A switching power supply is probably the most important part the system, everything else depends on it, but all the strikes were against it. It’s dull and boring and no one wants to do it. It’s considered trivial and so the least competent people were put on it and on top of all of that, there’s nowhere for them to learn anything, nothing’s written, because switching converters and regulators did not come out of academia, they came out of industry, because of the requirements of the space program. Everything was against it. People didn’t know anything, the least experienced people had to do it and there was nowhere they could learn.
Sum: What do you think is the cause of this, and what did they do to solve this problem?
RDM: As a consequence, industry management always fought a rearguard action against putting sufficient attention on the power supply, that is, they merely reacted to problems, always at the last minute, and it was always “fix” something. It’s improving but in the sense of reaction because inevitably the power supply was the cause of a large fraction of system failures and reliability problems, and yet management still did not really recognize that to any significant extent, and they would always want to minimize the cost and they would continuously go back and forth between two approaches.
One of them was “let’s buy the power supply outside from someone who specializes in it and so they do that, but then almost immediately they decide “well, they’re charging too much for this, we can do it cheaper ourselves” so they hire a power supply guy and will do it in-house. Well, again that power supply guy, even if he has some experience, still really isn’t able to run the whole project and compete with the specialist and so the in-house project has lots of money sunk into it and then the company centralizes the power designs from all their different divisions.
They make a Power Supply Central and they say they are going to design the power supplies for all of the product lines under the guidance of this expert they’ve got. Well, that doesn’t work either because they still don’t give it sufficient attention. It’s again “We have to have it so this is the cheapest way to do it” and so the in-house kind of thing usually fails as well.
And then they go back to buying it outside again and that nucleus group they had in the company dissipates and it all goes back into the separate divisions again, who either try to continue to design on their own, or they buy them independently outside. I’ve seen this cycle go maybe several times with major companies. The problem gets solved immediately but no one is learning anything. Management is always trying to put the problem out of sight.
Sum: Perhaps you might like to elaborate a little on how you got yourself into power electronics?
RDM: So this is how I got into it. Everyone gets into power electronics by accident. One kind of “accident” is, as I said, You’re the new guy in the company, you’re not doing anything important yet, so you get to do the power supply, and thatís how you begin learning about it and maybe going into a career.
My “accident” was that I did some consulting for TRW which happened to be in the spacecraft power group, and that’s where I first heard about switching converters, which were pretty new then, in the 1960’s, and totally unknown in academia. Totally. And so that’s where I learned how a switching converter worked and then you put a feedback loop around it to make a regulator and then the problems show up, because no one knew how to figure a transfer function for switching converters, which is the plant in the servomechanism control loop.
Sum: The switching converter is a nonlinear control system, I understand that it created a lot of problems for the people who put power converters on spacecraft. What is your view?
RDM: A switching converter is a DC transformer with an adjustable conversion ratio. That’s its purpose. It’s got to be an ideally lossless DC to DC transformer and it has to have a conversion ratio that is electronically adjustable so you can close the feedback loop around it and make it regulate the output. But the switching converter has a control to output transfer function, a function of frequency, and if you’re going to put a feedback loop around it, you have to satisfy Nyquist. You have to have a conversion ratio that is electronically adjustable so you control the feedback around it, so you can regulate the output.
There’s all this well-established feedback theory in which Nyquist has a stability criterion and there’s a bandwidth involved, there’s frequency response involved, and so standard linear feedback design requires knowing what the transfer function of the plant is. That’s the basis of the whole thing. Well, the plant in this case is a switching converter and no one knew what its transfer function was.
So here you have the situation where people in industry who have to build switching regulators by closing feedback loops around an unknown plant, so that’s when the consultant gets called. It’s oscillating, come and fix it, and so that’s what I did at TRW, starting in 1960, and in other places too.
Sum: Is that the beginning of the Caltech Power Electronics Group?
RDM: The first thing I had to find out was how do you come up with the transfer function for the plant, and industry of course doesn’t have time to do this so that’s why I started a research group at Caltech, although that wasn’t until 1970, actually, but that’s when I brought power electronics into the academic area at Caltech, and the first objective of the research group was to come up with a linearized model of the switching converter.
Sum: Can you say that Caltech is one of the pioneers in power electronics?
RDM: That began a trend, actually Caltech was not the first academic group to do that. Duke University already had done some of it although they tended to specialize in the magnetics part of it. But since then, a number of other places have started power electronics also. VPI (Virginia Polytechnic Institute, commonly known as Virginia Tech) was almost the next one, and now it is beginning to be recognized as an academic subject, but group startup is now strictly sort of rate limited, because to start a research group in a university, you have to have a professor with a Ph.D. in the field, and there aren’t any Ph.D.’s in power electronics except the little trickle that is being turned out by the existing groups. So it’s still strictly rate limited, but it is beginning to take hold.
Sum: But your regular analog circuit course is all you teach in Caltech, there were no power electronics, is that right?
RDM: Anyway, you notice that at Caltech I never taught a power electronics course because I was always doing this one that I’d already discussed. The analog circuits methods, which is a second level course. They have to have one previous, first level, course on active circuits. As long as they’ve had that it doesn’t matter what year they are.
Sum: So, instead of waiting for a Ph.D. in power electronics to come along to start a research group, you went ahead to help the practicing engineers by giving outside courses?
RDM: I began doing power electronics courses outside because there weren’t any, there were no books and all these designers were struggling to learn how to build switching regulators and so I started doing power electronics courses along with my colleague, Slobo Cuk. We were trying to pass the results of our research on to the engineers who could use it, because to my mind, the function of engineering research is not to write papers for other professors, it’s to make methods and techniques available to working design engineers.
Thus, our papers have tended to be partly tutorial even when they were presenting research results. As a consequence, they tend to be fairly long and that means we get hit with extra publication charges for overlength and all that, but anyway, to our mind that is our product and purpose.
We have two kinds of products. We have engineers themselves and we have technical papers that we hope are useful to engineers out there. So power electronics was the field where we did outside courses because that was the only way, there weren’t any books, and so we tried to help working engineers along that way. But because the power electronics was also used in the research, these design-oriented techniques that I’d been coming up with all the time were being developed to be applied to switching converters at TRW and other places, because that was both the research and the fertile ground for the development of these techniques.
It’s not an artificial “here’s a neat method, let’s see what we can use it for,” it was the reverse. We have to understand how this is working and we have to design it optimally, what is the best way to analyze it so we can do it, and that’s where the techniques came from.
Sum: Is it all new material for this power electronics course at this time?
RDM: It happened to be because of the purpose of coming up with better power electronics circuits, but the techniques themselves were applicable generally to analog design, so the course I did at Caltech continued to be the general analog design, and the outside ones were power electronics but they had a good chunk of this general analog stuff at the front end. My part of the outside courses was mostly divided into two pieces and the first piece was these analog techniques, and the second piece was the application of those techniques to switching regulators. Slobo did the magnetics part of it and more of the system design part of it, but I did the basic analog analysis techniques and the basic modelling of the switching regulator viewed as feedback system, and he dealt with the magnetics and the regulator properties in terms of design specifications.
Sum: I understand that you have been doing some of these power electronics courses in Europe too. How did that happen?
RDM: We were doing some courses in the Boston area which was one of our regular stops in September and Nils Backman came from SIFU in Stockholm. SIFU is a technical institute whose function is to provide continuing education courses for industry, and he was one of the instructors there, and so he came to take our courses with the intention of going back home and teaching similar courses himself at his institute.
Instead, he invited me, both of us as a matter of fact, to go over there and give our courses under the sponsorship of SIFU, which we did, and we did that fairly regularly, once or twice a year in fact for several years and I was beginning to feel that the analog kind of preamble was worth developing on its own for a broader audience, and so one of the courses I gave at SIFU was the first version under its present title “Structured Analog Design”. As a matter of fact Nils was the one who suggested that title “Structured Analog Design” and it was given for the first time in Stockholm. But I still did power electronics. I’ve been doing courses in Stockholm six-seven years now. After two or three years, the Structured Analog Design appeared in its own right. It actually was a three-day version of the introductory power electronics one which was two days, so when it stood alone, it went to three days.
Sum: Have you stopped doing power electronics courses now and concentrating on analog design?
RDM: Three years ago is the last time I did a power electronics course and then I formed my own company which is Ardem Associates and it’s simply a dba (“doing business as”), it’s a name, we have no employees, my wife, Val is the one who is the front office.
So three years ago we did public courses sponsored by Ardem Associates and then Part 2 came along 18 months ago or so, that also had its initial tryout in Sweden, but now it’s part of our sequence here and starting last September, Parts 1 and 2 were put on back-to-back.
Part 2 is a review of the techniques of Part 1, because Part 1 goes by mighty fast. It attempts to be more of a workshop so there’s more time to look at examples, maybe even suggest new ones, and apply the methods to more complicated problems.
Sum: Does it mean that your analog design course has to do with making the design task manageable?
RDM: That’s the bottom line of the divide and conquer type of approach, and the hardest part in learning to do that is how to break down the problems into manageable pieces. Anything that you can do to push away that brick wall of algebraic paralysis. That’s the downfall of 90% of design engineers, which is what I’ve identified as the crucial thing.
The course structure right now is last September I did the two courses back-to-back, but that’s awfully exhausting, even though I’ve given up the technical consulting.
K. Kit Sum, Switch Mode Power Conversion, Marcel Dekker, 1984, 336 pgs.