Loads of 15W SMPS blown up: help! 2

[GonzaloEE]

Hi all,

My company sells small 15W flyback SMPS for field appliances and surveillance equipment. One customer claims these PS fail while powering their magnetic doorlocks, even after adding a protection diode in parallel with the output.
Now we have dozens of failed PS for warranty repair, We're confused because I'm sure other customers are using these PS in much worse conditions with no problem, for up to 3 years in some cases.

Here's what we got:

Supply: Flyback SMPS, 15W, 220VAC in, 12V/15W out
Load: Magnetic Doorlock 12V, with 500mA avg./3A max consumption.
Protections: Customer is adding one 1N4007 diode in parallel with load (Id' suggest a schottky instead)

All boards have a TNY268GN switcher, an RCD snubber/clamp circuit for xformer primary, and a 2A diode bridge. Protection is composed of a 275V MOV and a 3A fuse, with no PFC nor EMI filter onboard. Most failed boards show the TNY and RCD snubber diode blown up.

I'm thinking of mains surges and load transients, though I'm not a power design expert. Any ideas for fixing this up will be highly appreciated.

Thanks in advance,

 

[IRstuff]

Seems to me that you need:

> samples of their door locks
> on-site testing of locks and SMPS in operation

TTFN

FAQ731-376

 

[GonzaloEE]

Thanks for your response. We have 2 samples, and powered the bigger one through a timed switch, to cut the solenoid current each second, with no diode and for 2 days. Out of 3 SMPS, none failed. I'm scheduling a field test now.

The model is a Safer IB-02T MAGLOCK 1200. Couldn't find their website.

Is there any practical and safe way to test these boards for mains surges?

Thanks in advance!

 

[IRstuff]

A plausible approach would be MIL-STD-461 conducted susceptibility; a bit of an overkill, possibly, but very defendable if it passes.

Have you tried a gross overload condition on these locks, i.e., they get jammed and draw current well beyond their spec condition?

TTFN

FAQ731-376

 

[VEBill]

Maybe there is something else on their local power that is causing these problems (in other words, nothing to do with the load).

It's not clear how the load flyback could get around to damaging the RCD, presumably in parallel with the primary input.

 

[BobM3]

Isn't a 275 V MOV a little close to the 220 VAC operating voltage?

 

[GonzaloEE]

Sounds good.
Thanks!

 

[GonzaloEE]

Thanks to all for helping.
IRstuff, the jammed lock test sounds good. I also came across with another thread discussing about that.

VE1BLL, I agree there's a little chance for any load surge for jumping across the xformer, and blow up parts at the High-side.

BobM3, the MOV on board is labeled ZOV 07D431K. Found their specs and they say 275VRMS max, 370VDC max, Vclamp range: 387V-473V, 1200A max/20us, and 28J/1ms.


Most failures involve the TNY (internal FET with 700V Vbreak) and the RCD diode (CMR1U-06M, fast recov., 600V 1A).

We often found swollen bulk caps (rated to 400V), broken MOVs, and low-side Schottky diodes blown. The mains fuse is almost always intact, so I'm thinking of mains surges.

Thanks to all!

 

[marginal]

hello

maybe you can try powerint.com website, they have forum and will be very good with your problem.

Did your smps have output short circuit protection?

I am surprised you have 3A fuse....i suppose flyback peak currents can be high but that is a big fuse for a 15w switcher.

did you use feedback compensation with your opto-isolator? (if you used an opto) even though tnyxxx uses on/off control mode , if you use opto, due to the opto pole, you need to compensate for it.

did you use a network anlyzer to analyse the loop gain and phase margins?

if the tnyxxx failed then i doubt its a mains surge problem, as the MOV would have caught it

did you have inrush protection......we had a customer who rang us up and said that our switcher failed when he turned it repeatedly on/off for 15 minutes!

 

[GonzaloEE]

Thanks for that of PowerInt forums. I'm checking it.
Well, it's a cheap board designed before I started working here, and Sales guys believe these issues don't justify further design upgrades in the short term -a common problem here.

Yes, there is a 10A glass fuse at the output, and the TNY have it own thermal/overload shutdown protection.

The 3A fuse is a good point. With a 80% efficiency and 220VAC input, the avg. input current is around 90mA (no PFC on board).

The low-side has an opto and a TL431 shunt Volt ref. for feedback, and TL431 Cathode-Ref pins are bypassed by a small 100nF cap (freq. response)
There is a parallel RC network between the opto and DC rail. R=51 ohm, C=33pF. There's also a 680ohm resistor in parallel with opto terminals, I guess for biasing the Volt. reference.

I haven't checked loop margins yet (first time I saw these boards). Didn't find any inrush PTC, though a number of MOVs show up cracked, probably aged.

Thanks for your help!

 

[marginal]

hi

that 100nF cap from cath-ref sounds big......that might be really taking your loop bandwidth down, giving poor transient response.....but then again it might need to be with a low CTR opto

you say it goes 3 Amps transient.....well that might be the problem., with slow loop response

also remeber that opto's dramatically loose their CTR with age which dramatically slows the loop response even more.

in truth, opto's are cheap, but pretty horrible ...not much longevity.

altenatives are more expensive tho'

eg uc1901 feedback transformer or put controller on secondary and drive mosfet from isol trxformer from secondary.....

the degrading of opto's with age is very considerable

 

[PHovnanian]

Is your customer's installation grounding any part of their latch circuit?

 

[Comcokid]

From the failure mode you're reporting, I suspect the issues is AC transients - not door lock solenoids.

As for a simple way of creating AC or primary side transients to test MOV's and such try this. Years ago I tested MOV's in a application by taking a large Sola isolation transformer- it weighed about 100+ lbs and was the size of a breadbox. I would connect the secondary to the device to be tested, and switch the primary side ON and OFF with a manual switch.

 

[marginal]

hi
Also, you wil obviously have LC type EMC filter, and maybe a CLC combo after the diode bridge....this can resonate at switch on without inrush limitation.......also the RCD snubber cap can resonate with the primary peakage inductance at switch on, and if its the wrong cap , it can be induced to bad resonation.

so if switched on/off repeatedly then it could produce the overvoltage.

maybe a mag lock got stuck and it made the power supply switch repeatedly

 

[GonzaloEE]

Comcokid, thanks for the idea on the surge test. The PowerInt forum says a blown up DC rectifier could lead to further failure. Is it possible to damage hi-side parts (primary clamp, TNY, etc) due to a blown up DC rectifier?

Marginal, the output has a PI (CLC) network with C1=470uF, L=unkown ferrite choke, and C2=470uF. The input doesn't show any CM choke or filter cap. Only found a couple of 4.7uF electrolytic caps, with their GND pins connected through a strange parallel RL network, with R=4,7k and a ferrite bead of unknown value.
Looks like the jammed maglock makes sense as a possible cause of failure. Any ideas for testing a stuck maglock?

I'm reverse-engineering the boards. Hope to post a schematic here ASAP.

Thanks to all for helping out!

 

[GonzaloEE]

PHovnanian, the boards have no grounding points. I'm not sure if the maglocks are grounded on field. I'll check that out with the client.
Thanks!

 

[IRstuff]

If the maglocks are at a different potential, for whatever reason, then they will equalize through your circuitry. Then, everytime your circuit is powered up, there will be a surge, which can eventually thrash your components.

TTFN

FAQ731-376

 

[GonzaloEE]

Thanks for your post. In a worst case, assuming the maglock didn't show a inverted voltage, I think it should trigger the PS overload protection.
You mean some kind of ballast-like inrush?

 

[IRstuff]

Not quite. It's more of a question than anything solid.

I've had problems with circuits powered from different AC mains and DC converters, where the "ground" potential for different DC circuits start out at different voltages. It was a practice to not tie the grounds together, so the potentials equalized through our differential receivers, cooking them after 3-4 power on cycles. One could get a couple hundred volt difference being equalized through some rather tiny transistors.

Has anyone done a complete failure analysis of the parts?

TTFN

FAQ731-376

 

[GonzaloEE]

Yes, I've also had ground-loop issues with serial xceivers. As told by the client, these maglocks are in series with our PS and some kind of switch. This is an offline flyback with isolated output, so I think that both the PS and maglock should float at the same GND level at all times.

Thank you.