It took decades for someone to figure out that a ketchup bottle works better upside-down. Now Arrayed Fiberoptics Corp. thinks it's found a similar breakthrough for optical connectors.

On Wednesday, the company announced a new design that's arguably the first major advancement for connectors in 30 years. 

Arrayed will be showing off the new design at the Conference on Lasers and Electro-Optics (CLEO) in San Jose, Calif., next week. 

Arrayed does pretty much what its name suggests: It sells 2D fiber arrays and 2D collimator arrays, connecting multiple fibers to optical equipment. 

The company does business from the corner of a small industrial park in Sunnyvale, Calif., in a two-room shop that, for the moment, doesn't even have Arrayed's name on the door. (It's not as if passers-by are going to come in and browse.) 

But it's been profitable since 2006, says president and founder Benjamin Jian, and continues to turn out products with a staff of just eight. 

Connectors are little plastic-and-ceramic widgets used for connecting one optical fiber to another. One end of each fiber snaps into either side of the little rectangular connector; think of them as highly precise Lego blocks.

They're not especially glamorous, but they're vital to optical systems in general and have helped make an empire out of a company like Tyco, now called TE Connectivity (NYSE: TEL). About 500 million optical connectors ship every year, Jian says. 

Over the decades, connectors have been improved in small nudges -- they've gotten smaller, for instance. But the basic design has stayed the same: Specifically, the fibers actually touch. Each one has tip that's polished into a convex shape so that the fibers make contact, like two circles kissing.

What's wrong with that, in Jian's mind, is that the insertion loss -- the light that doesn't make it to the other side, essentially -- varies randomly. If you unplug a fiber from a piece of equipment and then plug it back in, you might get a different insertion loss. That's normal, and engineers have learned to shrug it off.

Jian ran into this issue in 1998 as an engineer at another company, when he worked on a device that had 40 connectors. "I wanted to do a good job, so I checked and checked, and never got repeatable results," he says. "I went to my manager, and he said, 'Don't bother. This is typical of connectors.'"

So, Jian has created a non-contact connector. The fibers are still convex, but recessed, so that they don't meet. There's an air gap between them. 

The reason connectors aren't made this way normally is because some of the light coming out of the transmitting fiber will actually reflect back, disrupting the light source. It's not a lot of light -- about 4 percent, Jian says -- but that's well above the acceptable level of one part per million. 

To get around that, Arrayed uses an anti-reflective coating for the fibers. The resulting non-contact connector produces more consistent results and happens to have lower insertion loss in the first place. 

When Light Reading visited Arrayed, Jian gave a not-so-scientific demonstration, simply plugging and unplugging connectors repeatedly into a test bed. (It's so easy, even a journalist can do it.) A high-end, off-the-shelf connector showed insertion losses of 0.09 dB down to 0.04 dB with no particular pattern; it's not as if the number got better or worse with repeated tries. Arrayed's non-contact connector always showed insertion loss of 0.04 dB, sometimes dipping down to 0.03 dB.

The air gap has another advantage: The glass doesn't wear down over time. Jian compares it to a vinyl record, which gets a little bit damaged with every touch of the phonograph needle. (Kids, ask your parents.) The non-contact connector is more like a compact disc (hopefully without the risk of fungal rot).

Aside from the coating and the air gap, the non-contact connector isn't particularly radical. Each fiber slips into a ceramic ferrule that holds it in place, and then the two fibers are slid into a split ceramic sleeve where the ends meet (without touching, of course). 

It looks like a normal connector. In fact, the samples in Arrayed's lab were even the same color as the normal connectors, distinguished only by numbers penciled on.

So, would anybody care about a new type of connector? Light Reading ran the idea by Jordan Chaney, an engineer with Fujitsu Network Communications Inc.He says it's interesting enough on the surface to check out, although he was a little worried about dust and other contamination invading the air gap. 

But the insertion loss does seem good and the idea does seem novel, he says. 

Arrayed has built only hundreds of the non-contact fiber connector so far. Jian and his crew do the work themselves, running machines that polish multiple fibers at once and contracting the antireflective coating production to a thin-film manufacturer. If volumes get big enough, Jian would like to start using overseas contract manufacturers.

If it all works out, Jian might pursue the thing that made him rethink connectors in the first place.

"I have a bigger dream. I want to create a connector that has 100 inputs and 100 outputs," he says. "That doesn't exist, especially in a single-mode version." He admits, though, that a single-fiber connector is a much bigger market.[1]

[1] Craig Matsumoto, Managing Editor, Light Reading. [Online]. Available: visit Light Reading. [Accessed: 04-May-2012]