Xerox in Optical Switch Advance
Just this week they unveiled a significant advance in the development of the arrays of tiny tilting mirrors used in all-optical switches. They've come up with a new way of making micro-electro-mechanical systems (MEMS) which promises to be more manufacturable, more scaleable, and lower loss than existing technologies.
PARC is pretty late to the party. The MEMS scene is already crowded with at least a dozen other vendors aiming to make MEMS-based optical switches (see Optical Switching Fabric: MEMS). Some of them, like Cspeed Corp., have been working for three or four years and still haven't got products to market. But, according to Eric Peeters, who directs the Electronic Materials Laboratory at PARC, the Xerox technology is so revolutionary that it will allow them to leap-frog the competition.
"[Our MEMS chips] are a lot simpler to build, so we can crank out a batch in a week, as opposed to the month it takes in the rest of the industry," Peeters says. This will allow PARC to speed up the development cycle, he adds. In fact, Peeters claims that it only took six months to develop a process for making 1156-by-1156 mirror arrays with 100 percent yield.
Those figures are likely to get systems vendors very excited. Making the innards of large-scale optical switches is quite challenging, and achieving this without having to scrap any dud devices is almost unheard of.
So, how do they do it? The answer lies in the material used to make the mirrors. Unlike the rest of the industry, which uses silicon, PARC's mirrors are made from thin-film metals.
Metal is much easier to work with than silicon. It can be deposited by a simple process called sputtering. The metal films are pre-stressed, so that when the surface is patterned, the mirrors pop up on their own to relieve that stress (rather than being lifted up from the substrate by intricate micromachining). What's more, there's no need to machine hinges. The mirrors are bent instead of tilted, simply by applying a voltage.
"Anyone can build stress into a metal film -- usually it's there whether you want it or not -- but we've got good at doing it reproducibly," says Peeters.
However, the really big deal is not just the manufacturing, but the fact that the mirrors can be scanned through a much wider range of angles than usual. PARC is claiming twice the tilt angle of traditional MEMS. This is due to the fact that the structures can be raised much higher above the surface of the substrate.
This helps solve a key problem for optical switches -- that losses increase with switch size. That's because the light beam spreads as it travels. Steering the beam over a wider range of angles allows the beam to travel a shorter distance inside the switch, which reduces the spreading.
"We have already demonstrated scan angles that are considerably larger than all competitors that we know about", says Peeters. "Our modeling shows that we should be able to achieve angles that are so large that we could actually scale the size of the fabric to 10,000 ports. It's not clear to us that anyone wants this yet, but it's good to know we have the power under the hood."
Perhaps Peeters hasn't heard of Integrated Micromachines Inc. (IMMI), a startup that is also claiming it can tilt its mirrors through twice what the competition can do -- 50 degrees as opposed to a mere 22 degrees in developments by Calient Networks Inc. (see Calient Claims Breakthroughs On Optical Switches, Switch Startup Raises MEMS Questions and Dark Horse Joins Optical Switch Race). IMMI says it uses electromagnetic forces to turn its mirrors, delivering 1000 times more turning force than electrostatic (applied voltage) techniques.
Using Peeters' reasoning, it would appear that IMMI also has the potential to scale to monster switch sizes. But PARC would still have the edge, it claims, for three big reasons. One, as already noted, is the ease of manufacturing. Two, the electrical power consumption is low. Other approaches typically require huge amounts of electrical power, particularly for big switch sizes and large tilt angles -- it could be as much as 1000 watts. This is like having the switch sit in an oven, which can cause lifetime and reliability problems. In contrast, PARC's MEMS run off low power and low voltage. "The power is actually low enough so you could run the switch off a compact backup battery for quite a while," claims Peeters.
And three, the mirrors can be tilted more accurately, thanks to a proprietary feedback control mechanism, which was developed at PARC as part of a separate ten-year project. Peeter's reckons lots of other MEMS vendors are still struggling with the problem of aiming their mirrors accurately.
What happens next has not been decided, possibly because the future of Xerox itself is uncertain. Its financial position took a turn for the worse yesterday, when it warned of lower earnings and said that it had nearly exhausted its existing credit.
According to Peeters, one route is for PARC to partner with a company that has competency in making low port-count switches. PARC has the expertise in making the switching fabric, but needs to find the skills in fiber alignment and subsystem design, so that it can build the box around the switch core.
Another possibility is that Xerox will spin out a startup or form a joint venture, in the same way that it spun out SDL Inc. (Nasdaq: SDLI) twelve years ago. But whatever happens, it's going to happen fast, as PARC tries to catch up and overtake the other players in the MEMS switch space.
-- Pauline Rigby, senior editor, Light Reading http://www.lightreading.com