MOV fails and fuse blows at input of multiple 500W+ AC/DC SMPS

[dopeydavoid298]

Hi all!

I'm hoping someone has seen something similar or the solution is obvious.

We have multiple AC/DC SMPS (lab power supplies) at ratings at ratings of 400W, 1500W and 2X 600W all rated at "100-240" (max 265VAC in manual) that have had the fuses or MOVs fail (MOVs all measured <50ohm on the multimeter). These supplies range across two different manufacturers.

All supplies have CV/CC/CP settings and our grid voltage is 244-250 at the outlet. These incidents have occurred at different buildings on the site and in different configurations on the output.

In all cases, the trigger has been impulse current events on the 18-30V outputs. The first was inrush on the output of two devices that simultaneously died, and the next two were the 2x parallel connection attempts of 600W models (module 1 off and module 2 on, connect together, module 1 on, module 1 dies during balancing surge- Parallel works okay otherwise and follows manuals master/slave config).

In all cases the input fuses audibly explode (do not rupture) and the MOV either ruptures or measures low impedance. I'm suspecting input voltage ringing from the sudden application/removal of a huge load as our grid is high relative to overall operating range.

Does anything obvious come to mind? I can't see how else anything on the output of CV/CC/CP lab power supply can blow the input MOVs?
(I do have a device with 50kHz bandwidth that will convert the grid voltage to a 0-1V output for scope measurement.)

Dave

 

[KlausST]

Hi,

my assumption:
high voltage (overvoltage) --> MOV becomes conductive (for a short time) --> MOV gets stressed (either by multiple overvoltage situations or by overtemperature) --> MOV becomes conductive (continously) --> Fuse blows.

So the root question is: Where/why is there overvoltage.

With the current information it´s impossible for us to find out.
Thus we have to rely on you.

****
1) use a scope and measure the mains voltage. Trigger on overly high peaks.
2) tell us what else is installed in combination with the supplies. (your application)
3) tell us where you live, so we can have an idea about power grid stability
4) if you can get: use a mains analyzer. That counts and stores several exceptional mains voltage events.

Maybe take some photos of the wiring, especially at the mains side. Some details, some overviews.

Klaus

 

[Easy peasy]

you didn't say what the nominal ratings of the mov's were/are, high mains can degrade a mov to where it expires - or you had a local over volt event or two .....

 

[FvM]

"Exploding" fuse sounds like a design ignoring safety standards, e.g. insufficient short circuit current interruption capability of fuses. Or inappropriate fuse replacement.

Mains overvoltage and surges beyond MOV rating are most likely the primary fault. Would be interesting to know if the SMPS primary stage is designed with higer overvoltage margin than the MOV.

I fear there's no simple solution. A -10% auto transformer would be feasible.

 

[cupoftea]

In all cases the input fuses audibly explode (do not rupture) and the MOV either ruptures or measures low impedance. I'm suspecting input voltage ringing from the sudden application/removal of a huge load as our grid is high relative to overall operating range.

...yes, i think this is ithe key, they must have put the mov downstream of the mains-side LC input filter, and so load transient (refered to the input) makes the v(mov) ring up and the mov quenches it and dies in the process

 

[D.A.(Tony)Stewart]

When so many Joules of energy is discharged into a short circuit, the response is V= LdI/dt + I*ESR while the fuse has an thermodynamic rise time. When it explodes there is an almost immediate arc after ionization time in microseconds. Before that arc there is a potential that rises abruptly to continue this current from a discontinuity in resistance. That is of course the flyback effect. The power supply meanwhile sees a rise in voltage error and now demands more current from the input grid until it's input fuse or breaker blows. Yet before that, the MOV melts into a short circuit.

With the isolation inside the power supply, unfortunately this is capacitance coupling transferring this voltage transient back to the input, while the MOV starts to conduct. It also is rated to absorb a limited energy and is rated in Joules which you may compare with the energy stored in the supply capacitors. The fuse gap must be long in order to reduce the time duration of arc so that it has a higher holding current threshold.

This cascade of events is common and in order to decouple input from output the impedance needs to rise and the response time needs to be increased. The analogy would be a battery with a trickle charger rather than an active SMPS. Naturally this increase stored energy over a low ESR Cap by virtue of the energy density of batteries being a thousand times the capacitance per unit volume with greater aging factors is not a solution.

You might want to consider pre-balanced <1% and 10% pre-loaded PSU before tandem connection. I have found this to work on mainframe redundant PSU's with a load-sharing feedback link between PSU's. Otherwise, before this loading tandem current-sharing supplies with dual feedback tend to oscillate.

Consider a tandem MOV's with linear series R added equal to the MOV's Rs to prevent the thermal runaway process like the NTC effects in complementary BJT power Amps (MOV array).

Consider what plastic dielectrics with low ESR can shunt the abrupt voltage step currents with a damped filter to the short circuit load test.

Distribute the OCP fuse OVP overvoltage protection to prevent flyback ionization and quench some other way such as with TVS technology.

Try to simulate this singularity fault before any design changes.

--- Updated Sep 25, 2021 ---

It reminds me of the time a Sr Design Engineer used a gas tube AC OVP clamp on a microwave repeater. A lightning transient a block away caused a fire inside the metal box with a crater inside the PCB. He didn't have a fuse for the gas tube. I had to prove the negative resistance effect to him with a test jig and a hair trigger high voltage from 1 Megohm to collapse the gas tube voltage.

 

[dopeydavoid298]

Hi again all, sorry I'm slow to get back. There's too much happening right now.

I'd like to focus on the initial pair of units being damaged as I suspect something specific to the other paralleled supplies and the turn-on ramp up behavior. I also cannot take photos, nor can I be specific about the load, but I'll try.

We have an active load that draws about 20-30A at 28Vdc and 5A at 18Vdc from two input terminals (clean and dirty rails). The grounds are tied together and earthed. Power is supplied via two lab-style power supplies from the same manufacturer (400W and 1500W models), both with CC, CV & CP and input-to-output isolation of over 1000Vdc. To power these two supplies, different phases are used. 1 phase goes to a surge protected powerboard (I believe, cannot confirm) that then goes to the 400W supply and laptop. The other phase goes to the 1500W supply directly. Both supplies are set to automatically come on when the AC is switched on. AC is 244-250Vrms in Adelaide, Australia. This is a new building (2 years old). Electrical work WAS being done in a different room at the time...

Anyway, this was working for a while, then we decided to use a different three-phase breakout board. This new breakout board had more plugs so we did away with the surge-protected powerboards. Both supplies were still on different phases.

I turned on the mains AC at the 3-phase plug and about 5 seconds later we heard some bangs from the supplies. We turned the mains off and on again and found the circuit breaker in the building for that circuit went off (25A+). We had not yet brought the load up to full power (it was still in a type of standby mode that eats 40W).

On the large supply the fuse was fine and the '275' rated MOV had ruptured (it's 15A fuse was fine though).
On the small supply the fuse had popped and the same MOV was measuring low impedance and was popping subsequent fuses.

We have confirmed that:
-->The new 3-phase breakout board is wired correctly and has previously been used.
-->The load is fine.
-->Everything continues to be fine when operated from several smaller power supplies hooked up in parallel from a surge protected power board and a -10% transformer, (one of them being another of the 400W model type that failed earlier).

I did have a look at the grid voltage when the mass-paralleled arrangement was in operation from a single phase on a surge protected power board. I used an oscilloscope and some custom equipment as follows:

"240"V/10A outlet --> powerboard with no surge protection (measurement point) --> 10m extension -->
-->surge protected powerboard --> power conditioner --> power supplies --> load.

The power conditioner is a surge protection box with a few taps/relays to keep the voltage at "220V". In most cases it's at max and reducing the voltage by 10%.

As I'm using the ISO124 chip for the measurement, there is a lot of ripple at 500kHz and a 3rd-order filter to eliminate it, but amoung the mess I'm seeing 20us jumps of 60V. Occasionally, they line up with the peak of the waveform and could easily set off a MOV. My next test is to get a kettle and flick it on/off in different locations and with/without the surge protected power boards.

Dave

 

[D.A.(Tony)Stewart]

Your PSU's may not have a soft start and draw > 10x their rated power on startup. This capacitive load with low ESR creates a 3ph load imbalance and possibly makes things worse for the smaller supply. What are your goals (i.e. specs?) and what is your result? Please list all and Show block diagram with transient impedances where possible.

 

[Easy peasy]

You need to check the Neutral connections all the way back to the supply transformer ( measure the real ohms at 10ADC or AC at least ) and the quality of the earthing conductors all the way back to the supply Tx too and the N-E links on the main DB and all the sub DB's.

If the neutral lines are not good - then a large ( transient ) single phase load can cause more volts on the other phases.

We have had to beef up neutrals ( & earths ) in some locations to keep the N-E voltage down to < 5V ac rms ( 300 Hz with some 50Hz ).

What industries are nearby ? do they switch large loads / motors ?

Also HF ripple on the mains can kill a MOV due to the extra AC current it carries - sometime the ripple control network can do this,

sometimes someone testing on a nearby bench.

Has there been any lightning events in your area ? - ask the local MET office for times thereof

 

[dick_freebird]

Sounds to me like fuse opens and then the flyback
event punches out the MOV.

One bench experiment would be to apply a shorting
load to the fuse output, trigger off the release and
catch the true flyback peak to see whether you're
respecting ratings / deratings. Bearing in mind that
MOVs degrade with each "use" and if the problem
has been addressed by repeatedly replacing fuses,
the MOV may be getting degraded every time
leading to eventual shorting.

Might look to the input current limit / input OV
response to see whether the power module is
actually the cause of the flyback, like if the
inrush current protection is switched rather than
linear, maybe that's the chop. Or if inrush current limit
isn't really respecting the fuse ratings.

 

[D.A.(Tony)Stewart]

I suggest you balance your inrush loads or design an inrush limiter ( startup ICL varistor with time delay bypass relay) after you scope the characteristics of each. Then balance your phase loads. But your protection characteristics are self-destructive at present.

 

[dopeydavoid298]

We ended up going with a safe approach and have the following:

Surge/Filter/autoreg box --> 240/120 step down --> power supplies

Autoreg box:

LR2000 Tripp Lite | Line Protection, Distribution, Backups | DigiKey

Order Tripp Lite LR2000 (TL410-ND) at DigiKey. Check stock and pricing, view product specifications, and order online.

www.digikey.com.au

Step down:

MD-2400-E Triad Magnetics | Transformers | DigiKey

Order Triad Magnetics MD-2400-E (237-2075-ND) at DigiKey. Check stock and pricing, view product specifications, and order online.

www.digikey.com.au

Seemed logical as both supplies work down to 95Vac ("universal input"). This also has the benefit of reducing the inrush current I imagine?

This will be used on the field and with generators, so a Band-Aid approach is more practical than anything relating to fixing the building.

Sorry I can't be more technical, this is a work environment where working directly with mains circuitry directly is a no-no, so boxed modules is the only economical way to avoid administrative overloads.

I did read through everything, so thanks for all your input. Is the problem fixed? Probably. I'll let you know if things manage to blow up again and more attention gets paid to the project as a result.

Dave

 

[D.A.(Tony)Stewart]

I don't see how these choices will improve your situation.

These provide OVP (MOV surge protection and with auto-tap changer line conditioner and also UVP but not OCP ( over current or inrush protection) and step down transformer may help somewhat as an impedance conversion to reduce current peaks.

Remember 1500W ACDC supplies will store up to 10x this energy in ripple smoothing reactance and it is charging up this energy on power up that we believe triggered the fault. I had a similar experience as TE Mgr at Burroughs in the '80's when one of a dozen different ATE machines with a similar large ACDC PSU blew a power supply because all the large motor 14" HDD's in production caused a voltage sag from imbalanced loading with a transient over-voltage on the ATE which combined with it's inrush blew up the supply.

Line Stabilizers do not fix current transients, just steady-state voltage. But may fix OV transient which your blown MOV was chosen to do, perhaps undersized for [J]

 

[dopeydavoid298]

Heya Sunny,

I also forgot to mention that the two supplies (1200W and 400W now) are both off the same phase now, so that helps with the imbalancing.

The 240V/120V step down as you mentioned should reduce current peaks for the inrush condition. Using the 240V/120V step down also has the beneficial effect of guaranteeing the MOV in the auto-tap changer will go off far before the power supplies MOV (the former being able to absorb 1360J).

The auto-tap box also includes some LN/LG/NG capacitive filtering and the actual auto-tap regulation is still beneficial with overloaded "bouncing" generators in the field.

For OCP, if the 3-phase imbalance problem is removed and our OVP removes the effects of upstream external reactance, what can be done to reduce inrush current other than using the lowest operating voltage possible via the step down transformer?

In our environment we can only use "consumer" equipment, so component PCB level equipment is eliminated.

I can therefore think of:
->Manually bringing up a variac?
->Inrush limiting resistor that's bypassed with a relay? (I'm sure there is an off the shelf module for this)

Still, given the removal of the load-imbalancing problem, you'd imagine the equipment itself should be designed to survive whatever it get at this point?

How did you fix your issue with the HDDs?