the service interruption penalty, which is in particular important for business
access and mobile backhauling. The larger the number of monitoring threads (i.e., K) is,
the higher the OPEX saving obtains. Compared to the unprotected PON, the OPEX
deduction is more obvious in the DISCUS cases - (where dual-parental protection is
considered), as unprotected distribution part (including DFs and ONTs) becomes
dominating to trigger service disruption. Since the amount of DFs in dispersive case is
much larger than in collective area, the OPEX saving for the sparse populated area is
apparently higher than for the dense scenario.
 Figure 6: Failure related OPEX saving for considered scenarios normalized by the total OPEX related
to fault management of the unprotected PON in urban area with K=1.

4 Advanced technologies, mechanisms and standards for

physical layer supervision in PONs
4.1 Optical Time Domain Reflectometry (OTDR)
For troubleshooting the physical layer, Optical Time Domain Reflectometers (OTDRs)
are typically used. These devices periodically transmit optical signals modulated with
electrical pulses into the fibres to test, and receive the backscattered signal from the
fibre, thus characterizing the fibre through distance. There is a tradeoff between the
time width of the OTDR pulses (ns to us) and the measurement performance in terms of
maximum distance (dynamic range1) and two-point resolution (event dead zone2,
attenuation dead zones3). The shorter the pulses, the shorter the maximum distance
value, but the better then two-point resolution.
OTDRs can be used for Operation and Maintenance (O+M) of fibre networks in two
ways:
- De-centralized approach, which involves a technician dispatched to Outside Plant
(OSP) with portable OTDR when the network operation centre (NOC) detects an
alarm which may be caused by a fibre fault.

1 Dynamic range (dB): Ratio between the extrapolated point of the backscatter trace at the near end of
the fibre under test and the noise level reference at the fibre end.
2 Event dead zone: Distance between two opposite points (1.5 or 3 dB) down from the unsaturated peak

of a single reflective event.

3 Region after an event where the displayed trace deviates from the undisturbed backscatter trace more

than a given vertical value (0.1 or 0.5 dB).

 - Centralized approach, which involves using OTDR functions and remote active
supervision elements for diagnosing Optical Distribution Networks (ODN) from
the Central Office in case of active alarms. A variation of this approach is using an
OTDR in the Central Office where Optical Line Terminals (OLTs) of a PON are
located, and Fibre Bragg Grating (FBG) or Thin Film Filter (TFF) based optical
reflectors in the Optical Network Unit (ONU) side [3] . This allows discriminating
between a communication problem due to ONU failure and a physical layer
failure.

For the centralized approach, novel alternatives have also been proposed for
increasing system performance by partially avoiding traces overlapping behind power
splitters. These approaches are based on the use of Multi-wavelength OTDRs with
selectable wavelength of the test signal, in combination with special splitters [4] in the
outside plant.
Nevertheless, the backscattering coefficient in optical fibres is very low (around -80dB
in single-mode fibre for 1ns pulses) and in case of remote measurements, the measured
signal suffers a two way attenuation, thus OTDR receivers require a very high
performance with very low Noise Equivalent Power (NEP). Using analog detection,
OTDR receivers can typically achieve a maximum dynamic range value around 47 dB,
while 55 dB can be achieved using photon-counting OTDRs (10 us pulses) [7] .
A more challenging situation appears when a point to multipoint topology is under
test, such as the case of PONs, because the OTDR signal is highly attenuated by optical
power splitters and the reflections from different remote fibre branches are overlapped
in time at the OTDR receiver.
Both in-line and dark fibre supervision approaches can be used for OTDR supervision.
In-line supervision employs an optical signal within each PON for monitoring purposes.
On the other hand, dark fibre supervision uses fibres, unused for data transmission,
which are reserved for monitoring purposes, this dark fibres sharing as much as
possible the same fibre cables and ducts as the PON fibres.
In the case of in-line OTDR PON supervision, the standard waveband recommended by
ITU-T [3] is the U-band, which covers the 1625nm-1675nm range. The U-band is
specified for monitoring purposes when communication wavelength band extends up to
the L-band, thus this band is suitable not only with GPON and EPON, but even with
emerging XG-PON deployments and, in the future, with NGPON2 systems and beyond.
The hybrid Time Division Multiplexing/Wavelength Division Multiplexing (TDM/WDM)
LR-PON in DISCUS is looking beyond NG-PON2 to provide even higher capacities by
increasing the wavelength domain utilisation with a Dense WDM channel plan of up to
40 wavelengths each for up- and down-stream within C-band (1530nm-1565nm).
Standard U-band can then be used for fibre supervision in DISCUS LR-PONs.
In the case of using reserved dark fibres attached to PON fibre infrastructure for
supervision, there is no restriction in the waveband of the OTDR monitoring signal.
The following subsections examine in detail the external OTDR and transceiver-
embedded OTDR approaches and show how then can be used in the DISCUS reference
architectures.
4.1.1 External OTDR approach

The external OTDR approach consists of employing an OTDR device separated from
the PON transceiver.
When in-line supervision is chosen, this approach is typically done in a centralized way
by co-locating an OTDR, optical switches and wavelength multiplexers in the network
operator premises so that an OTDR device or card can be shared by several OLTs and
their respective ODNs.
Previous to DISCUS technologies, these implementation would be located in the old
LE/CO locations, thus the OTDRs and optical switches and multiplexers can be re-
utilized and shared for all ODNs supervision at the LE/CO, as shown in Figure 7.

Fig. 7. External OTDR located in old LE/CO for in-line ODN supervision.

With this approach, feeder fibres can be monitored as point to point fibres with the
OTDR, and distribution fibres can also be monitored, but overlapping between traces in
the OTDR receiver will take place.
For increased sharing of the OTDR between ODNs (thus reducing the number of
multiplexers and the size of the optical switch), multiplexers can be used before the
DISCUS 4x4 splitter; in this case, however, overlapping between feeder fibre traces will
take place in the OTDR receiver. Multiplexers can be avoided if 5x4 power splitters are
used instead of 4x4, thus reducing the number of elements and insertion loss, see Fig. 8
(b), where the additional input ports in the power splitters are marked in dashed lines.

(a) (b)

Fig. 8. External OTDR located in old LE/CO for in-line ODN supervision, using 4x4 (a) and 5x4 (b)
power splitters in the LE/CO.
 For all the former cases, it is assumed that the LE/CO keeps a management channel
with the Operations and Support Systems (OSS) of the network operator by other
means than the PON, so that the OTDR measurements can be reported and analysed by
the Operations and Maintenance (O+M) systems.
If dark fibre supervision is used, then the OTDR and PON signals are not multiplexed in
the same fibres, but they share the ODN ducts at a certain degree, see Fig. 9. In this
figure (lower picture), an example is shown where two dark fibres are used and they
are following different fibre ducts at a certain location of the ODN where two PON fibres
are segregated.

Fig. 9. Cost-efficient reliable PON architecture schematic with cable, fibre and PON topology layers
(upper) and detail of external OTDR located in old LE/CO for ODN supervision using dark fibres in the
LE/CO (lower).
The deployment cost of fibre infrastructure is the dominating part of the capital
expenditures (CAPEX) and should be minimized by a proper fibre layout. Taking
advantage as much as possible of the fibre infrastructure sharing, where
trenching/ducts can be shared by different fibre links, is a promising way to reduce the
CAPEX for operator. Fig. 9 (up) shows a cable layout design for PON which can enable
sharing the fibre infrastructure as much as possible. In Fig. 9 (up), different feeder fibre
(FFs) go through the feeder cable (FC) ring while the distribution fibres (DFs)
associated with the same PON goes through the distribution cable (DC) ring. A central
office (CO) typically supports more than one single PON. Hence, it accommodates more
than one OLT and connects to more than one Remote Node.
For the converged fixed and mobile network shown in Fig. 9, tree based logical
topology (i.e. point-to-multipoint PtMP) for passive optical network (PON) can be
mapped to ring based cable layout to enhance resiliency. This cable layout also opens
the possibility to have dark fibre monitoring scheme, where a fibre deployed in the
same cable but not used for any working PONs yet can carry out the supervision signals
for fault monitoring. In this way, the major cable fault (e.g. cut, bending, etc.) can be
easily detected. As the supervision signal goes through the dark fibre, it can be shared
by all the other fibres in the same cable and does not affect any working signals (in
either the spectral domain or time domain). On the other hand, to cover both feeder and
distribution sections we may need a lot of dark fibres. Then important future work will
be on how to minimize the amount of dark fibres to cover all the service areas to be
monitored.
External OTDRs can also be located in DISCUS M/C nodes for backhaul fibres and inter-
MC node fibres supervision, see Fig. 10. In the case of PON backhaul fibre supervision, it
is required to add the OTDR signal after the first amplifier, as the operation waveband
of amplifiers does not include the band reserved for OTDR.

(a) (b)
Fig. 10. External OTDR located in M/C Nodes for PON backhaul fibres and inter-MC node fibres
supervision using in-line (a) and dark fibres (b).

4.1.1.1 External OTDR case for Point-to-Point services using PON

In this section, we investigate a supervision scheme with external OTDR with
selectable operation wavelength, which could be compatible with the cost-efficient and
reliable fibre layout architecture for wavelength division multiplexing (WDM) based
passive optical networks (PONs) [8] . WDM based PON is widely considered for
business users and mobile backhauling, where high bandwidth and high security are
required.
Our scheme concerns measurement of optical time domain reflectometry (OTDR)
groups in several Distribution Fibres (DFs) at a time via embedded monitor ports in the
remote node. Compared with the conventional scheme where DF is monitored one by
one (i.e., single thread), our scheme can be referred to multi-thread case, which reduces
the total time to monitor the whole PON. Fig. 11 shows a DISCUS LE/CO and M/C node
with our supervision approach. A commercial OTDR device with external wavelength
adaptation module (EWAM) is installed at LE/CO which could be shared over a number
of optical line terminal (OLT) ports. Here, EWAM provides external wavelength
tunability of OTDR. It is also possible to use an internally tunable OTDR as an
alternative. The system operates on the basis of an OTDR pulsed signal sent from LE/CO
towards ONTs and backscattered or backreflected to the CO. The wavelength-adapted
OTDR signal is injected into the backhaul fibres by means of a filter at LE/CO and
reaches a passive RN. The considered cost-efficient and reliable fibre layout is
compatible with DISCUS protection for both OLT and backhaul fibres and could also be
done in DISCUS M/C node (see Fig. 11).

(a)

(b)

Fig. 11. WDM-PON optical supervision with OTDR+EWAM located in LE/CO (a) and in M/C Node (b).

The data and OTDR signals are separated by a filter at the RN, see Fig. 12b. Then, the
data signals are injected into one input port of the MxN arrayed waveguide grating
(AWG) (M≤N) [9] , while the OTDR signal passes a splitter which is connected to the
other (M-1) input ports of the AWG. Fig. 12a shows the operation principle of NxN AWG.
2:K splitter (N/K should be integer) is employed to broadcast the monitoring channels
in order to supervise multiple DFs simultaneously. Consequently, the required number
of wavelengths for monitoring is decreased by K times. For the example of a WDM PON
with 32 DFs and 2:8 splitter employed at the RN, 4 wavelengths are required for
monitoring. If the output ports of the splitter connect to the (x+1)-th (x>0) to the (x+K)-
th (x+K≤N) input ports of AWG, the assigned wavelengths for the monitoring are [λx+1,
λ(x+K)modN+1, λ(x+2K)modN+1, … λ(x+N-K)modN+1]+n∙FSR (free spectrum range) where n≥1. In the
example shown in Fig.10, x=1, N=32, K=8, and M=9. It should be noted if adapting this
scheme in single-thread case (i.e. K=1), the splitter is not needed. This system imposes
only minor power budget degradation for data channels caused by the insertion loss of
the filter at RN.
The backup feeder can be connected to any free input port of NxN AWG. Although the
assigned wavelength for a certain ONT in protection mode is changed, it will not
introduce extra expense since a colorless-ONT should support any wavelength used in
the same WDM PON. The example shown in Fig. 10 is with x=1, N=32, K=8, and M=10.
Furthermore, we should keep K≤8 to guarantee 1dB event sensitivity. It is because
higher splitting ratio decrease sensitivity due to the overlapping light backscattered
from all drop ports masking the events, e.g. 1 dB fault is detected in a fully-fiberized 1:8
splitter as 0.1 dB step on the reflectogram which is the event detection threshold of a
regular OTDR. As up to 8 drop lines are able to be measured simultaneously, the
monitoring time used for fault detection and localization of one PON could decrease to
12.5%.

(a)

(b)
Fig. 12. Principle of operation of NxN AWG (a) and Remote Node with wavelength plan schematic.
FSR: Free Spectral Range.
4.1.2 Transceiver-embedded OTDR approach

The basic idea of the embedded OTDR approach is to integrate the OTDR measurement
function into the data transmission system. The main advantage of this approach is to
link the measurement function with the data transmission. This is realized by using the
same wavelength by an on top modulation of the OTDR signal to the data stream. The
solution also fits very well if WDM PON architectures are used. The OTDR signal follows
the data stream through the network so that the OTDR result is linked directly to the
data path. The capability of the eOTDR solution in DISCUS LR-PON architectures will be
analysed in the following sections.

3.1.2.1 Embedded OTDR architecture

The main challenge for the embedded OTDR approach was to integrate its distributed
sub functions into the PON transmission system [9] . All configuration and analysis
tasks should be done in a centralised OTDR diagnosis and test centre. A standard
management protocol can be used to transfer the OTDR parameter to the central offices
and their OLTs. The OTDR results will be transferred via a standard file transfer from
the central offices to the OTDR diagnosis and test centre. The management information
base (MIB) of the transmission system was extended to provide the OTDR parameter
for the OTDR controller function at the OLT. The OTDR controller itself manages the
OTDR measurements from the data transceiver. The configuration and the OTDR
measurement data are transferred via the I2C Interface. The data transceiver was
extended by an ASIC (Application Specific Integrated Circuit) and an additional OTDR
receiver for the OTDR measurement as shown in figure 13.

Rx data
Lim Amplifier
Data
RX
10% tap WDM
Tx data Data/ coupler coupler
Laser
driver OTDR
Tx
Back scatter OTDR signal
OTDR
Tx enable Rx

TRx EEPROM

(TRx diagnostic)
LED
I2C Bus
Mixed Signal ASIC Quartz

Figure 13: Schematic view of OTDR SFP transceiver. The new components are highlighted in yellow
and blue.

4.1.2.1 Measurement method

The measurement method is based on a continuous modulation of the data stream. The
data stream has a transmission rate of 2.5Gbit/s. The OTDR signal consists of a PRBS
pattern and a minimal bit length of 50ns. The pattern length and the bit length can be
adjusted to the network topology or link length. A Pseudo Random Binary Sequence
(PRBS) 211 -1 pattern and a bit length of 100ns fits best for a 20km long fibre link. The
pattern transmission time should be longer than the round trip time through the fibre.
This is given in the above example.
The modulation depth typically is in the range of 10% in normal operation mode. The
additional OTDR signal has a very low impact to the data stream and can be neglected in
most cases. If the transmission system is disturbed in a strong manner the operator can
decide to increase the OTDR modulation depth up to 100% to accelerate the OTDR
measurement. Figure 15 shows the principle of the modulation scheme.
The back scattered OTDR signal is detected and averaged within the ASIC to improve
the signal to noise ratio. Typical average time are 4 minutes per measurement cycle.
The time is sufficient to ensure a dynamic range of 10dB.

optical
power
time

Figure 15: The graph shows the typical OTDR signal modulated on top of the data stream.

4.1.2.2 Results in PON architectures

The characteristic embedded OTDR performance depends on the modulation depth,
the bit length and the measurement time. The longer the measurement time is the
better the dynamic range is. After 4 minutes measurement time the dynamic range is in
the range of 10dB. This is sufficient to monitor the feeder fibre section. The visibility of
the distribution fibre section depends on the splitter ratio. The last meters of the drop
fibre section is not visible due to the less dynamic rage of the embedded OTDR system.
Reflective elements can be detected in all fibre segments of the network. Figure 16
shows the different OTDR traces of a typical network were a 1:4 splitter after a 4km
long feeder fibre. The graph shows measurement data of different fibre cut scenarios in
the distribution section. It is shown that it is possible to distinguish between fibre cuts
in different branches if the original network topology is known.

4.1.2.3 Embedded OTDR in DISCUS LR PON architecture

Amplifier can be used to extend the measurement length of the embedded OTDR
system. Some adaptations were needed to demonstrate the usability of amplifier in
combination with the existing eOTDR transceiver for DISCUS architecture. Circulators
and upstream amplification has been included in the downstream fibres in order to
allow the reception and amplification of the Rayleigh signal. The pattern itself had to be
enlarged to fulfil the roundtrip requirement as described in 4.1.2.1. This could be done
be using a longer pattern. The main restriction for the existing solution is the limited
sample memory within the ASIC. Two tricks were done to demonstrate the feasibility of
using amplifiers for OTDR measurements. The first one was to reduce the clock rate by
a factor of two. This allows a double length of the fibre link. The second trick was to split
the measurements in two parts. The first part samples up to the limit of the sample
memory and the second part samples with a delay related to the sample memory. Both
parts are concatenated to each other and then calculated with the original pattern. This
solution allows measurements up to 80km link length.

1
4
1.5 1
OTDR SFP 2.3 5
4.3 4

Change in Attenuation due to fiber breaks in different branches

1:4 Splitter, PRBS 2^10-1, 200ns, GPON OLT, 8sample filter

-1

-3
Attenuation [dB]

-5

-7

-9

-11

-13

-15
0 2000 4000 6000 8000 10000 12000 14000
Distance [m]

attenuation all attenuation without 1km attenuation without 4 km attenuation without 5 km

Figure 16: Embedded OTDR measurement results demonstrate the detection of fibre cuts in different
distribution fibre branches after a 1:4 splitter. The upper graph shows the network topology. The yellow
curve in the lower graph shows the OTDR result of a complete network. The green curve results if the
second branch is interrupted after 1.5 km behind the splitter. The blue curve results if the connector in
the third branch is open after 2.3km. The OTDR contribution of the former 5km is now missing. The red
curve shows the situation if the 4km of the fourth branch is missing.

Semiconductor optical amplifiers (SOA) were selected to run the test, because the
operating wavelength of fibre based amplifier does not fit to the transmission system.
Different amplifier arrangements were investigated to find out the best configuration
parameters. The tests were related to the reference scenarios described above. It could
be shown that the arrangement of two amplifiers in combination with a circulator as
shown in the second reference design for rural application shows the best result. A 80
km long link length with two cascaded amplifiers was measured with the embedded
OTDR solution. The network set-up is shown in figure 17.
 Figure 17. Measurement setup schematic for eOTDR measurements in amplified fibers
A typical eOTDR trace obtained in the former amplified fiber link is shown in figure
18. The connector losses and the splice losses along the graph between the nodes can be
easily assigned to the network structure.

Figure 18: typical eOTDR trace of a 80km long fibre link including two amplifier stages at 20km and
39.5km.
The feasibility analysis was done in a lab environment. The lab results must be
confirmed in a real PON as a next step. The experiments showed that the amplifier
strength has to be adjusted to the network topology.

4.1.2.4 Standardisation work

ALU supports and delivers contributions to the Broadband Forum working text WT-
287 on PON optical layer management.
 The purpose of the WT-287 is to describe ways to improve physical layer performance
and fault isolation capabilities for both ITU-T and IEEE PON systems.
Following requirements were included:
• The system architecture, main interfaces, and use cases of FTTx optical-link
management functions.
• Functional and performance requirements for FTTx optical-link management
functions.
• Implementation-dependent requirements (OTDR, RSSI-based optical-power
monitoring, etc.).
• Interface requirements (ONU-OLT, EMS-OLT, etc.).
Beside the technical discussion, the principal objective of the WT-287 is the
interoperability of the equipment, including compatibility with existing PON standards.

ALU main contribution was on the dedicated OTDR (embedded OTDR) description. A
proposal for the interface definition and the data format was made in cooperation with
other equipment supplier. It is agreed that the northbound interface of a dedicated
OTDR should support the SR-4731 standard for OTDR data transfer. The interfaces
between the OTDR test system and the OTDR controller as well as the interface between
the OTDR controller and the OTDR function are described independent from the SR-
4731. The WT-287 will be finalized in spring 2014. Figure 19 shows the entities for the
dedicated OTDR solution.

OSS / BSS

Optical Test & Diagnostics

Subsystem

B L

Link Monitoring OTDR

Management Controller

C I

OLT

PON Port

LInk OTDR
Monitoring Function

S/R

ODN

R/S

Link
Monitoring

PON Port

ONU

Figure 19: Reference architecture for the dedicated OTDR

4.2 Optical Line Supervision and Diagnostic in PONs
Optical Line Supervision is defined in [11] as the set of capabilities relating to the
measurement and reporting of the state of the optical link between OLT and ONT in
GPON systems. In order to support optical layer supervision, OLT and ONT transceiver
monitoring can be used for different purposes:
- Normal status monitoring, considering a buffer of historic data as reference in
a normally operating system.
- Degradation detection, performing preventive monitoring in order to detect
the potential faults before they affect the service and identifying the source of the
problem.
- Fault management, in order to detect, diagnose and localize faults.
The following performance monitoring items should be monitored in the PON
transceivers (both OLT and ONT): Temperature, voltage, laser bias current, transmit
and receive power. Typical ranges, resolution, accuracy and repeatability for each of the
former parameters is well specified in [11] , with typical responses times between 1s
and 300ns.
Similar functions are defined for digital optical monitoring functions in the standard
SFF-8472 for Small-form Factor Pluggable (SFP) modules, which can be embedded also
in Ethernet PON (EPON) transceivers.
IEEE 802.3ag provides connectivity fault management in EPON, which can be used to
remotely monitor a single link (remote failure indication, remote loopback and link
monitoring). Performance monitoring for Ethernet PONs can also be applied in the
framework of ITU-T Y.1731, which provides procedures and messages to perform
packet loss measurements, frame delay and delay variation measurements.
The ONT Management and Configuration Interface (OMCI) described in [12] is used by
the OLT for fault and performance management in the ONTs, this interface allowing the
OLT to request performance statistics from the OLS modules in the ONTs.
OMCI also allows ONTs to autonomously inform OLT about alarms and performance
threshold crossing. Examples of relevant alarms are “Dying gasp” or “ONU manual
power off” alarms, when the ONT reports imminent powering off due to loss of
electrical power to the ONT itself or because the subscriber has turned off its power
switch, respectively.
These alarms are relevant for operational savings to network operators, because in
case a subscriber contacts the call centre reporting a service interruption, technician
dispatches can be avoided if the OLT informs the OSS that the ONT has been powered
off and how. In case the service cannot be restored remotely, then the technician
dispatch can focus on the outside plant rather than inside the subscriber premises, thus
being able to locate and repair the faults more efficiently.
ITU-T also defines performance monitoring counters in order to facilitate PON
troubleshooting [13] . By reporting the number of corrected Forward Error Correction
(FEC) bytes in ONTs, it is possible to detect a signal degradation before it causes a signal
transmission interruption. Defects in the Transmission Convergence Layer of the PON
are also relevant for PON supervision. As an example, Loss of Signal (LOS) can be
detected at the OLT relating to an ONT whose upstream signal is not received.
These optical and performance monitoring approaches can help PON systems to detect
problems in ONTs which are not operating properly (rogue ONTs) and, therefore, can
cause transmission problems for other ONTs. Rogue ONTs may transmit in time slots
that have not been granted to them by the OLT due to hardware failures, thus the PON
needs a mechanism to detect this and also to mitigate the problem.
Unauthorized transmission, software, media access control and transceivers errors can
be detected in order to find the rogue condition causes and for prevention. Rogue
detection, isolation and mitigation mechanisms are also required in PONs. For example,
the OLT must be able to disable a rogue ONT. More details can be found in [14] .
DISCUS LR-PONs protocols consider the former recommendations in order to manage
the optical layer management of the PON and reduce operational costs as much as
possible.
A differential property of DISCUS LR-PONs are cascaded amplifiers. Each of them can
be treated as an extension element of the OLT and, according to ITU-T [15] , Optical
Amplifier based PON reach extenders should include a complete embedded ONT (eONT)
for management purposes. This eONT is connected internally by means of an optical tap
coupler facing the OLT side, see the following figure. The local controller can be used for
remotely managing and monitoring the performance of both upstream and downstream
amplifier sections and report the measurement to the OLT.

Fig. 20. Schematic of an optical amplifier based PON reach extender.

Another peculiar aspect of DISCUS is that multiple wavelengths are used in the PON
using WDM. For this case, specific optical monitoring parameters must be considered,
as defined by ITU-T in [16] .
 Current technology in optical transmission systems is able to provide the following
measurements, which should be included in the optical layer supervision of DISCUS LR-
PONs:
- Channel power
- Total power
- Optical signal-to-noise ratio (OSNR)
- Q-factor

Embedded monitoring functions for amplifiers will be included both at the input and
output, so that the amplification factor can be monitored, see Fig. 21.

Fig. 21. Schematic of possible monitoring positions in a generic optical network element.

4.3 Bidirectional Permanent Fibre Monitoring System Proposal

A Permanent Fibre Supervision System (PFMS) with bidirectional supervision signals
in a point to point fibre link is an enhancement to Optical Supervisory Channels (OSC) at
the Optical Multiplex Section (OMS) level in Optical Transmission Networks (OTN), with
the potential to detect fibre faults in a WDM amplified fibre link, such as OMS in OTN,
and also in the DISCUS architecture.
The PFMS uses two counter propagating WDM monitoring signals permanently
transmitted between both ends of a fibre link. The received power of the monitoring
signals at each of the fibre ends is continuously measured and analysed.
When a fibre fault (such as a fibre bend or cut) occurs at a certain location in the fibre
link, the fault will cause a simultaneous optical power attenuation of optical signals
transmitted in both directions.
 The time of flight (ToF) of the faulty event from the fault location to both ends of the
fibre link depends on the fibre length of the link, the fault location, and on the specific
group refractive index of the monitoring signals in the fibre.
When the attenuation caused by the fault event is detected at the processing units of
both receiver blocks of the PFMS, a pair of alarms will be generated, both alarms at least
containing the detection time of the attenuation at each PFMS receiver block with
regards a time reference available at both processing units and synchronized with
enough precision.
The delay between the reception times of the faulty event at the corresponding PFMS
receiver blocks can be used to calculate the location of the fibre fault, see [17] [18] for
analytical details.
We have proposed and experimentally tested this PFMS in single fibres (single mode),
see the following figure showing the experimental setup.

Fig. 22 Experimental setup for validation of the PFMS measurement technique. DFB: Distributed
FeedBack Laser. PD: Photo-Diode.

In order to test the performance of the proposal, fibre cuts were produced and the
results were saved and analysed offline in an oscilloscope. A measurement example
during a fibre cut is shown in the following figure.
 Figure 23. Measurement during fibre cut in a single mode fibre. La=4250m, Lb=500m, λa=1.53um,
λa=1.55um.

The time delay measured between both monitoring wavelengths was estimated using
a 3dB attenuation threshold, obtaining a value of 8.51 us, very close (12m
measurement difference) to the theoretical delay estimated using OTDRs for fibre
length measurement.
The following figures show the schematic of the PFMS for PON backhaul-fibre and
inter-MC node fibre supervision. Synchronized clocks between both ends of the PFMS
link and connectivity with the O+M systems are assumed, which is required in order to
gather the alarms from both ends of the PFMS and thus analyse the fibre fault.
The main advantage of this approach with regards external and transceiver-embedded
OTDRs is that it can go through amplifiers as an overlaid signal within the same
waveband than data signals, thus it has potential longer reach and simplifies integration
between data and supervision planes, avoiding also the need for additional elements in
the MC nodes and amplifiers. On the other hand, it demands more complex O+M
procedures (known fibre length, synchronization and gathering measurements from
both ends) and increases the cost of OLT transceivers and remote nodes.
 Fig 24. Bidirectional Permanent Fibre Monitoring System for in-line supervision of PON backhaul
fibres

(a) (b)
Fig. 25. Bidirectional Permanent Fibre Monitoring System for in-line (a) and dark fibre (b) supervision
of inter-MC fibres

4.4 Active Demarcation Points

The use of Fibre Bragg Grating (FBG) or Thin Film Filter (TFF) based optical reflectors
in the Optical Network Unit (ONU) side [3] is a well-known fully passive technique
which allows discriminating between a communication problem due to ONU failure and
a physical layer failure.
 The optical reflectors may be used as passive optical demarcation points at the
physical layer, so that the network operator can guarantee optical connectivity with
operator premises, thus leading to potential operational savings.
In the Japanese experience reported in [3] , Appendix I, filter embedded optical and
plug-adapter connectors are considered.
For avoiding inventory complexities, all the reflectors are designed to operate in the
same waveband (1650nm). In a highly-populated large split ratio PON, such as the
DISCUS LR-PON, an overlapping in time of all the reflection is expected, which can make
difficult or challenging the signal analysis of all the reflections and the assignment and
tracking of reflections to customers.
We analyse here the possibility of using active demarcation points that can be
remotely controlled and activated on demand, avoiding the overlapping of single traces
using OTDR and optical reflectors.
The basic idea of the active demarcation point was to control the remote device and to
build up a platform to perform some basic measurements on side of the demarcation
point [19] . The cost saving model of the active demarcation point is similar to the
embedded OTDR approach. Components of the transmission system should be reused
as much as possible. The same OTDR transceiver hardware at the OLT side is used to
perform the communication between the central office side and the demarcation point.
A low energy consuming protocol was developed to enable a battery power solution for
the demarcation point. The demarcation point device is in a sleep mode most of the
time. In this phase a very low energy consuming counter is running only. At the end of
the sleep phase the device change to a higher energy mode and is listening for time
information for the next communication phase. From this moment on a real time
counter starts at the demarcation point device. When the real time counter ends the
device starts listening for a valid message from the OLT. If no valid message is detect the
device goes back to the sleep mode phase and the counter starts again. In case of a valid
message the device performs the command and returns an answer to the OLT. The
simples command could be a status request. Figure 26 shows the timing behaviour of
two independent working demarcation devices.

Figure 26: Timing diagram for the demarcation device communication. The timer information from the
CO ensures that the demarcation point devices are synchronized for the common communication phase.
 Fig. 26 shows the measured timing signals of two connected demarcation devices (DD).
Inside a period larger than TSleep, the CO broadcasts to the DD a number of R
rendezvous signals RV, repeated at intervals TR = TSleep / R. The signals RV inform the
DD which happen to be presently awake if and when a communication “rendezvous”
will be arranged. If no RV signal is received, the DD goes back sleeping after the wakeup
time. However, if RV was received (DD1,2), the DD extracts the time stamp for the next
rendezvous with the CO, sets a high precision clock, and then goes snoozing for the
waiting time TSnooze. Snoozing DD awake exactly at rendezvous time and wait for
being addressed by the CO. On reception of its individual address each DD
communicates with CO and goes sleeping as soon as the next DD is addressed or after an
internal time-out.
The main hardware components of the demarcation point device are a data receiver to
extract the modulated timing information and the messages from the OLT. A µ
controller is used to process the received signals and to generate the response message.
A vertical cavity surface emitting laser (VCSEL) is used as a data transmitter relate to its
low laser threshold current. Figure 27 shows the test set-up to demonstrate the
operational functions. The figure shows also the modulated downstream signals and the
response signal at the OLT.

Figure 27: schematic view of the demonstration set-up for the demarcation point device. The
measurement graphs show the downstream signal at the DP receiver and the upstream signal at the OLT.

It is shown that the active demarcation point device is suitable as new network
element to separate the operator network from the user network.
5 Control plane-based link restoration in PONs
The key to fast restoration of service after a PON feeder fibre failure is to resume PON
Time Division Multiplexing/Time Division Multiple Access (TDM/TDMA) protocol
operation as quickly as possible. A range of potential PON supervisory and reactivation
methods for achieving this are described, for both 1+1 dual-homing protection and
dual-homed N:1 protection. The aim is to detect failure and reactivate all ONUs within
50 msec. This section concentrates on feeder fibre failure, but also begins to investigate
the complex interplay between feeder fibre failure and OLT failure.

5.1 1+1 Dual-Homing Protection

In the ultra-fast 1+1 dual-homing protection switching experiments for the LR-PON
protocol on Field Programmable Gate Arrays (FPGA) (Fig. 28), published at ECOC 2013
[20] , four potential reactivation methods were compared, following detection of the
primary feeder fibre failure.

1. Failure occurs: fibre cut/primary OLT is down

2. Upstream data flow stops
3. OLT detects failure and takes control
4. PON operational again

Fig. 28. Ultra-fast 1+1 dual-homing protection switching experiments on LR-PON protocol
5.1.1 Failure Detection

Hardware failure detection is performed by eavesdropping at the backup OLT (point

3). A receiver monitors upstream data at PON speed, and a counter/timer counts words,
re-setting when data arrives. The counter times out after 2.6 msec, which is sufficiently
long in a 100 km LR-PON to encompass any possible quiet window (1.161 msec) and
data already in the fibre at the moment of failure (1.1 msec).

5.1.2 Reactivation Methods

Table 1 gives the four reactivation methods and corresponding reactivation and total
protection times. The total protection time = hardware fault detection time + PON
reactivation time. Cold Start has a 100 msec failure response time-out instead. Results
show that the backup OLT can protect a 100 km LR-PON in as little as 4.1 msec, well
within the 100 msec required to avoid reactivating all ONUs individually, and within the
50 msec required to hold legacy leased line connections. Further explanation of these
reactivation methods are given in the “Reactivation Methods” section 5.2.2 below.

Table 1. LR-PON reactivation methods, reactivation times and total protection times.

PON Total
Reactivation Protection
Time Time

Cold all ONUs re-ranged 7 msec 107 msec

Start

Ideal no ONUs re-ranged, 1.5 msec 4.1 msec

differential feeder fibre delay
known

Range 1 ONU re-ranged 2.5 msec 5.1 msec

One

Flexible 1 ONU re-ranged in normal 2.5 msec 5.1 msec

operation
5.2 Possibilities for Dual-Homed N:1 Protection
5.2.1 Failure Detection

There are essentially three possible ways of detecting failure, all of which could be
implemented at either primary or backup nodes:
- optical power monitoring
- hardware failure detection
- PON protocol monitoring

Optical power monitoring could be used for example at the optical switch that
terminates the backup fibres as well as at the OLTs. The challenge here is that the power
monitors may need to detect presence/absence of light under extreme circumstances,
yet still provide a healthy OSNR. For example, when only DBRus are being transmitted
upstream (no payloads), most of a 0.125 msec frame has no light, or when there is only
one ONU switched on in the PON, most of a service interval has no light. If this proves
difficult, i.e. giving a potentially high number of false alarms, then the mean optical
power could be increased by adapting the LR-PON protocol to fill up each 0.125 msec
frame with upstream bursts, most of which consist of idle XGEM frames, to approximate
a continuous LR-PON signal. A Network Management (NM) ONU at the splitter location
could provide the dummy bursts.

Hardware failure detection could operate electronically as for the 1+1 protection
experiments above, eavesdropping at both primary OLT and shared backup OLT as
described in the following paragraphs.

LR-PON protocol monitoring is potentially very powerful for detecting fibre breaks. An
OLT could monitor the burst returns, e.g. of ONUs’ DBRu grants, and maintain a map of
which branches of the PON are functioning properly. For individual branches, the OLT
must be able to identify the PON branch from the Alloc-ID. For the entire PON (feeder
fibre break), location of individual Alloc-IDs is not needed. Timescales will depend on
the service intervals for bandwidth grants: SImax for fixed, assured and polling
bandwidths & SImin for non-assured and best effort bandwidths. These are likely to be
1 to a few msec over 100 km PONs. So presumably a feeder-fibre break could be
detected very quickly, in similar times to hardware failure detection. (As could missing
sub-groups of ONUs in PON branches). If all user ONUs are turned off, the NM ONU at
the splitter node would provide upstream bursts to ensure that feeder fibre breaks can
be detected at the primary and backup OLTs.

It is hoped to use standby OLTs for hardware failure detection and LR-PON protocol
monitoring. However, there are insufficient numbers of standby OLTs in N:1 protection
to dedicate each one to a single LR-PON. So it would be necessary to time-share each
standby OLT between multiple working OLTs. The number can vary, but depending on
the implementation the ratios could vary between 1:2 and 1:6 (we have published
results at OFC 14 [21] ). Hopefully optical switching, wavelength tuning and protection
switchover times will be sufficiently fast to allow 3 OLTs to be protected within 50
msec.

For feeder fibre failure detection, whichever failure detection method is used, it is
proposed that failure should be detected not only at primary node and backup node, but
also at a NM ONU located at the splitter/amplifier location where feeder fibres are
coupled together (on the ONU side of the splitter). When the primary feeder fibre
breaks, all three detections give the correct failed result. But the backup node, because
it is eavesdropping on US transmissions, detects a failure even when its own backup
feeder fibre breaks. So only the NM ONU and primary node can correctly discriminate
between feeder fibre and backup fibre failure. The NM ONU turns on the DS amplifier in
the backup feeder fibre, allowing a backup OLT selected by the backup node controller
to take over the LR-PON. All ONUs (including the NM ONU) re-synchronise to the
backup OLT and reactivate, before their LODS timers expire. The control plane does not
need to initiate protection switchover.

However, in truth the loss of LR-PON signal could be due to OLT failure instead of a
break in the primary feeder fibre. The NM ONU cannot discriminate between the two;
only the primary node knows it is an OLT failure. Furthermore, although feeder fibre
failure must be protected using a standby OLT at the backup node, it would be desirable
for an OLT failure to be protected by a standby OLT at the primary node instead, to
prevent the unnecessary re-routing of core network traffic.
There are three ways in which the correct OLT failure decision can be made. Firstly,
reactivation as per feeder fibre failure is started by a standby OLT in the backup node,
until the control plane has told the backup node to cease transmitting and reactivating,
and told a standby OLT in the primary node to take over. The NM ONU needn't be told
anything, so the initial reactivation doesn't need to be completed. It will be able to turn
back on the primary feeder fibre amp and turn off the backup feeder fibre amp, and all
ONUs will once again start reactivation. If it does have to be advised by the control
plane, then the initial reactivation will have to be completed first, but only the NM ONU
is allowed to communicate before the backup node standby OLT is turned off and
primary node standby OLT turned on by the control plane. In the second method,
reactivation as per feeder fibre failure is begun as before, but is stopped when the NM
ONU detects a standby OLT in the primary node transmitting DS, using a fibre tap in the
DS direction that bypasses the amplifier with a separate detector. It could then turn off
the backup feeder fibre amplifier and turn back on the primary feeder fibre amplifier.
Thirdly, reactivation as per feeder fibre failure is delayed by a time-out of suitable
duration in the NM ONU, to give the primary node controller time to select a standby
OLT to start transmitting DS in the primary feeder fibre, if the primary OLT has failed
(and a standby OLT is available there).

5.2.2 Reactivation Methods

Ideal reactivation requires no ONUs to be re-ranged after failure. Flexible reactivation

requires one or more ONUs to be re-ranged in normal operation. For both, it is not
sufficient just to measure the differential feeder fibre delay between primary and
backup OLTs, because dynamic bandwidth assignment (DBA) imposes constraints on
the upstream PHY frame offset values Teqdprimary and Teqdbackup, which are design values
for the lifetime of the primary and backup PONs. Firstly, these round-trip times should
ideally equal an integer number of service intervals (SIs), because if they fractionally
exceed an integer, an extra SI of delay can be incurred, and hence a larger bandwidth
overhead for the same Assured Bandwidth Restoration Time. Secondly, the SI itself
should equal an integer number of XGTC frames (0.125 msec). So the change in
equalisation delay ∆EqD must be calculated to take these into account. This is done after
the back-up OLT has taken over, and the ONUs have achieved re-synchronisation to its
DS frame and returned to Operation State.

After failure, if the round-trip time is measured at the backup OLT for ONUi, it is

Teqdmeasured ,i = RTD primary ,i + EqD primary ,i + drift backup ,i + DiffFeederFibreDelay (4)

RTDprimary,i is the nominal round-trip delay, including propagation delay and response
time, which the primary OLT used to set the EqDprimary,i value before drift occurred.
After failure, the ONUs continue to use their EqDprimary,i values, and the round-trip time
is increased by the differential feeder fibre delay. Therefore

Teqdmeasured ,i = Teqdprimary + drift backup ,i + DiffFeederFibreDelay (5)

The driftbackup,i values may differ from those that applied with the primary OLT, e.g. due
to Tx wavelength changes or further drift during failure switch-over. The design value
of Teqdbackup (for the lifetime of this backup PON) is assumed to be known in advance
(e.g. for 125 km). The change in equalisation delay needed to change the measured
delay to the design value is

∆EqDi = Teqdbackup − Teqdmeasured ,i (6)

Interestingly, if it is possible to trust that driftbackup,i, due to all possible causes, can at
all times be guaranteed to be within an acceptable range, even after failure, then Ideal
reactivation needs no measurement after failure and ∆EqDi can be pre-calculated as

∆EqDi = Teqdbackup − Teqdprimary − DiffFeederFibreDelay (7)

This applies to all Alloc-IDs having the same service interval, and can be pre-calculated
before failure, after initial fibre installation and measurements, and after Alloc-IDs’
assured bandwidth service intervals have been defined or updated. Any individual
Alloc-ID will be within the acceptable drift range around the upstream PHY frame offset
design value Teqdbackup.
However, if driftbackup,i could possibly be outside an acceptable range after failure and
re-synchronisation to the backup OLT, or if differential feeder fibre delay cannot be
measured with sufficient precision before failure, then ∆EqDi must be calculated by
eqtn.(6), using the measured round-trip time eqtn.(4), eqtn.(5) of one of the Alloc-IDs
(Flexible reactivation). Unfortunately, if only one ONU is measured, and the backup OLT
applies the result to all ONUs, it will cause even greater drift for other ONUs whose drift
is of opposite sign. ONUs that had been within the lower drift of window threshold DoWi
(e.g. +/-8 bits) may now exceed it. For these ONUs, the OLT will implement in-service
equalisation delay adjustment. At most 1,023 ONUs may need to be adjusted. The time
taken to do this needs to be estimated. However, if any ONUs had been in the process of
being adjusted just before failure, because their drifts exceeded +/-8 bits, their drifts
after restoration could now exceed the transmission interference warning threshold
TIWi (e.g. +/- 16 bits). Furthermore, if the wavelength of the backup OLT differs from
the primary OLT’s, or the wavelength of any ONU drifts slightly further during the
protection action, additional drift due to wavelength change can occur, increasing the
likelihood of exceeding the TIWi threshold. Any ONUs that do reach or exceed the
threshold become liable to mitigation actions such as deactivation, disablement or
diagnostic procedure (G.987.3 Section 13.1.6). There are several possible solutions for
avoiding this risk:
a) The TIW threshold could be ignored immediately after protection action, and in-
service equalisation delay adjustment performed on ONUs before restoring to
service. This might be risky if wavelength changes could be excessive.
b) Increase the TIW threshold. This is subject to the same risk as above.
c) Backup OLT measures just one of the ONUs. Primary OLT advises the backup
OLT of the difference between the drift of that ONUi and the mid-range drift of all
ONUs immediately before failure, and backup OLT applies correction for the mid-
range value. If identity of ONU to be measured is unknown by primary OLT, it
must send drift values of all ONUs to backup OLT, which can then calculate the
appropriate correction.
d) Backup OLT performs round-trip measurement Teqdmeasured,i of all ONUs,
calculates the mid-range drift value and applies correction to all ONUs. Hopefully
the additional time needed to measure all ONUs is within 1 frame (0.125 msec).
To accommodate any expected additional relative drift due to a different OLT
wavelength or ONU tuning accuracy after failure (if large enough to worry
about), a very short “quiet window” may be inserted (possibly just 1 byte)
between ONU bursts.
e) Backup OLT measures Teqdmeasured,i and calculates ∆EqDi for all ONUs, and
transmits them to each ONU individually. Hopefully this might take only around
2 frames. This provides an accurate value for the equalisation delay offset of each
ONU, which eliminates all drift errors.
5.2.3 Summary of Reactivation Requirements for Feeder Fibre Failure

Ideal reactivation:
Before failure, backup Nodes must hold:-
- all primary OLT service, MAC & traffic (US & DS QoS) parameters (PON-ID,
serial
numbers, ONU-IDs, Alloc-IDs, XGEM Port-IDs, SImax, ABmin, SImin, ABsur,
CIR, PIR, PBS, GIR, GBS), to avoid sending them after failure
- Teqdprimary design value (after initial measurement & calculation)
- Teqdbackup design value (after initial measurement & calculation)
- DiffFeederFibreDelay (measured after installation & before failure)

Before failure, backup Nodes can:-

- pre-calculate ∆EqD
After failure, backup OLTs must:-
- load all primary OLT service, MAC & traffic parameters from backup Node
- broadcast ∆EqD to all ONUs

Flexible Reactivation:
Before failure, backup Nodes must hold:-
- all primary OLT service, MAC & traffic (US & DS QoS) parameters (serial
numbers, ONU-IDs, Alloc-IDs, XGEM Port-IDs, SImax, ABmin, SImin, ABsur,
CIR, PIR, PBS, GIR, GBS), to avoid sending them after failure
- Teqdbackup design value (after initial measurement & calculation)
After failure, backup OLTs must:-
- load all primary OLT service, MAC & traffic parameters from backup Node
- measure round-trip time Teqdmeasured,i of one ONU & calculate ∆EqD for all ONUs
- broadcast ∆EqD to all ONUs
More complex procedures to reduce or eliminate the impact of drift, if needed, are
given in c) to e) above. Solutions d) or e) are perhaps preferable, because they require
no data to be sent from primary OLT to backup OLT after failure.

The control plane must co-ordinate the initial measurements and design decisions. Any
updates of SI values decided at the primary node and OLT must be communicated by
the control plane to the backup node before failure. No data needs to be sent from failed
PON to backup PON upon failure via control plane, except in c) above.
 When failure is detected, the control plane must be advised. For an individual OLT
failure, the primary node must inform the control plane. If only one or a small number
of OLTs have failed, the control plane may decide to use standby OLTs in the primary
node, to avoid traffic re-routing across the core network. For a feeder fibre failure (as
considered here in detail), whether supporting just one OLT or multiple OLTs at
multiple wavelengths, either or both primary and backup nodes must inform the control
plane. The control plane must then undertake re-routing of core network traffic to and
from the backup Node for the relevant VLANs, Pseudo-Wires, Label Switched Paths
(LSP) and whole wavelength channels.

5.3 Additional Drift Due to Wavelength Changes After Failure

If additional drift is given by fibre dispersion (e.g. 20 psec/nm/km) over at most 250
km round-trip, and maximum wavelength change is +/- 0.04 nm (0.4 nm channel
spacing assumed), then maximum additional drift is +/- 0.2 nsec ≡ +/- 2 bits at 10
Gbit/sec. Thus additional drift due to wavelength change after failure should be very
small, if wavelength accuracy can be kept within 5% of channel spacing.
6 Systems for fibre management in optical networks
Fibre networks and, more recently, Passive Optical Network (PON) deployments have
become a huge challenge to infrastructure operators that need to efficiently and reliably
manage a large number of fibre connections between active and passive equipment, not
only inside Central Offices, but in the outside fibre plant and inside customer premises.
On the other hand, PON service provisioning and maintenance suffers from specific
restrictions due to the inherent point-to-multipoint topology of these kinds of fibre
networks.
When an Optical Network Unit (ONU) is installed inside customer premises. some
mismatches between the customer premises identification and the Optical Line
Terminal (OLT) which provides PON service coverage to the specific customer may take
place due to PON inventory inconsistencies, which can be caused, for example, due to
mistakes in manual records during network construction or even during network
operation and maintenance. In these cases, it is required a mechanism to quickly detect
this mismatch and efficiently solve the situation, so that the delay in service activation is
reduced as much as possible and the ratio of service activations per day is optimized.
In these cases, the PON-ID maintenance capability defined in [22] can be used in order
to distribute the inventory information of OLTs to the ODN in the OAM channel of the
downstream optical signal. In this way, a handheld device can be used by technicians in
order to guarantee that the location for the customer provisioning is correct, and thus
proceed with customer premises equipment installation and service activation.
Another issue can take place when a service problem for an FTTH customer takes
place. In case a fibre fault is diagnosed in the drop section of a PON, the identification of
the fibre drop corresponding to the specific customer with service problems in the
power splitter, without the need to enter customer premises, is the most efficient way
to detect, locate and repair the fibre fault and restore the customer service.
In order to achieve this, advanced technologies are required linked to network
inventory systems and to the passive equipment used to build the PON ODN.
Radio Frequency Identification (RFID) technology has been applied to the inventory
management of the elements of copper cable networks and also of the optical
distribution network (ODN) in PONs. RFID passive tags can be installed in cable
adaptors and connectors, tags having a forwarder unit that returns a tag reply with
encoded ID data when an external tag reader transmits a radio frequency signal through
the forwarder unit. RFID elements can also be powered via an antenna using RF
wireless signals, which can be used for transmitting information to an integrated circuit,
which could save at least part of the transmitted information in a memory. Due to the
small size and low cost of the RFID elements and the possibility of encoding information
inside them and read it using wireless tag readers, they are suitable for managing
inventory systems in the construction and maintenance operations in electrical cable
patch panels and in PON deployments, by attaching RF tags in the cable connectors and
adapters. Unique cable end and panel port identifiers embedded within RF tags,
combined with means for reading these tags and comparing the connections between
cables with a system database, allows checking in an automated way if the proper fibre
connections have been physically established according to the system database.
 As an example, the following figure from [23] shows a schematic approach where a
Field Operation Tool, i.e. a handheld device, is guided during a fibre connection work
order in intelligent ODN equipment. Intelligent ODN equipment are passive devices that,
when attached to the Field Operation Tool, are temporarily powered and communicated
with the management system, which checks that the fibre connections are made
according to the specific inventory information of the work order in an automated way.

Figure 29. Schematic of the architecture for intelligent ODN network element management [23] .

The former approaches are suitable approaches for the management of LR-PONs in
DISCUS.
Moreover, an active splitter monitoring concept is proposed to detect the connectivity
status of the ONTs to the splitter ports. For maintenance work it is very important to
know exactly at which splitter ports the ONTs are connected. This piece of information
helps to avoid mishandling fibre connections. The idea is to measure the availability of
the upstream channel at the input ports of the splitter. By manipulating the upstream
laser output of the ONTs and measurement of the signal at the splitter it is possible to
detect the connectivity status of the manipulated ONT. Similar to the active demarcation
point monitor device the splitter is extended by a receiver for the communication and a
µ controller for the detection process. In addition, a remote optical powering solution is
implemented to provide sufficient energy to the splitter monitor.
A photo cell is used to generate electrical power. Figure 30 shows the elements of the
splitter monitor.
An ALU set-up could be used to demonstrate a process flow to detect the connectivity
status of the ONTs. The process flow could be adjusted to consume less power as
provided by a -10dBm optical input signal at 1430nm.
 Figure 30: sub components of the splitter monitor. The central office side an ODN management module
(OMM) is added to provide the energy for the splitter monitor via a high power laser at 1430nm. The
laser (TX) and the receiver (RX) are used for the communication between OMM and splitter monitor on
the right hand side. The micro controller (µC) is used to process the upstream detection at the photodiode
array (PD array).

For DISCUS, OMM could be at LE/CO in order not to go through SOAs. Powering from
ONT side could be also possible, at the expense of increasing ONT cost.
7 Summary and conclusions

This deliverable has presented an analysis of advanced technologies for optical

supervision and fiber management in Long Reach Passive Optical Networks (LR-PONs),
which is the fundamental approach of access network in the DISCUS architecture.
A generic model for estimating the operational savings of optical supervision in PONs
has been reported in Section 3, considering two cases: first, a model for calculating
savings of remote automatic supervision is reported; secondly, the savings of a
particular multi-wavelength OTDR supervision system for WDM-PONs are calculated,
showing a relevant potential saving in the case of the dual-parental protection studied
in the DISCUS access network.
Based on the primary reference architectures of the access network considered in
DISCUS, various approaches for integrating advanced technologies for optical
supervision in LR-PONs have been reported in Section 4. Optical Time Domain
Reflectometry (OTDR) has been studied both using external devices and including
OTDR embedded inside the PON transceivers. A new proposal for fiber fault detection in
fibers using a Permanent Fiber Monitoring System (PFMS) has been reported.
The performances, advantages and disadvantages of the various PON supervision
technologies for application in the DISCUS access network have been discussed. A
feasible framework of standards from ITU-T for PON diagnostic using alarms of active
elements in DISCUS LR-PONs, considering the peculiar aspects of optical amplification
and WDM in the fiber access network, has also been described.
In Section 5, a detailed analysis of ultra-fast link restoration approaches in PONs
considering 1+1 and N:1 dual homing protection for application in DISCUS access
architecture has been reported. For an engineered implementation of the analyzed link
restoration approaches, some questions would need to be addressed. Within DISCUS,
we will address them as necessary.
Finally, in section 6 two approaches for fiber management have been described. First,
Radio-Frequency Identification technology has been proposed within DISCUS to manage
the huge amount of fibers in a LR-PON access infrastructure, with the potential to keep
a correct and updated access network topology. Secondly, remotely powered (using
light) intelligent devices in the PON infrastructure are considered for advanced features
in PON fiber management within DISCUS.
8 References
[1] M. Ruffini et al., DISCUS: An end-to-end solution for ubiquitous broadband
optical access, IEEE Comm. Magazine, Vol. 52, No 2, Feb. 2014.

[2] Jiajia Chen, Patryk J. Urban and Lena Wosinska, “Fast Fault Monitoring
Technique for Reliable WDM PON: Achieving Significant Operational Saving”,
Optical Fibre Communications OFC, Mar. 2013.

[3] ITU-T L.66, Optical fibre cable maintenance criteria for in-service fibre
testing in access networks, 2007.

[4] P. Urban, G. Vall-Llosera, E. Medeiros, S. Dahlfort, Fibre plant manager: an

OTDR- and OTM-based PON monitoring system, IEEE Comm. Magazine, Vol.
51 (2), pp. S9-S15, Feb. 2013.

[5] P. Cochrane, A.R. Hunwicks and P.A. Rosher, Novel fibre break locator for
optical duplex transmission systems, Electronics Letters, 1987.

[6] S. Zemon et al., Decay of Transmitted Light During Fibre Breaks –

Implications for Break Location, Journal of Lightwave Technology, 12(9),
1994.

[7] P. Eraerds et al., Photon counting OTDR: Advantages and Limitations,

Journal of Lightw. Technol., Vol. 28, No 6, March 2010.

[8] Patryk J. Urban, Agerekibre Getaneh, Jean Pierre von der Weid, Guilherme
Penello, Temporão, Gemma Vall-llosera, Jiajia Chen, “Detection of Fibre
Faults in Passive Optical Networks”,. IEEE/OSA Journal of Optical
Communications and Networking, Vol. 5, Issue 11, pp. 1111-1121, Nov. 2013.

[9] S. Kamei, M. Ishii, A. Kaneko, T. Shibata, M. Itoh, “NxN Cyclic-Frequency

Router With Improved Performance Based on AWG”, IEEE/OSA JLT, vol. 27,
No. 18, pp.4097-4104, 2009.

[10] New monitoring concepts for optical access networks, Jörg Hehmann and
Thomas Pfeiffer, Bell Labs Technical Journal Special Issue: Next-Generation
Wireline Access Networks, Volume 13, Issue 1, pages 183–198, Spring 2008.

[11] ITU-T G.984.2 Physical Media Dependent (PMD) layer specification

(Amd2, Appendix IV), March 2008.

[12] ITU-T G.988 ONU management and control interface (OMCI) specification,
Oct. 2012.

[13] ITU-T G.987.3 10-Gigabit-capable passive optical networks (XG-PON):

Transmission convergence (TC) layer specification, Section 14: Performance
monitoring, supervision and defects, Oct. 2010.
[14] ITU-T G Sup. 49 Rogue ONU considerations, Feb. 2011.

[15] ITU-T G.984.6 Gigabit-capable passive optical networks (GPON): Reach

extension Rogue ONU considerations, March. 2008.

[16] ITU-T G.987 G.697 : Optical monitoring for dense wavelength division
multiplexing systems, Feb. 2012.

[17] D.S. Montero et al., “Fault location in full-duplex plastic optical fibre links
using synchronized decay time detection”, 22nd International Conference on
Plastic Optical Fibres (POF 2013), Sept. 2013.

[18] J. Montalvo et al., “Localización de Fallos en Enlaces Full-Duplex de Fibra

Óptica Mediante Detección Sincronizada”, VIII Reunión Española de
Optoelectrónica, OPTOEL’13, Jul. 2013.

[19] M. Roeger et al., “Optically powered low-energy demarcation device for

monitoring FTTx networks” Proc. NFOEC, NWC4, San Diego USA (2010).

[20] Seamas McGettrick, Lei Guan, Alan Hill, David B. Payne and Marco Ruffini,
“Ultra-fast 1+1 Protection in 10Gb/s Symmetric Long Reach PON,” Proc. 39th
ECOC, London, 22-26 Sept. 2013, Paper P.6.18.

[21] A. Nag, D. B. Payne, M. Ruffini, “N:1 Protection Design for Minimising OLTs
in Resilient Dual-Homed Long-Reach Passive Optical Network,” Paper
Tu2F.7, OFC 2014

[22] ITU-T G.984.3 GPON transmission convergence layer specification,

Amendment 3, April 2012.

[23] ITU-T L.64: ID tag requirements for infrastructure and network elements
management, Oct. 2012.
Abbreviations
ASIC Application-Specific Integrated Circuit
AWG Arrayed Waveguide Grating
CAPEX Capital Expenditure
CO Central Office
DBA Dynamic Bandwidth Assignment
DC Distribution Cable
DD Demarcation Device
DF Distribution Fiber
DFB LD Distributed-FeedBack Laser Diode
EPON Ethernet Passive Optical Network
FBG Fibre Bragg Grating
FC Feeder Cable
FPGA Field Programmable Gate Array
GPON Gigabit Passive Optical Network
LE Local Exchange
LR-PON Long Reach Passive Optical Network

LSP Label Switched Path

MC Metro-Core
MIB Management Information Base
MTTR Mean Time To Repair
NEP Noise Equivalent Power
NM Network Management
ODN Optical Distribution Network
OLT Optical Line Termination
O+M Operations and Maintenance
OMCI ONT Management and Control Interface
OMM ODN Management Module
ONT/U Optical Network Termination/Unit
OPEX Operational Expenditure
OSP OutSide Plant
OTDR Optical Time Domain Reflectometer
PD Photo-Diode
PFMS Permanent Fibre Monitoring System
PON Passive Optical Network
PRBS Pseudo-Random Binary Sequence
RFID Radio-Frequency Identification
RN Remote Node
SFP Small-form Factor Pluggable
SOA Semiconductor Optical Amplifier
TDM Time Division Multiplexing
TDMA Time Division Multiple Access
TFF Thin Film Filter
ToF Time of Flight
VCSEL Vertical Cavity Surface Emitting Laser
WDM Wavelength Division Multiplexing