Micro-electromechanical systems (MEMS) have emerged as a leading contender to serve as the foundation of next-generation all-optical switches. But vendors working in this field are leaving no technological stone unturned in the pursuit of enabling switch devices. Several companies have evaluated liquid crystals as the basis of switching components and report the technology soon should provide a viable building block for certain switching applications. In fact, modules based on liquid crystals should be available in the first half of next year.
Liquid-crystal technology has been used for various applications since the mid-'60s. While liquid-crystal displays form the most common application, technology variants have appeared in such arcane products as mood rings.
Liquid crystals, despite the name, are not truely liquid; they exist in a state between liquid and solid called mesophasic. The molecules have an elongated shape, often represented as a rod; under the proper conditions, the orientation of these molecules can be changed so that they face in a certain direction. The orientation affects the optical properties of the liquid crystal, which in turn affects the polarization of the light passing through it.
At least three companies--Corning Inc. (Corning, NY), Chorum Technologies Inc. (Richardson, TX), and SpectraSwitch Inc. (Santa Rosa, CA) have announced the arrival of liquid-crystal switch components or systems by the first half of next year. According to John Bayne, product line manager for optical-networking devices at Corning, liquid crystals perform well by a variety of yardsticks.
"It's really not any one particular attribute, but the combination of performance we believe is unmatched right now in the devices we've seen demonstrated, in that it has very low insertion loss--we're talking less than 6 dB to access 40 to 80 channels; it has outstanding uniformity, on the order of 1 dB across all the channels; and outstanding isolation, 40 dB or better," he explains. "You can probably find other technologies that maybe have lower insertion loss or another technology may have higher isolation. But we believe that liquid crystal puts them all together in a very nice package."
Corning has announced its liquid-crystal switch, called the PurePath, will accommodate 40 channels at 100-GHz spacing in a 2x2 format. Bayne says that Corning will ship samples in the first quarter of 2000 and ramp up a pilot manufacturing line at the beginning of the second half of the year. If customers want 80 channels at 50-GHz spacings, the company will be happy to comply. "That's one of the advantages of using liquid crystals--you can really get down to narrow channel spacings, on the order of 25 GHz or less," he says.
While liquid-crystal displays have enjoyed widespread use for some time, the application of the technology to telecommunications systems has lagged because the material has tended to exhibit comparatively poor performance due to moisture and temperature sensitivities, explains Dr. Richard Albert, senior scientist at SpectraSwitch. Dr. Albert's company plans to incorporate the technology into a variety of switches, including on/off, 1x2, 1xN, and matrix-switching arrays. Production should begin by the second quarter of 2000. Dr. Albert and other researchers in the field say that recent advances in packaging and other processes have solved the temperature-sensitivity problems and opened the door to efficient switches based on liquid-crystal materials.
"Liquid crystals have been around for a long time, and we're very confident that the failure mechanisms are well understood," Bayne explains. "We're very comfortable with how liquid crystals are going to perform in the telecommunications industry. But I think the industry itself is really going to want to see the performance and reliability data over time to understand that all the potential failure mechanisms have been identified and tested to the relevant standards."
According to Dr. Albert, most of the switches used with this technology combine birefringent beam splitters, liquid-crystal material, and either a grating or waveplate.
For example, Al Ranalli, Bradley Scott, and John Kondis of Corning described one such switch in a paper presented at the recent European Conference on Optical Communication in Nice, France. The paper described a 2x2 wavelength-selective crossconnect based on a liquid-crystal switch matrix and a bulk-optic architecture--the same architecture as the PurePath switch, Bayne reports. The switch accommodates two input and two output beams, collimated from each of four fiber ports. The beams are split according to the polarizations defined in what the authors called the "birefringent optical system."
A liquid-crystal spatial light modulator (SLM) provided by Meadowlark Optics enables switching to take place by rotating an optical field's polarization by either 0°° or 90° at the midplane. To separate the channels, the polarization-split input beams are incident upon a diffraction grating, polarized at the grating's most efficient polarization. After passing through the grating, the beams are selectively rotated in polarization, then passed to a lens in such a way that all the input components of a common wavelength are focused to a common pixel in the SLM linear array. One input component's polarization is made to correspond to one of the SLM's eigenpolarizations, while the other component's polarization is set to the complementary of the first. The system is then mirrored about the SLM so that switching is performed by polarization rotation at the SLM.
According to the Corning researchers, the device exhibits >30-dB isolation and <1-dB ripple over 40 channels spaced at 50 GHz. It provides what the Corning staffers called "excellent" in-band ripple and <7-dB insertion loss for any switch state. (Bayne reports that when the PurePath product is released, it should provide <6 dB of insertion loss.) The switch, however, did exhibit difference in peak transmittances between the two switching states, which the researchers attributed to spherical-surface irregularity in one of the polarizing elements. In fact, the system displays significant sensitivity to such irregularities because of the large (>6-mm-diameter) collimated beam that must be used to provide the required spectral resolution for a 50-GHz channel spacing. Edge effects in the SLM pixels also can adversely affect performance. But between-pixel vignetting and/or even larger collimated beams can mitigate these problems, the researchers said.
For its part, Chorum Technologies reported in Lightwave in November 1999 that its liquid crystal switch demonstrates minimal variability in insertion loss and crosstalk performance from 10° to 70° C while providing switching speeds well below 10 msec (see Figure). The company says it has tested both static and operational-switching elements for over 12,000 hours at 90°C with no measurable degradation in optical performance. That equates to a meantime between failures greater than the 20- to 25-year lifetime requirements for public-network deployment, Chorum staffers report.
Such performance data suggests that liquid-crystal switches should prove a viable technology for near-term switching applications. Several companies are said to be investigating liquid-crystal switches. Tellium Inc. (Oceanport, NJ) represents one such firm, which has evaluated liquid-crystal, MEMS, and other technologies as it works toward an all-optical switch.
"I think the big advantage with liquid crystal, at least as we have been exploring it, is the ability to add and drop different colors of light without having to demultiplex the whole multiwave signal," offers Nick DeVito, director of product marketing at Tellium. "I think the big advantage of MEMS is that it's probably a more mature technology and may be here a little bit quicker."
Both technologies need some work, according to DeVito. "I think one of the big challenges that we've been facing [with liquid-crystal technology] is temperature stability over a broad range of operating conditions. To a lesser extent, you have a similar problem with MEMS, just in terms of making sure you can control the mirrors and you can have good stability across the whole optical-light path," he says.
With those caveats on the table, DeVito appears to be leaning toward MEMS as the current favorite for Tellium's requirements. "The one that I think shows the most promise is MEMS. We've seen a lot of advancements in that area. I expect that probably in a year from now, we'll have devices that are scalable and reliable and efficient that we'll be able to use in carrier-grade products," he says. "The key difference right now is scalability; I just don't see the scalability in the [liquid-crystal] devices that are out there today."
"That is a valid point," admits Bayne. "This is not a device that will scale up to 512x512 or 1024x1024, at least the way it's currently designed. There probably are better technologies out there that will get you to the higher-channel-count optical crossconnect."
Bayne points out, however, that not every switching application will require massive port counts. He feels that switching systems will be required for applications with both high- and low-channel counts, as well as high- and low-port counts. "I don't think there's any one technology that will allow you to cover entire market space of optical switching and routing products," he says. "In each of those regions in the marketplace, you're going to see different technologies take a leadership role. And in the low-port count/high-channel-count market space, we think this [liquid-crystal] device will do quite well." As an illustration of matching technology to applications, Corning also has announced a switch based on thin-film filters for low-channel-count applications.
Bayne sees a wide variety of technologies, from liquid crystal to MEMS to planar solutions to thermo-optic devices, vying for niches in different markets. Thus, companies like Tellium will have plenty of options from which to choose. DeVito says the time of decision is at hand. "I think we'll probably make a choice sometime in early 2001, and then I think you'll see carrier-grade equipment available by the end of 2001, maybe the beginning of 2002," he predicts.
Even then, the task of switch vendors in building the pieces of the future all-optical network will not be completely accomplished. "Even with these devices, there's still one large issue that remains," DeVito explains. "That is reading the overhead channels and performing the actions based on the information derived from those channels. In order to do those kinds of functions, you still have to operate in the electrical domain. And so the next piece of device technology that we need is to be able to read the light and make decisions on the bytes that we see in the optical domain, rather than converting to the electrical domain to make that decision."
It appears that even with the advent of a variety of optical-switching devices, the all-optical network is still very much over the horizon.