JDSU's PLC-based ROADM
JDS Uniphase (San Jose, CA) this month will introduce what it claims is the first planar-lightwave-circuit (PLC)-based reconfigurable optical add/drop multiplexer (ROADM) that can be manufactured in volume. Unlike previous ROADM models, the device's manufacturing process and assembly methodology create price points low enough to make reconfigurability feasible at every node, the company says.
Adding reconfigurability to an OADM typically has meant adding a discrete switching component, which can be prohibitively expensive. Today's network nodes are therefore populated with fixed OADMs, even though these configurations require expensive truck rolls for bandwidth expansion. "ROADM deployment has been very limited and only when it's absolutely necessary," reports Jy Bhardwaj, general manager of JDSU's Waveguide Business Unit. "The cost has been the real inhibitor—until now."
With the PLC ROADM, JDSU has "gone the distance," contends Darryl Inniss, program director, optical components, at RHK (South San Francisco). "Now you have all the switches, the mux and demux. You have the VOAs [variable optical attenuators] and the photodiodes; they've put everything into one box."
The 32-channel-count PLC ROADM, for example, consists of two fully integrated, self-contained modules: a reconfigurable optical add module (ROAM) and a monitored demultiplexer module (Demux-T). This two-module system displaces a plethora of components, providing the functional equivalent of 32 switches, 104 taps, 103 monitors, 33 VOAs, and 96 mux/demux modules, not to mention the associated control electronics and fiber management. The JDSU ROADM incorporates optical-channel monitoring and dynamic channel equalization. "You can see where the value-add is," says Bhardwaj, "and why [ROADM] hasn't happened before."
Moreover, the device is based on technology with which everyone is relatively comfortable, reports Inniss. All the main functions of the ROADM are synthesized through the company's PLC-based arrayed waveguide grating (AWG) platform. Other technologies, including MEMS and wavelength blockers, still raise questions about reliability, yield, and manufacturability. "AWGs and PLCs have been around," says Inniss. "They've gone through extensive testing with many companies in many labs. It's not like you're introducing new technology. There are no moving parts. It's a very simple approach. [The PLC ROADM] should be a very robust product from those perspectives."
According to JDS Uniphase, the use of proven PLC technology enhances the ROADM's manufacturability, which has been a substantial stumbling block for previous incarnations. "Once we've carried out the fundamental design of an AWG or VOA, switch, or tap, we can actually print it again and again and again on the same silicon wafer," explains Bhardwaj. "It is not a question of having manual alignment, having lines and lines of people who are assembling any of these components; these are printed on wafers."
This assembly methodology is also critical to the development of a product that will hit the aggressive price points sought by customers, he adds. "We take the chip and screen at the chip level, and then we make subassemblies that integrate all the functions around the chip," he explains. "We screen at that level, and then we simply place all these prescreened subassemblies into a box and final-test the ROADM. Once the actual chips are screened out, the yield is very, very high. All the prescreening is fully automated, so the labor content is very low."
JDSU is bullish on the ROADM market. Assuming the devices hit the right price points and are easily provisioned, there is no reason why every single node doesn't become ROADM-based as opposed to OADM-based, says the company. "We think it will be an absolute necessity," asserts Bhardwaj, "just from what it will do in terms of simplifying the nodes, standardizing the nodes, and making DWDM look and feel and play like SONET. Reconfigurability is going to be standard within the next few years."
The widespread deployment of ROADMs isn't likely to occur overnight. According to Inniss, today's carriers face a cost/capacity decision. "It really depends on the carrier's architecture and how much traffic exists at different locations in the carrier's network," he says. "For some, it will happen in 2004-2005. For others, it may take until 2006–2007 before they get into a situation where photonic switches make sense."