There are usually 1,343 active volcanoes on Earth?

## Richard Phelan

| Published: June 3, 2010 – 5:20 pm

Yesterday, I was sorry to learn of the death of one of my dear college professors, Richard (Dick) Phelan. He was a good man, who lived a good life. His ideas will, one day, change the world. He certainly changed me and for that, I will be forever grateful.

I have mentioned Professor Phelan often to my engineering colleagues, and well, to anyone whoâ€™ll listen. He specialized in a niche area of mechanical or electrical engineering generally called â€ścontrol theory.â€ť

Professor Phelan realized that the way most engineers and scientists go about controlling things like the temperature of an oven or a houseâ€“ a thermostat, or the speed of a carâ€“ cruise control, is not nearly as good as it could be. And the reason is elegant and mathematically simple.

Thanks for reading this far, if you did. I will go on about Dickâ€™s big idea in the next few paragraphs. I hope you get that he had an idea that could change the world and allow us to do so much more with less. Richard Phelan meant a lot to me.

****

Dick Phelan and Pseudo Derivative Feedback (PDF)

If youâ€™re into this kind of thing, you know that most controllers work best with an integrator in the forward loop. For most of us, the classic example is sticking your hand in the shower to get a signal (hot, cold, cool, or warm) that your nerves, brain, and muscles use to set it just right. You adjust a faucet knob and wait. Your brain integrates the signal; you subtract what you feel from a suitable proportion or fraction of what you want, and eventuallyâ€“Â  Ahhâ€¦

An old trick, and a seemingly reasonable one, is to put a differentiator in the feedback loop. That way, the feedback signal gets stronger (uh, bigger), with not just the change, but with the rate of change. All good.

The scheme I described above is Proportional+Integral+ Derivative (PID) control. It is everywhere. I mean everywhere. Even the most sophisticated satellite attitude controller systems are PID. Oh yes, one can analyze the Laplace Transform heck out of those things, what with their spindly struts and drilled-out, mass-saving, trusses, busses, and brackets. But, when it comes down to it, often, very often, our colleagues abandon all the extra analysis and just optimize a few PID variables.

Hereâ€™s what Dick Phelan realized, pointed out, and fixed. With PID, weâ€™re integrating a signal we just differentiated. Any transducer noise or vibration often makes the feedback signal too big, or just wacky.

Instead of PID, try this: Feedback a signal in two places. On a control diagram, itâ€™s two summing junctions. Feed the signal back upstream of the integrator, and then, with a separate proportionality constant, downstream of the integrator. The signal downstream of the integrator acts as though we differentiated. But, there is no, or much less, noise.

Dick called his scheme Pseudo Derivative Feedback, or PDF. Iâ€™ve seen others refer to the same thing with an I in front of the acronym to indicate integration. Thatâ€™s not bad, but in tribute to Professor Phelan, I prefer PDF.

Now, Dick Phelan was a bit of character. Although one of his earlier textbooks gives the reader a solid understanding of Laplace transforms, later in life, he didnâ€™t think much of them. Consider these words from his textbook: Feedback and Control Systems, Cornell University Press, 1977:

â€śThe apparent justifications for past use of Laplace transforms were (a) it was a good way to keep students occupied in an intellectually stimulating way for quite a period of time, and (b) it established a jargon for the fraternity while effectively sophisticatingâ€“ in the true dictionary definition of the wordâ€“ a basically simple field of study into one that was awe-inspiring to the unknowingâ€¦â€ť

Well, control theory can get complicated, but itâ€™s the key to our future. We have to find ways to not just use less, as traditional environmentalists wanted us to do. Instead we have to find ways to be more efficient, much more efficient. We need to find ways to do more with less. PDF will help us engineers and designers do that. Thanks Professor Phelan. Letâ€™s change the world.

### 5 Responses to “Richard Phelan”

1. Nick says:

I had Prof Phelan for his controls class in 1985/86 time frame. He was a very good teacher, and I am sorry to hear of his passing on. He was always very enthusiastic about his controllers.

I never designed controllers, but had to pay attention to how structural effects and sloshing could be affected by control systems… i.e., bandwidth of controller and structural /slosh modes had to have a decade of separation.

I always wondered how the PDF controller would hold up to this frequency separation requirement, but never looked into it in depth. His controller seemed to be very high in bandwidth and was not meant to be used with flexible structures; but I admit I never looked into it much.

Professor Phelan was a truly good person.

2. Jose Amram says:

I never took a course from Dr Phelan but I bought his book, used, at a book store in Berkeley. I got his number, and ever after retiring, he took my phone call, spent a lot of time with me on the phone and sent me a package of notes with the errata and other handouts from the course, did not even ask me for postage! A career change kept me from the controls industry however it was refreshing to study a way to design real versus academic control systems. He will be missed.

3. Brian Skeels says:

I, like you, absolutely revere Prof. Phelan! He took me under his wing my junior year (1974-75) at Cornell, when I was staying after class struggling with the breadboard, trying to understand PID and PDF for the first time. I think initially he was trying to protect his investment of breadboad table top kits from being destroyed by a naive undergraduate who was probably ready to short circuit one of the Op amps inside the box. Anyway, I have used his method to solve many complicated controls problems involving ship stationkeeping and buoyancy, of which many PID purests said it couldn’t be controlled; and when shown PDF, dismissed it as “slight of hand”.

Sadly, its taken 30 years for the world to finally recognize Dick’s remarkable legacy. And funny thing now. Every time I see an Adobe Reader document (.pdf) I think of him… Bill, thanks for writing such a beautiful piece on him.

4. George Ditton says:

Dick Phelan actually was a real control engineer as opposed the the pole/zero academics through out academia. The PDF control algorithm he derived really does work and works very well. I know from personal experience, and it is simple and easily implemented, not withstanding derision from control engineers. I’m reminded of the Control Systems book wherein the author, and academic to be sure, after spending the whole book on the pole/zero paradigm at the end of the book recognizes that the real world is non-linear and that therefore the pole/zero methods don’t apply. Amazing! Phelan saw that the PDF algorithm would handle the non-linearities very well. And he was right!

I have used the PDF algorithm over and over and always with just a considered guess at the gains to begin with and it always just works!
I’ve studied Fuzzy Logic and how it can implement how people control things and I always come back to the PDF as being a lot simpler.

The standard PID control algorithm is a hack. It has always been a hack. Here is what it is. First it is a proportional control with an integral added to compensate for the offset and a derivative added to make it faster. That is the extent of it. But it is still a hack! Hell, if you even go through the math the PDF falls out. Not the PID.

Richard Phelan was lucky. He stopped the academic trail before he got to the PHD level and therefore his mind wasn’t frozen into the academic ideal and therefore he was able to see the real world and how it worked.

He was a great man for that.

5. Gary says:

Bill, thank you for posting this. I also appreciate your associated letter in the ASME Magazine. I appreciate your loss. I recall my collage engineering professors. The wiry ones always seemed to be the best ones.

I have spent many years developing control software for the natural gas transportation industry. This is the first time Iâ€™ve heard of the PDF controller. I find it interesting that this isnâ€™t a standard control block in most PLC/PAC programming languages. Fortunately modern controller languages let you expand their control functions through add-on instructions. I would like more information on the specific mathematic implementation so I can develop my own add-on instruction for the PDF controller which I intend to put into the public domain. Any suggestions on the best source?

Filed under: Consider the Following