x
Cable Tech

Broadlight Banks on PON Expansion

BroadLight Inc., a three-year-old access chip startup, is betting the farm on the ascendance of passive optical networking (PON).

Broadlight, which will be displaying its wares at the upcoming OFC Conference, has created a series of chips designed to help system vendors easily build PON equipment, which uses passive splitters to siphon bandwidth to multiple users. Broadlight sees growing demand for more economical high-speed fiber links to homes and businesses.

"Most companies in DSL are looking into PON," says Broadlight CEO Amnon Presler, who formerly was president of RAD Data Communications Ltd. To help the cause, Broadlight offers a three-part PON solution that includes transceivers, media-access-control (MAC) chips for key PON elements, and software that conforms to specs adopted by the International Telecommunication Union (ITU). Presler says Broadlight's ability to offer all three PON elements sets it apart from other players. He also stresses the chips Broadlight's selling are geared to the very latest ITU specs that incorporate so-called Broadband PON (BPON), which adds wavelengths to PON architectures based on Asynchronous Transfer Mode (ATM) to better handle multimedia traffic (see Optical Access, page 3).

Presler says Broadlight is in trials with five companies, though he won't name them.

Broadlight's strategy seems an exercise of raw nerve, considering the competition, which includes the likes of NEC Electronics Corp. and Oki Optical Components Inc. Word has it that even more big chipmakers with Asia/Pacific connections are eyeing PON in the wake of market reports that sales are up and startups are getting funded (see Optical Gear Rising, Report Says and FlexLight Bags $3M More). There also are signs that the possibility of a PON surge hasn't escaped Intel Corp. (Nasdaq: INTC) (see Intel Invests in Optical Solutions and Intel Invests in Passave).

But will tiny Broadlight, with just 30 employees, get attention from customers? Presler is convinced Broadlight's timing couldn't be better. Government-subsidized fiber builds in Japan and Korea are leading to PON deployments not matched in the U.S. or Europe -- even though Presler has hopes for a larger stateside market by 2004. Carriers in these countries are open to ATM PONs, not Ethernet ones, Presler says. And they're ready to spend.

Indeed, Presler claims to be readying a bid for one of Japan's largest system vendors, a contract for parts that could bring in hundreds of thousands of dollars annually for many years.

Presler's got other big plans. His company is investigating ways to reduce the cost of PON components by integrating them and making them more efficiently. The startup is partnering with a Korean maker of optical transceiver parts that would give Broadlight the means to mass-produce cheaper PON parts in Asia.

"We're close to signing a contract for a joint venture that will include our electronic intellectual property and their mass production capabilities," Presler says. But he won't divulge any names.

It's an ambitious plan, and a challenging one. On the downside, the competition is fierce and the market in flux and unpredictable. The siren song of PON continues to fascinate and elude startups (see PON Pushers Huff and Puff). On the upside, funding's in the works, Presler says, to build on the $10 million the startup's already gleaned. A third round, including strategic investment from a major U.S. customer, should materialize before midyear, he says.

Bottom line? If PON prospects improve, it's likely to stimulate the market for parts like Broadlight's. But whether the recent flurry of positive PON news reflects a solid and global trend remains to be seen.

— Mary Jander, Senior Editor, Light Reading

For up-to-date information about the coming OFC Conference, please visit Light Reading’s Unauthorized OFC Preview Site.

ke6udl 12/5/2012 | 12:27:16 AM
re: Broadlight Banks on PON Expansion PON Relativism

It is always the same. Whether buying a car, TV, set of headphones or cell-phone. Relativism determines price; options may vary. In the world of consumer products, where expectation meets the road, bigger screen TVs and more powerful boom boxes command higher prices. It has to be this way, otherwise why would anyone invest in going to the next higher level. Same holds in the world of fiber optics. The price of higher-power, wider-bandwidth transceivers will always command a price premium no matter what the volume. Still canGÇÖt get something for nothing even these days. Face it; optics guys like to eat too!

PONs raise the cost of fiber relative to copper not because it is a fiber infrastructure, but because they make the cost of the CPE/CLE equipment significantly higher than it might be otherwise. It isnGÇÖt the PON protocol that is at fault, itGÇÖs been done and redone many times over the years with the same results. Rather it is the pure P2MP nature of PONs that is the problem whereby it forces higher cost optics to be used when only a fraction of that carried bandwidth is consumed at any given site. Think of it as a Humvee trying to masquerade as a Honda, or think of it as asking for OC-48 LR-2 optics at OC-3 IR prices. And like the Satellite dish, most of us wonGÇÖt buy it unless it is priced below the threshold of pain.

PONs arenGÇÖt future proof, nor are they passive as they attempt to fulfill the 1st promise as evidenced below:

1. ITU-T G.983.1 Paragraph 5.6.2 Optical Interfaces states: The optical properties of the ODN shall enable the provision of any foreseeable service, without the need of extensive modification to the ODN itself.

While optical wavelength transparency, reciprocity and fiber compatibility are sited as properties to be preserved, fiber-count to ONUs (feeder-fiber-count) must also be added to the list fiber paths, both real and virtual ultimately set the limit on future PON technology extensibility. Given that the subscriber cross-section of an ONU (in terms of total homes passed) is fixed, bandwidth upgrade in a PON requires more and smaller optical multiplex devices at the curb which need to be fed by feeder fibers, the very thing PONs are supposed to reduce.

2. Appendix 1 of ITU-T G.982 introduces the idea of adding optical amplifiers and perhaps DWDM to the access solution as ways of surmounting the high-loss of passive power splitting/combining to preserve split ratio as bandwidth demand increases. How can we continue to fool ourselves that these technologies, which we find difficult justifying even in the metro space today, make sense in the access arena, especially when we have the necessary tools and technologies to compete favorably against copper in new installations today.
optical_optimist 12/5/2012 | 12:27:10 AM
re: Broadlight Banks on PON Expansion Post # 1 is excellent !!

PON is starting to remind me of ISDN and ATM to the desktop in the 1980s/1990s. It is always three years away from "mass market" per the prognosticators. The RBOCs keep trying to get
the costs down, but the costs cannot get much lower without significant volume deployment.
This is not longer a reliability issue. The PON technology is reliable enough for RBOC deployment
in business and residential, or at minimum deserves more comprehensive in-service trials.

PON also has no compatibility with TIRKS/NMA, which absolutely amazes me. The RBOCs participate in the FSAN consortium religiously and yet there is no strategy for integration into the legacy systems. I think (??) all of the PON network management standards are CORBA-based with
no TL-1 requirement.

Outside plant splitter engineering is another area in which no cookie cutter solution exists today. Again can you model a splitter in TIRKS today ?? I doubt it.

PON can work for business and residential, but the RBOCs have to accept the fact that PON is not copper-based and incremental revenues and real operational cost savings are needed to drive the business case. These drivers exist but there is still an element of risk-taking that the RBOCs do not seem willing to take in the foreseeable future.

If the RBOCs do not embrace PON, PON Will be nothing more than a cottage industry in North America. Europe, Asia, and Latin America are more likely to see major deployments.

--- Out

P.S. I would welcome feedback on TIRKS/NMA for PON. Am I missing something ?? Is there some grand solution in the works for next generation OSSs to support PON in the RBOCs ?
doco 12/5/2012 | 12:26:57 AM
re: Broadlight Banks on PON Expansion I can't agree with ke6udl at all. Then again - I have taken a pretty big drink of the PON koolaid.

To refute a couple of his/her points.

Correct - more powerful or higher bandwidth optical components are going to command higher prices. But, the price of the laser/receiver are just one part of the overall system. All the electronics behind those optical components is another, bigger, piece. Also, a point to multi-point system requires less transceivers than point to point systems. For example - to serve 32 homes you would need 33 lasers and recievers for a PON system, but 64 for a point to point system. That is true even for a copper based system. By using point to multipoint configurations you have less ports in the central office, allowing for greater density.

As to the cost of those actual optical components, in a PON that goes shorter distances, you need power to get past the optical split, but dispersion isn't as much of an issue as when you go 100's of km. Also jitter requirements at the edge of the network can be relaxed some. These can drive down costs over SONET components.

As to the quotes for G983.1 - I didn't verify them directly, but I wouldn't doubt that there are words like that in the standard. That being said, those who are creating PONs are NOT putting active electronics in the field, and not running enough fibers for point-to-point systems. It is typical of any standard defined by committee to have phrases that make everyone happy. Just because there are there doesn't mean that is how people actually implement things. The implementations of several different vendors all capitalize on the passive components in the field. This avoids needed powered boxes in the field, reducing maintenace costs.

The differences in maintenance costs, floor space in the central office costs, and the incemental revenue that can be added from additional services can make PON systems viable.
palomar7 12/5/2012 | 12:26:54 AM
re: Broadlight Banks on PON Expansion doco, to conter-refute some of your points:

re: "But, the price of the laser/receiver are just one part of the overall system. All the electronics behind those optical components is another, bigger, piece."

Good points, but the electronics behind either a PON OLT or PON ONU are more complex technically and operationally, as well as higher cost (compared to pt-to-pt) due to the TDMA request/schedule/grant protocol, the coarse and fine ranging and synchronization protocol, and due to burst-mode receiver at OLT and burst-mode transmitters at ONUs.


re: "Also, a point to multi-point system requires less transceivers than point to point systems. For example - to serve 32 homes you would need 33 lasers and recievers for a PON system, but 64 for a point to point system."

Also good points on the surface, but the cost and performance specs for those transceivers are higher for PON, on both sides of the connection. As you mentioned, the (14-20dB) higher transmit power requirement means higher cost, and the burst-mode operation means higher cost and complexity, as well the specs (almost across the board) are harder to meet for PON transceivers. Also, OLT transceivers are different from ONU, whereas for pt-to-pt they are the same. So scale economies will be harder to achieve.

Another point w/r/t PON cost is that only at 100% take rates per OLT do the economics meet theoretical (i.e. OLT cost for a 32 split PON divides by 16 (for say 50% take rate), not 32).


re: "By using point to multipoint configurations you have less ports in the central office, allowing for greater density."

When you say this you imply that Pt-to-pt is home-run. I favor a point-to-multipoint active approach, where you have a high density switch within 3-5km of homes, then a single fiber of any distance you need (tens to hundreds of km) that aggregates at the same or greater ratio than PON, back to the headend/CO/PoP. Some municipalities I have talked with intend to put those switches in schools or fire houses, so no CEV or new AC power runs are necessary. Apparently that 3-5km distance works out well with typical school/firehouse densities.

re: "The implementations of several different vendors all capitalize on the passive components in the field. This avoids needed powered boxes in the field, reducing maintenace costs."

Although this statement has intuitive appeal, it remains to be proven in practice. With zero problems over their entire life, all-passive will be cheaper because switches have some level of periodic maintenance regardless of problems. However, when you get problems in individual drops, multi-point passive is a painful troubleshooting and resolution problem because you don't have access to individual lines. The cable guys have more problems with the passive, shared coax portion of their network than any other. I realize it's not exactly the same, but many of the issues and basic troubleshooting issues are the same.

re: "The differences in maintenance costs, floor space in the central office costs, and the incemental revenue that can be added from additional services can make PON systems viable."

Maintenance and floor space I addressed earlier. Added service revenue is extremely limited and not scalable to the degree claimed by PON vendors. Compared to pt-to-pt active, PON does not even compare for ability to easily add future services and revenue streams. If more that a couple people per OLT want services requiring big bandwidth, the rest are out of luck. As a service provider, I can't imagine putting yourself in that position of having to turn down revenue unless you are 100% certain the scenario will never happen. It seems to me that as a service provider you have every incentive to make it happen and assure it can happen painlessly and on large scale.

Perhaps the most telling argument I have seen against PON is that every early-market shared infrastructure has ultimately evolved to dedicated connectivity. In those previous early markets, the dedicated connectivity was not there (economically) when deployments first began. PON, however, has sat in the "deployment waiting room" for a couple decades, so in today's market, an entry position of dedicated connectivity does not carry that penalty.
ke6udl 12/5/2012 | 12:26:49 AM
re: Broadlight Banks on PON Expansion doco:

I drank from the PON straw myself, so my opinion is tempered in reality along with decades in this industry on both the system and component side of the equation. That said, I am only expressing my opinion so that together we can at least approach the truth through honest dialog.

While it is true PON halfs the laser count, it requires many more high performance devices than competitive solutions. Second, PONs require high performance optics where they can least be afforded/amortized, and that is at the subscriber CPE/CLE, and by my calculation, the subscriber optics more than doubles the system hardware cost. To understand where I am coming from consider an active star topology, the same topology as PON, so fiber plant remains the same.

The active star only requires two high performance lasers, between the OLT and the curb-side active hub, referred to here as the ONU. But high performance in this context is somewhat of a misnomer since these lasers are used P2P and don't have to push through the 18dBs or so of a 32:1 splitter plus fiber loss, etc. And as we all know, power comes at a premium. And don't forget, PONs need that high power laser at all the subscriber locations.

So what about the subscriber links? We all know the rationale behind the star is to minimize and share fiber. so we want the active hub of the splitter/combiner point to be located as close to the subscribers as possible. But to preserve existing design rules, we probably need to assure subscriber drops consistant with CSA or extended CSA (Ex-CSA) guidelines per GR-909-CORE. That means we would like our FINAL subscriber links to support 3.6km at a minimum and 7.2km max. If we further assume every man is born with entitlement to 100Mb/s or 155Mb/s bandwidth, at least until he is an adult, we have several commercially available technologies to choose from.

1) 850nm VCSEL lasers over Corning Infinicor SX+, which will do GbE over 3km. 155Mb/s =>8km. (verified)

2) 1310nm planar lasers and VCSELs over SMF-28.

In the first case 155Mb/s transceivers @ $25 and less are here today, and are projected to drop even more. So the cost of these optics on both ends of the subscribers link is now cheaper than the passive splitters and combiners in the PON ODN, and you have 155Mb/s secure, symmetrical, simple. No $500 optics at the home. No need for complex electronics at the home to assure network fairness and security.

In the second case only one vendor has 1310nm VCSELs, but is getting close to commercial release, so we have to wait and see what they bring to the table in terms of cost competitiveness. For years I've heard of low cost 1310nm FP lasers, but I have my doubts they will compete with VCSELs long term.

Jumping ahead:

I've looked at the cost of doing the active star fed by 2.5G, feeding 48 subscribers with 100Mb/s pipes. The total equipment cost shared from OLT to subscriber (sharing active ONU cost) came in at about $150.00. Since the subscribers have dedicated links, I assume the optics in the home resides in something you purchase from Sears, Radio Shack, Circuit City, etc., long term, or in some kind of set-top today. The cost I've considered there is the cost of the subscribers transceiver and the framer sitting behind it, that presents a standard interface to the set-top or appliance. And I'm looking beyond 2006 here when digital TV is the norm, when CATV as we know it shifts to VOD paradigm.

Powering the active ONU:

Most important issue relative to active star that requires flexibility on the part of the network owner and local PUCs. But the new technologies like VCSELs and smaller IC geometries have reduced the power requirements while increasing bandwidth. VCSELs run on 1/5th or less power than their PON counterparts. ONU power per subscriber comes in at about 2 watts these days. In my cost estimates, I've assumed a ONU-48 being fed via siamese drops from 2 to 4 of the closest subscribers. PUCs would issue power credits to those subscribers to cover the cost of power. Under this kind of scenario, I find the active star hardware component at <$700 per subscriber. Don't know how network provided power for ONUs would change the equation.

Well, that's it for now. But let's keep up the dialog. Just want to make sure we understand how we got where we are, and make sure we know where we are going and why.

Mr. ke6udl

Biggest issue
doco 12/5/2012 | 12:26:39 AM
re: Broadlight Banks on PON Expansion
Ok - I can't back down now I guess.

To counter palomar7's counter (or something like that :)

The electronics to handle the additional scheduling burdens of a point-to-multipoint I'm convinced are not nothing - but on the order of a dollar or two. The biggest problem is probably the burst mode receiver at the OLT. That does have considerable more expense - but it is spread out over 16-32 customers. I believe it could be a lot cheaper if some of the timing specs were relaxed (i.e. given more than 3 byte times between bursts) However, some of the entrenched players in the standards comittee like raising the bar for others, and the RBOCs like to go for the after that extra 1% in efficiency gains. So - I don't agree with you completely - but don't disagree completely either. I just don't think it is that big of an issue.

As for count vs cost for the optics - to decided that we would need to lay out real numbers from real suppliers and then add it up for each way. I don't have those numbers at my disposal. I think we can agree that there are tradeoffs here.

I'll agree that trouble shooting a passive system is more complex than a point-to-point. However, it is not the same as a coax based system. In a PON system you can go to the splitter and start unplugging ports. In a coax system - either like the cable company has, or like the old 10Base2 systems - you have a bus where a problem can be introduced anywhere and the only way to solve it is to keep doing binary splits of your network until you find the problem. In the field that requires driving, opening different boxes, etc. for each test. Much harder than a PON where you have one splitter point to go to.

As to you last paragraph or two - I think you are trying to have your cake and eat it too. First you argue that having active node and then going point to point from there gives you better debugability. I'll agree with you on that one. If you can get space in your fire-house or school with AC, power, etc. for free - all the better. (I don't believe that you will - it will cost something)

But now - you argue that a PON can't scale. A PON will allow you to sell big bandwidth customers just as much as having a node at the school house. You have some big pipe to the school house and it splits it's bandwidth to x number of customers. If you have too many customers wanting too high of bandwidth your pipe to the school house is clogged. Same as on a PON. You probably have a bigger pipe to the school than you would on a PON - but you also probably have more customers on that pipe, so a greater probability of having too many to support what you have.

The telephone companies have been managing this issue for decades already. They watch developement and leave some spare capacity available to handle it. If demand picks up - they add more capacity. Where a couple of T1s once served a small village, they upgraded to T3's and now OC3's.

I don't see that managing a PON is any different in that regard than managing an active node in the field doing point-to-multipoint. An active node probably has a larger scale - which can mean that demand grows much faster than anticipated, but it can also mean that you have more customers to spread cost across. It's a give and take and I don't see either as a real winner.
ke6udl 12/5/2012 | 12:26:26 AM
re: Broadlight Banks on PON Expansion "New York has too many phones?"

Eduardo Gelbstein is a Senior Special Fellow of the United Nations Institute for Training and Research and a contributor to the United Nations Information and Telecommunications (ICT) Task Force and to the preparatory work for the World Summit on the Information Society that will take place in Geneva, Switzerland on 10-12 December 2003.

Presentation: Connectivity for a better world
Writing emerged about 5,000 years ago. Other than optical flags and telegraphs operating over short distances, information moved at the speed with which it could be physically transported until 1843 when Samuel Morse demonstrated his telegraph. By 1865 the telegraph had become an international phenomenon and an international
organization was created to deal with its global spread: the International Telegraph Union (ITU). Its present name replaces the word GÇ£telegraphGÇ¥ by GÇ£telecommunicationsGÇ¥.

Since then, it took the world 100 years to connect 1 billion telephones. The Internet acquired 600 million subscribers in just 7 years. Their distribution however is not even: New York City has more telephone lines than the whole of rural Asia and London has more Internet subscribers than the whole of Africa. This uneven distribution is called The Digital Divide.

Lack of connectivity is one of the factors defining the Digital Divide. Economic, social and cultural issues are just as important. The complexity of this matter should be no excuse to ignore it. Nations remaining outside the Information Age will be unable to participate in the 21st Century economy and will thus face long-term poverty and potential instability. This has become an ethical as well as technical issue.

There is ample evidence that information and communications technologies play a major enabling role in sustainable development and many countries have moved from being producers of commodity products to become major exporters of hi-tech products and services. This evidence should help defining all actions to deal with the digital divide as opportunities, rather than problems.



palomar7 12/5/2012 | 12:26:11 AM
re: Broadlight Banks on PON Expansion doco, thanks for the well-thought comments. I've got a few follow-ups:

re: The electronics to handle the additional scheduling burdens of a point-to-multipoint I'm convinced are not nothing - but on the order of a dollar or two.

Yeah, that's probably a fair comment, given today's silicon scale. It's a pretty complicated design and validation job though (for a flexible scheduler anyway), and it's open to vendor value-add, but it probably won't add tremendously to cost. I'd guess closer to 10 bucks for a moderately flexible scheduler, but we're down in the "pennies" relatively speaking.

Your comments got me thinking about another area. Some front-end functions at the ONU, such as link layer security at 1-2.5 Gbps, adds pretty significantly to cost. I should say "can add" since I haven't really looked at what all is required for GPON "security." I'm sure you are aware of the big premium for Gb-capable security processors compared to 100Mb processors that do essentially the same functions. Since every PON ONU is a broadcast receiver, its first few processing stages all have to operate at the max rate. Depending on what fields are secure and what algorithms are implemented, this can be an expensive proposition. I can't really provide any hard data here, so this was just a passing comment. If you can add details, please do.



re: "burst mode receiver"

I agree with most of what you say, and like anything the complexity is lost once it's solved and in production, and the cost declines as it becomes commonplace. I just tend to think it won't reach those milestones for BPON because BPON is already obsolete. GPON is pretty bleeding edge, and meeting many of the specs is anything but "mainstream" today. Will it ever get the market share to realize the cost declines is the ten dollar question.


re: I'll agree that trouble shooting a passive system is more complex than a point-to-point. However, it is not the same as a coax based system. In a PON system you can go to the splitter and start unplugging ports.

I recall within the past 6-9 months a debate in IEEE EFM about outside plant fiber connectors, and the dialog included RBOC reps. What I recall is there were no certified temperature/humidity-hardened connectors. Condensation and freezing thereof were issues mentioned. Has that (lack of OSP-certified connectors) changed? If not, these would all be fusion spliced rather than connectors, and that was the angle I was approaching the troubleshooting problem from. If there are outside-plant-certified connectors I buy into what you said.

Another interesting variation on this theme is distributed splitter PONs (lots of x2, x4, and x8 splits, which makes it even more of a pain due to the distance between them). Yet another variation that--if I recall--Bell South seemed fond of, is having all the splitters in the CO. Though this kills all the fiber savings arguments for PON, it does eliminate a lot of other concerns that I have. The "splitter aggregation point" that Corning and others have proposed inherits many of these properties.


re: In a coax system - either like the cable company has, or like the old 10Base2 systems - you have a bus where a problem can be introduced anywhere and the only way to solve it is to keep doing binary splits of your network until you find the problem. In the field that requires driving, opening different boxes, etc. for each test. Much harder than a PON where you have one splitter point to go to.

Addressed this scenario, and how it can apply to PON, in my comments before this. This points to one of the problems: there are so damn many variations, and even a seemingly small one can change a problem to a non-problem, or uncover a new one. When this is comprehended, it's pretty hard to believe any cost or performance generalization claims any vendor makes. They never give sufficient detail on what their claims are based on.

re: As to you last paragraph or two - I think you are trying to have your cake and eat it too. First you argue that having active node and then going point to point from there gives you better debugability. I'll agree with you on that one. If you can get space in your fire-house or school with AC, power, etc. for free - all the better. (I don't believe that you will - it will cost something)

Yes, agreed, it will have some cost, but it will be insignificant compared to a CEV (unless you have to build something), and will not be visible as street furniture, nor are there rt-of-way issues. This approach is probably only applicable for a municipality.

I'd like to comment on your scalability comments later. Gotta run. Thanks for the interesting dialog.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE