CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

Behind the scientific gobbledygook method

CNET News.com's Michael Kanellos may not be confused with Einstein, but he deserves a break after a particularly cruel intellectual ordeal.

The annual International Solid State Circuits Conference is one of the more interesting get-togethers I attend.

Why? Because I have absolutely no idea what's going on.

On Wednesday at 11:15 a.m., for instance, ISSCC attendees could choose among:

• A lecture on a one-cycle lock time slew-rate-controlled output driver presented by Hynix;

• The IBM blockbuster: "A 16Gb/s Source-Series Terminated Transmitter in 65nm CMOS SOI;" or

• UCLA's "11b 800MS/s Time-Interleaved ADC with Digital Background Calibration."

In the end, I opted for the speech on the "2D Microfluxgate Earth Magnetic Field Sensor" because it was sandwiched between a presentation on a microgyroscope with automatic mode matching from Georgia Tech and another on a gravimetric FBAR circuit operating in liquids with flip-chip packaging from Infineon. I figured it would be easier to stay put than lose my seat.

The average person can barely fake any sort of comprehension in the proceedings for very long. The person next to you--usually a chip designer or a professor at a university--will invariably try to strike up a conversation.

What do you say? "Er...gravimetric FBAR. Are those the boots Richard Gere wore in American Gigolo? Do you know where the Loss Compensated Distributed Amplifier Smackdown is? Is it in Cypress II?"

A universe ruled by the whims of the customer is one rife with chaos.
It just doesn't work to just blurt out, "So where's the value add?" and then go back to your nap.

The really enthralling part of all this, of course, is getting (once again) tangible and overwhelming proof that there remains a fairly large population of people in the world far smarter and more dedicated than I will ever be--or you, for that matter. No offense--don't you think your time would be better used devising a wide locking range regenerative frequency divider like J.C. Chien from National Taiwan University?

If you listened to a lot of the chatter from bloggers or analysts in recent years, tasks such as improving computer performance and ramping up processor speeds were supposed to become irrelevant. Instead, leading companies in the future would be set apart from competitors by improving industrial design, introducing products rapidly, or coming up with ways to create a tighter emotional bond with consumers.

A symbolic death knell for specs came when Intel and Advanced Micro Devices submarined clock speed on their chips.

A universe ruled by the whims of the customer, though, is one rife with chaos. One day they want SUVs; a few months later they are clamoring for hybrids. Lost: yes, but who knew?

Ponder this for a moment: Why did YouTube win the online video war? It emerged around the same time as similar companies such as Veoh Networks, and the sites function in a similar manner. Yet YouTube has become a worldwide titan while Veoh is best known as a place to download skateboard videos or kung fu movies that fell out of copyright. The same applies for MetaCafe.

I asked a venture capitalist once how he picked winning consumer start-ups. He shook his hand in a pantomime of shooting craps and threw the imaginary dice. "Whoosh," he said.

Admittedly, customers need to be kept happy. But if forced to choose between devising fine-grain redundant logic using defect prediction flip flops (thank you, T. Nakura of NEC) and trying to figure out why Tony Danza does better in the 3:30 to 4:30 time slot than in prime time, I'd choose the first. It just seems so much less nebulous.

Besides, research is simply fundamental. Although accountants might view it as an expensive task that can be outsourced, none of the easy-to-use technological marvels that everyone takes for granted today would be possible. Those cell phone videos of celebrities having sex couldn't be captured without digital signal processors, a type of chip first unfurled in a 1980 paper presented at ISSCC by Bell Labs.

In a decade, personal in-car radar systems may be common. And the two papers presented by University of Southern California professors on the subject at the conference this year (one is titled "A monolithic 4-Channel Beam-Former in 0.13 CMOS Using a Path-Sharing True-Time-Delay Architecture") could well be the foundation for that.

Gotta run. I have to see a man with DLL jitter reduction techniques.