Current and next-generation PONs:
A technical overview of present and
future PON technology
Elmar Trojer, Stefan Dahlfort, David Hood and Hans Mickelsson
The authors describe the evolution of GPON and associated trends in the
industry. They briefly introduce and compare current PON technologies
and describe options for reach-extended GPON and next-generation ac-
cess systems.
Introduction to passive
optical networks
Video components in the service bundle
call for higher bandwidth, forcing most
telecommunications operators to contem-
plate upgrading or completely renewing
their copper-centric legacy access networks.1
Gigabit-capable passive optical networks
(GPON) and Ethernet passive optical net-
works (EPON) are two standards that open
the door to new opportunities both for ven-
dors and operators.
Major vendors, including Ericsson, have
added PON technology to their broadband
access portfolios, and operators around the
world have shown considerable interest in de-
ploying this technology in combination with
VDSL2 (fiber to the cabinet, FTTC) or as res-
idential access (fiber to the home, FTTH).1
The three major PON standards are BPON
(broadband PON, currently only of historical
interest), GPON, and EPON.4-6
BPON and its successor GPON are
ITU-T recommendations sponsored by
FSAN, a vendor and operator committee.7
EPON is an IEEE option developed by the
TERMS AND ABBREVIATIONS
BPON Broadband PON
CDR Clock detection and recovery
DSL Digital subscriber line
EFM Ethernet in the first mile
EPON Ethernet PON
FEC Forward-error correction
FSAN Full-service access network
FTTC Fiber to the cabinet
FTTH Fiber to the home
GPON Gigabit-capable PON
IEEE Institute of Electrical & Electronics RE Reach extender
Engineers
ITU-T International Telecommunication SDH Synchronous digital hierarchy
Union – Telecommunication
Standardization Sector
NGA Next-generation architecture
OA Optical amplifier
OAM Operation, maintenance and
administration
64
IEEE Ethernet in the First Mile (EFM) initia-
tive. Given that operators are driving GPON
standardization via FSAN, the GPON stan-
dard reflects operator needs more directly
than does EPON.
Although all three systems work on the
same principle, there are several differences
between them (see Table 1).
GPON evolution
After some minor initial deployments of
BPON, the industry realized too late that a
BPON-based optical distribution network
(ODN) could not be incrementally upgraded
to any next-generation technologies. In short,
the logistics of upgrading an entire PON
simultaneously were daunting, and the cost
of installing a parallel upgrade PON was
prohibitive. Out of this experience grew the
requirement that it must be possible to make
incremental next-generation upgrades on the
same ODN.
There were, and still are, many candi-
dates for next-generation PON systems.
The ITU community thus drafted G.984.5,
which reserved wavelengths for use by next-
generation applications without constraining
ODN Optical distribution network
OLT Optical line termination
OMCI ONT management control
interface
ONT Optical network termination
ONU Optical network unit
PDFA Praseodymium-doped fiber
amplifer
PON Passive optical network
QD-SOA Quantum-dot SOA
RPT Remote protocol terminator
SOA Semiconductor optical amplifier
TDFA Thulium-doped fiber amplifier
TDM Time-division mulitplexing
VDSL2 Very high-speed DSL 2
WDM Wavelength-division
multiplexing
them. By contrast, the EPON community
saw bandwidth limitations as the most seri-
ous problem and immediately began work
on 10Gbps EPON (802.3av) as the successor
to 1Gbps EPON.
The economics of real deployments also
led ITU-T, encouraged by FSAN, to start
working on a reach-extension standard for
GPON, provisionally designated G.984.re.8
This specification allows optical amplifiers
or signal regenerators to be used
• to extend reach to 60km;
• to increase the split ratio; or
• to achieve a combination of both.
The next-generation architecture (NGA)
defines two stages of evolution (Figure 1):
NGA1 and NGA2. NGA1 is compatible
with GPON deployments in accordance with
G.984.5. Compatibility with a GPON reach
extender is also expected, but has not yet
been explored in detail. It is anticipated that
NGA1 systems will be commercially avail-
able around 2010. Some NGA1 candidates
are:
• XGPON1, which supports data rates of
10Gbps downstream and 2.5Gbps up-
stream;
• symmetric XGPON2, which supports
10Gbps downstream and upstream; and
• WDM option to overlay PONs and point-
to-point connections on the same fiber
infrastructure in G.984.5 enhancement
bands.
Recognizing that G.984 definitions will not
be suitable forever and new developments
will obviate the need for backward compat-
ibility, ITU-T has not constrained NGA2
by the GPON ODN. It is anticipated that
NGA2 products will be available around
2015. NGA2 may use a new fiber network,
introducing in particular the ability to use
dense wavelength-division multiplexing
(WDM) splitters instead of power splitters
to separate users via different wavelengths on
the same ODN.
Long-reach PON
Basic GPON systems support a maximum
physical reach of 20km on a 32-way split
or 10km on a 64-way split. Although this
seems sufficient for most deployment cases,
the industry seeks an extended link budget
for two reasons:
• Longer reach: Reach-extended PONs sup-
porting 100km would allow thousands of
today’s central offices to be consolidated to
a handful of huge metro nodes, thereby
simplifying the network architecture and
Ericsson Review No. 2, 2008
 reducing operating expenses.9 A modest 8VeVX^in
increase in budget would make it possible
to reach remote customers.
• Increased split ratio: A split ratio of 64 or
even 128 would reduce the cost per sub- C<6'
scriber of a PON system.10 PON technol- C<6&^cXajY^c\ :#\#]^\]Zg"gViZI9B
adc\"gZVX]dei^dc 9L9B!
ogy is being driven by the need for higher 89B!
bandwidth, but low take rates also mean d[ D;9B!ZiX#
JhZ bdc i
L9Bdei^dcid XdbebZc
that the business case for high-rate services ZcVWaZhiVX`ZY<$M<EDC!
(
Zfj
^
is a future promise rather than a current E"id"EdkZgaVn!dgWdi]
'
fact. As a consequence, operators want to cXZ
iZ
serve a large number of basic users from Zm^h M<"EDC
<EDC & 8d Je/'#*<!*<!&%<
each PON by sharing common equip- 9dlc/&%< Hea^iiZg[dgC<6'
ment. edlZghea^iiZgdg
hdbZi]^c\cZl
There are two basic ways of extending reach
EdlZghea^iiZgYZeadnZY[dg<^\VEDC
(Figure 2). A single-sided extender enhancing cdgZeaVXZbZci$cdVYY^i^dc
the optical transceiver function of the optical
line termination (OLT) keeps the ODN pas- Cdl s'%&% s'%&*
sive and allows for a limited budget increase
(Figure 2, top). Figure 1
An active mid-span extender (reach ex- Coexistence and upgrade tracks as discussed in FSAN.
tender, sometimes designated RE), as cur-
rently drafted in G.984.re, would allow the
optical trunk line to be extended to achieve
a total reach of 60km. This represents the
maximum logical reach of the GPON proto-
col layer (Figure 2, bottom).
To succeed, a reach extender must comply
with the following fundamental require-
ments:
1. It must be cost-effective – that is, there
must be a business case that supports the
deployment of reach extenders (cost ben- Figure 2
efits over keeping exchange buildings). Reach-extension topologies.
2. It must be transparent to existing GPON
OLTs situated on the network side and op- 8aVhh8
tical network terminations (ONT) on the DCJ )%`b
the user side. JC>
3. It must be available for early GPON de- G$H
H^c\aZ
ployment. D9C h^YZY DAI
ZmiZcYZg
Figure 3 compares different approaches to en- >;9 H$G HC>
hancing the performance of reach extenders. DCJ G$H Edhh^WaZ
Forward-error correction (FEC) techniques JC> ZmiZcYZYDAI
can extend reach by typically 5km. Likewise,
optical amplifiers (fiber, semiconductor or VH^c\aZ"h^YZYZmiZch^dc
distributed) can increase the link budget to
60km or more.
8aVhh7%¶'%`b 8aVhh7 %¶)%`b
Optically amplified PONs DCJ )%¶+%`b
Optical amplifiers (OA) can be used in
JC>
GPON either to boost transmit power and
G$H B^Y"heVc
receiver sensitivity on the OLT side or to D9C ZmiZcYZg DIA DAI
work as an in-line amplifier in mid-span >;9 >;I H$G HC>
reach extenders. DCJ G$H
A typical GPON RE based on amplifica- JC> D9C Dei^XVaY^hig^Wji^dccZildg`
tion comprises DIA Dei^XVaigjc`a^cZ
>;9 >ciZg[VXZ[^WZgYgde
• wavelength filters to separate upstream
WB^Y"heVcZmiZch^dc >;I >ciZg[VXZ[^WZgigjc`
and downstream signals;

Ericsson Review No. 2, 2008 65

TABLE 1: MAJOR PON TECHNOLOGIES AND PROPERTIES OLT receivers. To achieve carrier-class
performance, new OLT blades would have
Characteristics EPON BPON GPON to be designed, which violates rules 2 and
Standard IEEE 802.3ah ITU-T G.983 ITU-T G.984
3 for REs.
• Cost: Optical components (SOAs and fib­
Protocol Ethernet ATM Ethernet, TDM
er amplifiers) are prohibitively expensive
Rates (Mbps) 1250 down / 1250 up, 622 down, 155 up 2488 down, 1244 up (rule 1).
8b10b-encoded
Span (km) 10 20 20 Regenerated PONs
Split-ratio 16 32 64 In a regenerated PON, the signal is refreshed
by means of optical-electrical-optical conver-
sion. Figure 4 illustrates a typical 3R RE
(3R: reamplify, re-shape and re-time the
signal). The 3R RE contains OLT and ONT
transceivers as well as clock detection and re-
covery (CDR) units for upstream and down-
• amplifier for each direction; and nologies in an ODN. Note: he numbers in stream directions (a burst CDR is needed for
• optical bandpass filters in the upstream parenthesis indicate the relative performance the upstream).
and possibly downstream directions for of the different amplifier characteristics. For management and alarms, the RE may
noise limitation. As can be seen, semiconductor optical am- contain an optical network unit (ONU)
Fiber and optical amplifiers can be used plifiers (SOA) are the most promising breed that terminates physical layer OAM and the
to extend the reach of a GPON system of amplifiers. Nevertheless, SOA-based reach OMCI (ONT management control interface)
with a 64-way split to 60km. Thulium- extenders exhibit high noise figures (espe- management channel. The embedded ONU
doped fiber amplifiers (TDFA), quan- cially in the O-band) and are comparatively is part of the management model currently
tum-dot semiconductor optical amplifiers expensive. On the other hand, REs based on under discussion in G.984.re.
(QD-SOA) and Raman amplifiers are em- optical amplification are technically achiev- REs based on signal regeneration also ex-
ployed in the downstream band (S-band), and able but practically questionable due to hibit some problems:
praseodymium-doped fiber amplifiers transparency and cost: • Burst recognition from ONT: For the RE to
(PDFA) and QD-SOAs are employed in the • Transparency: OAs for GPON bands are decode upstream bursts correctly, basic
upstream band (O-band). not transparent. Even when equipped with GPON framing parameters, such as the
Table 2 gives a high-level summary of the filters, noise results in a reduced-extinction preamble and inter-burst gap, must be ex-
pros and cons of different amplifier tech- ratio, which is incompatible with standard tended, which results in a loss of protocol
efficiency and transparency (rule 2).
• Software incompatibilities: Delay and optical
layer measurements lose their standard-
ized meaning. Likewise, troubleshooting
becomes more complicated, because new
TABLE 2: OPTICAL AMPLIFIER SUMMARY alarm correlations and suppression must
( NOT SUITED, SUITED, VERY WELL SUITED) be added, which increases complexity and
cost.
Raman • Issues with standard OLT: A 3R RE renders
xDFA QD-SOA (non-linear some OLT burst receiver functionality and
Characteristic (fiber amplifiers) (optical amplifiers) amplifier) OLT dynamic range capabilities worthless.
Band Upstream Downstream Upstream Downstream This further motivates the need for special
(O-band) (S-band) (O-band) (S-band) OLT blades (rule 2).
Maturity
Remote protocol termination
Size (2~3) (2~3) (0.5~1) (0.5~1) (1) Purpose-built for a TDM-shared optical tree,
Cost (4~10) (3~8) (3) (2.5) (0.5~1) the G.984 protocol was never intended to
Power
compete with robust protocols such as syn-
(2~4) (2~4) (0.5~1) (0.5~1) (0.5)
consumption chronous digital hierarchy (SDH) or 802.3
Ethernet on point-to-point links.
Gain
Reach extension is nothing more than
Optical adding a long point-to-point trunk line
bandwidth to the point-to-multipoint GPON ODN.
Noise On the trunk line, however, the point-to-
multipoint specialization of the GPON
Overall
protocol is a disadvantage in terms of receiver

66 Ericsson Review No. 2, 2008

6eea^XVWaZiZX]cdad\n

GZbdiZegdidXda
iZgb^cVi^dc

7jghi(GgZ\ZcZgVidg

8aVhh9
Dei^XVaVbea^[^Zg

8aVhh8 dei^Xh

;:8

DCJ9;7

DCJ;E

8aVhh7

% &% '% (% )% *% +% ,% -% .% &%% &&% &'% Figure 3

IdiVagZVX]`b Reach-extension options for PONs.

synchronization, dynamic range, burst mode

power management, noise margins and jit-
ter budget. Moreover, the benefits of proto-
cols designed for point-to-point transmission
links are not available. Hence, the remote
protocol terminator (RPT), which circum-
vents these problems.
In terms of delay, budget, split ratio,
maintenance, existing silicon, existing opti-
cal components, and software, an RPT looks
exactly like a GPON OLT to the shared part
of the ODN (the TDM tree). But instead of
the PON protocol, the RPT uses the right
tool for each exchange. For best results, Figure 4
10Gbps Ethernet uplinks are recommended, RE block diagram of a 3R regenerator.
although one may use SDH depending on
market interest. Standard Ethernet protec- 8dcigda
VcYYViV
tion schemes, such as link aggregation group 8dcigda
VcY
(802.3ad) and rapid spanning tree protocol Y^V\cdhi^Xh <EDCDCJ 8dcigdaVcYY^V\cdhi^Xh
(802.1w), exist for a redundant system. Hd8
The RTP has been designed to host next-
generation access technology options, such
8dcigda
as higher speed TDM-based PONs, WDM- DCJ VcYY^V\cdhi^Xh DAI
8dcigda
based PONs, or both. igVchXZ^kZg VcYYViV 8dcigda igVchXZ^kZg
VcYYViV
Estimates show that the cost, size and
power consumption of a protected remote IM 789G GM
site RPT are the same as for a regenerator but
considerably less than for an optical amplifier
(rule 1). In contrast to other RE approaches, >;I 9ViV 9ViV >;9
GM 89G IM
an RPT is based on existing technology,
standards, and components (rule 2) and can
be delivered within the normal product
development time frame (rule 3). Table 3 8adX`
compares G.984.re REs and the RPT.

Ericsson Review No. 2, 2008 67

 &%<<EDC JH JH 9H
&*+%
&)*%
&('% Dei^dc(
8L9Bdei^dc JH Dei^dc&"' &**%
<#.-)#* &)%%
&(%%
&*-% &+'*
&((% &)*%
9;7dei^dc Dei^dc&"& Dei^dc'
JH
<#.-)#* &'.% &)&*
&*(%
&*+%

&(+% &*%% :7 :7k^YZd

<EDC$:7 9H &**% &*+*
;Edei^dc JH
&'+% &)-% :7 :7Y^\^iVa
&*(.
&(+% &*-%
7EDC JH GZhZgkZY 9H
&'+% &)-%
Figure 5
& ' ( ) * + , - . &% && &' &( &) &* &+ &, &- &. '% '&
Wavelength allocation for GPON (inclu- &',& &'.& &(&& &((& &(*& &(,& &(.& &)&& &)(& &)*& &),& &).& &*&& &*(& &**& &*,& &*.& &+&& &+(& &+*& &+,&

ding G.984.5) and NGA on the CWDM grid &(+% &)+% &*(% &*+* &+'* &+,*
D"WVcY :"WVcY H"WVcY 8"WVcY A"WVcY J"WVcY
(G.694.2). &'+%

Next-generation access Shared capacity of at least 10Gbps in the WDM. In particular, it includes
Now that GPON has been standardized and downstream and 2.5 or 5Gbps in the up- • wavelength ranges to be reserved for future
is in production, the optical access commu- stream direction. use (Figure 5) – G.984.5 specifies three op-
nity has begun discussing candidate technol- 3. NGA1 must be highly flexibility in terms tional enhancement bands with option 1
ogies for next-generation access. The high- of coexistence (GPON and G.984.re reach in the E band (water-peak band), option 2
level requirements for an NGA1 system are extenders), upgradeability to higher split in the C- and L-bands, and the RF band,
already clear, as follows: ratios, and support of special-purpose as option 3, for future services provided
1. Fiber-lean scenario – that is, coexistence overlays (for instance, business services). video overlay is not deployed;
with working GPON on the same ODN. 4. Volume cost comparable to GPON. • wavelength-blocking characteristics for fil-
NGA1 must support upgradeability, one ters that protect the GPON downstream
subscriber at a time. Service overlay a la G.984.5 signal in the ONT/ONU from interference
2. Major improvement in performance over G.984.5 defines wavelength ranges reserved from the new bands; and
GPON in terms of rate and reach or split. for additional service signals overlaid via • GPON upstream wavelength-reduction
options, to free spectrum in the O-band
for future services. In all likelihood the
distributed feedback (DFB) laser option
will be most widely deployed.
TABLE 3: COMPARISON OF G.984.re REs AND RPT XGPON and overlaid PONs
There are basically two ways of increasing
G984.re RPT the capacity of a TDM-based PON system
Ready for early GPON de- No Can be developed expedi-
for NGA1 in compliance with the main re-
ployment tiously – no unknowns quirements.
Figure 6 shows that one can speed up a
Interoperability with existing No ONT – OK. OLT – OK with
OLT/ONTs some modification
regular 2.5Gbps GPON system to 10Gbps
(XGPON) and overlay it on a separate
Can be used for 20/40/60km In theory. Any number of OK
extension and 64- or 128-way issues must be resolved
G.984.5 wavelength. One might also over-
split lay four colored 2.5Gbps GPON systems via
WDM multiplexers on four different wave-
40 km trunk OK OK
lengths on the same ODN to obtain 10Gbps
100 km trunk Not without change to delay OK overall system capacity.
time definitions
The complexity and cost of an XGPON
Compatible with existing No OK system is in the high-power, high-speed opti-
G984 series cal transceivers, which use electro-absorption
Hardened for outside plant TBD TBD modulated lasers. The ODN itself remains
deployment unchanged. The cost of an overlaid PON
Eye-safe Depends on choice of OK is dominated by the multiplication of OLT
technology equipment and the interim upgrade of the

68 Ericsson Review No. 2, 2008

ODN with wavelength-selective splitters set of GPON to make the technology more I9BhZg^VagViZ

(hybrid splitters, which comprise low-ratio flexible in terms of deployment and services HnhiZb
power splitters and four-channel WDM and to make the fiber infrastructure “future- XVeVX^in

multiplexers). proof.”
Ericsson has put considerable effort into The long-reach option, as currently speci- )%<Weh
&L
)%<Weh
)L
)%<Weh
&+L
pursuing the XGPON track and is currently fied in FSAN, will make it possible to use +)%<Weh
demonstrating an experimental high-speed active reach extenders to increase the span
&%<Weh &%<Weh &%<Weh
GPON system that, following the principles of the system to 60km. This approach can &L
V
)L &+L
of coexistence and low-cost design, supports be employed to reach remote customers or to &+%<Weh
10Gbps downstream and 2.5Gbps upstream. host more users on a PON. '#*<Weh '#*<Weh '#*<Weh
In light of an RPT, the key parameters for In terms of coexistence, GPON is fully &L )L
W
&+L

a next-generation system are high port den- prepared for the future. Next-generation sys- '#*<Weh &%<Weh )%<Weh

sity, small footprint, and reduced power con-
sumption.
Conclusion
Current GPON technology is a powerful op-
tion for deep-fiber broadband access. Serious
efforts to standardize and develop this tech-
nology have steadily extended the feature
tems – for instance, XGPON, overlaid PON, L9BX]VccZaXdjciL
or some other technology, will work on the
GPON fiber infrastructure, allowing opera- Figure 6
tors to capitalize on their investments for de-
cades to come.
Upgrade potential of NGA1 downstream
capacity.
Ericsson’s upgrade strategy is very clear in
terms of long-reach and next-generation op-
tical access using remote protocol termina-
tion and XGPON.

REFERENCES

1. Bernstein Research: Fiber: Revolutionizing the Bell’s Telecom Networks. May 2004,
2. Baker J., Cagenius, G., Goodwin, C., Hansson, M. and Hatas, M.: Deep-fiber broadband
access networks. Ericsson Review, Vol. 84(2007)1, pp. 4-8
3. G.993.2, Very high-speed digital subscriber line transceivers 2 (VDSL2). ITU-T, 02/2006
4. 4 G.983.1-5, A broadband optical access system with increased service capability by
wavelength allocation. ITU-T, 03/2001
5. G.984.1-5, Gigabit-capable Passive Optical Networks (GPON). ITU-T, 03/2003
6. IEEE 802.3ah, Ethernet in the First Mile Task Force, www.ieee802.org/3/efm/
7. Full-service Access Network (FSAN), www.fsanweb.org
8. G.984.re draft, Draft G-PON optical reach extension. ITU-T, 12/2007
9. Davey, R.P., et al.: Designing long-reach optical access networks. BT Technology Journal,
Vol. 24 No. 2, 04/2006
10. Trojer, E., et al.: Optical Access Network Evolution, NOC 2007, Stockholm, 06/2007

Ericsson Review No. 2, 2008 69