A: Most universities switched from an analogue discipline to digital, because digital was (considered) magic. At the same time, you had Mentor Graphics, Synopysis and Cadence (which all make semiconductor design tools) focus completely on digital design. Now what happened is that analogue has become the emerging industry and a few universities, like Georgia Tech, Washington, Stanford and Berkeley, recognise the value of analogue...and we get most of their PhDs. Will the black art aspect change as analogue grows in importance?
The analogue tool industry has to catch up. Even when they do, however, analogue is still a real tough science. One of our guys, Larry Lewicki, says "Brian, I could design an (analogue to digital converter) at 30MHz. If you asked me to take it to 60MHz, it would take me one week. If you asked anyone else and gave them the documentation it would take them at least a year." That's the difference. You just know how to do it. What's driving consumption?
The only things on the face of the planet that use zeros and ones are microprocessors and digital signal processors (which manage digital signals in mobile phones). Even a digital satellite signal rides on an analogue carrier. You can't send a zero or a one. It doesn't have any meaning out there. It's fine to do zeros and ones for spreadsheets, and that's why the PC uses the least amount of analogue. But we're not doing spreadsheets anymore. We're doing digital photography. We're downloading images and graphics from the Internet, and we're doing more and more stuff wirelessly. All of that is analogue. What are some of the coming analogue ideas for wireless?
At Berkeley they are looking at a 10 gigabit-per-second radio with an onboard variable length inductor that can change its personality. You walk into the red carpet room at the airport and your PDA or your personal computer starts sniffing the air to see if there is a 2G, or a 2.5G, or 802.11b network. It covers the spectrum and it picks the cheapest path to the IP backbone and configures itself to be that radio. Let's say you're doing something that's voice intensive. It will still keep sniffing to see if another protocol gets introduced that is even cheaper. What other projects is National working on?
We have this vision that your smart card will have biometric information. It will have your bank account and your passport and your medical data -- but only your thumbprint can activate it. So you shove it into a slot and it says, "Yeah, this guy's passport is real." Or "yeah, this guy's got the bank account to back this up." Or maybe you just load it up with $50,000 and burn it off over time. But if you ever lose it, (the card) doesn't have your thumbprint so it's useless. We're working with a particular technology where you don't leave your thumbprint; you rub it. We've been working on that for around six months. Chip prices aren't extremely high for most analogue products, though.
The (average selling prices) are low, but the margins are high. We're talking about chips where you can get 9,000 die per wafer. If you get 9,000 good die, 50 cents looks like a pretty good ASP. We have one chip were you can get 39,000 die per wafer. We had to invent our own saw to scribe it. Once you've got that kind of technology and get pretty good yields, every fab can produce perfect wafers. We created this technology called chip scale packaging, where you epoxy the whole wafer, then saw it and throw away any chips that don't yield after packaging because you are so confident in your (quality). Like nearly every other manufacturer in the world, National is rapidly moving into China. Why now?
One of the reasons we decided to build a test and assembly plant in China is that we ran out of capacity in our test and assembly plant in Malacca, Malaysia. Forty-eight percent of our business came out of Asia this past quarter. It used to be 10 percent. Over half of Taiwanese manufacturing has moved to China. I have a prediction that reunification will happen sooner than anyone expects it to and it will be driven by Taiwan. There are a million Taiwanese expatriates in China. Look at Foxconn (a Taiwanese-based contract manufacturer). They have 70,000 employees. They are headquartered in Taipei. They have only 150 employees in Taipei. National has tried to promote Internet appliances for years, without much success. Why don't these sell well?
The problem is the marketing. For 24 years, we've been told by Intel that the only thing that matters is megahertz. And so you have salesmen working at Fry's and Good Guys who can only talk about how many gigahertz a box has. So everybody said, "Why should I pay more for something that is the subset of the PC?" But is there really a market for something like that?
We have seven PCs per 10 houses in the US. That doesn't mean 70 percent of homes have a PC. We're probably still at the point where 45 percent of the families in the US don't have a PC. I've given my parents at least four ThinkPads and one Dell. I'm trying to get them up and running so I can send them emails and pictures and stuff. My mother's last feedback was that she was depressed and despondent because she feels like she's stupid. She cannot look at attachments I sent her. She can't get stuff on the Internet. With the bill for semiconductor fabrication facilities to process chips from wafers with 300-millimetre diameters running close to $3bn, how does a company like National stay competitive?
It's becoming almost impossible for the smaller semiconductor manufacturers, with smaller being anything less than an Intel, or a Texas Instruments or an NEC to afford these 300mm fabs on your own nickel. The good news is that not all of us compete. For example, National and LSI don't compete in anything, so what you are going to see are more and more partnerships on these 300mm fabs. We have a partnership with TSMC (Taiwan Semiconductor Manufacturing Co.) They do the advanced process technology and we use their fabs for the prototyping and the early volume. When did National begin to outsource manufacturing to foundries?
On 0.5 micron (in the mid-1990s), I had an epiphany. My process technology guy at the time told me: "You know, from this point on, semiconductor guys aren't going to really add that much value to the process (manufacturing) technology roadmap. It's the equipment suppliers that determine the process technology roadmap." If you think about it, we all buy commercially available off the shelf sputterers, steppers, etchers and it is really the equipment suppliers that determine what that process looks like. In reality, the drivers of process technology roadmaps are the Applied Materialses, the Nikons and Canons. When it came to copper, only IBM had to invent their copper-sputtering machine. The rest of us could buy it off the shelf. The interesting thing is that we all have to recognize that the equipment suppliers play a more and more important role in the process technology roadmap. Foundries, though, often suffer from a reputation for not being on the cutting edge. Is that the case?
Since TSMC is now one of the largest consumers of capital equipment in the semiconductor industry, the equipment suppliers are going to listen to them more and more. It is a myth that TSMC is behind the rest of the industry. They are as state of the art as anyone.
See Chips Central for the latest headlines on processors and semiconductors.
To find out more about the computers and hardware that these chips are being used in, see ZDNet UK's Hardware News Section.
Let the Chips Central editor know what you think by email. And sign up for the weekly Chips Central newsletter.
