Controls done the right way by Focus Embedded

Key Points

  • Several highly complex forms of mathematics have evolved to describe feedback control.
  • A high percentage of people who attempt to design control systems have no idea that the math that describes them even exists - and they therefore can't predict whether they'll be stable or have acceptable transient response.
  • The more complex a software controls algorithm is, the less deterministic its time to completion becomes. For this reason, a complex software control loop can have enough jitter in it to cause a system instability somewhere else.
  • Sometimes mild system instability is actually desirable. But you have to know when - and how far you can push into the right half of the s-plane before you have a problem.

Industrial and Controls Design

Control systems design seems to be the one place in the world of new product design and development where more people want to fly by the seats of their pants than any other.  And because control systems theory says (quite clearly) that there are limitless ways in which minor changes in the real world can move poles in the complex plane, it’s no wonder so many people unwittingly create systems with poor transient response at best and unanticipated system instability at worst.
That RS422, RS432, RS485, I2C, Ethernet - TCP/IP, CAN, or Fieldbus data link between your plant and your plant controller has some latency.  Do you know what it is and how that’s introducing a phase delay into everything you eventually do to control the plant?
In layman’s terms, if you buy a bunch of sausages and let them sit for a day under refrigeration before you cook them, the spices mix with the meat, and they’re tastier.  But let them sit for a month and see what happens when you eat them.  And by the way, do you want to eat them when the best anybody can say to you is that they’ve been in the refrigerator for some amount of time between a day and a month.
Proper observation of the rules of linear time invariance (LTI) theory and a knowledge of the advanced mathematical techniques for predicting system stability (Root Locus of Evans, Routh-Hurwitz Matrices, etc.) can make the difference between a system performing optimally and one being overdamped, underdamped, or flat-out unstable.  In extreme cases you may find that instability is actually a desirable thing so long as the controlled system doesn’t spend too much time in an unstable operating regime, since it’s instability that actually goes hand-in-hand with fast transient response.
(The Wright Brothers recognized that their 1903 Flyer was inherently unstable, but they concentrated their efforts on the directional controls problem that would allow the pilot to correct its behavior. Samuel Langley chose to try to make his Aerodrome stable and then focused all his efforts on making a more powerful engine.  And in the end, the Flyer lifted off from the sands of eastern North Carolina and the Aerodrome went straight from its launcher on top of a houseboat to the bottom of the Potomac River.)
Whether your problem is servo control, motor control, some complex chemical process control, or any of a million different challenges in the area of feedback control of a dynamic system; you owe it to yourself to find the people who are going to solve the theoretical half of your problem first and then immerse themselves in the practical problem of reducing it to real-world circuits or software.  Diving in and whipping out circuits without knowing the mathematical underpinnings of what they do is a formula for trouble.
And “think first, then act” is the Focus Embedded way.