x

Network Photonics: A Corvis Copy Cat?

Network Photonics Inc. yesterday announced its first products -- a couple of subsystems for use in what the startup calls “wavelength switches” (see Network Photonics Unveils First Products).

The subsystems combine the functions of an all-optical switch with those of a wavelength mux/demux -- providing a way for light to be switched from input to output fiber without undergoing any optical-electrical-optical conversions.

The concept isn't new. It's what Corvis Corp. (Nasdaq: CORV) has been preaching all along. Now a growing number of vendors are singing the same tune -- and claiming they can achieve the same thing at a fraction of the cost and in a fraction of the space.

Whether they can or not is moot, from two points of view. First, nobody's buying any of this gear at the moment, so the issue is somewhat academic. Second, Corvis deliberately designed its switch so that more compact, less costly cores could be swapped in at future dates, according to David Smith, its VP of hardware engineering.

In other words, there's nothing stopping Corvis buying subsystems from Network Photonics or another supplier for future versions of its switch. Sources also say that Corvis itself is developing a second-generation core for its switch and that this has characteristics comparable to those of Network Photonics' subsystems.

Still, Network Photonics says that its subsystems are ready for use right now, so it's worth digging into the details.

One is called the CrossWave 1200 and is for use in reconfigurable add/drop muxes; the other is called the CrossWave 2200 and is for use in crossconnects.

Both models have a mere four ports, which makes them sound quite small. However, these ports are for whole fibers, and each fiber can handle up to 96 wavelengths. So, in reality, the switches can handle a total of 1.92 terabits a second in a single 3RU chassis. That makes them the world's highest density optical switches, according to Steve Georgis, Network Photonics' president and CEO.

Corvis's switch handles even higher capacities -- six fibers, each handling up to 320 wavelengths carrying 10 Gbit/s apiece, according to Smith. However, the current version of Corvis's switch occupies four racks of equipment. Mind you, this is a complete switch, not just the switching subsystem.

The guts of Network Photonics' switches include gratings in silica and a single row of microscopic (50 micron square) tilting mirrors made using micro-electro-mechanical system (MEMS) technology. The startup dubs the setup "1D MEMS" (see Network Photonics Lifts the Lid). One grating splits the light into the standard ITU grid of wavelengths, so each wavelength strikes a different mirror. The mirrors are tilted to direct the beams to the appropriate output ports, where they're recombined by a second grating.

Georgis says the setup delivers excellent optical performance: maximum loss from input to output fiber is 6 decibels; channel spacing of 50 GHz; passband of 32 GHz at -0.5 dB; polarization dependent loss of 0.2 dB; and switching time of 250 microseconds.

However, Beni Kopelvitz, senior VP of R&D with another optical switch subsystem startup -- Lynx Photonic Networks -- questions whether the switching subsystem is non-blocking, bearing in mind that there's nothing stopping two input ports trying to route same-color wavelengths to a single output port.

List prices are $40,000 for the CrossWave 1200 and $80,000 for the CrossWave 2400, according to Georgis, who says this equates to roughly $500 a channel. It's important to remember that this is just for the subsystem. Original equipment manufacturers (OEMs) would have to surround this with a lot of extra hardware and software, Georgis acknowledges, pushing the cost per channel up to around four times this figure.

Corvis's Smith says he expects prices to fall even further. "We would like to see the final per-port cost to the customer to be down towards $500 to $1000," he says.

Network Photonics isn't alone in developing subsystems that combine wavelength multiplexing and demultiplexing with all-optical switches. Corning Inc. (NYSE: GLW), JDS Uniphase Inc. (Nasdaq: JDSU; Toronto: JDU), and SpectraSwitch Inc. have developed products in this field, based on liquid crystal technology (see JDSU Parades Picky Part, Optical Switching Gets Flexible, and JDSU: MEMS Aren't Just Memories). JDSU is also developing a larger-scale wavelength switch based on MEMS technology. So is Agere Systems (NYSE: AGR).

Georgis says that Network Photonics' subsystems are being tested by half a dozen OEMs and three major North American carriers. He also says that some carriers are inviting RFPs (requests for proposals) for upgrades to their long-haul backbones -- and those RFPs call for the kind of subsystems he's making.

Of course, RFPs and tests don't guarantee orders. But Georgis says his 139-staff company has enough cash -- $50 million -- to sit out the recession for quite a while. "We will exit 2004 with money in the bank, even if we've had no revenue by then," he says.

— Peter Heywood, Founding Editor, Light Reading
http://www.lightreading.com
dispersion 12/4/2012 | 10:32:05 PM
re: Network Photonics: A Corvis Copy Cat? Re: "But Georgis says his 139-staff company has enough cash -- $50 million -- to sit out the recession for quite a while. "We will exit 2004 with money in the bank, even if we've had no revenue by then," he says."

What a great use of $50M - pay 139 people for a couple of years waiting to sell a few boxes. Maybe he should return the money to his investors, most likey they would get a better IRR...

Better yet, Network Photonics could start a new VC fund, they may need some new technologies in a couple of years...
PIC guy 12/4/2012 | 10:32:05 PM
re: Network Photonics: A Corvis Copy Cat? I will appreciate if someone can educate me a bit on the total cost of the Crosswave 1200 ends up being close to $200k.

I thought all the software and hardware comes with the subsystem. Does it mean they don't have amps and gain equalizers etc which have to be added on.

Thanks
lightpimp 12/4/2012 | 10:31:58 PM
re: Network Photonics: A Corvis Copy Cat? I think he meant depression...


$50 million -- to sit out the "depression" for quite a while. "We will exit 2004 with money in the bank, even if we've had no revenue by then,"

Their VCs must love hearing this! Keep it burnin' Its just paper anyways! LOL! Gotta love it!
vikm007 12/4/2012 | 10:31:56 PM
re: Network Photonics: A Corvis Copy Cat? It is better to be out this business causing misery and open dry cleaning shops with this money..VC's will be happier as the company is almost guaranteed to make money...why be glamored by high tech..low tech rules these days.
Peter Heywood 12/4/2012 | 10:31:54 PM
re: Network Photonics: A Corvis Copy Cat? The list price of the CrossWave 1200 is $40K. However, the CrossWave 1200 is the equivalent of an engine for an automobile. It would form a key part of a switch sold to carriers, but OEMs would have to add a lot of other stuff, including amps etc as you say, plus all the software furnishing the control plane.

Georgis says that all of this other stuff typically means that the switch sold to carriers - the equivalent of the whole automobile in my analogy - would cost four times the cost of the switching subsystem.
manoflalambda 12/4/2012 | 10:31:49 PM
re: Network Photonics: A Corvis Copy Cat? Hey all,

I see that the CrossWave 1200 has only one drop port and one add port for adding/dropping bands of wavelengths. Is this of interest to vendors or is there a preference for access to all 96 lambdas?

Also, how do either of these products prevent like colours from overlapping? Or does any external product/software need to handle this?

Salute,
Manoflalambda
MEMS_DAWG 12/4/2012 | 10:27:45 PM
re: Network Photonics: A Corvis Copy Cat? I am curious as to the associated cost of support electronics for deploying a (MEMS) optical switch?

MD
rs50terra 12/4/2012 | 10:27:27 PM
re: Network Photonics: A Corvis Copy Cat? Crosswave1200 seems to be a very peculiar OADM. It has only one local port for Drop. This suggests that if you want to drop more than one wavelength at your location, you need to add a demultiplexer. Am I missing something, or is there something wrong with their design?
rs50terra 12/4/2012 | 10:27:26 PM
re: Network Photonics: A Corvis Copy Cat? Peter,
You are correct in the case of the 2400, which is part of a large cross-connect. Although, in this case, when you start adding modules to create a large cross-connect you have two problems (only?): cost and attenuation.
If you want to build a Clos network to create a 32 X #2 network, the cost of this box will run in the millions and the attenuation will force you to include internal amplifiers.
I can't really see how a CrossWave 1200 based OADM can fly. The CrossWave 1200 alone is $40,000. Add some additional components (for example a demux, some VOAs for the Add) and software and your cost goes up to $50,000? Let's mark it up for some 30% Gross Margin and we reach a list price of some $70,000.
Where do these guys live? Haven't they heard about the new, sane economic environment?
It's not 1999 baby! It's 2002. Wake up!
manoflalambda 12/4/2012 | 10:27:07 PM
re: Network Photonics: A Corvis Copy Cat? Crosswave1200 seems to be a very peculiar OADM. It has only one local port for Drop. This suggests that if you want to drop more than one wavelength at your location, you need to add a demultiplexer. Am I missing something, or is there something wrong with their design?

Hi. I gathered the same thing from their design. They must select one or N of the wavelengths and keep them on the same port. Perhaps not enough space to allow individual lambdas on either the add or drop side?

Salute,
Manoflalambda
HOME
SIGN IN
SEARCH
CLOSE
MORE
CLOSE