What need to consider in Frequency Compensation of a Switching Power Supply ?

[whitemoustache]

Hi

I read some books and papers, including Switching Power Supply Design ( by A.I.Pressman ) and Switching Power Supplies A to Z ( by S.Maniktala ) , but still puzzled.

Generally, those books and papers mention what need to consider in Freq Comp for a PWM system with an EA. However, most other cases are not mentioned, for example, PSM and PFM cases, EA works as a integrator...

R.W.Erickson's Fundamentals of Power Electronics contains more, but it does not mention so much about Freq Stab.

Is there any books discribe such a subject in detail ?

I wander what a system designer do when he is designing such a board. Does he calculate and abey some kind of rules, or just try and summary ?

Besides, I heard that some kind of systems need not Freq Compensation. For example, a PSM system with a Hys Comparator needn't that, and someone else says that an LED driver need not Freq Comp because it applies constant avg current ( I totally cannot understand this one ) .

Similarly, some chip datasheets declare that their system need not Freq Compensation Component at all ( for instance, iW3620 ). So what's the difference between them and other chips ?

Best Wishes and Thx:-D

 

[crutschow]

This is a useful guide for loop design of standard type switching power supplies.

For other types of switching regulators, the amount of type of compensation depends upon the design of the device.

The iW3620 is a current controller and does not have an inductor and capacitor in the output that requires compensation for the resonance of those components as a typical switch regulator does.. It apparently has enough built-in compensation to keep the loop stable for that particular configuration.

The spec sheet for the device you want to use will usually tell you want you need to know about what compensation the device may need.

In general if there is an inductor and capacitor in the output of the switching regulator and there is feedback from the output to the input to regulate the voltage or current, then you will need some form of compensation to achieve gain and phase margin to keep the loop stable. Without that inductor and capacitor, such as some LED current regulators, then little on no added external compensation may be needed.

 

[whitemoustache]

Hi!
Does "an inductor and capacitor" mean the components in the red circles ? Or do you mean something else ? ...
Besides, is every PSR system, which contains no feedback from the secondary side, needs not freq compensation ?
I am a fresh guy and I really don't know why ~~~ Is it just a knowledge from experience, or a mature theory obtained from deduction in books and papers ?
Thank you very much

 

[whitemoustache]

View attachment 74406
Hi!
Does "an inductor and capacitor" mean the components in the red circles ? Or do you mean something else ? ...
Besides, is every PSR system, which contains no feedback from the secondary side, needs not freq compensation ?
I am a fresh guy and I really don't know why ~~~ Is it just a knowledge from experience, or a mature theory obtained from deduction in books and papers ?
Thank you very much

 

[mtwieg]

In general you need to know the open loop transfer function of the converter under various conditions (maximum/minimum line/load). The transfer functions for common topologies and control schemes can be found easily by searching through reference books and application notes. Some more exotic converter control schemes (transition mode converters, variable frequency converters, etc) can be harder to figure out, but all of the models can be figured out with state space averaging, if you know your math well enough. Once you know your open loop transfer function, you select a desired crossover frequency and phase margin, and design the appropriate compensation.

Other feedback methods like hysteresis (or bang bang controllers) don't really need frequency compensation, and are almost always inherently stable, but will usually suffer poorer overall regulation, greater ripple, slower response times, etc.

 

[whitemoustache]

Thank you for helping!
State space averaging is not so easy, and no paper of Critical Continuous Model is found. Maybe I can calculate it and send a papaer.
Additionally , state space avg seems more likely to give a close loop function...
For bang-bang case, are hysteresis and hystereic are both correct ? I just see that in a pdf the day before yesterday, but have not seen such a term before. Does that means a PSM system applying a hys Comparator ?
If not , could you please introduce that for me ?
Thx and best wishes!
:-D

---------- Post added at 13:25 ---------- Previous post was at 13:25 ----------

Thank you for helping!
State space averaging is not so easy, and no paper of Critical Continuous Model is found. Maybe I can calculate it and send a papaer.
Additionally , state space avg seems more likely to give a close loop function...
For bang-bang case, are hysteresis and hystereic are both correct ? I just see that in a pdf the day before yesterday, but have not seen such a term before. Does that means a PSM system applying a hys Comparator ?
If not , could you please introduce that for me ?
Thx and best wishes!
:-D

 

[crutschow]

Does "an inductor and capacitor" mean the components in the red circles ? Or do you mean something else ? ...
Besides, is every PSR system, which contains no feedback from the secondary side, needs not freq compensation ? ....................

The transformer you are highlighting contains no inductance that is a significant factor in the loop stability. I am referring to the output inductor and filter capacitor in a typical switching regulator.

Not having feedback from the secondary side does not necessarily mean you don't need compensation. You can have feedback from the primary side which may require loop compensation, depending upon the type of converter.

 

[whitemoustache]

I am referring to the output inductor and filter capacitor in a typical switching regulator.

Do you mean that:
If a system needs compensation, then there must be a comp ind in the output. Then if there is no comp ind there, we can speculate that it needs not compensation ?

Do I misunderstand ?
If so, what do you exactly mean by that? :shock:
If not, it is still not clear that what kind of systems need compensation....and it's not clear that if a system needs not compensation, that's because of its topology or the chip ?

For instance, iw3620 and mp4021 neither have a comp ind in the secondary side, does that means that topology needs not compensation ? ... And why not ?8-O

Thx very much! :razz: