CompoundSemi Online - The Original Compound Semiconductor Industry Newspaper

September 20, 2007...The never-ending question is whether compound semiconductors and silicon-based semiconductors are friends or enemies. Of course, when you put CS materials on top of silicon, like Nitronex and IQE know how to do, the question is moot. Obviously they marry well and produce lots of smart, hybrid children. But when you read two different news articles that come to totally different conclusions regarding compounds vs. silicon as transistors or power amps, well, that gets many an industry insider's juices going. It can also confuse industry outsiders.

The articles that caught my eye both recently appeared in IOP's Compound Semiconductor magazine, our CS industry colleagues 'across the pond.' The first article announced that IBM has intentions of producing CMOS power amps within the next three to five years and that they'll be worthy of replacing some of the many GaAs PAs that currently enjoy residence in a number of WLAN front end systems and cellphones. Good old Big Blue. Once a tremendous computer hardware giant, now just a squeaky wheel. While longtime IBM friend Bernie Meyerson helped popularize SiGe materials and devices (which we claim as part of our CS world since it includes two ore more elements), the reality is that, these days IBM is mostly known for its software expertise. It dabbles now in devices, mostly at the R&D level with traditionally with limited follow-up or excitement. So when IBM says it intends to bring forth a CMOS 7RF SOI (Silicon on Insulator) RF process that will pose a serious competitive process to GaAs, I just yawn and say... we'll see. Not that they couldn't do it, but... why bother? Unless, of course it works well and then we can do like we did with SiGe, and claim it. After all, SOI indicates it involves one or more elements.

The process is their seventh gen (thus the "7RF" moniker) 0.18 µm CMOS process which would compete with the CS industry's current 0.25 µm plus processes. It could only pay off if IBM can make the end products seriously cheaper and better than GaAs. And thanks to incredible market penetration by leading CS industry companies like RF Micro Devices, Anadigics, and TriQuint, GaAs is now entrenched in the market. IBM's goal is a one-chip phone, but I'm sure RFMD can do that in MBE-produced GaAs if it isn't already doing so. The IBM guy spearheading the project is Ken Torino, IBM's director of foundry products. He says that in the current corporate schedule, volume production using the new process will begin in Burlington, Vermont in '09. I love his quote: “We're pretty much working with the ‘who's-who’ of the industry and it's in cooperation, not in opposition.” Yeah, right. It sounds like pretty much traditional head-on competitive talk to me. He went on to say... “It might be introduced to wireless LAN first, only because it's easier, but everybody recognizes that the big win is the huge volumes of cellular phones. It depends how ambitious the people with power amp design expertise are to move into CMOS, but it would be possible to do it in 3-5 years.”

We've heard that talk from siliconers before. "...Possible in 3-5 years." The key line is: "It depends how ambitious the people with the PA design expertise are to move into CMOS." Much like government program directors, by necessity, systems designers are loathe to taking risks with a new technology. They've always been risk-adverse. Once they finally get their teeth into a relatively new technology, they keep on chewing and don't let go. After 30 years, GaAs is finally entrenched in their design culture. GaAs isn't likely to move over willingly (not even for GaN).

The joy of joys is the uplifting message I found in the second article where Intel's director of technology strategy and Semiconductor Industry Association (SIA) roadmap whiz, Paolo Gargini, finally admits that the compounds have a chance of gaining overdue respect and recognition... by 2015. We've only been knocking at he door begging for equal status at SIA for about twenty years. So what's another eight or so years.

Speaking on behalf of SIA, what Paolo Gargini pointed out in a briefing at IBM's Madison Avenue boardroom in New York (ironic coincidence) was that what the semiconductor industry as a whole (compounders and siliconers) needs to deliver in order to keep Moore's law moving on track, is the higher electron mobility that III-V devices can deliver. Plus a threefold higher processor speed, or power consumption that is ten times lower. Good company man that he is, he pointed specifically to Intel's P1266 process replacing silica dielectrics with high-k materials and polysilicon electrodes with metal. He also underscored that the already breaking away from all-silicon material technology. He said, “Disruptive and paradigm shifting technologies are needed to keep Moore's law going.” Lovely man.

These latest pearls of wisdom from an Intel person remind me of when I was working inside Emcore in corporate communications, way back in the late '90s when VCSELs were first making their mark. Some guys from Intel came out to Emcore's Mode operation in Albuquerque, New Mexico. Mode was one of the major VCSEL pioneering companies, which were surprisingly hard little devices to explain. Their potential applications were only just being tapped. Emcore and others certainly knew how to make them, but what was lacking were practical applications to make high volume production profitable. VCSELs are such tiny devices that they can be produced much like the advanced MOCVD-based LEDs that now dominate solid state lighting applications. The trick is getting the volumes up so the costs can go down. What Intel came to tell Emcore... and these were R&D whiz kids from Intel telling us this... was that VCSELs were being looked at seriously by Intel as next gen memory devices. Talk about potential volume...

At the time, Intel swore us all to secrecy about that being their intent. Of course one loose-lips [then, now ex-] Emcore VP went right out and blabbed it all over, so Intel ended up walking away from the scene, never to be heard from since. Until now. Well, maybe they've talked about compounds since, and certainly we've seen Intel around compound semi circles in one capacity or another, but for all intent and purpose, they canned the VCSELs for memory devices at that timeframe. A similar dead end move by Emcore, was when some of the advanced thinkers wanted to get serious about making oxide reactors, materials and devices. Emcore senior management said nix to that idea. (They also said nix to InP, which in retrospect was probably a good move.) It sounds like MOCVD technology developers who, at the same timeframe, elected to continue pursuing oxide technologies made a very good decision. We'll just have to wait and see if Paolo Gargini's hunches pay off, and the dreams of compound semi advocates really do come true.

Fairchild Subsidiary, System General Wins Taiwan Invention Award
Compoundsemi News Staff

CyOptics Introduces its 1310nm FP DFB TOSAs
Compoundsemi News Staff

Epic Communications Introduces Even Smaller Front-End Module
Compoundsemi News Staff

TDI Introduces Nonpolar GaN on Sapphire
CompoundSemi News Staff

Micromem Takes MRAM from Lab to GaAs Fabricator
CompoundSemi News Staff

Avago Introduces Wireless Transceiver for Wireless Base Station Market
CompoundSemi News Staff

SRC, NIST, and NRI Enter Partnership for Future CMOS FET Replacement
CompoundSemi News Staff

IBM Claims Single-Chip RF Solution for Mobile Devices
Compoundsemi News Staff

Anadigics Ships Wireless WLAN PA to NXP
Compoundsemi News Staff

Compounds and Silicon: Complementary or Competitive?
Jo Ann McDonald, founding editor

Fairchild Subsidiary, System General Wins Taiwan Invention AwardCompoundsemi News Staff

CyOptics Introduces its 1310nm FP DFB TOSAsCompoundsemi News Staff

Epic Communications Introduces Even Smaller Front-End ModuleCompoundsemi News Staff

TDI Introduces Nonpolar GaN on SapphireCompoundSemi News Staff

Micromem Takes MRAM from Lab to GaAs FabricatorCompoundSemi News Staff

Avago Introduces Wireless Transceiver for Wireless Base Station MarketCompoundSemi News Staff

SRC, NIST, and NRI Enter Partnership for Future CMOS FET ReplacementCompoundSemi News Staff

IBM Claims Single-Chip RF Solution for Mobile DevicesCompoundsemi News Staff

Anadigics Ships Wireless WLAN PA to NXPCompoundsemi News Staff

Compounds and Silicon: Complementary or Competitive?Jo Ann McDonald, founding editor