Is gain/phase analyzer needed?

[eem2am]

gain phase analyzer

Hello,

I am working for an SMPS company in Chicago.

I am testing the SMPS’s.

Lately I tested a 200W Buck converter with:-

Vin = 120 – 360VDC
V(out) = 50V
I(out) = 0 to 4A
Switching Frequency = 100KHz
Think its current mode but do not have much access to schematic.


The Engineering Manager told me that they do not have a gain/phase analyser, and that they do not use them.

The SMPS Engineer told me that he just does his best to avoid SMPS instability, by putting a very large value of capacitance on the output. –In this way his crossover frequency is so low that he believes he virtually assures himself a good gain and phase margin.

I am told (by the SMPS Engineer) that SMPS’s often go out to the customer, and then end up needing to be returned because of “problems” which they then work on fixing.

The power supplies that I see here often show instability –oscillations of a few KHz on the output, accompanied by a whining of the wound components.

We are going ahead with doing thermal testing (running the SMPS at –10 deg C through to 80 deg C) on the SMPS even though it has oscillations on its output at 25 deg C at certain input line and output load conditions.

Would I be right in saying that the company is seriously handicapping itself by NEVER using gain/phase analysers ?

 

[max0412]

gain phase analysis

I think you are too hung up on compensation and stability etc. It makes for a nice thesis topic , but usually if you follow the manufacturers suggestions for compensation mathematically and tune values by test (dynamic) under your expected load and temperature you shouldn’t have any problems.

90% of SMP’s on the market are directly copied from the application note or reproduced with minor modifications.

I’ve pulled some apart to clean or repair that lack common mode chokes look like they were soldered by twelve year olds, missing expensive components filtering, inrush suppression and TVS etc. This was obvious from looking at the silk screen and noting that the components weren’t there! These are supposed quality PSU’s from some of the biggest names in the business. You buy components so you know why they omit things like Chokes TVS etc they are expensive. I suspect they send samples to regulatory bodies for testing with all proper components and then when it comes time for mass production they omit them. Prime example my 350W computer PSU has no choke I found that out when cleaning it. It solved the mystery as to why a TV operating off the same circuit branch would get fuzzy reception when the PC was on.

In the real world you seldom have the time or accesses to the specialized (expensive) equipment that some PHD geek working for a large semiconductor manufacture has with a limitless budget.

Foremost on your mind should be safety anyways to techs who may work on it to the end user. Try and in vision every possible way a pea brain could either install or abuse your product and then ensure if it’s going to fail it fails safe!

[Edit] You may have noticed that all the people saying you need a analyzer are the ones who make them. You know Ridely makes them? I’m not knocking them if you build PSU’S by the hundreds of thousands yes they are a worth while investment.

My 2 cents for what its worth.

 

[eem2am]

funktion gain phase analyzer

I appreciate you getting back max0412

if you can always get away without using a gain/phase analyzer then its a waste of time i would agree.

-But i've got oscillating , unstable SMPS's in front of me, and the company are saying to put them through environmental tests anyway.

If you don't use a gain/phase analyzer, you end up having to use a too big output cap for example, which brings a whole host of issues and problems of its own -in this case even using a very big output cap hasn't stopped oscillations.

If an SMPS is oscillating, it seems to me that this is high time to get a gain/phase analyzer so it can be seen whats actually going on.

I know what you mean about "PhD thesis" but using a standard piece of test equipment such as a gain/phase analyzer (albeit £12000 worth) isn't going down that road...i don't think?

 

[max0412]

ridley ap300

Assuming the company has been in business for a while. How have they gotten by without a VNA for all that time? I haven’t used one since school.

If you are just supposed to do environmental testing note your results and forward them to the design engineer. Are you supposed to do any diagnostics and perform necessary corrective action? Maybe it’s someone else’s (more experienced) job to review the environmental results and perform corrective action if necessary. Do they not have standard test procedures you fill out? Are you even certain that what you are seeing is instability?

I wouldn’t keep harping to your employer about the impossibility of doing your job without a $12000 + piece of equipment .He might decide to pay for it out of your pay-cheque.

 

[eem2am]

gain phase setup

I'm not sure they'll sack me and buy a gain/phase analyser with the pay they save...........In USA companies can get the wages of workers who have engineering qualifications payed by the government.

Sounds good but i find it means companies tend to employ people and then just leave them stagnating as a "spare pair of hands if ever needed"

Anyway, oscillating output on static load must EITHER be feedbacl loop instability or subharmonic oscillation (subharmonic oscillation due to poor slope compensation). ?

These two are distinguised by the fact that subharmonic oscillation is depicted by a gate drive signal which is long_pulse then short_pulse repeatedly....whereas feedback instability makes the gate drive signal vary less starkly.

Do you confirm this?

 

[max0412]

ridley matlab smps

Anyway, oscillating output on static load must EITHER be feedbacl loop instability or subharmonic oscillation (subharmonic oscillation due to poor slope compensation). ?

But how can you be sure without your Gain -phase analyzer?

What is the person who designed this saying about your conclusions ?

Hysteric controllers under certain conditions have funky gate drive variations similar to what you describe.

I find it hard to believe someone is going to give you an SMPS for environmental testing and your first test setup at 25C shows all the things you describe. So one of two errors are possible your test set-up is incorrect or whoever designed that SMPS should consider a new vocation.

 

[TekUT]

phase gain analyzer

**broken link removed**

May be this can be useful for your testing.

Bye

 

[eem2am]

ridley ap200

well to be honest, if i may be excused to take your joke seriously............

.....i dont see how an SMPS can be properly designed without a gain phase analyser.

Especially since this one is for load_sharing with an identical other SMPS.

And with load sharing , the load share loop must be a decade at least below the main loop of the converter, so how would one know what to set the share loop at if one hadn't analysed the main loop?

and also, you can get your crossover frequency very low by using massive caps on the output...and assure good stability......but then in event of short circuit this cap will heat up badly and be destroyed quicker because of its size....so with a gain/phase analyser, you can more ideally size the output cap.[/quote]

 

[max0412]

gain phase analyzer application

Well if you consider it so critical to successful SMPS design Buy one!

We agree to disagree.

There are thousands of SMPS’s out there that have been working for years without the use of a gain phase analyzer. Compensating SMPS is well documented analyzing waveforms for various load and line transients are well documented to estimate gain/phase. Almost all controller manufacturer offer a tool to facilitate loop compensation, power stage component selection provide detailed test procedure to determine the stability of your PSU etc. Etc. Etc.....

If someone chooses not to use the tools “freely available” whether it be manufacturer provide software or freely available literature well then maybe they should buy a gain phase analyzer.

Not every anomaly you see are think you see can be attributed to inadequate phase margin or insufficient slope comp.

And you still never answered my question “what is the person who designed this SMPS have to say about your conclusions. I’m not saying your conclusion is incorrect without being there it would be difficult to say either way. I am saying one doesn’t need an expensive toy to build stable reliable SMPS’s.

I’m sorry if my joke offended you.

 

[FvM]

www.ridleyengineering.com/

I am saying one doesn’t need an expensive toy to build stable reliable SMPS’s.

That's also my opinion in this regard. I'm generally testing switching supplies with simple pulsed loads, in most cases load resistors switched by a FET switch, sometimes pulsed current sources. These tests have revealed all stability problems up to now. Basically, sufficient phase margin should be achieved by design, but in some cases, you have parts with undocumented properties and in any case, you should finally verify the design.

The said pulsed loads have also proven their worth for productions tests.

 

[eem2am]

excel phase margin current mode converter

FvM:-

In your case, have you read the article
"Step load transient testing"
..(also known as "Transient response and loop gains of power supplies") to be found in the "Articles" section of switchingpowermagazine.com under the heading "Control Design" ?

(you have to register and log in)

-This article categorically and completely demonstrates that step load testing of SMPS's is inadequate.

 

[FvM]

ap300 phase margin

This article categorically and completely demonstrates that step load testing of SMPS's is inadequate.

Some time ago, scientists proved that the earth is a disc...

When I hear words as categorically and completely I'm strongly reminded to this kind of erroneous beliefs. These guys possibly showed something meaningful, it doesn't change the fact, that the said tests fully worked in my applications respectively turned out adequate according to my intentions. Hopefully, the article doesn't completely disprove this possibility.

Do you have a realistic example of a hidden SMPS stability problem, that doesn't reveal in a pulsed load test?

 

[max0412]

ap300 used bode

Well if I made freq/analyzer every other method would be inadequate too!!

It helps with sales.

Maybe you missed this at your switching power magazine site. Look under parts (equipment), lo and behold what do you know they sell freq/analyzer (Ridely).

If you miss it Here is the direct link.


**broken link removed**

 

[eem2am]

max0412,

Ridley does not sell gain/phase analysers.
Ridley runs courses on SMPS etc, because they teach SMPS stability, they recommend the AP range of gain/phase analysers. "AP" is not Ridleyengineering.com.

Its not money that makes Ridley talk like he does -even if he does profit (i dont know)...there is VERY LITTLE profit in selling gain/phase analysers for SMPS use...(more at RF)...in fact , only two companies in the world make gain/phase analysers for SMPS use(ie DC to about 30MHz)

One is via Ridley (actually "AP" eg the AP300)...other is Venable

FvM,

There are some superb examples in that article showing how a smps can be right on the border line of instability but the load step would miss it.

The article converterted me....as it prooved it was so............there was SOLID proof...it is there for you to read

 

[FvM]

There are some superb examples in that article showing how a smps can be right on the border line of instability but the load step would miss it.

I'll possibly look for it, occassionally.

To me, SMPS stability is mainly a practical design topic. I must confess, that I never experienced a case of the said kind through the years. Thus I doubt, the problem may be fabricated artificially, without much practical relevance. As far as I understand, gain/phase analysis is basically a small signal method, correct me if I'm wrong. So I may want to fabricate an contrary example of a dynamic instability, that only reveals with a considerable load step. Actually, I experienced a similar behaviour more than once.

To add another view, time domain behaviour of the process value, e.g. in reaction to a set point or disturbance step is also a common quality criterion in control engineering. I'm not aware of an example where a bad designed controller shows a good step response.

P.S.: "Regarding Ridley does not sell gain/phase analysers"

The Ridley engineering website distributes a switching power magazine article of the master himself, that tells about AP200 analyser:

This instrument is marketed and supported exclusively by Ridley Engineering.

Not a problem basically, but interesting anyway.

P.P.S.: To avaoid a possible misunderstanding. I know, that frequency domain measurements are also valuable in SMPS design and test, I'm using them, too. I believe, that R. Ridley does basically a good job in propagating systematic methods in SMPS design and emphasing the role of control loop analysis.

You can however achieve good (and possibly similar) results with rather different tools.

 

[max0412]

Ridley does not sell gain/phase analysers.

As already pointed out by FVM yes he does.

Ridley runs courses on SMPS etc,...

I know all about Ridely (Dixon ,Brown etc) he is a good EE and a shrewd business man it wasn’t meant as jab.

I’m not saying it’s not a good tool to have as I already said but I say it again if you are say Antec or similar and your primary sales are SMPS yes it would likely be a worth while investment. But I don't think it's necessary nice maybe but not necessary.If on the other hand you are just doing (limited) PSU’s for specific products (you know the load ) I think they are a waste of money. Have you ever noticed on some PSU’s that they have a label stating that it is for use with the product it’s sold with only? Or similar. That is all they will guarantee correct operation with.

Put another way there are a *ell of a lot of other things I’d buy before that.

 

[eem2am]

if a gain/phase analyser is considered not necessary............
..then do you calculate and plot bode plots using say Excel or Matlab?

If you do.....wouldn't it be best to check your results with a "genuine" plot from a gain/phase analyser?

...If you dont make out bode plots....then how do you size your feedback compensation components as well as output caps and any inductors etc?

you can of course use huge output caps and severely lagging feedback so that you almost gaurantee stability....................but then you open up a whole new can of worms in poor transient response as well as big short circuit currents out of the oversized output cap as well as heating in the output cap etc etc and on.........

 

[gigi61]

I have recently got a Venable 350, unfortunately without software.
Could somebody help me, please?