In my last column, I showed you how to build an analog computer that simulates the motion of a car over its wheel when stimulated by a "bump" (electronic design, March 31, p. 20). Could these systems be simulated with Spice? Sure, these days they can. Yet analog computers have been around since the 1950s, and Spice has only been around for 25 years, meaning the heyday of analog computing lasted well over 20 years. It's true that special-purpose simulation programs for analog systems were written for digital computers in the late 1960s and early '70s. So, we can't say, "analog computers would still be here if only...." But simulation using digital computers became a lot easier when you could use Spice. You didn't have to write your own Basic program.
Do analog computers always "tell the truth"? Not a whole lot better than Spice. There are 1000 ways that Spice can tell lies, and there are hundreds of ways that an analog computer can lie—but they are different lies.
Did I simulate the system shown in my last column with 48 K2-Ws? No, but I did use one K2-W plus approximately six sections of LMC6082. It worked.
Using Your Analog Computer In fact, there are many real-world systems that can be simulated pretty well with analog computers. My truck speed controller (electronic design, Nov. 6, 2000, p. 146) was an analog computer. It worked surprisingly well and gave me no trouble. My PID controllers for temperature control have worked well. The ball-on-beam balancers (electronic design, Analog Applications Issue, Nov. 20, 1995, p. 50) that I built were analog computers. And my final, best version was a hybrid (analog and digital) computer. Many analog circuits that I design are rather like analog computers. I never assume these circuits are going to work right the first time; I always lay them out to make it easy to tweak, adjust, and adapt.
Maybe I can show some results on my Web site. Go to www.national.com/rap and look for analog computing near the top.
Now that we have an analog computer, how do we use it? The strong point of an electronic analog computer is that it's fairly easy to manipulate a variable to see what the system will do. (That's a major advantage of all analog systems.) These days, you can change a coefficient of a Spice parameter in just a few seconds and see the results in (perhaps) just a few seconds more—best case. Thus, digitally facilitated computation can be reasonably interactive. But it wasn't always that way. We old-timers can remember when a "Spice-deck" was really a huge stack of punched cards, and if you turned in your simulation job, it could take hours to get a result. Asking for three variables might take days, if you had enough priority to get good computer time. Analog computers have always had real-time interaction between the twist of your wrist to change the coefficient, and the response on a CRT. So, you can get a good feel for what is happening in your system.
Stay tuned for Part 4, which will come in one of my May columns. I'll be talking about various methods for adjusting the analog computer's coefficients. You'll also get to see George Philbrick's ternary gain-setting schemes.
Comments invited! rap@galaxy.nsc.com —or: Mail Stop D2597A, National Semiconductor P.O. Box 58090, Santa Clara, CA 95052-8090
If I am not mistaken, this person has almost solved the problem in analog. I know that the first requirement of a competent regulator is that it contains within it a model of the system to be regulated, and that it then has something to do with being that systems complement in some mathematical sense, so that in theory the model building is the hard bit. The original statement of the theory is here: http://preview.tinyurl.com/5oxms4 but I don't have the maths to turn it into an actual method for you, I think you need to couple the real and fake systems in such a way as to get them to cancel out disturbances, or rebuild the model as a sort of anti-model, which I think is the more efficient version.
Josh W -November 13, 2008
I have been reading here and there about the "glorious past" of analog computers, and I would like to show another example of analog computing still widely used by many companies. Imagine the development of a digital system control of an aircraft, an electric power plant, power transmission line or even high power electric motor. When all digital simulation of the controller and the plant are done, there comes the time of actually building and testing the prototype of the controller. Its performance can only be verified in real-time, but testing using the airplane, power plant or other large scale object might be too risky or costly. In this case the analog simulator solving in real-time the differential equations of the plant can be used. The simulator is connected to the controller actuators, and feeds back the controller with continuous-time sensor signals thus fooling it into thinking that this is the real thing. This gives an enormous head start to software/hardware engineers who now can see their real digital algorithm and their real controller in action, and can diagnose any software/hardware glitches at a very early stage of the project without risk of expensive failures. This approach converts the serial hardware-software development process into a parallel one, because the software development can start (and be finished) without the plant being available. I have been using this methodology for years with sizable time and cost savings. This example of analog comuting is as pure as it can be and is still alive and well.
Zbigniew Wolanski -February 12, 2004
Your Comments:
Enter the text from the image below
Please refresh the page if you have trouble reading this text.
Search Electronic Design
Email Newsletter
Sponsored By:
Electronic Design UPDATE provides readers with late-breaking news, opinions from industry experts, and timely technology stories. It's a unique opportunity to get your product message in front of engineers, engineering managers, and corporate managers while they're reading about critical information online.
Reprints
