Electronic Design

Fiber Comes Home

A cheaper version of the traditional optical network, the passive optical network is the ultimate bandwidth solution.

Fiber optical networks are what we would deploy today for combined telephone, TV, and Internet access if we were building a broadband communications system from scratch. Optical networks have the bandwidth to handle any current application and anything we may want in the future. However, optical networks are expensive and have even greater installation cost.

Optical fiber and component prices have come down over the years but copper cable still wins. Such high costs have forced us to make do with the ancient existing copper twisted-pair telephone infrastructure and the newer hybrid fiber-cable TV network. But those systems also have limits that we are reaching gradually. The conventional pots twisted pair has just about come to the end of its useful life despite its amazing success with DSL. True broadband requires something faster. The cable TV people currently hold the U.S. lead with broadband access because the coax cable will take signals out to nearly 1 GHz. The solution for the RBOCs (regional Bell operating companies) appears to be a less expensive version of the traditional optical network known as the passive optical network (PON).

A PON is a point-to-multipoint tree or star-like topology network using a fiber with no expensive optical-electrical-optical (OEO) intervening electronics. The electronics only appear at each end of the link, one at the carrier's central office and the other at the subscribers end. Passive splitters (couplers) distribute the signals to as many as 32 nodes (64 maximum in some systems) at distances up to 20 km, in some cases farther.

These connections are also referred to as the Optical Distribution Network (ODN) and the "last mile" or the "first mile." They can also be called fiber to the premises (FTTx) where x means the home, business, curb, or user. It is one more attempt to get broadband into the homes and small office/home office (SOHO) businesses. Now, the telecom carriers are finally getting serious about broadband by beginning what is expected to be the gradual replacement of the old twisted pair with PONs.

The survival and continuing competitiveness of the regional Bell operating companies (RBOCs) is at stake. These companies have been losing subscribers slowly for the past decade as more of them move to wireless or alternative telephone connections such as VoIP on a cable TV system. As VoIP rolls out, the traditional carriers stand to lose even more, not to mention the inability to take advantage of the many exciting and profitable broadband applications like video-on-demand (VOD) and gaming. While DSL has kept many of the carriers viable, it has almost run its course. The newer ADSL2+/2++ is helping to speed up the networks but the range is very limited. Fiber provides all the bandwidth for whatever services may be needed now and in the future.

The U.S. has a pattern of falling behind the rest of the world and missing major opportunities in electronic technology. Asia already took away consumer electronics. Cell phones are another loss for the U.S., again with Asia leading the way and even Europe being farther ahead. Now it's broadband. With its ultra restrictive and even stupid telecommunications regulations and policies, the U.S. is again falling behind. But now with the major carriers like SBC, Bell South and Verizon recognizing their own survival is at stake, maybe we can catch up. With PONs on the verge, the affordable triple play (voice, video and data) may be within reach.

The PON concept has been around for decades with the first PONs built during the early 1980s in Europe. High costs prohibited much use in the U.S. and the carriers turned to DSL and cable TV systems for broadband connections. Yet, development has continued over the years and several standards have emerged.

The first, APON or ATM-based PONs, was originally developed by the Full Service Access Network (FSAN), an organization of telecom service providers. This standard uses the mature asynchronous transfer mode (ATM) packet transmission system employed by virtually all telephone carriers. Data is transmitted in 53 byte packets with 48 bytes plus a 5-byte overhead. The base speed is 155.52 Mbits/s but that is scaled to 622.08 Mbits/s in most systems. A more advanced version, called broadband PON (BPON), is now the primary target standard for most US PON systems. It includes video transmission and can handle Ethernet. ITU-T standard G.983.x cover both APON and BPON.

Figure 1 shows the major elements of a BPON system. At the central office (CO) of the carrier is the Optical Line Terminal (OLT). This handles communications with the remote sites, known as the Optical Networking Unit (ONU) or Optical Networking Terminal (ONT). In between the OLT and ONU is the ODN, made up of passive optical devices used as splitters. Also known as combiners because of their dual bidirectional nature, splitters simply divide the light power between two or more outputs.

Passive splitters are available in 1:2, 1:4, 1:8, 1:16, 1:32 and even higher level versions. Usually power is split equally among the outputs but there are asymmetrical splitters that can divide the power in an uneven ratio as demanded by the system. Splitters are cheap, highly reliable, and require zero maintenance. In most existing or proposed systems a 1:4 splitter is used at the OLT and then 1:8 splitters are used in the field to get up to 32 users on a fiber. Although BPON supports up to 64 users, 32 will probably be the normal upper limit in early systems.

At the CO, the OLT uses wavelength-division multiplexing (WDM) for transmitting and receiving. Data and voice is broadcast to all users on a 1490-nm laser at a download rate of 622 Mbits/s. In COs where video is distributed, a separate 1550-nm laser path is provided. It is multiplexed onto the fiber with the voice and data in a combiner. The video transmits in the standard analog format used by cable TV systems. Upstream traffic for the user to the CO is handled by time-division multiplexing users on to a 1310-nm channel back to the OLT. Careful timing is controlled by the OLT to ensure that each user gets a time slot regardless of the distance differences that will be inherent. The aggregate upload speed is 155 Mbits/s.

The other major PON standard is Ethernet PON (EPON). Also known as Ethernet in the first mile (EFM), this is an IEEE standard under the Ethernet umbrella (802.3ah). This protocol was formally ratified in June and it uses a topology and method similar to BPON. The network configuration is similar to the BPON ODN in Figure 1. However, all information--voice, data and video--is transmitted in standard Ethernet packets with up to 1518 bytes at rates up to 1.25 Gbits/s. EPON uses a 1490-nm laser for downstream transmission of voice, video, and data. The video is digital rather than analog as in BPON. The range is up to 20 km with 32 users maximum per fiber. Each user receives a final data rate of 10 or 100 Mbits/s. Upstream data is also by Ethernet packets at up to 1.25 Gbits/s on a 1310-nm laser.

EPON is the choice of FTHx in Japan, Korea and other Asian nations. A simpler protocol with less packet overhead than ATM, it comes cheaper and fully compatibility with IP networks. On the downside, it has not yet fully dealt with the QoS and provisioning issues. Numerous vendors are already working on or have solved these problems with proprietary designs. While the major U.S. carriers will no doubt be first to adopt APON/BPON, EPON will likely find a niche and grow in some areas. BPON and EPON are expected to coexist in North America in the future.

Finally, GPON or Gigabit PONs, is a superset or extension of BPON that achieves symmetrical 1.25- and 2.5-Gbit/s rates. The FSAN group has defined a set of standards that the ITU-T is expected to adopt later this year as the G.984.x standard. GPON is designed to use encapsulation, making it protocol agnostic and able to support both traditional TDM and any packet-based protocol.

Because of the cost of laying fiber, PON deployments are expected to begin in Greenfield developments (new construction) and multi-dwelling units. Overhead aerial wiring will be used to minimize cost. Laying underground cable is very expensive. To get PON to the home, a hybrid technique is expected to be used. This method runs the fiber to a box at the curb, then uses the standard local loop twisted pair to carry the connection inside the home, using an advanced version of DSL. These short copper links using VDSL or ADSL2+ can produce rates up to 50 Mbits/s, adequate for many broadband applications.

At this point, PON technology is well defined and is expected to eventually replace the old copper infrastructure. Market research firm iSuppli expects major PON deployments to begin in 2006. The firm predicts cable modem and xDSL to remain dominant through 2009, accompanied by very strong growth of PON over a 10- to 15-year period beginning in 2006.

The fact that the network is passive and needs little or no work beyond the installation has made PON very attractive. Better still, the semiconductor vendors have provided a broad range of chips to make building the central office and customer premise equipment faster and easier than ever. While some early PON equipment vendors used ASICs to implement the OLT and ONT units, the latest crop of specialty chips should expedite growth in this market and lower costs.

One of the first available PON chips was Freescale Semiconductor's MC92701, introduced late last year. This BPON termination layer device is designed for service in ONT boxes at the customer's premises. It works with a Freescale PowerQUICC communications processor and several external memory chips to produce a complete ONT box. The chip complies with the ITU-T G.983 standard and supports dynamic bandwidth assignment (DBA) to achieve quality-of-service (QoS) control and peak bandwidth allocation. A clock and data recovery (CDR) circuit and clock PLL are on-chip and operate from an external 19.44-MHz crystal. The chip is in production now.

Freescale's latest BPON chip, the MPC8340BPON system on a chip (SoC), is shown in Figure 2. This device is based on Freescale's e300 PowerPC core processor which is the foundation for the company's recently announced MPC8349E PowerQUICC II Pro communications processor family. The MPC8340BPON's e300 core runs at 266 MHz and includes a 32-kbyte instruction cache and a 32-kbyte data cache. Also included is a DDR SDRAM memory controller, three 10/100 Ethernet controllers, dual UARTs, I2C buss support, serial peripheral interface (SPI), and interrupt controller, a general purpose I/O, ATM adaptation layer 5 (AAL5) and Utopia interfaces. It handles all ATM cell processing with full operation, administration and maintenance (OAM) support.

The BPON section of the chip is derived from the earlier MC92701 device. On-chip CDR and clock PLL works well with typical triplexer optical modules. The MPC8340BPON chip reduces parts costs and reduces time to market for BPON set top boxes. It comes in a 27 x 27 mm package. Initial samples will be available in the fourth quarter of 2004.

Broadlight is another company addressing the BPON space by making optical transceivers for both the OLT and the ONU ends of the system, as well as Media Access Control (MAC) chips and an ITU-U G.983.3 APON software stack. The company's XN230 device is a BPON compatible ONU customer premise chip.

More recently announced is Broadlight's XL230 BPON-based controller chip for the CO (Fig. 3). It supports the 155- and 622-Mbit/s downstream rate as well as the extended rate of 1.244-Gbit/s downstream. A burst CDR and the SERDES are included on-chip. A patent-pending dynamic bandwidth allocation feature permits better utilization of upstream bandwidth. It has selectable level 2/3 Utopia interfaces. And the integration with Broadlight's transceivers is completely glueless. This chip supports up 63 ONUs.

Broadlight is working with Broadcom to develop complete solutions and reference designs. For example, their XN230 chip works with the Broadcom BCM6349 residential gateway chip and the Broadcom BCM3341 VoIP chip. More combinations are on the way.

Ikanos Communications bets that a copper link will be the biggest choice to reach the home. The company's line of fast DSL chip sets provides much higher performance than ordinary DSL chip sets. The Fx chip set is designed for use in BPON systems where the downstream termination is at an ONU which then communicates with the customer ONT by way of the standard local loop copper. The Fx 7030 chip set delivers up to 70 Mbits/s downstream and 30 Mbits/s upstream while the Fx10050 chip set provides 100 Mbits/s downstream and 50 Mbits/s upstream. Symmetrical speeds can be achieved at the lower rate in each chip. Both of these devices feature four ports as well as ATM Utopia level 2 and Ethernet (XMII) interfaces. These chip sets work in ONU, OLT, switches, routers and fiber concentrators.

The Fx10050S chip set is designed to work in the subscriber located equipment and features 100 Mbits/s downstream and 50 Mbits/s upstream or 50 Mbits/s symmetric. It works with the Fx7030 and the Fx10050. It has a built-in host for Ethernet applications. All of these devices have an integrated front-end (IFE) with integrated line driver, variable gain amplifier, low noise amplifier, related filters and other discretes that greatly minimize the need for external components.

While some companies are targeting the BPON, others are going after the international market for EPON. Passave is an example. Their initial target markets are in Japan and elsewhere outside the U.S. Already they have shipped tens of thousands of PON ports into the Asian market. Their primary products are the PAS5001 and PAS6001-B. The PAS5001 integrates an Ethernet Media Access Controller (MAC) with the EPON protocol management on a single chip. It is designed for CO OLT applications. Fully 802.3ah compatible, it supports up to 127 ONUs and features an ARM9 core with all the standard interfaces. It includes built-in 128-bit AES encryption. Their ONU reference design is shown in Figure 4.

The ONU side of the system can be implemented with the PAS6001-B. The interfaces include full duplex transmit and receive 1.25-Gbit/s TBI interface, full duplex 10/100 MII or 1000-Mbit/s GMII/TBI for connection to standard switch IC or PHY chips. Both chips feature programmable dynamic bandwidth allocation.

Centillium Communications is also addressing the EPON space. Its recently announced Colt chip is designed to implement the OLT at the CO in an EPON system. Centillium's Mustang chip targets the customer premises ONU boxes. Both chips are fully IEEE 802.3ah compatible and use an on-board MIPS processor. The Colt OLT features 802.1d bridging, 802.1q configurable VLAN, IGMP (Internet Group Management Protocol) for multicast and broadcast filtering, built-in encryption and decryption, and dynamic bandwidth allocation. The Mustang chip is for ONU customer boxes and includes the basic features of the Colt. It also features MII/GMII UNI ports. The clock and data recovery (CDR) and SERDES are also on-chip.

These protocol chips work with a transceiver chip like the Centillium Zeus or Apollo. The Zeus physical media device (PMD) is designed for point-to-multipoint operation and includes a fully integrated laser diode driver, limiting amplifier, a CDR for 155 and 622 Mbits/s and a dual quantizer. The data rate range is 125 Mbits/s to 1.25 Gbits/s. This burst mode device works with APON/BPON or EPON systems, in either OLT or ONU. It also features digital diagnostic monitoring compliant with the SFF-8472 standard. The Apollo chip is the PMD for point-to-point optical systems

Centillium's Unicorn chip is a broadband service processor (BSP) used to implement complete bridges or routers in customer premises equipment. Besides its fast MIPS processor, it also contains a DSP for voice-over-Internet-protocol (VoIP) operation. It supports four voice channels and has a TDM bus for PCM. The Unicorn has IPsec encryption in hardware as well as full DES, 3DES, and AES encryption . Software for packet header processing and IKE is provided. The chip includes two Ethernet 10/100 ports, USB, UART, GPIO, and SPI interfaces. All chips have standard EJTAG/JTAG support built in. The Apollo PMD is sampling now and samples of the Colt, Mustang, Unicorn, and Zeus are expected later in Q3.

The passive optical suppliers are also drooling over the prospect of a big PON rollout. Most already have splitter/combiner products that work but many are creating new devices for this potentially lucrative niche. While most splitters in PONs are expected to be fixed devices, Lynx Photonics Networks is betting that remotely controlled programmable splitters will have their place in the larger networks. By using static splitters, the network will be very inexpensive and highly reliable. Careful design is the key to selecting the split ratios and putting the splitters where necessary. This requires research and planning that may later prove to be different from reality as the system grows.

Demographic changes and fluctuating demand will make it tough to pin-point in advance just what the network should look like. Then as the demand grows, it may still be necessary to make changes, additions, and upgrades in the system, thereby offsetting much of the reason for the PON in the first place. Laying new fiber and making enhancements will be expensive and result in system down time not to mention the long provisioning time required. The new Lynx Photonics Networks programmable splitters will solve the problem.

The new Lynx LL-3560 is a remotely controlled and reconfigurable PON splitter that lets the carrier reconfigure a PON to fit constantly changing needs. By placing such devices at strategic locations within the network, like the CO or regional distribution center, the carrier can modify the light power distribution between the various branches of the network during a network build out. Provisioning becomes fast, easy, and inexpensive. The LL-3560 uses Lynx's Planar Lightwave Circuit technology with thermo-optical switches and splitters. Split ratios can be fine-tuned to 1% resolution.

It is anyone's guess where the PON trend will take us but I am optimistic. The traditional telecom carriers are definite ready for it as are the consumers who finally see the convergence. I predict a success here, over time. But the forthcoming broadband wireless access (BWA) business is developing quickly and we will see it as early as next year. I bet it will be only minimally competitive to PON but a real stimulant to the broadband field. BWA will fill a real need in the small towns and rural areas. Watch for my special report on the broadband wireless coming later this year.

Broadlight Inc. www.broadlight.com
Centillium Communications Inc. www.centillium.com
Ethernet in the First Mile Alliance (EFMA) www.efmalliance.com
Freescale Semiconductor www.freescale.com
Full Service Agreement Network (FSAN) www.fsan.com
IEEE www.ieee802/3efm.org
Ikanos Communications www.ikanos.com
International Engineering Consortium www.iec.org/online/tutorials/
International Telecommunications Union www.itu.int/ITU-T/
iSuppli Corp. www.isuppli.com
Lynx Photonic Networks www.LynxPN.com
Metro Ethernet Forum www.metroethernetforum.org
Passave Inc. www.passave.com
Passive Optical Network Forum (PONF) www.ponforum.org
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.