Global Optical Access Systems Based on ATM-PON

56 FUJITSU Sci. Tech. J.,35,1,pp.56-70(July 1999)

UDC 621.391.6:621.395.3

Global Optical Access Systems Based onATM-PON

VMitsuhiro Yano VKazuo Yamaguchi VHaruo Yamashita(Manuscript received January 20, 1999)

This paper discusses a global optical access network architecture to consolidate aneconomical ATM-PON (Asynchronous Transfer Mode-Passive Optical Network) that iscompliant with the international standard, a flexible access node configuration with agrowable IP functionality and legacy service support, and a regional transfer network.The GFR service support functionality and dynamic ATM-PON bandwidth allocation tocope with IP traffic growth are discussed. Also, since cost-effectiveness is the keyfactor in the access network, we also present an optical device and assembling tech-nology. The main parts of this architecture and technology have been successfullyhardware-implemented.

1. IntroductionThe explosive increase of traffic in data

communications is a clear sign that we have trulyentered the multimedia information age. Variouskinds of access networks such as wired (metallic,coaxial, fiber) and wireless networks will supportusers’ demands both economically and efficiently.Among these networks, optical access systems areconsidered to be essential to support the type ofhigh-speed, wide-band, and convenient multime-dia services shown in Figure 1. To cope with theever-increasing needs and traffic, the developmentof global optical access networks based onequipment that is mass-produced and thereforeinexpensive is expected.

A global optical access network will be con-figured to efficiently support data traffic such asInternet traffic as well as constant bit-rate trafficsuch as that of the legacy service. Also, ATM tech-nology with data or IP support functionality willcontinue to play an important roll in the accessnetworks as the key technology for multimediaservice support. This is because the ATM canhandle heterogeneous access schemes uniformly

and provide several required QoSs (Quality ofServices).

To configure the above access network, theATM-PON (Asynchronous Transfer Mode - PassiveOptical Network) 1),2) is considered to be the mostpromising candidate. Currently, efforts to con-struct a world-standard economical access systemare being promoted in the FSAN (Full ServicesAccess Network),3) the ATM-Forum, the ITU-T, andso on. As a result of these efforts, the recommen-dation for the ATM-PON physical layer wasformally approved as G.983.1 in the ITU-T SG15meeting on October 1998. This ATM-PON pro-vides a transport vehicle to support FTTH/B(Fiber-to-the-home or building) and FTTC/Cab(Fiber-to-the-curb or cabinet), which accommo-dates xDSL metallic interfaces currently definedbased on the ATM. Thus, we can expect to see theconstruction of global and economical optical ac-cess systems that are based on a worldwidestandard and mass-produced ATM-PONs.

In this paper, we describe an ATM-PONglobal optical access network architecture and itsequipment and focus on IP support functionality

57FUJITSU Sci. Tech. J.,35, 1,(July 1999)

M. Yano et al.: Global Optical Access Systems Based on ATM-PON

in the access node and ATM-PON layer. Also, thekey technologies to construct an economicalaccess system, i.e., optical devices and assemblingtechnologies, and several examples of their imple-mentation are described.

2. Global optical access system2.1 Flexible optical access network

and node architectureAn optical access network should be config-

ured economically and flexibly to accommodatevarious services such as POTS, ISDN, leased lines,CATV, multimedia services of the Internet, VOD,and so on. To attain this objective, we considereda consolidated access network with an ATM-PONfor economical optical subscriber access, an accessnode of the ATM switch core engine that supportsvarious kinds of interfaces and upgradable IPfunctionality, and a ring network as a regionaltransfer network for efficient bandwidth utiliza-

tion. All functionality is based on global standardsthat will be applicable worldwide.

Figure 2 shows a global optical access net-work and node architecture that Fujitsu hasproposed. For subscriber access, a single star, anactive double star, and a passive double star to-pology are supported. For the UNI interfaces,various kinds of interfaces such as legacy inter-faces and IP optimized interfaces are expected tobe applied. In the access node, various interfacesare terminated and data are routed with appro-priate signal processing, for example, QoSassurance, to designated interfaces. Multicastingfor video distribution and IP support functional-ity such as the GFR (Guaranteed Frame Rate)service4),5) and LAC (L2TP access concentration)for dynamic selection of ISPs (Internet ServiceProviders) are expected to be included. On thering network, bandwidths shared by several nodesare used efficiently, and the DWDM (Dense Wave-

Figure 1Access network for multimedia information era.

VideoVideolibrary

Office

Home

Outdoor

Head end(CATV)

Computercenter

Core network

Accessnetwork

Optical Internet

Videolibrary

58 FUJITSU Sci. Tech. J.,35, 1,(July 1999)


length Division Multiplexing) scheme can be ap-plied for large-capacity needs.

2.2 ATM-PONA PON (Passive Optical Network) topology

provides cost-effective shared media for multipleusers, and an ATM transport scheme is suitablefor supporting various kinds of multimediaservices. An ATM-PON with both of these char-acteristics can configure an economical opticalaccess network.

A schematic diagram of an ATM-PON isshown in Figure 3. An ATM-PON consists of anOLT (Optical Line Terminal) and ONUs (OpticalNetwork Units) which are connected via opticalfibers and a 1:n optical star coupler. The trans-mission scheme of downstream traffic (OLT toONU) is continuous TDM (Time division multi-plexing), and ONU receives designated cells. Thetransmission scheme of upstream (ONU to OLT)

traffic is burst mode TDMA (Time Division Mul-tiple Access), and the ONU transmits dataaccording to timeslot allocation information thatis generated by the OLT and included in the down-stream frame. Thus, the OLT in the ATM-PONallocates upstream bandwidth or timeslots flexi-bly to each ONU. Single fiber WDM (WavelengthDivision Multiplexing) is employed for bi-direc-tional transmission to reduce costs. A burst modetransmission using a 1.31 µm wavelength for theupstream and continuous mode transmission at1.55 µm for the downstream are applied.

The key issues are to accommodate burst orIP traffic efficiently as well as the legacy servicesand to produce cost-effective optical modules/devices. These items are discussed in Section 3.2and Chapter 4, respectively.

3. IP traffic accommodation technologyThe remarkable growth of data and IP traf-

Figure 2Global optical access network and node architecture.

Router

ATM45M/156M

Router

ATM-DSU

ATM-DSU

ATM-PON(ONU)

xDSLTE

Telephone (FTTC/Cab)

INS64/1500VOD/CATV

Router

Leased line

ATM-25M

ATM-PON(ONU)

#1 Card

#2 CardPONLT

Multimedia card slots

(FTTH/B)

Optical splitter(Passive)

ATM-PON

(Active)

Double star

Single starOLT

LT

LT

PONLT

ATM SW

INF

Rin

g IN

F

Ring NW

Management system

IP supportfunctionality(e.g., GFR, LAC),Multicasting, etc.

BackboneNW

POTSN-ISDN

IP

ATMVODCATV

RN



fic requires economical accommodation of the ac-cess networks and the backbone networks. Thefollowing describes an efficient service supportfunction in the access node and the ATM-PONlayer for such bursty traffic as Internet traffic. Thekey functions to accommodate Internet trafficeconomically will be an ATM-based traffic concen-tration or bandwidth sharing both in the accessnode and the ATM-PON layer.

3.1 GFR support functionThere are several candidate ATM service

classes, for example, ABR (Available Bit Rate),UBR (Unspecified Bit Rate), and GFR (Guaran-teed Frame Rate), to support bursty traffic suchas Internet traffic.

We consider that most users want a servicewhich provides a kind of guaranteed QoS withoutany special function. The GFR is designed to ac-commodate burst traffic efficiently with some levelof service guarantee and without complicated in-

teraction between users and networks. Therefore,we adopted the GFR as a new service supportfunction to enable users to explore the Internetover ATM access networks economically.

From the viewpoint of resources and band-width allocation, the resources that guaranteebandwidth for the MCR (Minimum Cell Rate) ofeach connection are reserved and the residualresources are dynamically shared by users. Wedeveloped a GFR function that includes a group-based FIFO queuing concept.6) It provides an MCRguarantee capability by using the EPD (EarlyPacket Discard) technique. A novel scheme basedon a virtual tagging technique is also implement-ed to achieve fair sharing with the FIFO queue.

For implementation of GFR, an extra ATMcell buffer memory is required to perform resourcesharing and MCR guarantee. Based on a flexiblenode architecture, we can easily add GFR featureson the OLT by using the loop-back function andthe line-interface-plug-compatible card.

Figure 3Schematic diagram of ATM-PON.

ONU #1

#1

ONU #2

#2

ONU #3

#3

ONU #n

#n

Userpremises

ONU : Optical Network UnitPR : Preamble bits

OLT

OpticalLineTerminal

Passiveopticalsplitter

32 branches

1 fiber(WDM)

#n

#2

#3

#1P

R

PR

PR

PR

Guardtime

ATM cell

#n #2 #3#1

��

Downstream: 1.55 µm(Continuous TDM)

155 Mb/s or 622 Mb/s

PLOAM cell conveysupstream time-slotallocation information(#1, #3 ,#2, ...., etc.)

Upstream: 1.31 µm(Burst mode TDMA)

155 Mb/s



Among several alternatives, we adoptedFIFO queuing for the GFR support because it hasa simple and compact implementation whichmakes it possible to efficiently use the buffer spacebetween multiple VCs. In this FIFO queuing, wehave introduced a group-based GFR buffer or aper-group FIFO queuing. Each group buffer con-sists of a FIFO and independently provides GFRfor VCs accommodated in the group. The readoutcontrol for the multiple queues is carried out inproportion to the allocated group bandwidth, i.e.,the so called “weighted round robin method” isused.7) Also, the concept of grouped VPs is applied.

In order to guarantee the minimum cell rate(MCR), a tagging function is used in conjunctionwith the FIFO queuing. When the input rateexceeds the MCR given for each VC, the ATM cellsare tagged as CLP (Cell Loss Priority) = 1 orCLP = 0. An EPD function is also implemented topreserve the AAL5 frame structure even in a con-gested state. To support the EPD, two thresholdsfor the queue length are set. The low thresholdguarantees the MCR, and the high threshold is toavoid cell discarding by buffer overflow.

Figure 4 shows the evaluation of MCR guar-antee at an ATM layer. In this evaluation, twoVCs share a group bandwidth of 20 Mb/s. TheMCR of both VCs is 2 Mb/s. The input rates of themeasured VC are 3, 5, and 10 Mb/s, and the rateof the background VC is controlled from 0 to 100Mb/s. As the background traffic increases, theoutput rate of the measured VC decreases. How-ever, even if it is heavily congested, for example,when the total input is over 100 Mb/s, the outputrate of the measured VC stays above the MCR.

Since we adopted the FIFO queuing, the ex-cess resource is shared in proportion to the inputrate of each VC. A user who sends more dataframes could obtain many resources. Moreover,the long or tandem connection traversing multiFIFOs tends to get fewer throughputs comparedto short connections. To achieve fair sharing forthe FIFO queuing, we adopt a virtual taggingtechnique. In this scheme, the frame tagging is

performed based on a virtual MCR (V-MCR), noton MCR. Here, the V-MCR is proportional to theMCR and the total V-MCR is equal to the outputrate of the FIFO. Thus, the available resource isshared by active connections in proportion to eachMCR. To avoid excess untagging caused by the V-MCR instead of the MCR, we also introduce avirtual CLP. The real CLP is set in correspon-dence to the real MCR, and the virtual CLP is setaccording to the V-MCR. The virtual CLP is onlyused for the input control of the FIFO and isnever transmitted across networks. Our proposedscheme can achieve fair sharing in proportion tothe MCR, whereas the conventional FIFO couldnot achieve fair sharing between a long and shortpath.

3.2 Dynamic PON bandwidthallocation on an ATM-PONIn the ATM-PON, the OLT (Optical Line

Terminal) allocates upstream bandwidth ortimeslots to each ONU (Optical Network Unit).It is possible to accommodate constant bit-ratetraffic like the legacy service by allocatingtimeslots periodically. However, for data trafficgenerated by PCs, routers, Web servers, etc., it isnot efficient to allocate timeslots periodically be-cause it is not constant bit-rate traffic but burst

Figure 4Evaluation of MCR performance at ATM layer.

0 10 20 30 40 50 60 70 80 90 1000

2

4

6

8

10

12

Background traffic (Mb/s)

Mea

sure

d re

ceiv

ed b

it ra

te : 10 Mb/s

: 5 Mb/s

: 3 Mb/s



traffic. Several studies8)-10) have been reported onthis issue. We have studied dynamic bandwidthallocation in the ATM-PON as a solution for thisissue and confirmed that the dynamic allocationmethod is effective for IP burst traffic accommo-dation.

Below, we describe a simulation whichshowed the effectiveness of this method. Weassumed two kinds of user traffic via UNI; one isconstant bit-rate traffic, which is generated bylegacy services like POTS/N-ISDN, and the otheris data traffic, which is generated by PCs or LANs.To accommodate the two different kinds of trafficeffectively, we prepared individual transmission-waiting FIFO buffers for each kind. In the legacyservice, the cell delay, cell delay variation, and cellloss rate should be kept to a minimum. To accom-plish this, cells for legacy traffic must haveprecedence over data traffic in the service multi-plexer block in the ONU. Moreover, the bandwidthused for transporting the legacy service is, at theleast, pre-assigned to decrease the cell delay oflegacy traffic. The bandwidth of the pre-assignedband is set by the operator and is not changed bythe bandwidth request/allocation algorithm. Byusing a dynamic bandwidth allocation method toshare the shared band among connected ONUsfor the data traffic, we can get a statistical gainand accommodate data traffic effectively. Thebandwidth of the shared band is the total up-stream bandwidth minus the bandwidth of thepre-assigned band. The ONU generates band-width requests in accordance with the queuestatus in the waiting FIFO, and the OLT dynam-ically allocates some bandwidth to each ONU fromthe shared band according to each bandwidth re-quest. Legacy service traffic such as POTS trafficwas simulated in a periodic cell generation. Thebandwidth of this traffic was 0.76% of the totalupstream bandwidth. A three-state source modelwas used for simulating burst data traffic, theaverage bandwidth of which was 9.24% of the to-tal upstream bandwidth for each ONU. The peakcell rate of the burst data traffic in the three-state

source model was 25.6 Mb/s.In the simulation, we assumed that eight

ONUs are connected to one OLT interface and allONUs have the same conditions. The transmis-sion rate, frame structure, and cell length arebased on FSAN or ITU-T G.983.1 specifications.The transmission rate is 155.52 Mb/s for each di-rection, and one timeslot is 56 octets for theupstream. An upstream timeslot can transfer auser cell (which consists of an ATM cell and athree-octet PON header), a PLOAM cell, or someshort cells. Short cells can be used for the MAC(Media Access Control) channel for such actionsas bandwidth requests. One frame is constructedfrom 53 upstream timeslots. The propagationdelay from the ONU to the OLT (Tp) is assumedto be 30 timeslots (i.e., 17.3 km equivalent). Thebandwidth for user traffic is 52/53 of the total up-stream bandwidth in the following cases.1) Condition 1: Bandwidth request using short

cellsThe ONU uses short cells to request up-stream bandwidth. The OLT allocates thetimeslot of the shared band in proportion tothe notified queue length from each ONU.Assuming that one short cell has a seven-octet length and 32 ONUs can connected toone OLT interface, eight ONUs can send eachshort cell in one timeslot and the time inter-val of short cells is four frames.

2) Condition 2: No dynamic bandwidth allocationAll upstream bandwidths are pre-assignedto each ONU equally. This condition wassimulated for reference.

Figure 5 (a) shows the required length ofthe transmission waiting FIFO buffer in the ONU.

The dynamic bandwidth request/allocationmethod can drastically reduce the size of the trans-mission-waiting FIFO buffer required for datatraffic as compared with that in the case of nodynamic bandwidth allocation.

Figure 5 (b) shows the cell delay distribu-tions for each condition. The buffer length for



evaluating cell delay is assumed to be long enoughto ensure that no cells are lost. Compared to Con-ditions 1 and 2, the cell delay without the dynamicbandwidth request/allocation is much larger thanthat with the dynamic bandwidth request/allocation.

These simulation results indicate that thedynamic bandwidth allocation on the ATM-PONlayer is useful for accommodating burst data traf-fic efficiently from the viewpoints of the requiredbuffer length in the ONU and the cell delayreduction in data traffic.

4. Optical device and assemblingtechnologies

4.1 Optical transceiver on ATM-PONA burst mode transmitter/receiver is required

to not only keep costs low but also to achieve aperformance that is high enough for flexible in-stallation. Techniques for the optical transceivermodules in the ONU and OLT on single fiber WDMATM-PON systems that emphasize a feed forwardmethod for the ONU transmitter and a newlydeveloped PD slow tail-current compensation tech-nique for the OLT receiver are described below.11)

4.1.1 Burst mode transmitterTo enable a simultaneous response, we

adopted an automatic power control (APC) free

technique by using a feed forward method whichcontrols the laser diode (LD) drive current accord-ing to the ambient temperature instead of by APC.Using non-bias current modulation to obtain ahigh extinction ratio, the turn-on delay increaseswith temperature due to an increase in the LDthreshold current (Ith). Pulse duty deviations dueto the LD turn-on delay time (td) degrade the OLTreceiver sensitivity.

To prevent this effect, we have developed afeed forward method to control the duty.Figure 6 (a) shows the principle of this method.The drive current pulse duty is adjusted to com-pensate for an insufficient td (T) according to thetemperature to keep the duty of the optical out-put constant. Figure 6 (b) shows the opticaloutput waveform for a 155.52 Mb/s non-bias mod-ulation using the above duty compensationmethod. We confirmed that the optical outputwaveform duty deviation was kept to nearly zero.

4.1.2 Burst mode receiverA burst mode receiver is required for high-

speed, high-sensitivity, and a wide detectablepower difference between burst cells (loud/softratio) for flexible installation. On the other hand,a higher transmission efficiency requires a reduc-tion of the PON overhead (OH) and a short guard

: Condition 1

Condition 1

Condition 2

: Condition 2

Cel

l los

s ra

te

Fre

quen

cy

Transmission waiting FIFO (cells) Cell delay (ms)

1.E-06 1.E+02

1.E+04

1.E+06

1.E+07

1.E+08

1.E+00

1.E+00

1.E-02

0 200 400

(a) Required buffer length in ONU (b) Cell delay distribution

0 5 10 15

Figure 5Dynamic bandwidth allocation in ATM PON.



time between burst cells.However, after the input light to a photo di-

ode (PD) is turned off, the PD exhibits a slow tailcurrent due to its minority carriers, and thiscauses bit errors at the next cell if the guard timeis not enough to decrease the PD tail current witha high loud/soft ratio. This characteristic of thePD is a problem in burst mode receivers. To solvethe PD tail current problem, a circuit compensa-tion technique is needed. We therefore developeda PD tail current compensation method that elim-inates the slow response current and a high-speed

automatic threshold level control (ATC) circuit todetect the decision level.

The slow tail current response after the in-put light is turned off is similar to the behavior ofa CR integrator. Therefore, a compensator can bemade from a simple CR circuit connected to thePD anode which detects these phenomena and avoltage follower (VF) whose output is connectedto the preamplifier (PRE) input through a resis-tor. The operation of this tail current compensatoris shown in Figure 7 (a). If a high-power signalis input to the PD, the VF output level changes as

Figure 6Principle of duty compensation and optical output waveform.

(b) Optical output waveform

100% 100%

td (T)

td (70°C)

td (0°C)

Optical outputDrive current pulse

T = 0°C

T = 70°C

(a) Principle of duty compensation

Ta = 0°C

Ta = 25°C

Ta = 50°C

Ta = 70°C

(H: 1 ns/div.)

Figure 7Principle of tail current compensation and PRE output waveform.

PRE

Rf

Voltage follower

VCR

PD

IPD

IE

Input signal

Photo current

CR detection(VCR)

Eliminated current

(IE)

PRE inputcurrent

(a) Principle of tail current compensation (b) Waveforms with and without compensator

(V: 40 mV/div., H: 1 µs/div.)

Without compensator

With compensator



the CR circuit charges/discharges, passing theeliminated current through the resister accord-ing to the voltage difference between the VFoutput level and the PRE input bias level.Figure 7 (b) shows the PRE output waveform.

The developed ATC circuit consists of a high-speed peak detector, DC-feedback (FB), and a 1/2circuit. The peak detector quickly senses theinput signal’s high level while the DC-FB holdsthe input signal’s low level. The 1/2 circuitgenerates the mid level of the input signal as adecision level. Figure 8 shows the results ofexperiments with an IC implementation of thiscircuit using a 12-bit PON-OH (guard time: 4 bits,preamble: 8 bits). The new tail current compen-sator achieves a loud/soft ratio of 27.4 dB andimproves the minimum sensitivity for a bit errorrate of 10-10 by 5.0 dB.

4.2 Assembling optical devicesThe manufacture of cost-effective, high-per-

formance, mass-produced equipment requiresinexpensive optical devices. To meet this require-

ment we have been developing innovative tech-nologies for optical devices.12) These technologiesinclude a technology for plain assembly of opticaldevices such as electronic devices and anintegration technology for optical circuits forvarious functions.

Key issues to be solved are as follows:1) For cost-effective, mass-produced devices, we

must achieve an innovative reduction in thenumber of components such as the opticallens and components for assembly and an

Figure 9Optical device assembling technique.

Corner-illuminatedphotodiode

Optical absorptionregion

Angled surface

Waveguide

Silicon substrateInput light beam

(c) Corner-illuminated PD

Taper-waveguide integratedlaser diode (LD)

Output light

Waveguide

Silicon substrate

(b) Taper-waveguide integrated LD

Laser diode (LD)

Solder padMarker

V-groove

Optical fiber

(a) Passive alignment

Silicon substrate

Figure 8Experimental results for receiver sensitivity.

-36

-34

-32

-30

-28

-30 -20 -10 0

Cell 1 average input power (dBm)

Cel

l 2 m

inim

um s

ensi

tivity

(dB

m)

Cell 1 Cell 2

-6 dBm -33.4 dBm

27.4 dB

With compensator

Without compensator

5 dB

-



innovative assembly process (particularly forthe alignment for optical semiconductordevices).

2) For high-performance devices, one key issueis effective assembly of optical devices forwavelength division multiplexers.To satisfy the first requirement, we have

developed an innovative assembly technology foroptical semiconductor devices such as laser diodesand photo-detectors. Conventionally, the align-ment is carried out with the semiconductor devicesactive, i.e., the lens is adjusted with the opticalbeam. This technique has some disadvantages: itrequires expensive alignment components, takesa long time to align, and requires complicated fa-cilities. To overcome these disadvantages, we havedeveloped an innovative assembly technique called“passive alignment”. As shown in Figure 9 (a),

optical alignment is carried out with only a me-chanical alignment of the laser diode to a markeron a silicon substrate and mechanical fixing of thefiber to a V-shape groove etched on the silicon sub-strate. We have achieved a highly uniform opticalcoupling efficiency with this technique.

Moreover, to realize a high coupling efficien-cy, which will lead to various applications of thepassive alignment technique, we have developednew optical semiconductor devices. For laserdiodes used as a light emitter, a taper-waveguide(Figure 9 (b)) has been integrated into a laserdiode. This structure is equivalent to an integrat-ed lens in a laser diode, and it provides a highoptical coupling efficiency without the need for anoptical lens. For a photodiode used as a light re-ceiver, we have realized a high optical couplingefficiency for corner illumination with an angled-

Sub-INF

OLT

Inter-INF

PON-INF

GFR-card

Inter-INF

Intra-INFSingle star

Active star

Passive star

ATM-PON

Remotenode A

TM SW

Management system (Control/Alarm)

Residual 2.4 Gb/s portsavailable for extendedfunctionality such asRing-INF, CATV, existing legacy-INF,LAC, etc.

Operator interface (10 Base-T)

Figure 10OLT configuration.



surface (Figure 9 (c)). This structure is more suit-able for a surface mounting configuration than theconventional one.

To realize a high-performance function for thesecond requirement, we have developed a hybridintegration assembly technology for planar light-wave circuits (PLCs) and optical semiconductordevices on a silicon substrate, which we call thePLC platform technology. An embodied implemen-tation is described in Section 5.4.

5. Hardware implementation for globaloptical access systemThe global optical access system architecture

described above was applied to fabricate an eco-nomical and flexible access system. The mainhardware implementation to be made commercial-ly available is described below.

5.1 Equipment implementationFigures 10 and 11 show the configuration

and a photograph of the fabricated OLT, res-pectively. A total of 40 ATM-PON interfaces,subscriber-line interfaces based on SDH, or intra-office and/or inter-office interfaces based on SDHcan be accommodated per unit. Using a 150 mm

330 mm printed circuit board for these inter-faces achieves high-density subscriber accommo-

Figure 12Prototype GFR card.

Figure 11Photograph of OLT.

dation, which leads to an economical node. Theseinterface cards are slot-compatible (slot-inter-changeable) and they are configurable with botha non-redundant and redundant scheme forintra-office and inter-office interfaces. The ATM-SW core engine provides a 20 Gb/s throughput andsupports the QoS for CBR and UBR services andalso the GFR service with an add-on GFR card.Some residual 2.4 Gb/s ports of the ATM-SW areprepared for supporting the global optical accessnode functionality described in Chapter 2; that is,for supporting the ring interfaces, CATV interfac-es, inter-working interfaces for existing legacytelephony services, or IP-specific functionalitysuch as LAC. For the ATM-PON interface, a 60byte-PON cell (7 byte-overhead and 53-byte ATM

Interface slots Note)

(Total of 40 interfaces)

Extended slots

SW(0)/SW(1) Cont/Alarm

Note: Total of 40 interfaces for ATM 155 Mb/s-SubscriberINF (non-redundant), PON-INF (non-redundant),ATM 155 Mb/s/622 Mb/s Intra-/Inter-office INF(redundant), or GFR card (redundant)

Interface speed

Max. number of VCs

Max. number of groups

MCR

Output rate

Total buffer size

Power consumption

622.08 Mb/s

4096

Item

2048

0.1 to 6.0 Mb/s for each VC

0.1 to 149.76 Mb/s for each group

29127 cells

Specification

18 W

Table 1Main features of prototype GFR card.



cell) structure was used in this implementationand a recently defined ITU-T G.983.1 compliantinterface will be introduced with a dynamic band-width allocation method.

5.2 GFR implementationTable 1 shows the specifications of the

prototype GFR card, and Figure 12 shows a pho-tograph of it. By using a 150 mm 330 mmprinted circuit board with an attached sub-board,this card has been made plug-compatible with a622 Mb/s interface card. Simply plugging this cardinto any interface slot gives the GFR service func-tion to the OLT. Using the GFR card, 4096 VCscan share the bandwidth with guaranteed MCR.

Field Programmable Gate Arrays (FPGAs) in con-junction with control memories perform the mainfunctions of GFR such as CLP tagging, EPD, andinput/output control of buffer memories.

5.3 Optical modulesFigure 13 shows a photograph of the ONU/

OLT modules. These modules are 55 49 8.5mm. By using an external WDM filter, they canbe applied to single fiber WDM ATM-PONsystems. The OLT module is composed of a con-tinuous mode transmitter and a burst modereceiver, and the ONU module is composed of aburst mode transmitter and a continuous modereceiver. Figures 14 (a) and (b) show the blockdiagrams of the OLT transceiver module and theONU transceiver module, respectively.

Table 2 summarizes the performances ofboth modules without a WDM filter. These mod-ules have been applied to commercial systems.

5.4 PLC platform technologyAs shown in Figures 15 (a) and (b), various

optical functions, for example, a wavelength divi-

Figure 14Block diagrams of OLT and ONU optical modules.

ONU module

1.3 µm Burst mode transmitter

FF Duty LD DRV. LD

Controller

Temp. detector

1.5 µm Continuous mode receiver

AGC

RetimingPRE

PIN-PD

DATA INCLK IN

DATA OUTCLK OUT

(a) Block diagram of ONU module

OLT module

1.5 µm Continuous mode transmitter

1.3 µm Burst mode receiver

PRE

LD LD DRV. Duty FF

Controller

Temp. detector

PDATC

RESET

Peak 1/2

DATA OUT

Tail currentcompensator

LIM

(b) Block diagram of OLT module

DC-FB

DATA INCLK IN

Figure 13OLT/ONU optical modules for ATM-PON.



sion multiplexer and a splitter were fabricated byusing PLC platform technology. For further im-provement, we have developed technologies suchas the encapsulation of optical semiconductor de-vices with plastic materials. For the applicationof the plastic encapsulation techniques common-ly used in LSIs to optical semiconductor devices,we developed a double plastic-layer technique.Around optical semiconductor devices, we use atransparent silicone resin which passes lightsignals and relieves thermal stress. We then

encapsulate with epoxy resin to improve humidi-ty resistance. The optical modules based on thePLC platform technology will be used in the nextcommercial systems.

6. ConclusionWe presented a global optical access network

architecture and equipment to provide an econom-ical and efficient multimedia service support,focusing on the ATM-PON and IP support func-tionality such as GFR service support and dynamic

Transmitter

Wavelength

Mean launch power

Continuous "0" output level

Extinction ratio

Pulse duty deviation

Receiver

Minimum sensitivity

Maximum overload

Receivable power difference

Operating temperature range

Supply voltage

Size

Power consumption

Items

Burst mode

1297.1 to 1327.8 nm

+0.6 to +0.9 dBm

-80 dBm

–

3 %

Continuous mode

-37.2 dBm

-3.0 dBm

–

0 to 70°C

+3.3 V ± 5%

55 × 49 × 8.5 mm

0.8 W

ONU

10 to 60°C

+3.3 V ± 5%

55 × 49 × 8.5 mm

0.9 W

Continuous mode

1553.3 to 1558.6 nm

+0.6 to +0.9 dBm

–

11 dB

–

Burst mode

-33.5 dBm

-6.0 dBm

27.4 dBm

OLT

Table 2Main features of OLT/ONU optical modules.

(b) Photograph of PLC module(a) Configuration

Figure 15PLC platform technology.

Corner-illuminated photodiode

Taper-waveguide integratedlaser diode

Optical circuit (PLC)

Alignment marker 1 cm



bandwidth allocation in the access node. We thendescribed the key technologies for constructing aneconomical optical access system; namely, a high-performance burst mode transmitter/receiver,innovative techniques for reducing components,and the PLC platform technology. This architec-ture and technology has been successfully appliedto commercial systems. A dynamic bandwidth al-location mechanism and fully PLC-based opticalmodules are currently under development for thenext commercial systems.

References1) Y. Takigawa, S. Aoyagi, and E. Maekawa:

ATM based Passive Double Star systemoffering B-ISDN, N-ISDN, and POTS.GLOBECOM’93, pp.14-18, Nov. 1993.

2) N. Terada, A. Kondou, H. Uematsu, K.Hamada, H. Ishii, Y. Takigawa, and S.Ishikawa: Field trials of MPEG-2 basedCATV and VOD system using ATM-PONaccess networks. ISSLS 98, pp.153-159, Mar.1998.

3) Full Service Access Network (FSAN) Initiative.URL (http://www.labs.bt.com/prof soc/

access ).4) ATM-Forum: Traffic Management Working

Group Baseline Text Document. BTD-TM-01.02, July 1998.

5) I. Omiya, S. Naka, and Y. Maeda: ATMSHARE-LINK SERVICE BASED ONATMPON TECHNOLOGY. GLOBECOM’98,Access Networks Mini Conference, pp.187-192, Nov. 1998.

6) J. Tanaka, M. Goto, K. Nisimura, M. Okuda,and T. Ishihara: Implementation of GFR

Service Class Into ATM-Based AccessSystem. GLOBECOM’98, Access NetworksMini Conference, pp.21-26, Nov. 1998.

7) M. Katevenis, S. Sidiropoulos, and C.Courcoubetis: Weighted Round Robin CellMultiplexing in a General-Purpose ATMSwitch Chip. IEEE Journal on Selected Areasin Communication, 9, 8, pp.1265-1279 (Oct.1991).

8) H. Ichibangase and F. Hidani: A Study onAccess Method about ATM-PON. In Proceed-ings of The 1997 IEICE Spring Conference,pp.440, Mar. 1997.

9) M. Yoshino, S. Touma, and S. Tsutsumi:DYNAMIC BANDWIDTH ALLOCATIONFOR ATM-PON SYSTEM HANDLINGTCP/IP TRAFFIC. GLOBECOM’98, AccessNetworks Mini Conference, pp.27-32, Nov.1998.

10) M. Miyabe, M. Kasa, K. Tajima, T. Shinomiya,and H. Yamashita: A Study of Dynamic Band-width Allocations for ATM-PON. IEICETRANS. COMMUN., E81-B, 12, pp.2364-2370 (Dec. 1998).

11) K. Mori, T. Akashi, Y. Tochio, H. Nobuhara,K. Tanaka, N. Nagase, Y. Akimoto, H.Kitasagami, K. Shimizu, T. Yamane, A. Abe,and M. Kawai: 155.52Mb/s Optical Transceiv-er Modules for ONU/OLT on ATM-PONSystems. ECOC ’97, Vol. 3, pp.363-366,September 1997.

12) K. Terada, K. Tanaka, K. Miura, and M. Yano:High Power and High Sensitivity PLC Mod-ule Incorporationg a Novel Passive AlignmentMethod. J. of Lightwave Technology, 16, 1,pp.66-72 (Jan. 1998).



Mitsuhiro Yano received the Ph.D. inElectronics Engineering from the Uni-versity of Tokyo, Tokyo, Japan in 1973.He joined Fujitsu Laboratories Ltd. in1973 and has been engaged in re-search and development of opto-elec-tronic devices and packaging tech-nologies. He is a member of the Insti-tute of Electronics, Information, andCommunication Engineers (IEICE) ofJapan and a senior member of the IEEE.

Haruo Yamashita received the B.S.degree in Physics from the Universityof Tokyo, Tokyo, Japan in 1978 and theM.S. degree from Tokyo Institute ofTechnology in 1980. In 1980 he joinedFujitsu Limited, Kawasaki, Japan, andsince 1983 he has been with FujitsuLaboratories Ltd., where he is currentlya Research Fellow in the NetworkSystems Laboratories. He has beenengaged in research and development

of broadband optical access and backbone systems. He is amember of the Institute of Electronics, Information, and Com-munication Engineers (IEICE) of Japan.

Kazuo Yamaguchi received the B.E.and M.E. degrees in Metallurgical En-gineering from Yokohama National Uni-versity, Yokohama, Japan in 1971 and1973, respectively. From 1973 to 1996he was with Fujitsu Laboratories Ltd.,Kawasaki, Japan. In 1996 he movedto Fujitsu Limited, Kawasaki, where heis currently General Manager of theOptical Access Systems Division. Hehas been engaged in development of

broadband optical access systems, including STM-PON andATM-PON. He is a member of the Institute of Electrical andElectronics Engineers (IEEE) of the U.S. and the Institute of Elec-tronics, Information, and Communication Engineers (IEICE) ofJapan. Mr. Yamaguchi received the Young Engineer Award(Gakujutsu-Syoreisho) from the IEICE of Japan.

Global Optical Access Systems Based on ATM-PON

Documents

Transcript of Global Optical Access Systems Based on ATM-PON