& cplSiteName &

Scion Seeks to Slice Components Costs

Light Reading
News Analysis
Light Reading
4/19/2002

Has Scion Photonics Inc. made a big leap forward in reducing the cost of photonic integrated circuits? The components it exhibited at last month's Optical Fiber Communication Conference and Exhibit (OFC) suggest it might have.

For a kickoff, Scion created quite a sensation by showing a multiplexer chip containing 17 (yes, seventeen) separate Arrayed Waveguide Gratings (AWGs), all fully fiber-interfaced, on a single piece of silicon. It's probably the largest photonic integrated circuit in the world. What's more, Scion had connected up all the devices to prove that they worked.

Perhaps nobody actually needs 17 AWGs on a single chip right now, but that's not the point. The point is that Scion was able to showcase some impressive manufacturing expertise -- and that expertise will enable Scion to make high-performance photonic integrated circuits at a fraction of the cost of those from other vendors, according to Venkatesan Murali, the company's enthusiastic COO.

But it was more than just a technology demonstration, Scion also showed real products, which are available now, according to the company. They include VMuxes, comprising an AWG integrated with multiple variable optical attenuators; optical channel monitors; and "colorless" AWGs that can be reused for different groups of wavelengths in the C-band (also called cyclic frequency AWGs).

These parts started shipping to two customers in the last few weeks, Murali revealed to Light Reading. And he expects Scion to sign a major customer before the end of the current quarter.

Scion's sales pitch is very well oiled, despite the fact that it claims to have no sales and marketing people in the company at all. Here's the version we heard from Murali:

Integrated components based on AWGs have come under pressure recently, he says. Prominent companies like Agere Systems (NYSE: AGR) have dropped AWGs in favor of traditional thin-film filter muxes (see Agere Favors Thin-Film Filters). And an alternative technology, the bulk diffraction grating, is starting to look attractive for high-channel count systems, the niche where AWGs used to reign supreme (see Lightchip Launches 'AWG Killer').

Scion's proposing to reverse that trend, by making AWGs so cheap they'll be the technology of choice even for very low channel counts (see Scion Offers $100/channel AWGs). "We know that we can start taking on the thin-film filter community at eight channels," Murali claims.

How is it going to do this? By leveraging the economies of volume manufacturing. While technically Scion is a spinoff of disk-drive head manufacturer Read-Rite Corp., which provided the investment and facilities for the company to get started, it appears that the show was soon taken over by a bunch of people from Intel Corp. (Nasdaq: INTC). And, it has to be said, Intel does knows a thing or two about semiconductor manufacturing.

A materials scientist by education, Murali worked at Intel for 13 years in its microprocessor and memory products section, gaining experience in semiconductor wafer fabrication, advanced assembly and fabrication techniques, and optics. Scion's two VPs of engineering and its VP of materials were also at Intel and have similarly distinguished pedigrees. In fact, it's worth noting that four out of the six key executive positions at the company are connected with manufacturing.

The upshot of all this expertise is that Scion gets yields that are double the industry norm, according to Murali. Japan's NTT Electronics Corp. (NEL), a key vendor of AWGs and photonic circuits, gets only one or two dice per wafer, he contends (see Japan's NEL Targets Metro). Scion, which fabricates its devices on eight-inch wafers, claims 50 dice per wafer. That translates directly into much lower manufacturing costs.

The only other fabs that use eight-inch wafers are U.K foundry Optical Micro Devices Inc. (OMD) and Intel's new photonics fabrication facility (see Intel Intros Photonics Unit). Most other AWG vendors still use four-inch wafers, which have one quarter the area.

One interesting nugget of information about the startup -- and this isn't apparent from the company's Website -- is that at least some, if not all, of the ex-Intel executives spent their later years at Intel in its VC division, Intel Capital. As one of the most active VCs in the industry, Intel Capital has invested in more than 400 companies, a sizeable number of them in telecom, giving Scion's executives access to a vast amount of intelligence about potential competitors in the industry.

On the technology front, Scion seems to have plenty of innovations up its sleeve. For instance, it has developed a evanescent-field VOA (variable optical attenuator), which is similar in principle to the technology that Molecular OptoElectronics Corp. (MOEC) and ProtoDel International Ltd. use for their VOAs. Scion applies the idea to a waveguide rather than to a piece of fiber (see ProtoDel Pushes Passives for a description of the fiber version).

Evanescent-field VOAs consume less electrical power than competing technologies, which are usually thermo-optic. A 40-channel VMux from Lightwave Microsystems Corp., for example, requires 17.5 Watts, Murali contends, whereas a similar device from Scion consumes only 8W. (On its Website Lightwave actually claims power consumption of under 20 W at 65 ºC or under 25 W at O ºC.)

Another important area is software, says Murali. He says Scion has written its own software to simulate the optical, thermal, and mechanical properties of the devices before fabricating them. As a result, the devices are always right first time, Murali claims. That's one of the key reasons the company can crank out devices from start to finish in as little as two weeks. Its competitors, he contends, take anything from three weeks to several months to perform the same service.

John Midgely, CEO of Lightwave Microsystems isn't buying it: "If something sounds too good to be true, then it probably is."

However, we could only find one chink in Scion's story. As Lightwave's VP of marketing Ferris Lipscomb puts it, "It [Scion's technology] is fundamentally low cost, but only in high volume. And nobody's running in volume right now.

"In the boom times people would order blanket orders of 50, or even 100 a month," he adds. "Now people are ordering exactly what they need." Even in the good times, medium volume was as high as it ever got in terms of pricing.

Lightwave says that its prices are below $200 per channel. Prices have come down drastically in the last 12 months, according to Lipscomb. But, in general they haven't reached $100 per channel -- Scion's target.

He goes on to note that true price comparisons are still tricky: "People toss numbers about. But until you look at the specification and the volume required, it's really impossible to compare. For every customer the specification is negotiated, almost on a custom basis."

Scion's main shareholder is Read-Rite. The other investors are Integral Capital Partners and Tyco International Ltd. (NYSE: TYC; London: TYI). Scion's funding to date totals $30 million.

— Pauline Rigby, Senior Editor, Light Reading
http://www.lightreading.com

(18)  | 
Comment  | 
Print  | 
Related Stories
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
Page 1 / 2   >   >>
redface
redface
12/4/2012 | 10:34:02 PM
re: Scion Seeks to Slice Components Costs
Does anyone know how Scion can create planar waveguide VOAs with only maximum 1 dB insertion loss? See http://www.scionphotonics.com/....

My understanding is that normally waveguide VOAs are based on Mach-Zenhder interferometers and have high (2 dB) insertion loss. Scion's announcement of low loss VOA matches that the low loss VOA array from Gemfire (which is based on leaky polymer waveguide).
beachboy
beachboy
12/4/2012 | 10:33:56 PM
re: Scion Seeks to Slice Components Costs
The market is so bad for products like AWGs, not just because no-one is implementing more than a few lambda chunks at a time, but also since the price/lambda is being slaughtered by people like Scion (even for a few pieces) making the overall market size small even if volumes were on track.
borneo
borneo
12/4/2012 | 10:33:55 PM
re: Scion Seeks to Slice Components Costs
Scion, like many other hyped component ventures (e.g., remember Nanovation) will have to really deliver on their claims to be credible. And therein lies the rub. The market is, at best, a year away from any measureable expansion. Hence, while Scion may indeed have the goods will they still be around when the cycle goes positive again?
fibervision
fibervision
12/4/2012 | 10:33:48 PM
re: Scion Seeks to Slice Components Costs
I was at OFC & saw the 17 FI MUX display. Yea, sure it was useless, but that wasn't the point. If the data displayed was correct, every single device worked. In other words, It will be hard to justify a 4" process fab. But, time will tell if Scion's claims (and company) can survive the slowdown. In this case, bigger is not better.
dwdm2
dwdm2
12/4/2012 | 10:33:48 PM
re: Scion Seeks to Slice Components Costs
"... Scion created quite a sensation by showing a multiplexer chip containing 17 (yes, seventeen) separate Arrayed Waveguide Gratings (AWGs), all fully fiber-interfaced, on a single piece of silicon. It's probably the largest photonic integrated circuit in the world. What's more, Scion had connected up all the devices to prove that they worked."
-------------------------

This is just about the vaguest statement I have ever seen. It sounds a lots of fantasy without saying anything meritworthy.

So there are 17 AWGs on a single chip. Exactly how big is the chip? No clue (well, biggest in the world !-). What are all these chips being used for? Hope not just a decoration. Altogether how many channels we are talking about? Don't ask me!

If all AWGs are on a single chip, why fiber interfacing? Elsewhere they have interfaced PLCs on chip with on-chip waveguides -- lot less mess, higher reliability.

If all they are making is a basic mux, why 17 AWGs need to be connected to each other (interfaced) via fiber?

Can someone please help us understand...perhaps Pauline herself?

Thanks.
LightSeeking
LightSeeking
12/4/2012 | 10:33:41 PM
re: Scion Seeks to Slice Components Costs
Redface,

It is theoretically possible to limit the MZI insertion loss to less than 1 dB or even less than that. The question is that whether Scion actually achieved it in real devices along with the other parameters they listed.

LS

- "Does anyone know how Scion can create planar waveguide VOAs with only maximum 1 dB insertion loss? See http://www.scionphotonics.com/....

My understanding is that normally waveguide VOAs are based on Mach-Zenhder interferometers and have high (2 dB) insertion loss. Scion's announcement of low loss VOA matches that the low loss VOA array from Gemfire (which is based on leaky polymer waveguide)."
waterbuzz
waterbuzz
12/4/2012 | 10:33:40 PM
re: Scion Seeks to Slice Components Costs
I give an A+ to the marketing group at Scion. As for the company as a whole....PLEASE. Give me a break. These guys are a whole lot of nothing. They can't deliver squat! The fact that they had to build a 17 piece, waste of time demo, tells me that they are clearly not focused on the customer but rather some virtual science fair. I would have been more impressed with products that people can actually buy today.

$100 channel is so much to do about nothing! I'd would pay a $1000 to see Scion deliver a true high performance DWDM at $100 channel and actually establish a long term business. To come out at $100 channel clearly indicates these guys are amateurs. I will boldy say that you cant deliver top-class DWDM at $100 channel today. Sorry, these guys dont know what they are getting into. Please tell us in 1 year how you are doing and how many orders you have.

In the meantime, please keep the market clear of your utter nonsense noisemaking -it's very annoying.
dwdm2
dwdm2
12/4/2012 | 10:33:39 PM
re: Scion Seeks to Slice Components Costs
"Scion's proposing to reverse that trend, by making AWGs so cheap they'll be the technology of choice even for very low channel counts (see Scion Offers $100/channel AWGs ). "We know that we can start taking on the thin-film filter community at eight channels," Murali claims."
-------------------

In optical industry the cheapest is not always the best. Copetitive price is a major issue (not the cheapest!), however performance and reliablity are of paramount importance.

The real test would be if they can go IPO and sustain on a real customer base. Burning VC money to impress others in a science project manner is not what gonna take them where they'd like to be!
Pauline Rigby
Pauline Rigby
12/4/2012 | 10:33:26 PM
re: Scion Seeks to Slice Components Costs
I can shed a bit of light on this. (It looks like some stuff got cut from my story -- it was getting a bit long).

Murali told me that Scion uses an evanescent-field VOA, similar in principle to the side-polished fiber VOAs offered by Molecular OptoElectronics Corp. (MOEC) in the U.S. and ProtoDel International Ltd. in the U.K. But in Scion's case they've made it out of a waveguide rather than a fiber.

This type of VOA consumes less electrical power than competing technologies, which are usually thermo-optic. There's a power consumption comparison in the story. I believe that this type of VOA also has a low insertion loss because it's an in-line technology.

There's a sketchy description of how it works in this story about Protodel: http://www.lightreading.com/do....

[email protected]
Pauline Rigby
Pauline Rigby
12/4/2012 | 10:33:25 PM
re: Scion Seeks to Slice Components Costs
Hi dwdm2,

I think Scion was making a point about yields with its 17 AWG demo. The only way to prove beyond any doubt that all the devices from the same piece of silicon are working devices is to leave the silicon intact.

Scion didn't use waveguides to connect the individual AWGs on the chip because they were never meant to be used together for a single application. The wafer would then be cut up into inidividual AWGs or possibly broken up into pairs.

Having two AWGs on the same chip is quite useful apparently. In many applications, you need two AWGs, one to act as mux and one to act as demux. By putting them on the same chip I guess it saves on packaging costs. In this use, both AWGs must be fiber-interfaced separately.

Yes, the chip was huge. Again, I think that was the point. Some integrated circuits -- an 8x8 non-blocking switch was quoted as an example -- are so complex that they won't even fit on a single 4 inch wafer, and that's before you worry about whether you can make the thing accurately enough (yields again).

[email protected]
Page 1 / 2   >   >>
Featured Video
Upcoming Live Events
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events