While systems vendors have readied their GPON offerings for the RBOC showdown, they’ve had to do so without as much help as expected from their chip suppliers when it comes to burst-mode receiver electronics. Tighter timing parameters and a wider dynamic range make the GPON burst-mode electronics more technically challenging than their BPON and EPON counterparts. The business case for the merchant silicon vendors also may be proving less than attractive. The result is that module vendors have found themselves left to their own devices-literally-when it comes to the development of GPON transceivers.
The burst-mode receiver is housed in the optical line terminal (OLT), located in the central office (CO). The OLT communicates bidirectionally with the optical network units (ONTs) located at each subscriber’s premises, with typically a maximum of 32 to 64 ONTs per OLT. The upstream transmission is based on time-division multiple-access (TDMA) technology, whereby each ONT is allocated a different timeslot in which to transmit data back to the OLT in the CO. Every time a burst of data arrives from the ONT, the burst-mode receiver in the OLT must synchronize with the clock in the burst before it can decode the data within it.
“Generally speaking, the faster your transmission rate, the harder it is to do,” admits Christoph Pfister, director of NeoPhotonics’ global access business (San Jose, CA). The EPON specification is more relaxed, enabling the use of standard SONET transceivers that have been adjusted to meet EPON requirements. “What the IEEE did was say, ‘We will give you a lot more time to do the clock synchronization, to synchronize to the new pulse,’” explains Pfister. “That’s one way of making it work at the higher line rate, but what you give up is efficiency, which means you give up usable bandwidth.”
Put another way, GPON’s higher throughput efficiency is the result of tighter timing parameters. The FSAN specifies 6 bytes of preamble between bursts of data in a GPON system, whereas the IEEE specifies a more relaxed 128 bytes of preamble between bursts in EPON systems, reports Ley Mee Hii, director of FTTH product line management at LuminentOIC (Chatsworth, CA).
That said, it’s not just the bursty upstream traffic that makes the burst-mode receiver so challenging to manufacture; it’s the combination of the burst mode and a wide dynamic range of 15 to 28 dB. “In the real world, this is the difference between a very close ONT with a small split ratio versus a very far ONT with a high split ratio,” says Pfister. “In an extreme case, the next pulse could be 15 dB lower than the current pulse.” The FSAN actually relaxed the specification last fall to provide this range in an attempt to accommodate both what is technically feasible today as well as what would be required in the future.
The wide dynamic range places particular stress on the transimpedance amplifier (TIA), a critical component of the burst-mode receiver along with the avalanche photodiode (APD) and the limiting amplifier.
In addition to the problem of complexity, the business case for the burst-mode receiver may not hold much appeal for IC suppliers. Volumes are an order of magnitude lower for the OLT versus the ONT. Moreover, the burst-mode electronics themselves could sell for as little as $1 from an Asian source, notes Jag Bolaria, senior analyst with The Linley Group (Mountain View, CA).
“Someone making this where it’s not for their own use is really up for a big investment for a low-volume merchant product,” adds Karen Liu, research director of components at Ovum-RHK (Boston, MA). “I think this biases it toward being done internally by, say, a transceiver vendor or even a system board maker. This is a natural tendency when something is low volume,” she adds.
Transceiver vendors can testify to the accuracy of Liu’s assessment. Perhaps because of the twin effects of a difficult problem and little economic incentive to solve it, more than one module vendor reported that merchant silicon houses that had promised to offer GPON burst-mode receiver electronics failed to do so. As a result, the transceiver vendors faced little choice but to do it themselves-a task that certainly presented its challenges.
“One of the biggest problems is that, let’s say, Maxim develops a TIA and a limiting amp and Freescale develops one also,” explains Hii. “The problem we are seeing is that the TIA of Maxim cannot work with the limiting amps of Freescale. If you look at the standard, traditional chipset, I can use anyone’s TIA and I can use anyone’s limiting amp, and I can make a module out of it. The chipsets have to be able to work together,” she says.
Nevertheless, the module vendors have persevered. LuminentOIC, for example, demonstrated an OLT burst-mode receiver at the recent OFC/NFOEC conference. Developed in-house, LuminentOIC’s OLT transceiver uses readily available, off-the-shelf components, according to Hii. “We use a standard chipset and a microcontroller within the module to handle some of this decision-making. It’s more of a discrete solution,” she says, adding that LuminentOIC’s transceiver meets the stringent timing parameters of the ITU-T G.984 GPON specification.
The folks at NeoPhotonics have devised a similar plan to develop the technology in-house using discrete components. According to Pfister, they also have been approached by one of their larger systems customers to jointly develop the burst-mode receiver, and they are implementing this strategy as well. “In parallel, what everyone is doing is keeping an eye on these critical components, especially the TIA,” he reports. “There are several companies who have committed to doing that again because they think it is a problem they can solve eventually.”
Even though GPON systems can function with workarounds, transceiver vendors like LuminentOIC would use a burst-mode receiver IC if one were available, says Hii, who cites the board space that would be saved. The availability of a burst-mode receiver chipset also would enable increased competition, which, in turn, would lower prices closer to the $1 Bolaria envisions, she says. “One of the problems we are seeing is people who are coming out with this burst-mode chipset, the cost of the chipset is very, very high.”
One could argue that there doesn’t seem to be a lot of incentive for the chipset vendors to develop burst-mode receiver electronics-particularly given its comparatively low volumes, low price points, and technical challenges. “It isn’t clear if they see a big enough market to go off and spend the money to spin the silicon,” agrees Bolaria. “It’s one of those chicken-and-egg scenarios where you need a commitment from somebody to deploy in high volume. That gives them enough of a return to go start spinning their silicon.”
“I could imagine a scenario where a company chooses to do this because it lines up well with their strengths,” adds Liu. “Basically, they can use it as bragging rights. There’s something to be said for going after a market that other people don’t go after. And if you have it available, there’s a good chance people would use it.”
NeoPhotonics’ Pfister says he is confident that the chipset vendors will come out with satisfactory devices in the not-too-distant future, motivated in large part by Verizon’s commitment to deploy GPON in volume next year. “They are upping their forecast for this year,” he says, “and people are listening. Now people are seeing that this is a real market, this is a reasonably tough problem to solve, and whenever you solve tough problems, you can charge for it.” The cost pressure isn’t quite as high on the OLT side versus the ONT side, and this too may prove attractive to the IC vendors, notes Pfister.
Among those expected to be in the mix are Maxim Integrated Products (Sunnyvale, CA), Vitesse Semiconductor (Camarillo, CA), TriQuint (Hillsboro, OR), Freescale Semiconductor (Austin, TX), and AMCC (Sunnyvale).
The competition, if it materializes, also could be intense from the Asian IC vendors-especially the Japanese, who have particularly strong process technology. “These are such low-cost parts, you actually end up getting people in Taiwan doing them, too,” adds Bolaria.
Interestingly, a host of vendors have publicly announced their intentions to enter the GPON media access controller (MAC) market, where price points are a lot more attractive. The MAC sells for between $20 and $30. Even the traditional EPON folks like Passavé, recently acquired by PMC-Sierra, have announced their intention to enter this market. At OFC, AMCC representatives noted that they were nine months away from final product for chips on the ONT side and 12 to 15 months away for the OLT side. They are still working out which feature sets their customers will demand, said a source at the show.
For his part, Bolaria does not see the burst-mode receiver problem inhibiting the adoption of GPON. “If you think about it,” he says, “the chips weren’t necessarily there for early BPON trials either. People ended up building FPGAs and ASICs to solve the problem.”Meghan Fuller is the senior news editor at Lightwave.