“We’re not like other semiconductor companies,” says Bob Dobkin, co-founder and CTO of Linear Technology Corp, “for the most part we have a unique set of products because we combine circuit design, silicon design and testing.”
Dobkin is an analogue legend having designed the industry’s first three-terminal adjustable voltage regulator, the first bipolar low-dropout regulator and has designed a couple of dozen circuits which each pulled in over $100 million revenues.
31 years after starting Linear Tech he’s still designing circuits.
“I’ve spent my working life doing what I like,” says Dobkin, “I’m an artist and I paint silicon.”
“Most of the engineers at Linear like working on linear circuits, they like what they’re doing,” adds Dobkin, “making linear circuits is like putting a puzzle together. If I retired from Linear I’d want to go and make circuits.”
“We don’t make a new product unless it’s going to be creative,” says Dobkin, “we don’t want to make a product which is already out there. We’ve got to be significantly better and interesting.”
One thing that makes LTC different is having control over every part of the process.
“We make over 90% of our own wafers and can match the process to the circuit design,” adds Dobkin, “we have hundreds of different processes to make different products. We test and trim all our products.”
That allows LTC to make products which have unique capabilities like huge temperature graduations on different parts of the chip or very low voltage driving very high performance which can’t be matched by competitors.
Dobkin gives a for instance. “We had an 80V regulator. TI decided to copy it. They came out with a data sheet. Later they changed it from 80V to 60V. We interviewed the guy who had got the job of trying to copy the regulator. He said it was more trouble trying o copy the regulator than starting from scratch.”
“We’re very careful to keep our customers happy; we’re very strongly attached o our customers,” says Dobkin, “we want to have a personal relationship with our customers. Customers can call in and get to talk to the design engineers. 30% of our products come from interactions between LTC engineers and customers.”
“Most of the design engineers have been here for 15-20 years,” says Dobkin, “they always get calls asking them to leave Linear – but they don’t. Even the junior engineers don’t want to leave.”
“Our people are very important. They share in the profit right from the start. ‘You make me rich, I’ll make you rich’.
Dobkin doesn’t see much in the way of start-ups and, of them that do exist, “None have their own fab,” he says, “most of the new start-ups are in RF which takes a lot of people. I don’t see a company starting now doing the same things we do.”
“We recruit when our sales allow us to. We’ve had very good luck recruiting out of schools and growing our own engineers,” says Dobkin, “we think it takes years to be a good analogue engineer – 5 to 10 years before they can start to do their own circuits.”
Dobkin still sees some new potential recruits. Asked what he looks for, he replies: “I ask them technical questions. I give them circuits to analyse. Some have only two transistors. I look for two things: one, the answer, the other how they approach the problem. If they don’t know where to start they don’t get hired. If they get excited by the circuit, that’s the kind of guy I want.”
Asked if the move to integrate more would mean the analogue industry would have to use finer process, Dobkin replies: “We don’t go to finer processes to put more on the chip we go to finer processes to get more performance. A mask set for 40nm costs $1m. You have to sell a lot of ICs to get that money back.”
He doesn’t totally subscribe to the view that integrating more is the way ahead for the analogue industry, believing that making better building blocks is an equally good way to go.
“Each of our products will sell between $500,000 to $1m a year,” says Dobkin, “our products last ten years – so we can still get our growth that way. We’re not expecting $30m from one product.”
MEMS is not something LTC will be going for anytime soon. “We’ve looked at MEMS and it seems you can buy an accelerometer for less than 15 cents – that makes our interest go way down,” says Dobkin.
Asking Dobkin who he admires among LTC’s competitors in the analogue industry results in a short list.