RCD trips; What is considered "normal"?

[Halcyon]

Yesterday, I had an RCD trip which feeds a server rack. The rack contains the usual IT equipment; Servers, switches, UPS etc...

I understand that an abnormal condition led to the tripping of the RCD, and I've had this occur before with the same/rack circuit, but I can't pin it on any particular device as over the years, things have been removed/added. Over the past 5-6 years, I've maybe experienced this 3 times, so it's a very random and spurious event, not something that occurs frequently.

My question is, could this be considered "normal" and simply shrug it off, or do I have a bigger, developing problem to try and troubleshoot? Could something such as static build-up or some other odd, random event cause this, as opposed to some piece of equipment going faulty?

[TimFox]

Leakage current to PE in multiple line filters?  Either reactive current (too many capacitors in parallel) or inappropriate leakage paths?

[Ian.M]

Line filter Y capacitor self-healing events?

Also consider the effect of large common mode transients on surge surpressors that clamp to ground.

[Whales]

RCD could be faulty.  They do age and become problematic, they're not designed to last forever.

Any chance it's also an AFI?  Those are famous for EMI sensitivity.

EDIT: Are you using any shielded CAT6 that leaves the room?  Or other comms cable with an earth ref, like HDMI?

[Halcyon]

RCD could be faulty.  They do age and become problematic, they're not designed to last forever.

Any chance it's also an AFI?  Those are famous for EMI sensitivity.

EDIT: Are you using any shielded CAT6 that leaves the room?  Or other comms cable with an earth ref, like HDMI?

The RCD has been in service for about 8 or 9 years, but if it becomes more of an issue, I'll consider replacing it.

Only comms leaving the cabinet (and the room) is unshielded twisted pair (and fibre).

[tooki]

RCD could be faulty.  They do age and become problematic, they're not designed to last forever.

What would fail? The heart of an RCD is the current transformer, which basically nothing can go wrong with, and then a coil (powered by any current mismatch between the L and N windings of the current transformer) that trips the breaker.

[Siwastaja]

With "a lot" of earthed switch mode power supplies, their Y capacitor leakage currents add up. You need to think about it when choosing how many and which power supplies are put behind an RCD, or when choosing the RCD sensitivity. What is the trip rating, 6mA, 30mA, something else?

Chances are, you could have expected leakage close to the rating, so that any small disturbance brings it over the edge. If so, you need to change for a bigger RCD (if allowed by code), or divide the loads into multiple groups, each protected by their own RCD.

[SeanB]

As you typically are bound by regulations on RCD trip current, you probably would be best off measuring leakage current in the rack, and if it is over half the rated 30mA, then it is time to do a second feed to the rack, with it's own RCD. Then this also allows you to put devices with redundant power connections onto each supply, so that a single trip does not fail them all. Same for UPS devices as well, one per power feed, and  then as well your monitoring will need ot alert for a failure of a single supply.

[Halcyon]

I believe it's rated for 30mA but I'd have to double-check. I believe that's the maximum allowable in Australia, but stand to be corrected if I'm wrong.

I might have to get an electrician out just to do some tests and if required, run another circuit.

[mikeselectricstuff]

RCD could be faulty.  They do age and become problematic, they're not designed to last forever.

What would fail? The heart of an RCD is the current transformer, which basically nothing can go wrong with, and then a coil (powered by any current mismatch between the L and N windings of the current transformer) that trips the breaker.

They do definitely fail - domestic types are built down to a price, so subject to things like failing electrolytics.
They aren't designed to deal with regular tripping - they generally energise the trip coil with mains voltage via a small triac, and rely on the fact that it trips to interrupt the current before the triac smokes - any mechanical wear that prolongs the switching time can kill it.
And being connected to mains are subject to transients etc.

[AVGresponding]

Modern ones (nine years old is modern enough) are controlled electronically,  and they have the usual failure modes as Mike has pointed out.

It's also worth noting a 30mA RCD is unlikely to trip at as high a current as nominal; that's the maximum allowable tripping current. If you read the data sheet you'll find they typically trip at around 18-25mA.

My vote goes to degrading caps in the mains filters of the IT gear though.

[mikeselectricstuff]

Yesterday, I had an RCD trip which feeds a server rack. The rack contains the usual IT equipment; Servers, switches, UPS etc...

I understand that an abnormal condition led to the tripping of the RCD, and I've had this occur before with the same/rack circuit, but I can't pin it on any particular device as over the years, things have been removed/added. Over the past 5-6 years, I've maybe experienced this 3 times, so it's a very random and spurious event, not something that occurs frequently.

My question is, could this be considered "normal" and simply shrug it off, or do I have a bigger, developing problem to try and troubleshoot? Could something such as static build-up or some other odd, random event cause this, as opposed to some piece of equipment going faulty?

It might be useful to actually measure the leakage current (ideally clampmeter, alternatively a meter in line with the earth), as with multiple devices, it may be that you're running quite close to the trip threshold, so things like mains voltage excursions could push it over the edge.     

[Halcyon]

My vote goes to degrading caps in the mains filters of the IT gear though.

Which is entirely possible. One of the servers has been in almost 24/7 operation for something like 18 years (running modern Linux however, fully patched). If I was to start looking anywhere, it would be that server.

It's just one of those things, why replace it when it's working fine (although I do have a replacement plan for it already). I've only replaced one power supply in it during its life.

[m k]

Do you change UPS batteries before they fail?

[AVGresponding]

Do you change UPS batteries before they fail?

If you have an effective PPM program you do! In my experience, this is the minority...

[Xena E]

My vote goes to degrading caps in the mains filters of the IT gear though.

Which is entirely possible. One of the servers has been in almost 24/7 operation for something like 18 years (running modern Linux however, fully patched). If I was to start looking anywhere, it would be that server.

It's just one of those things, why replace it when it's working fine (although I do have a replacement plan for it already). I've only replaced one power supply in it during its life.

It happens.

Transient hits the filter caps, the film is compromised, then the foil blows clear.

If you're running near the RCD threshold with leakage anyway it activates and you're left wondering what the problem is.

The answer is that there's probably no actual problem.

[indeterminate]

If the rack is not in a domestic environment then there is no need to have the circuits supplying it protected by a RCD in Au/Nz
There are sum other requirements and limitations on locations but you will only find RCD's in the circuits for the wall outlets in a data center
all the rack supplies are CB only.

[AVGresponding]

If the rack is not in a domestic environment then there is no need to have the circuits supplying it protected by a RCD in Au/Nz
There are sum other requirements and limitations on locations but you will only find RCD's in the circuits for the wall outlets in a data center
all the rack supplies are CB only.

The requirements have tightened up in the UK, and I wouldn't be surprised to see Aus and NZ follow suit. It won't prevent a non-RCD install for dedicated IT systems/rooms, but it will make it a lot more difficult for the installation manager to justify (in order to pass a compliance inspection or in the event of legal action following an incident), as there will be more and more hoops to jump through.

[Siwastaja]

It's also worth noting a 30mA RCD is unlikely to trip at as high a current as nominal; that's the maximum allowable tripping current. If you read the data sheet you'll find they typically trip at around 18-25mA.

Yeah - which is also why misdiagnosing a "broken" RCD is pretty common - when problems occur, the RCD is replaced, and the new one might have a few mA larger actual trip current, problem goes away, and the old RCD is declared "faulty". Until, after a few years, the trip current drifts for whatever aging-related reason. Or, the opposite, old RCD is replaced for any reason, and the new one happens to be a few mA more sensitive, and now the electrician who does not actually measure the leakage currents and read the specs, thinks the new one must be faulty because it trips unlike the old one.

In other words, when RCDs trip, one should really measure the leakage current.

[jonpaul]

Hello: Very common issues.

Whats the curreent Rating and trip curve of the breaker?

What is Load average and peak amperes?

1/ Check load and CM current leakage at the breaker

2/ Replace breaker

3/ Check for overload ?

4/ Sectionalize servers to use several breakers.

5/ Use PDU (power distrib unit) with per load monitoring.

Just my experience, since 1970s...

have an absoultely fantasitc day !

Jon

[Siwastaja]

Just my experience, since 1970s...

Maybe this is the problem, because after 1970's, this new fantastic invention called RCD started to gain traction. While it is arguably older, it still was unknown to many before 1990's, so I understand your confusion. RCDs and circuit breakers are two completely different components; this thread is about RCD tripping, not breaker. I suggest you open a new thread in the Beginners section if you need more guidance.

[Haenk]

Not sure if legally required, but I think testing the RCD every 6 months or 12 months is highly recommended, so they are likely not made to last forever. So if it's a decade old, maybe replace it is a good idea anyway.
But I guess you have some leaking filter caps adding up, no easy fix for that - either remove the RCD or replace the PSUs. If it happens on a frequent base, you might add RCD to each of the devices to narrow down the failure. Not a cheap solution, though.

[soldar]

With "a lot" of earthed switch mode power supplies, their Y capacitor leakage currents add up. You need to think about it when choosing how many and which power supplies are put behind an RCD, or when choosing the RCD sensitivity. What is the trip rating, 6mA, 30mA, something else?

Chances are, you could have expected leakage close to the rating, so that any small disturbance brings it over the edge. If so, you need to change for a bigger RCD (if allowed by code), or divide the loads into multiple groups, each protected by their own RCD.

Yup. Just divide the load. You have N circuits protected by one RCD. Get another one and divide the circuits between the two. If one RCD keeps tripping you can then move a circuit from that one to the one that does not trip. Or further subdivide. After a while you should have a fairly good idea of where the problem originates.

[nvmR]

Don't forget to check dc and ac current on the ground line.
Do note the value on the meter after you press tare, and compare to that value.