Agere Favors Thin-Film Filters

Agere Systems' (NYSE: AGR) announcement of a 50GHz mux/demux highlights the fact that the company is choosing thin-film filter technology over Arrayed Waveguide Gratings (AWGs) for wavelength multiplexing (see Agere Ramps Up 3D MEMS).

Until now, the consensus in the industry seemed to be that TFFs, AWGs, and a third alternative, Fiber Bragg Gratings (FBGs), all had their own application niches. But Agere's strong push into TFFs suggests that it now sees one of those technologies coming forward as an overall winner.

"Thin-film-filter-based passive technologies are less expensive than fiber Bragg gratings or AWGs for 1x4, 1x8, and 1x10, the configurations most in demand by customers," says Tom Christie, director of strategic marketing for Agere's optical core and metro networks division, in the press release issued today.

Originally Agere had both thin film and AWGs under one roof, but the company quietly ditched its AWG business about 18 months ago, according to Dan Wilt, product line director for components. "We had some issues with that component, and we haven't yet relaunched it," he says.

Could those issues relate to the fact that Agere's biggest customer was no longer buying its AWGs? Lawrence Gasman, an analyst with Communications Industry Researchers Inc. (CIR) seems to think so.

Agere had originally developed AWGs for internal consumption while still part of Lucent Technologies Inc. (NYSE: LU). But Lucent appears to go elsewhere for its AWGs now. According to Gasman, who is currently co-writing a big report on wavelength multiplexers, Lucent's main AWG suppliers are Hitachi Cable Ltd., NTT Electronics Corp. (NEL), and Lightwave Microsystems Corp.

Agere also realized that it was competing with itself, says Wilt. At medium channel-count numbers, AWGs and TFFs can each fit the bill, so Agere would not necessarily gain more customers by offering both types of product.

And Agere isn't the only company to back away from AWGs recently: Nortel Networks Corp. (NYSE/Toronto: NT) also killed its AWG development work (see Nortel Nixes Passive Components). Nortel, however, doesn't makes its own TFFs, but these are a commodity product on offer from dozens of vendors, including JDS Uniphase Inc. (Nasdaq: JDSU; Toronto: JDU) and Oplink Communications Inc. (Nasdaq: OPLK).

Agere outlines the advantages of TFFs over AWGs in its press release. In a nutshell, TFFs have better insertion loss, crosstalk, and stability with temperature than AWGs, it claims. A typical figure for insertion loss is 3.8 dB, it claims. For comparison, AWGs vary between 4.0 dB and 8.0 dB, depending on the vendor and the application they have been designed for. And only the products with losses at the higher end of the range have optical passband characteristics as good as TFFs, Agere notes.

Against TFFs is the fact that they are more complicated to assemble and package than AWGs. Nevertheless, Agere claims the cost of producing them is still much lower.

It should also be kept in mind that AWGs are a lot more than just another way of splitting and recombining wavelengths. They're a starting point for making a wide range of integrated optical components such as mux-variable optical attenuators, optical channel monitors, dynamic gain equalizers, and optical add/drop multiplexers (see Photonic Integrated Circuits).

Agere's 50GHz mux/demux component, called the W7500, is available now in both samples and production quantities. It rounds out Agere's product range, which also includes 100GHz and 200GHz versions. The new device will be showcased in a live demonstration at the Optical Fiber Communication Conference and Exhibit (OFC).

— Pauline Rigby, Senior Editor, Light Reading
http://www.lightreading.com For more information on OFC 2002, please visit: www.nottheofc.com

Page 1 / 3   >   >>
dwdm2 12/4/2012 | 10:50:23 PM
re: Agere Favors Thin-Film Filters While conceptually simpler, filter based WDMs are complicated to manufacture. Precision fiber alignment and bonding is necessary to stabilize the passband for each ITU channel.

Also, for every channel, a different filter corresponding to its passband is necessary. Therefore, the center wavelength of each passband differs from the ITU frequency at random.

For multiple channels, the filters are cascaded; and since the losses are cumulative, this yields highest loss for the last channel in a module.

To adjust the insertion loss uniformity within tolerable range, attenuation is added to low loss channels to bring them closer to the highest loss channel.

Since many filters are involved, packaging is tedious and prone to various kind of failure. The overall yield of filter based WDMs is only about 40%, making this technology not the best choice for WDM applications!

Nevertheless, photonics integrated circuit (PIC)
"http://www.lightreading.com/do..." technologies are still not matured enough to claim major market share, therefore, filter based WDMs enjoy a sizable market share.

This trend is bound to change. AWG technology is maturing fast. The biggest driving force is the market. As the demand rises, TFF will have hard time to keep up.

Right now the TFF folks not only control the technology, they also control the market. However, the majority of the startups are shooting for integrated approach - one kind or another. This tells us that PICs will win eventually.

How soon? That's the question!


captain kennedy 12/4/2012 | 10:50:21 PM
re: Agere Favors Thin-Film Filters Any comments on crosstalk or temperature stability? Which technology has lowest system impact?
mess 12/4/2012 | 10:50:19 PM
re: Agere Favors Thin-Film Filters This is of no surprise to me.
Few weeks ago I talked with one folk from Nortel and he told me that Nortel doesn't use AWGs. They use TFF.
When you look at the reports of public companies like Bookham you see that they don't sell AWGs.
Currently there are 39 vendors that offer AWGs but their market does not seem to rise, why ?
Can anyone add more to it ?
--- Mess
Physical_Layer 12/4/2012 | 10:50:18 PM
re: Agere Favors Thin-Film Filters TFF is well know to have a steeper roll-off, so the crosstalk performance is better. They are also much less sensitive to temperature. AWGs are very temp sensitive but you can (with cost) add temp control which solves the problem. Some vendors are doing athermal AWGs (not sure if they are commercial yet) but you usually take an insertion loss hit to get an athermal device.
Petabit 12/4/2012 | 10:50:18 PM
re: Agere Favors Thin-Film Filters Ignore the technology. As far as the system vendors are concerned, they don't care whether the devices are TFF, AWGs, or whatever.

There are two classes of WDMs, ones that work in series, and those that work in parallel. TFFs are serial devices, which means that you can 'pay as you grow' and just sell as many channels as you need. AWGs are parallel devices and you need to buy all your channels day one.

If you really believe that your system is going to carry 40 channels day one, then use a parallel device. Otherwise, a serial technology like TFF or FBG will win.

The new low stress deposition techniques for TFFs have pushed they very far beyond FBGs in performance, which is why you are seeing TFF becoming the dominant filter technology.

aurora 12/4/2012 | 10:50:17 PM
re: Agere Favors Thin-Film Filters Petabit writes: "AWGs are parallel devices and you need to buy all your channels day one."


Generally true. But a band filter approach would allow me to build the system now and add the AWG's as needed.

Don't AWG's give us much denser packaging? One reason I suspect AWG's will be appropriate in applications closer to the core. Especially since the core is likely to have Amplifiers nearby to overcome the AWG's insertion loss.

But I'm not an optics engineer. Corrections welcome.
Component_Guy 12/4/2012 | 10:50:17 PM
re: Agere Favors Thin-Film Filters Nice to see Agere finally get some return on one of their investments. If I remember right, they paid ~500M for Herrmann two years ago, about $7M per employee.

Dr.Q 12/4/2012 | 10:50:16 PM
re: Agere Favors Thin-Film Filters To be balanced, and not be too long winded in one post, here's a comparable look at TFF's

1) Size - the TFF chips are made in a batch process as well. The wafer is diced to chips that are ~ 1mm x 1mm, thus several hundred can result from a good wafer.
A single device in a cylindrical package is about 30 mm long, 6 mm diameter.

2) Wavelength control-- Active monitoring in the wafer growth allows very good control of the center wavelength of the wafer. There is some variation across the wafer, but individual chips can be annealed to shift their center wavelength. Fine tuning of the wavelength is done at packaging, by tilting the chip in a controlled angle.

3) Loss -- The necessary optical coupling is from fiber-fiber through free space, which can be done with <0.1 dB. The chip has some background loss typically, 0.1 ~ 0.5 dB. This results in a single device with far lower loss than an AWG.
The loss is cumulative, however, when you cascade many of these to make a WDM (as 'dwdm2' pointed out in Post #1)

The thermal stability is, as noted before, very good. 2 pico-meter/C (0.002 nm/C) is currently feasible.

- Dr. Q
Milano 12/4/2012 | 10:50:16 PM
re: Agere Favors Thin-Film Filters Petabit is right: on a cost per channel basis, AWG should win the day, however in the current environment the challenge is to build a system on a pay as you grow basis. Who needs to pay for 40 lambdas day one? Nobody needs a 5-year payback on a new system. Most systems out there grow only at a few lambdas per year.

The band approach would be worth a thread by itself. Though you may be able to grow more progressively, you end up with quite a few AWGs in the system, increasing both the initial cost and the cost per channel.

Physical_Layer 12/4/2012 | 10:50:15 PM
re: Agere Favors Thin-Film Filters Dr Q - fantastic summary, thanks for posting it.

To all: care to comment on the suitability of Echelle Grating technology? Clearly it is less mature than AWG technology (although they've both been around for a long time), but there appear to be major chip-size advantages which should improve yield and make temp control easier to do. Any other comments?
Page 1 / 3   >   >>
Sign In