1 watt DC-DC power supply module reliability issues

[jeffl_2]

This is my first post here but I've been involved in electronic circuit design for many decades. As part of a current design effort I take an unconditioned AC voltage, put it through a bridge and large capacitor to get maybe 9 volts, then I pass that through a linear regulator to get 5 volts, then I send that to a (typically Chinese-made) 1 watt isolated dual +/-12 volt DC-DC power converter module. (Please note I'm NOT looking for circuit design help so I'm not posting schematics or anything else that isn't relevant.) The trouble is on the (typically) second or third power cycle the device outputs will be zero and the module will be burned out, EVEN IF I HAVEN'T EVEN HOOKED UP THE OUTPUTS YET, the linear regulator (typically LD1117) is never harmed, nor is a second module across the SAME input that's generating isolated +5 for another purpose! The vendors I've identified for this type of product are aimtec, CUI, Mornsun, gaptec, Mean Well, XP and Cincon, so far I've blown out two from each of the first two vendors (I don't even have parts from all these brands, still waiting on my distributor for delivery for some of them). A representative part # (aimtec) would be AM1DS-0512DJZ. So far when I talk to any of their "support people" all I get is "what is your company name and website?" as if big customers are the only ones "entitled to" application and reliability support. It seems to me all these parts basically come out of some windowless factory in China and NONE of these vendors know anything about the product design or application! The only "clue" I have is SOME vendors say the modules OUGHT to be "protected" from "undervoltage conditions" but there isn't an actual spec at which they admit the module would self-destruct! Is there some "minimum level of product" below which it just isn't advisable nowadays to expect any level of reliability or support AT ALL? (These aren't for DOD application so I don't have an infinite budget, it's a consumer product with an extended warranty so I can't afford gold-plated stuff but it DOES have to be generally reliable.) When did the market change so that power supply modules can be so goll-darned unreliable but somehow they still meet the requirements of the market? Has anyone had the same level of frustration with this type of product, and how did you resolve it?

 

[MacGyverS2000]

I can't tell what you're on about... are you asking for help with the issue (despite wanting to post a schematic), or are you merely letting off steam? PSUs aren't child's play to design, as some think, and surge currents can play a big role in reliability. 1W is tiny in the grand scheme of things, which also means it's likely a cheap Chinese commodity design with no real protection. Help us help you, if that's what you're after, by posting a schematic... "representative" product numbers are near useless as every product is designed differently.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[jeffl_2]

You need to have a little patience with me here, I'm not a power expert and I still don't plan on becoming one, and even if I made myself into one I have ZERO interest in reverse-engineering some piece of Chinese crap and trying to make it work! Before all this "price first, last and only" attitude came in with the mass flow of modules from the Chinese mainland, there USED to be a little bit of an "ethic" that if you bought something like a power converter module from a reputable website (in my case Digikey, although obviously buying from them isn't a guarantee of anything) you could AT LEAST stick one on your prototype "breadboard", and at least BEFORE hooking up the load you ought to be reasonably certain you could SEE an in-specification voltage on the output terminals, and even cycle the power 2 or 3 times without causing the silly thing to blow up in your face. That is NOT the case today, and I'm guessing that my primary "defense" is to identify all the power module vendors who do assembly on the Chines mainland and just start by disqualifying ALL of them. That's because I've basically already written off the bottom six vendors by price since either their parts have already self-destructed OR they have demonstrated that they have no "credible" product support on either their website or by email or phone. You see I'm having to "guess" that what's causing the failure is an "undervoltage condition" because clearly it can't be OVERvoltage since the linear regulator precludes that. Exactly NONE of these modules have an actual specification about this on their data sheets, now a couple of them (gaptec being one) says on one of their app notes that you need to "protect against" undervoltage, but nowhere do they actually define what that means so I can design a way to protect against it, and assure that the parts are protected! Yes I'm a circuit developer but in another context I'm more or less an end-user since these are all sealed modules, and I feel I have a right to expect that the applications engineering was already done SOMEWHERE by SOMEBODY. What do they expect me to do to get something reliable, purchase a couple thousand of these parts and get my own enviro chamber, blow all of them up and establish how they all failed? All these outfits have already decided they aren't going to share with me the schematic for the internals of the module, EVEN IF I WANTED TO how am I going to do any failure analysis without it? This has never happened in my career (although that's varied and "in theory" I'm retired) that such a high percentage of different vendors' parts fail FOR NO KNOWN DOCUMENTABLE REASON, maybe this is the "brave new world" but I really don't want any part of it!! And if that's "just a rant" I'm sorry!

 

[itsmoked]

Sorry you're having these issues.. I use many of those modules in various products and while maybe one out of a couple 100 crap out I don't ever have your level of failures.

The most common failure though is for the oscillator to not start up. I think a soft supply to them causes this. I also think that that can cause burnout because unrestricted current sits there running thru only part of the switcher topography.

I suspect the problem in your case is your linear regulator. It's probably causing a soft start-up that's screwing with the DC-DCs. I bet if you switched on/off your DC-DCs after the linear they'd not fry IF there's a large bulk cap after the linear.

Keep in mind all DC-DCs have a huge startup demand. If that hammers your linear then the linear limits the inrush for a while. While it struggles to recover bad things will be happening to the DC-DCs.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[jeffl_2]

So these converters don't have an enable input, you're saying I ought to consider a small P-chan MOSFET to snap it on (maybe driven by zener or sup chip)? Why wouldn't the mfrs mention that somewhere?? (why wouldn't they design that in?) So long as I keep RDSon low should be OK?

I have had a "largish" (47uF) cap at the linear reg output trying to kill noise and keep supply Z low. Maybe that's slowing startup.

What's "startup current" for how long and best way to meet it, I assume it can be ahead of enable FET, maybe I leave the cap in there?

Primary reason for the linear reg is the tight published spec (4.5-5.5v) for input. In the absence of other info I have to try and meet the numbers that I can see, if I don't need it it would be great to be rid of.

 

[itsmoked]

I'd put a .1Ω resistor in front of it with a scope to see what the current demand is and another scope view of the voltage out of the linear. See if you see a problem and if you do you'll likely gain the insight you need to remedy it.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[jeffl_2]

Uh, I think you're implying I could be better off explicitly selecting a regulator which deliberately OMITS current limit and all these other limits/controls (thermal etc.)? So it doesn't restrict the current that the converter needs at startup? And that's what I had to learn the hard way?

Older design, simple and cheap? Can do...

 

[itsmoked]

No, not at all. Pick the one you want with good availability and watch what happens on the input. If the input is dicey make mods to get the input to play as well as needed. In other words, don't try things based on speculation! Investigate the input situation and mitigate any pathological events being seen. Being setup for study you are automatically setup to see how mitigation efforts are working.


Changing subjects, most of those small switchers have minimum load requirements that need to be provided. Most all designs with them I've done have needed passive loading added to meet the spec. Particularly watch the negative rail because with a +/-12V unit you frequently load the + side easily but the negative side often just negatively biases some opamp or something else light. Add resistors where needed to meet the minimums.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[jeffl_2]

Yes I hear what you're saying...my scope capabilities aren't really where I need them to be right now. Hope to upgrade soon. I do have a simple design to replace that regulator with parts that I have in stock that would easily accommodate a stiff startup current surge if I have to. And I'm not "distributing" that output anywhere, this happens to be an all-analog design, so I'm not all that concerned that it has to be all that great as a logic supply or anything. But it isn't that the regulator is "just on the edge of" providing the surge, these low-cost dual output devices are VERT HARD to get reliably started, the singles are no trouble at all!

So I spent a little while researching part #'s and data sheets. It turns out several of the vendors that I guess I wasn't looking at very hard because I didn't know them too well (XP Power, Delta, Recom, Murata, Traco) are actually quoting 3 year warranties (some of them anyway) and MIL-HDBK-217 reliability hours. And parts from these vendors are available at prices that are quite competitive with some of the ones that I had been considering (well if you can get them at all, this is all quite relative, even power converter modules are in short supply these days). And it's weird, you've got some distributors supplying low-quality parts at high markups out of one group of vendors, and a different group of distributors supplying much higher reliability at lower markups from a different group for the customers actually building something for sale. Or that's how it seems anyway! Now this doesn't mean the reliable parts don't have a turn-on surge, but it does seem like a better field to go hunting in, and they SHOULDN'T self-destruct quite as easily if they're supporting a 3 year warranty!

 

[jeffl_2]

BTW I just meant the trigger on my Kikusui COS6100M is unstable, I just found there's manuals online and an adjustment for that is available so it's probably "just another project"

 

[jeffl_2]

So here's my observation: the single-output devices are designed to work OK regardless of what's powering them. But beware those duals, a hypercompetitive market environment has forced vendors to use a design that demands that they be on a supply with the ability to supply a very high inrush current to the converter (at least for the 5 volt model). If you don't have that then you must pick a vendor who says on their data sheet "3 year warranty" because they use an entirely different design! Got it now? Good!

 

[sunnysky]

Boost converters without a soft start feature might demand 10x more power or more to startup than it is rated for output. Your choice of sources are poor and I would expect failure.

If your 9V source can meet the minimum input voltage of the boost converter under these demands often reserved for batteries, that should be your source. But beware supply regulation specs must be observed and recognize there may be compromises and it may still fail.
Otherwise you are making a self destructing oscillator without sufficient awareness of requirements.

 

[jeffl_2]

Well I can report the reason I was having problems was with the cheaper converter modules I was getting TWO failure modes superimposed, one was input overvoltage, the other was output overcurrent (this wasn't intentional nor a miswire, I had just forgotten that during initial testing without an extra supply present the existing one would draw extra current). I'm still kind of reeling from this, I could have a linear regulator with a 5 volt output and if the input to that regulator went over maybe 9.3 volts the converter at its OUTPUT would fail, which barely even makes sense! Once I changed over to a converter with a 3 year warranty the overcurrent issue went away but the input being hypercritical about overvoltage actually remained. It wasn't until a actually put a 5.2 volt TVS direct;y across the converter input that the problem went away, even a 6.2 volt zener that I had lying around wouldn't save them. I've done quite a bit of circuit development and this is the first time I've ever seen a part this eager to find a way to destroy itself, it's just about incredible but I was very careful about these observations and will stand behind every observation that I've made here.