Trying to learn digital control for switching power supplies.

Thread Starter

vichiker

Joined Oct 26, 2012
2
Hello Everybody. My name is Victor Castrellon. I am a power electronics engineer. I am am trying to learn how to design the controller stage of switching converters is the digital domain. I know how to do it using classic control theory. Can anybody gave me a clue about an easy to use kit for the to learn it? Thank you
 

Papabravo

Joined Feb 24, 2006
21,094
Not sure what you are referring to "using classical control theory". That would imply systems with continuous linear time invariant dynamics. An SMPS is definitely not one of those. In the digital domain there is one bit of information that is useful in decision making, and that is if the output voltage is above or below some threshold. There is also one digital output bit, which controls the switch. If the output voltage is above, you open the switch and if it is below you close the switch. You might also have time delay elements with is another non-linear element to consider. So I'm at a loss to understand what you think you are missing.
 

crutschow

Joined Mar 14, 2008
34,201
I assume you are referring to the analog PID feedback loop of a typical PWM switching regulator.
Papabravo appears to be describing a hysteretic (bang-bang) control loop which is a different animal, and doesn't generally require any loop compensation.

This PID loop can be emulated with a digital controller, and that is sometimes done.
An alternate approach is to use a Fuzzy Logic control loop in a microprocessor.
I don't know of any "easy to use" kit to learn these, as they are both reasonably complex.
Suggest you do a google search for these topics.
 

schmitt trigger

Joined Jul 12, 2010
858
What he means is that a conventional feedback loop compensation creates the required poles and zeroes utilizing resistors and capacitors.

But newer systems realize the same function with a DSP core.
Similar to realizing analog vs digital filters.
 

schmitt trigger

Joined Jul 12, 2010
858
BTW, I actually interviewed for a company in Silicon Valley which was developing such solutions.
The interview involved solving some strenuous mathematical functions, which were above my knowledge.
 

Papabravo

Joined Feb 24, 2006
21,094
What he means is that a conventional feedback loop compensation creates the required poles and zeroes utilizing resistors and capacitors.

But newer systems realize the same function with a DSP core.
Similar to realizing analog vs digital filters.
If this were a linear system described by a transfer function, then I would agree with you. Problem is that transfer functions are not useful in systems with nonlinear elements.
 

crutschow

Joined Mar 14, 2008
34,201
Problem is that transfer functions are not useful in systems with nonlinear elements.
Are you saying that linear transfer functions are not useful in PWM switching regulators, because they are.
The PWM block is considered a linear gain block (modulaion control voltage to PWM average output voltage) by assuming that the switching frequency is too high to significantly affect the frequency response of the feedback loop, which is generally well below the switching frequency, due to the LC output filter.
 
Last edited:

Papabravo

Joined Feb 24, 2006
21,094
Are you saying that linear transfer functions are not useful in PWM switching regulators, because they are.
The PWM block is considered a linear gain block (modulaion control voltage to PWM average output voltage) by assuming that the switching frequency is too high to significantly affect the frequency response of the feedback loop, which is generally well below the switching frequency, due to the LC output filter.
Yes, if the system can be linearized around an operating point then you can do that. Is that still true at startup and under all load transient conditions?
 

crutschow

Joined Mar 14, 2008
34,201
Is that still true at startup and under all load transient conditions?
Not if the PWM would try to go past 100% duty-cycle output.
But the linear model allows you to simulate and test the loop stability for normal line and load transients.
That's how I designed a buck regulator for work, and the loop stability of the actual circuit was quite close to the simulated value.
Startup and overload conditions, causing non-linearities such as integrator windup, are common in all PID feedback systems and must, of course, be allowed for.
 

Papabravo

Joined Feb 24, 2006
21,094
Not if the PWM would try to go past 100% duty-cycle output.
But the linear model allows you to simulate and test the loop stability for normal line and load transients.
That's how I designed a buck regulator for work, and the loop stability of the actual circuit was quite close to the simulated value.
Startup and overload conditions, causing non-linearities such as integrator windup, are common in all PID feedback systems and must, of course, be allowed for.
I'm good with those clarifications.
 
Top