Help Fix RCCB Power Supply Issues

[Marcolao]

Hi, I got a problem and seeking for help, could anyone do me a favor.

The power supply of network data centre is supplied by a UPS. The UPS’s output connects to 2 nos. of 800A MCCB individually. And the circuits are protected by:
( Form >> to )

A) 800A MCCB >> 100A MCCB >> 100A AI switch >> ( circuit 1 & 2 )

1) 63A Type C MCB >> 63A AI switch >> 32A RCBO >> Power socket
2) 63A Type B MCB >> Floor box (40A RCCB + Power socket)

B) 800A MCCB >> 100A MCCB >> 100A AI switch >> 63A Type B MCB >> Floor box (40A RCCB + Power socket)

The case is that, a fault occurred at data server that supply by the power socket of circuit A(1). The 32A RCBO was tripped. When I tried to reset the RCBO, it tripped again immediately. At the mean time, an unpredictable scenario was happened. Some RCCB (around 6 nos.) at circuit A(2) and circuit B were tripped at the same time immediately. That caused around 7 nos. of servers were downed. Could anyone do me a favor to give a hint? Thanks you so much.

 

[jim hardy]

what's a RCCB?

i don't understand the acronyms.

A UPS is a very different power source than a transformer
and is sensitive to "Inrush Current", as can be drawn by large computer power supplies.

I don't know a thing about servers except that they have power supplies with huge filter capacitors.
The initial surge of current to charge those capacitors can be sensed as an overload by an inverter's protective features
or cause its voltage to sag a lot.

though one would think anything reasonably modern would be co-ordinated.
what i describe was big trouble back in 1980's.

today's equipment is equipped to limit inrush current
sometimes by a thermal device called "Inrush current limiter"
that takes a few seconds to cool down and return to 'active' status
it's just a NTC thermistor


just a thought

old jim

 

[Marcolao]

Thank you so much, The cause of tripped of the 32A RCBO had been found and the fault part had been isolated. I just would some hints on that why other RCDs tripped at the same time.

P.S. RCCB (residual current circuit breaker) is a kind of RCD

 

[jim hardy]

... The cause of tripped of the 32A RCBO had been found and the fault part had been isolated. I just would some hints on that why other RCDs tripped at the same time.

P.S. RCCB (residual current circuit breaker) is a kind of RCD

RCD? I'm really obsolete, don't know what that is either... but never mind that.

HINTS ? well i feel terribly inadequate here , but a drowning man needs a hand not a reference to Archimedes' principle.

Not knowing your equipment, i am guessing.
So consider this sequence and judge whether it possibly fits your system.
If i am completely off the mark it's no surprise to me. but here goes:

BACKGROUND: UPS is typically an inverter
Inverters have not much overcurrent capability
Therefore in my day UPS outputs were protected typically by fast fuses,
(perhaps nowadays they use fast circuit breakers instead ?)
Loads powered by inverters must not be allowed, when starting up, to trip the fast breakers/fuses. To that end they must be either:
...naturally low-inrush,
...or provided with inrush current limiting circuitry,
...or never all turned on at once , instead turned on in sequence,
...else their starting (inrush) current may well trip the fast fuses/breakers or overload the inverter...

POSTULATED SEQUENCE:
>>> Several power supplies are connected to same UPS source, each through its own circuit breaker..
>> those power supplies are equipped with thermistor "Inrush Current Limiters" which limit the charging current drawn by filter capacitors at power-on time.
>> A fault in a piece of equipment throws a dead short across source.
>> That short makes the UPS voltage momentarily sag to zero and remain zero until the breaker clears the fault. Fault clears at following line cycle zero crossing. Depending on where in the line cycle the fault occured, the voltage could be depessed for as long as a half cycle, eight or ten milliseconds (i don't know if you're 50hz or 60 hz). Actualy more like 2/3 cycle worst case..
>> during that half cycle the input filter capacitors in all power supplies powered by the UPS do not get charged back up because UPS voltage is depressed. In fact they discharge partway.
>> Next half cycle, after the fault has cleared and UPS voltage has recovered, all power supplies take an unusually big gulp of current as their filter capacitors charge back up - recall they missed a meal last half cycle and are now hungry for current.
>> That big gulp of current was enough to trip some of your fast breakers.

>> maybe not on first fault, which could've happened near a zero crossing so subsequent "gulp of current" wasn't inordinately large. But when you reclosed on it, Murphy's Law says that was at worst possible instant - right after a zero crossing so next one is the full 8 or 10 millisec away giving filter caps more time to discharge, and of course you got the blame .


>> reason the "inrush limiters" did not prevent the big gulp of current is they did not have time to cool down. A half cycle is not long enough. They limit current when cold and allow full current when warm, so they are warm during normal operation and cool back off when you remove power..

>> of course any automatic start sequencer was probably unaware of the power sag so stayed inactive,
...
that's the best "hint" i can offer; look into the inrush current of your supplies - compare that to overload capability of your UPS and trip curve of your breakers.

Nuclear industry learned this lesson in 1970's. Computers of that decade had inrush characteristics that were just awful.
There are some IEEE papers on the subject written by David Bratten, then of Solidstate Controls Corp and a well known industry expert..

If this turns out right, rejoice - Mother Nature wanted your industry to learn it too.

good luck. i would be interested to hear what you find.

old jim

 

[Marcolao]

No MCCB or MCB was tripped, only the the RCCBs were affected. That caused around 7 nos. of servers downed. Finally, a fault equipment at circuit A(1) had been isolated, and all circuits had been reset. My question is that ;

1) Why the other RCCBs at the final circuit of the separated circuits were tripped during I reset aforesaid RCBO?

2) How the fault affect separated circuits?

 

[jim hardy]

No MCCB or MCB was tripped, only the the RCCBs were affected. That caused around 7 nos. of servers downed. Finally, a fault equipment at circuit A(1) had been isolated, and all circuits had been reset. My question is that ;

1) Why the other RCCBs at the final circuit of the separated circuits were tripped during I reset aforesaid RCBO?

well, in your 'attached thumbnail' i see where the faulted equipment was connected.

but i do not see anything labelled "RCCB" , just a note below "floor box".
Where are the RCCB's that tripped when you tried to reset?
Can i assume it was "32A RCBO " in lower right that you tried to reset?
What's difference between a 32ARCBO and a 40ARCCB?

2) How the fault affect separated circuits?

i guessed at inrush following a voltage sag, as i described in post above.

are these RCB* thingies so called "Ground Fault Circuit Interruptors" ?
that could open a whole separate train of thought.
GFCI's are sensitive to certain frequencies and i hope your switching power supplies are electrically "quiet" in that range.

just answered my own question : http://en.wikipedia.org/wiki/Residual-current_device

i guess these RCB's are "residual current" devices ?
do you know which sensitivity they are?
Computer installations are allowed to use medium sensitivity, you may have high sensitivity installed and be getting nuisance trips. Or the earthing system may need some cleanup work

- as i said i don't know your equipment. best i can do is suggest these hints.

old jim

 

[Yuri B.]

There are RCBO's in only one out of seven 100A branches ? (in all others - RCCB's ?)
At first I would checked for loose connections everywhere - up to the 2500A ACB.

 

[Marcolao]

Further circuit diagrams are attached for reference. Thank you so much.

 

[jim hardy]

thanks for those diagrams.

wow- to me 400 kva is a big inverter. I'm accustomed to 20KVA and smaller.
I'll assume one that big is three phase. Looks like it in drawing.


you said: ""..only the the RCCBs were affected...""

only the 'residual current circuit breakers'... well that suggests (but does not prove) current in the earthing conductors...
which brings up two more questions:
what did you find in that faulted piece of equipment? Was the short circuit between hot line and earth, or between hot and neutral?
and
Are these RCCB's single or three phase? Are you servers powered line to neutral or line to line?

For a line to line or line to neutral fault , '32A RCBO' should have tripped on its overcurrent sensor not on residual current. For a line to earh fault it should have tripped on its residual current sensor, which trips at a much lower current.


So next question is - " How much current does your server power supply allow to return through its earthing conductor in normal operation? "
That depends on its internal capacitance, and most of that is likely in a "line filter" or a surge protector.
Can you make a test power cord with an ammeter in the earthing conductor and measure that return current? Shouild be in low milliamp range, or high microamp.. Compare that to the residual trip setting of your RCBO's. If it's a surprise and close to your setting, you may want to reconsider how they're set.
Just perusing literature i see typical settings between 3 and 500 milliamps.

The next question will be "How much current does your server power supply allow to return through earthing conductor under abnormal conditions? "
but that one will take some thnking and probably some testing to figure out..
Are there surge protectors on the power wires in the servers?