When you design a power supply, how deeply do you apply control theory? :x

[D.A.(Tony)Stewart]

Control Theory is useful to correlate step load response , noise and impulse response with Bode Plots. Since Inductors are so non-linear with temperature and load and the gain changes significantly with load, that theoretical simulations and design are only as good as they are validated with the previous tests against required specs. Every design, ought to be checked, but they often are not for all kinds of load response to Bode PLots. But this and Nyquist plots are the best for analyzing control issues, but in the end the choice of the inductor and capacitors is the biggest non-linear characteristic in simulation. Inductors have leakage, non-linear BH saturation that change with voltage, temp and current. Capacitors change with voltage, temperature and have inductance and ESR. A resistive load is not the best real-world test.

Bode plots are the best for optimizing bandwidth to reduce ripple and phase-margin induced overshoot when you have control over your variables. Unless you have done this, it's hard to predict how it will perform with various loads unless they are all tested. Control points need to be considered at the design stage to see if circuit adaptation is required to monitor and inject without influencing the design with injected load. Otherwise time domain tests are done in several dozen categories. If you have not done cross-regulation, inout regulation, load-regulation, temperature regulation, load step response at various step sizes and measure Cpk of the margin against specs, you have no idea how good it really is.

 

[AMSA84]

But those kind of optimization and check response it is not applicable to those ICs market that I was referring back. Right?

I asked that because the control theory that come in books and we learn at university it is not related to those ICs.

 

[BradtheRad]

But grabbing on the first topic, that control theory is only important then for those who design THAT kind of specialized ICs, right?

To produce a successful commercial IC, it requires there be an absolute genius somewhere on the R&D team, who knows more than control theory. He also knows (or can figure out) how to overcome the hurdles that crop up.

These hurdles may or may not be directly related to switching the coil at the proper frequency/duty cycle.

So the team must construct more control circuitry, above and beyond the obvious standard components.

They may not document all they do. At least not openly. There is such a thing as trade secrets.

 

[ZekeR]

But grabbing on the first topic, that control theory is only important then for those who design THAT kind of specialized ICs, right?

Control theory is important for anybody who does real engineering.

 

[AlienCircuits]

Control theory is important for anybody who does real engineering.

This is a subject I'm noticing a lot. So many people seem to design feedback systems without truly understanding feedback systems. If I ask a person why they used a certain part or value, they should be able to tell me why. Its disturbing and this is when I hear people say things like "I just know how circuits work, I think in a different way that doesn't explicitly use math" or "This is real engineering, not that book crap you learn". I seem to catch people who made a design without having an answer for their choices often.

It bothers me so much and inspires me to make sure I do as much engineering as possible in a design. I try not to rely on luck, guess work, and trial-and-error. Sometimes these shortcuts can get you progress past a hairy engineering analysis, and then you can then easily apply engineering again, but I try to make those isolated practices.

 

[D.A.(Tony)Stewart]

wrong.. control theory is used by all expert who want the best performance for a wide variety of loads and understand why there is ringing for step load response.

If you have 1st order control system, you dont need any theory, but I challenge you to find a SMPS that is.
If you follow a great cookbook recipe and use exactly the same parts, layout input and load, then you are pretty safe without understanding how it works. But if something goes wrong.... who will you call?

 

[ZekeR]

I think we can all agree from personal experience that it is possible to survive as a manager without being good at it. Just because somebody carries a manager title and is expected to manage people doesn't mean he or she is actually competent at doing it. Sure, they might do everything a manager does (writing reports, submitting forms, giving performance reviews, scheduling, budgeting), but they aren't actually managing their personnel.

Likewise, incompetent engineers are common. It is often difficult for supervisors to know who is competent and who is not, because there are many aspects of engineering which do not fall under the "real engineering" category, such as administrative tasks, making powerpoints, and politicking. (Not surprisingly, incompetent engineers are usually the best at politicking, as it is their only hope for survival so they'd better be good at it.) Most incompetent engineers look pretty good on paper.

Am I going to say that an engineer who doesn't understand control theory is incompetent? You bet I am. Does this mean that >50% of all engineers are incompetent? Probably. But that doesn't mean they're evil, or that they don't do work. They just happen not to do real engineering, that's all.