POWER SUPPLY FIRE

no fan, and more then 1 year in academia, no idea what the cumulative exposure was.

Steve

No matter what the circumstance of the supply was, the breaker did NOT perform as designed, and that FACT is quite obvious.

It sounds very much as though something fell behind it and burned with a very hot flame, the electrical damage was secondary.

If there is no sign of melted aluminuim from the top part of the panel:

My own experience of universities has meade me extremely cynical and you may never get to the bottom of this. The best you can hope for is that they blame your friend's design without taking any further action against him which might risk exposing the truth.

:On Tue, 17 Feb 2009 10:59:10 -0800 (PST), snipped-for-privacy@yahoo.com wrote: : :>x-no-archive: :>>

:>> If they left the 30KV supply on for an extended time, would static :>> electricity draw these metal powders into the power supply? I am :>> thinking "thermite" here. :>>

:>> -- :>> Joe Leikhim K4SAT :>> "The RFI-EMI-GUY"© :>>

:>

:>

:>or did the supply have a cooling fan that may have drawn in the :>dust... :>

:>how long was the supply in use in this environment before the :>incident? :>

:>

:>Mark : : : No matter what the circumstance of the supply was, the breaker did NOT :perform as designed, and that FACT is quite obvious.

But even assuming the 20A CB failed to trip that doesn't explain why the panel mount fuse on the 30kV generator didn't protect the unit from fire. I am guessing that the panel mount fuse would not have been more than a few amps rating at most since the HV output was only 30W.

I think that using a SMPS (without output protection) to supply the HV switcher might be the cause of the problem.

If the switcher or the HV generator circuit failed and caused a high load on the SMPS output then the panel mount fuse would not have blown irrespective of whether the 20A breaker was faulty or not. This is because of the overload protection characteristic of the SMPS is generally set to 110% of the rated maximum load current rating of the supply. When the output is severely overloaded they go into "hiccup" mode to self protect, and this will not cause the input fuse or any thermal-magnetic circuit breaker on the 110Vac input side to trip. During hiccup mode, and depending upon the severity and type of overload, the SMPS can still supply power to the load between hiccup pulses. If the SMPS was very conservatively rated and was capable of much higher output than the HV switcher required for normal operation, the situation could be even worse. In my experience the only time an input fuse on a SMPS blows is because of a shorted bridge rectifier or switching FET or transistor or other catastrophic failure. Because the original fuse could not be found it is not possible to know what happened to this device.

It might be possible for the HV module (because it is relatively low power) to have continued to be operational in some manner during "hiccup" mode of the SMPS even though it was placing an abnormal load on it. I can't imagine just what fault condition in the HV module might cause this without inspecting the device, but it might be possible. If so, then there may be a means for creating excessive heating or some other flammable condition, but without seeing the thing I can't imagine what it might be.

I know it doesn't help now but the DC output of the SMPS which feeds the HV switcher should have been protected with a smart electronic circuit protector which tolerates short term duration overloads lasting milliseconds but which will trip within seconds of any overload between 1.3 - 1.8 times the selected current rating. This would have disconnected the HV switcher from the output of the SMPS if there was any hint of an excessive load on the SMPS even faster than the SMPS could react to protect itself.

Such devices are represented by the E-T-A ESX10 series of devices.

It might be informative to also read their white papers

You are assuming that the fire was entierly electrical... Perhaps the current was interrupted well before the current needed to trip those breakers was reached, and the damage was from the (secondary?) fire?

It has already been suggested that since that part of the supply was conveniently missing, that the "fuse" that was installed may have been what is commonly known as a code violating "bypass". like a nail or other such stupid behavior.

I do agree with your other post that a fire may have caused the electrical shorting, not the other way around. With a 30kV source active, electrons have a way of finding their way into some strange "drain points".

The word creep comes to mind, and that word can include carbon trails, and most certainly can include combustion level temperatures once it finds whatever attracted it. Even at only 1mA.

I do not think that interlink needed any protection as switchers are always made such that a full short on the output does NOT cause any fires. The supply (switcher) must "fold back" to get any certifications at all. Most switchers (open frame) have no protection at all, and the integrator has to incorporate front side and load side protections, if any, themselves.

Exactly. It also will not, nor can not cause a fire, or other excessive overheating.

Which is as it should be. The DESIGNER MUST select a fuse that DOES blow when his 30kV @ 1mA source gets loaded beyond its rating. Problem is that most folks that only need 1mA at those voltages, also need very low ripple numbers, so that gets achieved by way of a RC filter on the output. So even a dead short on the output will not load up the supply any further at all in many cases. That is why I mentioned for him to make sure that the multiplier caps selected were small enough that not too much energy was being stored, yet not so small as to increase ripple on the pre-filtered output. That also makes the front end requisite of the HV gen section less, increasing overall efficiency.

On an AC fed device, front end fuses usually are there to protect the AC entry into the unit, and most DC side stuff is isolated from the AC, yet protected by that fuse.

So yes, I digress and agree that internal protections would further protect it. I don't think it caused a fire though.

One should examine the SMPS, because if it caused the fire, it would also be severely damaged, not just the corner of the system case.

I still trust SMPD makers a bit more than that. I do not trust missing elements of a fire, and that alone would get it thrown out of a suit action.

Whatever. His HV side, working or not, did not cause this, as long as the SMPS was made to normal currently used standards. That is the main reason why I suspect that something else on that bench had a fire, as you earlier mentioned may have happened. The 30kV *may* well have been the ignition source AT the 30kV end, but that would be THEIR fault for leaving it up on a set up that had no protections. That set-up is their responsibility, and if it was a bench experiment setup, then leaving it up, unattended would ALSO leave them as culpable for any failure mode as well.

One does not go around leaving 30kv HV sources up and running in a setting where it has not been properly, permanently installed such that the HV output is sealed from arcing all the way to the load.

It is beginning to appear as if the fire caused the electrical wires to burn as well, and the hot conduit can soften or melt electrical wire insulation. THEN a short would occur and blow the breaker.

So the fire may not have been caused at the front side of the supply at all.

Unattended HV "bench set-ups" are a pretty stupid behavior to start with.

I question whether they have any case at all, as unattended HV sources that are not sealed all the way to the load are not the fault of the supply maker if the set-up catches fire.

That itself was quite stupid of them. HV? Nobody home? Not permanent? DUMB!

Another theory: What if, SOMEHOW, the fuse was IN PLACE, but the voltage involved caused an arc across the fuse terminals? Just a thought.

BTW, I got this idea from

:On Wed, 18 Feb 2009 09:10:24 GMT, Ross Herbert :wrote: : :>I think that using a SMPS (without output protection) to supply the HV switcher :>might be the cause of the problem. : : : I do not think that interlink needed any protection as switchers are :always made such that a full short on the output does NOT cause any :fires. The supply (switcher) must "fold back" to get any certifications :at all. Most switchers (open frame) have no protection at all, and the :integrator has to incorporate front side and load side protections, if :any, themselves.

All of the SMPS I have encountered do not fold-back during overload on the output. Some may incorporate fold-back current limit on certain output rails but these are usually specially designed units.

More often. they go into pulse-by-pulse (hiccup) current limit mode during overload conditions. When overloaded (110% of rated output) they will shut down and then re-test the output to see if the overload condition still exists. During this re-test period they can pump out considerable current into the load before shutting down again if the overload is still present. The amount of current that can be delivered into the load during re-test will be much greater if the rated output of the SMPS is excessive to the requirement of the load.

If the HV switcher was exhibiting a partial failure which caused internal overheating of components during overload hiccup mode, then this could be the reason for the fire starting.

I reiterate, SMPS when in hiccup mode, will NOT cause a mains input fuse or circuit breaker to activate. That is why there MUST be a sensitive overload protection device included between the output of the SMPS and the input to the HV switcher to completely disconnect the two sections. If the HV generator was able to produce excessive heating while the input SMPS was in hiccup mode then this may have cause the fire which burnt out the rest of the unit. Any fire hot enough to melt the aluminium panel would certainly destroy any fuse or holder mounted on it.

I must point out that my suppositions are those of the devils advocate. I can't see the items described by the OP so I can't possibly know what actually transpired. However, the situation I have desribed has to be considered by the OP and the parties involved as part of the "what if" considerations as to the possible causes of the fire.

Never more than the same 110% cut-off value for normal operation.

But the input current of the power supply was protected by a small fuse, which was intact, and it would have opened long before there was enough heat in the line cord to cause its insulation to melt and the conductors to short together. A typical 18 AWG line cord will survive a solid short circuit long enough to open a 20 amp breaker on the distribution panel, but it could fail if there is a prolonged overload.

A 20 amp breaker is not guaranteed to trip until current reaches about 25 amps, and may take as long as an hour at that. Even at 50 amps, it may take two or three minutes and still be within specification. It will typically take at least 8x and as much as 12x rating to trip instantaneously. You did say that it tripped? So it could supply as much as 5000 watts for a couple of minutes, or 10 kW for a few seconds. I don't think that is enough energy to vaporize an aluminum lid.

I recall forensic analyses in the NETA magazine (InterNational Electrical Testing Assn), but the closest I could find was this:

These links may be helpful:

A friend had a problem where a SLA battery for one of his products apparently caught fire and caused damage. He determined that it must have been from an outside source of combustion, such as a gasoline-soaked rag. I think something similar happened in your case, and there do seem to be signs of tampering with evidence.

Paul

I suspect that fault started there (at the outlet) with an arcing fault, followed by the 20 A breaker failing (closed) and the real damage was done by over current through the 400 A feed. That is enough oomph to do the kind of damage seen. But this is just my wild guess based on clearly insufficient data.

:> :On Wed, 18 Feb 2009 09:10:24 GMT, Ross Herbert :> :wrote: :> : :> :>I think that using a SMPS (without output protection) to supply the HV :> switcher :> :>might be the cause of the problem. :> : :> : :> : I do not think that interlink needed any protection as switchers are :> :always made such that a full short on the output does NOT cause any :> :fires. The supply (switcher) must "fold back" to get any certifications :> :at all. Most switchers (open frame) have no protection at all, and the :> :integrator has to incorporate front side and load side protections, if :> :any, themselves. :>

:>

:> All of the SMPS I have encountered do not fold-back during overload on :> the :> output. Some may incorporate fold-back current limit on certain output :> rails but :> these are usually specially designed units. :>

:> More often. they go into pulse-by-pulse (hiccup) current limit mode :> during :> overload conditions. When overloaded (110% of rated output) they will :> shut down :> and then re-test the output to see if the overload condition still :> exists. :> During this re-test period they can pump out considerable current into :> the load :> before shutting down again if the overload is still present. The amount :> of :> current that can be delivered into the load during re-test will be much :> greater :> if the rated output of the SMPS is excessive to the requirement of the :> load. :>

:> If the HV switcher was exhibiting a partial failure which caused internal :> overheating of components during overload hiccup mode, then this could be :> the :> reason for the fire starting. :>

:> I reiterate, SMPS when in hiccup mode, will NOT cause a mains input fuse :> or :> circuit breaker to activate. That is why there MUST be a sensitive :> overload :> protection device included between the output of the SMPS and the input :> to the :> HV switcher to completely disconnect the two sections. If the HV :> generator was :> able to produce excessive heating while the input SMPS was in hiccup mode :> then :> this may have cause the fire which burnt out the rest of the unit. Any :> fire hot :> enough to melt the aluminium panel would certainly destroy any fuse or :> holder :> mounted on it. : :But the input current of the power supply was protected by a small fuse, :which was intact, and it would have opened long before there was enough :heat in the line cord to cause its insulation to melt and the conductors to :short together. A typical 18 AWG line cord will survive a solid short :circuit long enough to open a 20 amp breaker on the distribution panel, but :it could fail if there is a prolonged overload.

I don't recall the OP clarifying what the SMPS input fuse rating was. If you understand how SMPS work you can short an output rail and it will go into hiccup mode while the short exists. This condition, provided that the input devices and the switch mode IC are functioning normally, will NOT blow an input fuse of the correct rating. This fuse only protects against a catastrophic failure on the input side of the SMPS (such as shorted diode or MSOFET). I am assuming something like a 4A slo-blo might have been used in this case. So assuming the SMPS was able to continue functioning with a partial overload in the HV switcher (not a full blown short circuit) which it was supplying, it may have been able to pump a large enough current into the faulty HV switcher to cause it to overheat or somehow cause ignition of componets, which then progressed and burnt up all the other components including the input fuse. If the input fuse was destroyed by the fire then it would have been disconnected from the 110V supply before the 20A circuit breaker could trip.

It is possible that the fire was started on the output side of the SMPS, and not in the HV switcher, although this would be most unlikely in my experience.

: :A 20 amp breaker is not guaranteed to trip until current reaches about 25 :amps, and may take as long as an hour at that. Even at 50 amps, it may take :two or three minutes and still be within specification. It will typically :take at least 8x and as much as 12x rating to trip instantaneously. You did :say that it tripped? So it could supply as much as 5000 watts for a couple :of minutes, or 10 kW for a few seconds. I don't think that is enough energy :to vaporize an aluminum lid.

I never saw anything which indicated that the 20A CB tripped. In fact I think the OP said it didn't iirc.

: :I recall forensic analyses in the NETA magazine (InterNational Electrical :Testing Assn), but the closest I could find was this: :

I can recall an instance several years back where a client experienced a spontaneous fire on the motherboard of a new IBM desktop PC. The SMPS, being

From the OP's first post:

"20A industrial circuit with a GFI, backed by a breaker into a bus fused at 400A. Breaker tripped, NO GFI trip."

The SMPS in this unit was probably no more than 100 watt, and more likely about 60 watt, and fused at about 1 or 2 amps, because the load was:

"We're talking two 30 watt HV modules systems here"

And only one was used at any time. It really sounds like something on top of the supply ignited and reached a temperature where the aluminum may have burned, although the lack of Al oxide is puzzling. Apparently there was a consequential fire involving other flammable materials that may have caused the line cord to burn. But it's also puzzling why the conductors in the conduit were melted on both sides of the outlet:

"Wiring was conduit from the panel to a Ul approved lab grade outlet strip built along the wall, wiring for 4 feet on either side of the outlet in use was burnt/destroyed."

I'd love to see the pictures...

Paul

d.

ve

ed

I decided days ago not to ask for photos because I doubt the attorneys would approve. But the photos of this would of course be very interesting.

Ask anyway - Let _them_ assume the burden of refusing your request.

Has it reached litigation? Maybe you could solve it over a conference table.

Good Luck! Rich