|
"Dobby" as many colleagues call him, and I got together at the Applied Power Electronics Conference (APEC) at the Disneyland Hotel in Anaheim, CA. We met in the usual cell-phone way -- we agreed to meet in a general spot (near the APEC registration booths) at an appointed time. Then one of us would call the other giving a more specific location while waving the free hand to draw attention. It's GPS on the cheap. I'm glad we did it this way instead of trying to find each other by covertly reading the name badges on passersby. I never would have seen a badge with Bob Dobkin on it because he was wearing one with "Mickey Mouse" (it was after all the Disneyland Hotel). Such was my first impression of the co-founder and CTO of Linear Technology - a real laid back kind of guy.
P: Efficiency was the buzz word at this conference -- efficiency of computers/desktops for the server market and the telecom market. What's your take on this topic?
B: You hit a hot button of mine; because everybody talks about efficiency but it's the wrong thing to talk about. They need to talk about power loss. They're related but you focus on the wrong thing when you talk about efficiency. Efficiency changes with the voltage. So, it depends on what the output voltage is. Power loss is pretty much independent of that. For example, let's say I have a supply with 1V at 10A and it's 80% efficient. If I change that output voltage to 2V, it's now 90% efficient. But it dissipates the same amount of power. Power loss is the same at 2W. If you look at what's happening in a switching regulator, there's quiescent current, that's input supply related, there's switching losses, that's frequency related and input supply related, and there's IR losses, which is the resistance losses, that is output current related. There's no inherent relationship to output voltage. If you look at a bunch of switches on the different company data sheets, and you look at the power loss at the different output voltages, they are all the same. From the engineer's standpoint when you are designing these devices, power loss tells you how big your power is and what you think it's going to be. If you're running on a battery system it tells you how much wasted power you have. Efficiency is a good number but power loss is a more important number. They may be talking about efficiency but what they are really saying is that they have low power loss. For example, you are dealing with efficiency versus power loss when you have a badly operating design. When you are on standby power you still have memory to keep alive -- you don't really care whether it's efficient or non-efficient, you just want to know what the minimum power loss is at that point, to keep the memory alive. So power loss tells you how long your battery life will be.
The problem is that the efficiency is a function of output voltage so if the manufacturer wants to make a switching regulator good, he specs it at a higher output voltage and it's now more efficient. And if a competitor says the output voltage is adjustable between 1 and 4V he will spec it at 3.3 because he will get a really good efficiency number there. Someone that specs it at 1.2V will get a worse efficiency number.
It's not wrong to talk about efficiency but it doesn't give you the whole picture. So, if we get a call from a customer and he says that they are looking at our part but a competitor's part is 5% more efficient, you need to know that the competitor is specifying the output at 3.3V and we are specifying it at 1.2V.
Bob Dobkin, Founder and CTO of Linear Technology doing what he likes best
P Let's talk some about LTC as a company, the direction of the company and even about some of the recent changes at the top.
B As I'm sure you noticed -- Dave Bell (former President) left the company.
I am still the CTO but I'm now taking a more active role helping to fill in along with CEO Lothar Maier. Linear Technology is and always has been a high performance company and when things turn into low performance and commodity, then we are getting out of our business.
P Is there a connection between Dave leaving and this change?
B I think Dave's perception of where he wanted the company to go and where it was going wasn't quite the same and he wanted to go off and do some other things. He left on very friendly terms.
P So where do you see LTC going, what markets?
B We will continue to be in the markets that we are in now. The market is still huge. The IC market is what --100 - 200 billion dollars? The part we're in is about $30 B and LTC's part is about $1B. So we have plenty of room to grow. One of the things we like to do is take product that is low cost commodity and figure out how to make it so much better that companies will want to buy it and pay more for it because it is no longer a commodity item. So LTC will go in and look at a customer's system and figure out an efficient solution. It's good for the silicon and it's good for the customer. And if it's really good, then others will copy us in a few years. We never get design wins when there are super solutions around. We try to give the customer value so they get their money's worth.
P I heard some rumblings that LTC may be considering going private?
B No, that's not true. It may be something coming from Wall Street. Anyhow, I haven't heard it. So if I haven't heard it then if anyone's talking about it then they're not getting it from the right people.
P Why would LTC want to go private?
B Well, because we are very profitable and we have a lot of cash. So if we were to go private we could service our debt from our income.
P So, what you're saying is that it's entirely possible for the company to go private.
B Yes, but our line of credit is at 5% so I don't think we want to use $500 million in profit. We couldn't service all the debt but we could do part of it.
P What's your take on what power design engineers should focus on?
B Designers should focus on power loss, manufacturability, and depending on the system size, they may want to include monitoring which is on the PMBus now. I think the digital power view is being over-hyped
I remember when we started analog ASICs vs digital ASICs. LSI and VLSI were going great guns with digital ASICs. There's a place for digital power but by far the most important portion of it is the telemetry. I don't think a customer cares if the system is analog or digital - as long as it works. In theory, digital can help you do some sophisticated things -- you can change the frequency response as a function of load steps and change it back again. But is it going to be worth it? And what kind of problems will it generate? If you have a well controlled system that doesn't oscillate and then you start playing with it and you change the frequency response of the system, are you going to get certain kinds of loads that will make it oscillate for a few cycles, or settle down? You don't know. But if the digital system works and it's not too expensive then the user doesn't care.
P What's LTC's take on supporting digital power designs in the future?
B We will certainly support the telemetry sections and that will probably be supported within switching regulators. But that's a development project and we're looking at all the options -- so whatever makes sense. So if it costs more to do the same functions digitally than in analog then we won't do it.
P Another aspect of power efficiency is related to thermal issues. What kind of designs is LTC looking at to help engineers solve their thermal challenges? I'm thinking about servers and telecom and wondering if that's an area that you are looking at?
B Yes, it's an area we're looking at. Most of the switching supplies are pretty efficient. We have some modules that are very small and very efficient and we can optimize the FETs, the switching regulators, and control the frequency and the inductor to get very good efficiency for its size. The resistance of the FETs is just a few milli-ohms and the whole module running at high output current is only dissipating a couple Watts. I think that there are some architectures that need better efficiency, especially at low output voltages. We are looking at those for modules. We're looking at getting the frequency up without dissipating a lot more power. And that means having good FETs, a smaller inductor without bigger wire sizes so your I2R losses aren't large.
P Do you also have to look at integrating devices?
B For what we're doing we don't have to integrate the power FETs on the same chip as the controller. But you do have to have it within millimeters of it. So you can put it in a module package. You can put it in the same package and bond them together and that's close enough. Because there's a place you run into that problem with the inductor and the lead inductance in and out.
P Looking way out into the future - what kind of disruptive technology do you see that will help us solve these problems? The idea that generated this question had to do with the telecom and server industry and they need a tremendous amount of power and now we have to deal with cooling systems which are becoming more expensive than the technology in the room. It seems that something else must be done to solve this conundrum.
B If I knew the disruptive technology, we would be making it and I wouldn't be telling you. But I don't see it. There are some people that will sit here and tell you that we need to get switches really small but that increased losses will go to resonance. But there is a down side because you get large circulating currents from which you get losses. The designs for which this is hardest to do this is down at low output voltages. There are technologies that are good for it and LTC doesn't necessarily make them all as chips because some of them require a lot of magnetics and an integrated solution with magnetics, chips and FETs we could do as a module but selling it as a chip it gets too hard to teach the users how to use it. LTC was one of the first companies doing switching regulators and it was a huge uphill climb to teach users of boards how to put a switching regulator on the board. They had never done it and an inductor was something they had never seen (unless they opened a radio). In analog you don't use inductors, you only use switching regulators, and we had to teach them. By doing that we learned some things too. If you make it too complicated then your customers don't want to look at it. So some of the solutions will always be modules and you integrate the portion that makes sense to integrate. And, you couple it through the inductors to end up with a solution that helps solve the problem. Linear Technology is mostly an IC company and we now have some modules -- so if it makes sense to move in that direction then we will. There are still a lot of places that need linear regulator solutions. We now have some for 1.2Vin, 1Vout that is all bipolar (because you can't get FETs to turn on at that low a voltage). Additionally, we have LDOs up to several amps. Those products are some of the things that customers look to LTC for because they are very rugged, and we have some new ones coming out for transients in the automotive area.
P So, always having something in the pipeline is what makes your company vibrant.
B Not only having something in the pipeline but also having something in the pipeline that your customers want. But there's always a limitation. You can always put as much power in a linear regulator as you want to dissipate, and if you have a board mounted design then you just a couple Watts to dissipate. So, if you have a switching pre-regulator at 2.5V you can get reasonable amounts of power at 1.8V and still not go over the power limitation for dissipating it on the board. And if you're at 1.8V and want to go to 1.5V then you don't want to do it with a switcher design.
P So are you looking at retiring any time soon?
B This is my life and I like what I'm doing. It's a creative outlet. I look at designing like learning a foreign language - you have this thing that you want to translate and you know the basics, and you look up the other pieces and you start writing some equations to figure out what the things are doing. But all that knowledge has to be part of you before you write in that language. And designing for ICs is like writing poetry -- you have to know all the bits and pieces to make it work. It has to be part of you. It is fun but it is hard work too. And (laughing) when you design an IC you have hundreds of thousands of people who know exactly who to call (remember the datasheet?) if something goes wrong with that IC.
Indeed, it is like writing poetry.
About the author
Robert Dobkin
Date of birth, place of birth 10/6/43, Philadelphia, PA
Education:
Attended Massachusetts Institute of Technology
Career:
National Semiconductor 1970 -1981
Title: Director of Advanced Circuit Development
Linear Technology Corp. 1981- present
Title: Founder/Chief Technical Officer - responsible for new product development
Other activities:
* Pilot
* Holds over 90 patents for linear ICs,
* Authored more than 50 articles and papers
* Man of the Year 1976, Electronics Magazine
* Electronic Design, Hall of Fame
* EDN Innovator of the Year, 1994
|
