SMPS stability

[Fluorescence]

hello,

With these fora having recently debunked myths concerning

1. Power factor correction and efficiency
2. Right half plane zero of boost converter

...i wonder if the situation concerning ensuring the stability of a switcher power supply using time domain (step response) methods only is valid ?

That is, as opposed to using Bode Plots etc (frequency domain) to help ensure switcher stability.

After all, Bodes are for linear systems and switchers often have non-linear operation.

Here is an application note which gives support for the time-domain method....(please see pages 48, 49)

http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1031,C1061,P1266,D4176

[Its application note AN19 from Linear.com]

 

[Fluorescence]

Sorry i forgot to add the background to this post...

A few years ago i was at an interview at a branch of a large and well known switcher company and the Engineering Manager (who i notice often contributes technical articles to power electronic periodicals) said that no matter how many calculations they did to ensure SMPS stability, the SMPS often went unstable.

He said it was exasperating because the way to correct it, was often to randomly dab in some capacitance at some place in the circuit and it this would stabilise it.

-Made me think that the Bode and Nichols charts etc are of limited use ?

 

[xnuke]

I'll be the first to say that I don't have a thorough knowledge of SMPS, as I work in industrial process control, but I did my MS thesis on control of a specific architecture of SMPS.

I don't see a "method" in the app note so much as I see iterative trial and error, and no satisfactory measure of a margin to instability is proposed. It sounds to me like a lot of time and effort would go into trying to demonstrate stability this way, and I don't believe it would be worth it or a that guarantee of stability can be made by their approach. I know of no area in control theory where stability can be demonstrated by examining time domain response. Robust control design methods should be used to account for circuit model uncertainty. Of course, their is always a performance vs. stability trade-off.

I think the small-signal linear frequency domain methods have advantages if the operating point of the SMPS isn't varying widely, despite the switching behavior. Small-signal operation does result in satisfactory linear models of circuit behavior in many cases. (I used state-space averaging to describe the small-signal behavior and linearize the SMPS in my thesis.) It's the large-signal operation where the nonlinearities make the control challenging. Of course, everyone wants their power supply to be able to deliver 0-100% load current at the desired voltage at all times, which makes things difficult.

xnuke
"Live and act within the limit of your knowledge and keep expanding it to the limit of your life." Ayn Rand, Atlas Shrugged.
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Fluorescence]

Thankyou xnuke..

"I see iterative trial and error, and no satisfactory measure of a margin to instability is proposed."

..the supporters of this time domain analysis would say that they would get sufficient phase margin by changing compensation components until an optimised transient response is obtained.....then they double the compensation cap value for example, so that they are sufficiently "backed off" from the critical region where instability and oscillation may start.

In fact, experienced engineers can get a good idea of phase margin by observing the transient response.

Its rare to find control experts in the switcher companies that i've been to. At one company, they wanted to build a huge power supply involving a lot of SMPS's acting together to give the (huge) output power.

The Director refused to let the existing SMPS engineers start the job, as he said a control expert would have to be employed to model it first, due to its size and complexity. In the year that i was at this company, no such control engineers could be found to be employed, and the project stagnated.

Given that there seems to be such a shortage of control expertise in SMPS, i wonder if most SMPS companies end up resorting to time domain methods, after initially doing frequency domain techniques to start it off?