Optical components

Startups Carry VCSEL Flame

Pioneering tunable laser startup Coretek may have been closed down after its acquisition by Nortel Networks Corp. (NYSE/Toronto: NT), but that doesn't mean other vendors have given up on the idea of developing tunable Vertical Cavity Surface Emitting Lasers (VCSELs) (see Coretek Is Closed).

Indeed, a new startup -- BeamExpress Inc. -- has recently sprung up in Europe. It is featured in a new report on tunable lasers, published today on Light Reading: Tunable Lasers Revisited.

BeamExpress was founded in late July 2001 by Professor Eli Kapon at the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland (see Swiss Launch Light Emitter Startup). In October last year, it revealed its intention to manufacture tunable lasers and fixed-wavelength laser arrays (see BeamExpress Intros VCSEL, CEO).

It takes on established startup Bandwidth9 Inc., which appears to be the only vendor offering a tunable VCSEL for sale today.

Both companies want to capture the market for metro transmitters. "There isn't much choice in the metro," says BeamExpress's CEO Jean-Claude Charlier. "You either use long-haul technology, which is expensive, or you use inexpensive plastic enterprise transceivers, which don't do the job properly." Tunable VCSELs could find the middle ground, offering a combination of medium cost and solid performance.

The technological challenges faced by these two startups are not to be sniffed at. Making a VCSEL that operates in the DWDM band around 1550 nm is not trivial, because indium-phosphide -- a material that generates light at 1550 nm -- is difficult to design with (see Laser Blazers). Adding tunability presents an extra challenge.

The two companies have contrasting approaches to the same goal. Bandwidth9 uses a technique called "pseudomorphic growth," which is a way of depositing one material on top of a dissimilar material in a single epitaxial process (see Bandwidth9 Claims Laser Breakthrough). It uses this technique to combine an indium phosphide gain (light-producing) region, with a gallium-arsenide-based Bragg mirror.

BeamExpress, on the other hand, uses wafer bonding, which is a high-temperature way of fusing two materials together. The materials are first produced in separate processes.

Not everyone believes that the wafer bonding approach is a viable technique for mass production. According to Markus Ortsiefer, CTO of German VCSEL startup VertiLas GmbH, says it's difficult to bond whole wafers uniformly, so the technique is generally only used for small pieces of semiconductor. "Wafer-bonding has no future," he says. "It's too complicated, trust me." VertiLas has an alternative method of making 1550 nm VCSELs, but only plans to make fixed-wavelength, not tunable devices.

But BeamExpress stands by its technology. "We have a partnership with a company developing wafer bonding kit, so we can produce very high-quality fused interfaces," claims the startup's CEO Jean-Claude Charlier.

The resulting laser has an output power in the region of 1 mW, typical for a VCSEL but insufficient for metro and long-haul applications. To overcome this drawback, BeamExpress plans to pump the laser optically, using a second laser. In this respect, its technology resembles that of Coretek, which also used optical pumping.

However, BeamExpress says that its laser will be better than Coretek's in a couple of ways. Coretek boosted the output power of its laser by integrating an amplifier into its package, in addition to using optical pumping, whereas BeamExpress claims to achieve high output powers without an amplifier. "The amplifier is the most expensive part by far, and we don't need one of those," says Charlier.

Bandwidth9, on the other hand, chooses to boost the transmission distance of its laser by integrating a modulator from CyOptics Inc. (see Bandwidth9 Picks CyOptics Modulator). This kind of amplifier delivers some gain and has what's called "negative chirp," which increases the resistance of the signal to dispersion, allowing it to go farther with lower power.

The other area in which BeamExpress claims an advantage is in the design of the MEMS (micro-electro-mechanical system) cantilever and mirror, which are used to tune the device. "We have achieved a record low control voltage of just 4 volts," Charlier boasts. Coretek and Bandwidth9 need much higher control voltages than this, he contends -- up to 40 V.

Bandwidth9 acknowledges that the tuning voltage is 40 V. But, says the company's VP of marketing Jake Weise, this is with almost no current. Additionally, he notes that the latest product, MetroFlex G2, takes +/- 5 V user-supplied inputs and generates 40 V internally. (It was difficult to verify Coretek's tuning voltage, since the company no longer exists.)

Of course, it should be pointed out that Bandwidth9 has been around a lot longer than BeamExpress and has notched up design wins with some top tier customers, such as Nortel (see Bandwidth9 Scores a Coup). It is now on its third-generation product and has had time to get further with reliability testing -- a process that cannot be speeded up. BeamExpress will have its work cut out to compete successfully.

— Pauline Rigby, Senior Editor, Light Reading
crystalgrower 12/5/2012 | 12:53:41 AM
re: Startups Carry VCSEL Flame Usually when people tell you a technology has no future, it means they're too narrow minded to think there are aspects of the technology which they do not understand!

Over the years I've heard "educated" people say this about, MOCVD, InP lasers, and VCSELS!

Isn't science about having an open mind?!?!!
darkone 12/5/2012 | 12:53:35 AM
re: Startups Carry VCSEL Flame Science is about having an open mind. Engineering is about designing the best compromise solution. Business is about selling the cheapest products that will meet the market requirements.

Wafer bonding appears to be a simple technology, but after lots of time and moeny it is still not a good solution for bonding InP (active region) to GaAs (DBR layers). The basic problem is that the surface of a grown wafer is not flat. The UHV hydrogen method seems to work well for polished wafers, but it appears to have a very poor yield on wafers with layers grown (MOCVD or MBE) on them.

They would need a great (read expensive) engineering effort to modify either the UHV cracked hydrogen bonding system to deal with nonflat wafers or a novel (again likely expensive) way of growing wafers that produces flat surfaces.

I am sure we don't understand a lot of things about wafer bonding, but I feel it is unlikely that they are going to sort it out.

deer_in_the_light 12/5/2012 | 12:53:29 AM
re: Startups Carry VCSEL Flame Any opinion on maturity of MEMS wafer bonding processes ?
ken9988 12/5/2012 | 12:53:22 AM
re: Startups Carry VCSEL Flame We received several calls and resumes from ex Bandwidth9 employees who were laid off recently. They said that Bandwidth9 is shuting down their Atlanta office and laid off 35% more people. Their CEO, CFO,...CxO all are runing to hunting for a buyer, otherwise shuting down will be inevitable.

Can anyone confirm this?

darkone 12/5/2012 | 12:53:19 AM
re: Startups Carry VCSEL Flame It seems to work fairly well. Since they are just using silicon without grown layers the wafers can be polished flat. Also, silicon cleaning is a more mature technology. And they can use anodic bonding with quartz wafers. I have also seen some interesting work on silicon carbide encapsulation.

BobbyMax 12/5/2012 | 12:53:01 AM
re: Startups Carry VCSEL Flame Although VCSEL is a powerful technology in that it helps bring down the cost of optical networks. In spite of its many applications: Telecommunication Switches

Data Communication Networks (Gigabit Ethernet (LAN), Storage Area Networks, High Speed I/O (Gbic),and Optical Storage, etc. In spite of its high potential the VCSEL technology has not advanced. In fact, quite a few VCSEL companies have gone out of business. One of the main reason deems to be manufacturing difficulties and cost.

Bar Code Scanners
lightbeer 12/5/2012 | 12:53:00 AM
re: Startups Carry VCSEL Flame I have a friend who was let go from BW9 recently, and he says that the management is trying to find a buyer for the company. His comment was "no one is buying product at this time".

He also mentioned that BW9 changed their product focus from the metro (no use for tunables at this time) to regional networks to try to capture the sparing market, but their product had such poor reliability that it defeated the purpose of replacing multiple fixed wavelength lasers with tunables. Hope this bit of gossip helps.
The_Escapist 12/5/2012 | 12:52:22 AM
re: Startups Carry VCSEL Flame What do you get if you buy BW9? Backstabbing management that lays off people actually capable of making a device that has any hope of being reliable?

People with no discernable talent beside making Powerpoint slides and manipulating spreadsheets to calculate the costs of manufacturing the unmanufacturable?

Unfortunately, you mostly get the above. And a few really brilliant engineers that have kept the place going. I only those guys land on their feet. The rest of them, well...

I'm suprised the VC's haven't taken back the little of their money that is left.
ken9988 12/5/2012 | 12:52:16 AM
re: Startups Carry VCSEL Flame Escapist, you sound like an ex-Bandwidth9er. Can you shed a little more light on this company - Why their product is unmanufacturable; Why the cost is till high? Any real customers?

From their CEO's statement, he is still very cocki - Graham says, Bandwidth9 CEO: "It's not a question of interest; it's a question of whether they can afford us," http://www.lightreading.com/do...
BlueWater66 12/5/2012 | 12:52:14 AM
re: Startups Carry VCSEL Flame I've never worked at Bandwidth9, but there are a number of public patents and other documents that show what they do.

As I understand it, the base structure 1550nm VCSEL has a huge number of epi layers. On top of that, there is a large MEMs "arm" that goes up or down to adjust the wavelength. Growing something that large has to be a nightmare. Plus, their output power is probably very very low. They are also targeting the directly modulated market. As you increase and decrease the drive current on this device, thermal effect and other things will probably cause the structure to move around or bounce. This might cause chirp issues. A guess.

So, (1) very low yield, (2) a bitch to actually fab with very expensive MBE processes and packaging, (3) low output power, (4) chirp

This is a typical large silicon valley VC type play. They pump huge amounts of money into a technology push company. I suspect they are choking on the burn rate, exit options and lack of new outside investors to sustain the value. If they are looking at an inside round, then life stinks. It becomes a nasty zero-net-sum process... hence looking for a buyer who will at least return some of their investment.

BUT- The comments above are very negative from a down market perspective. In reality, I actually like the technology. If they could pull in another $100M and keep working on it, it would be a really cool device. Huge companies like Intel pump billions into 90nm/300mm fabs to make chips that have a huge number of gates. This is a push concept, but backed by someone with the money to do it. And it works. With enough money, these sorts of integrated tunable lasers would eventually work and offer a real value proposition. But I think the start-up market won't currently support this.
Sign In