A closer look at the volt RCD tripping issue

[acmotor]

We did some tests the other day on g4's volt and a selection of EVSE units. Not a fancy setup, just looking at what is going on.

All units passed the Oz standard 'test and tag' tests as would be expected. (insulation, continuity, leakage) Of particular interest, there was almost no earth leakage (<1mA).

The volt itself 'passed' the test and tag tester checks (as a non double insulated appliance). But still caused the occasional RCD trip. The delay and soft start of the volt charging meant that the test and tag tester had completed its checks before the volt charger had powered up. (note here... the test and tag system was unable to see the continuing volt earth leakage fault.)

A 4.7ohm 10W resistor was fitted in the earth line so an oscilloscope (floating) could monitor the earth current. (4.7V = 1A earth current)
A clamp probe could have been used, however we expected to need to see high frequency spikes so went with a simple resistor.



The volt charge was then started a few times. These spikes were not the J1772 contacts going in/out. There is a handshake delay before a contactor in the volt pulls in causing the spike.
We are looking at around 2A earth current here.


This more significant event caused an RCD trip...
Well it is 1.5A for 2mS. The RCD should trip.


An iMiEV was checked with the same setup and earth currents were significantly less. e.g.


OK, so the volt RCD tripping is real and caused by spikes at the start of charging.
The spikes are generally in pairs, one positive and one negative. The negative one (in this setup) seemed to be the offender that tripped the RCD. (but probably just the larger one anyway.)
The leakage is not continuous but is short (mS) and large in amplitude
(Amps) no wonder it trips RCDs.

So, what to do.
Ideas please guys.
A common mode ferrite core inductor for the active and neutral may help to balance the spike current in the mS range so the active to earth spike is seen on the neutral as well to partly satisfy the RCD.

edit: topic spelling !

[coulomb]

We did some tests the other day on g4's volt and a selection of EVSE units.

Great test, AC.

The volt itself 'passed' the test and tag tester checks (as a non double insulated appliance). But still caused the occasional RCD trip. The delay and soft start of the volt charging meant that the test and tag tester had completed its checks before the volt charger had powered up. (note here... the test and tag system was unable to see the continuing volt earth leakage fault.)

Interesting. I think that the "soft start" is merely ramping up the charge current slowly, not some electrical circuit in or ahead of the charger. But the delays mean that you can't test the charger itself with a standard tag tester.

This more significant event caused an RCD trip...
Well it is 1.5A for 2mS. The RCD should trip.

The magnitude and duration of the current is indeed significant. As you suggest, this has to be the source of the problem.

So, what to do.
Ideas please guys.

I suggest that this is current into "Y" capacitors at the input to the charger. These are from lines (active and neutral) to earth which are supposed to conduct switching transients to earth and thereby not radiate the switching transients over the power cord.

I did some rough calculations using E = 1/2 . C . V^2 and using a square 1.5 A for 1.5 ms pulse to estimate the energy (at 0.3 mJ). I used 150 V as an "average" voltage across the capacitor. Maybe 200 V might have been more suitable, closer to the RMS voltage of the mains. I'm assuming that the second, much longer spike occurred at closer to the peak of the mains sine wave. This calculation resulted in about 0.3 uF for the Y capacitor, which is near the upper range of what I would expect. If I use 200 V for the mains voltage, I get 0.45 mJ and 0.02 uF.

So maybe the Holden charger has slightly larger Y capacitors than other chargers, and this makes it more susceptible to RCD trips. So the fix for Holden may be to supply smaller Y capacitors in future, if they can still pass EMI tests. But that's not really suitable for Volt owners.

A common mode ferrite core inductor for the active and neutral may help to balance the spike current in the mS range

That's worth trying, I think. It should not reduce the sensitivity to real earth leakage much, as the common mode inductor would have little effect at 50 Hz.

A single inductor in series with the active might also be effective, as it would reduce the amplitude of the spike. But such an inductor would make the power factor very slightly worse, whereas the common mode inductor should have little leakage inductance, and so affect the power factor less.

I think it doesn't matter if the inductor saturates with 10 A AC through it when charging. It should be sized to not saturate with the almost 2 A spikes that can be seen. I suspect that a few turns of wire, with careful attention given to polarity (those things always confuse me), in a toroid about 50 - 70 mm diameter, would do. But that's a wild guess without having done any magnetic calculations. An inductor rated for 10 A (at DC) is about 28 mm diameter, and would be too small to get ~ 2x5 turns of 1.5 mm^2 insulated wire through comfortably.

Would something like that fit inside the Holden EVSE? It would go at the power cord (not J1772) end, of course.

[ Edit: had size of toroid badly wrong. ]

[weber]

Yeah. Good work acmotor. Am I understanding correctly that the same operation most times produces the 40 us doublet we see in the first photo (and doesn't trip the RCD), and only occasionally produces the 2000 us negative spike we see in the second photo (and trips the RCD)?

These two waveforms are so different I would expect them to have completely different causes. The skinny doublet is almost certainly the derivative (i.e. high-pass filtering, AC coupling) of a skinny single spike elsewhere, and is of no concern. But the fat negative spike has a real DC component. Can you theorise about how the conditions differ when the lone spike is produced, and test those theories? Or does it seem purely random?

As well as what Coulomb mentioned, it might relate to battery current rushing in to charge an output capacitor on the charger. In that case you would only expect the RCD trip to happen if it had been off charge for long enough for that cap to discharge. Is that what happens?

[coulomb]

... the 40 us doublet we see in the first photo ...

I meant to say that I thought that the doublet was due to contact bounce on the contactor. But microseconds apart is quite possibly far too fast for a mechanical bounce.

... it might relate to battery current rushing in to charge an output capacitor on the charger.

But why would that current flow in the earth lead?

In that case you would only expect the RCD trip to happen if it had been off charge for long enough for that cap to discharge. Is that what happens?

With the Y capacitor at the input to the charger, their current would depend both on what value the mains is at when the contactor comes on, and also what residual charge the capacitor is at; this would depend on what part of the cycle the contactor switched off last time, and how long since last time. I doubt you could predict anything from that, unless you could measure the capacitor voltage directly.

[acmotor]

Good you guys came on board with this. All thoughts welcome.

GM in the US has not sorted the issue (though RCDs and 230V are less common there). There has even been a suggestion by GM that a non RCD circuit should be used.

Holden in Oz has made no progress/has no idea even though the fault has been reported right through to the US by numerous users.

The RCD trip doesn't happen every charge (or re-plug).
It is not time since last plug in dependent. i.e. 10 seconds re-start can also trip RCD.
The 'doublet' with varying amplitudes occurs on most power ups. The pic was a worst case without trip compared to the iMiEV worst case (a fraction of the size). They are the same size charger (3.3kW) so the iMiEV spike was used as the reference. (The iMiEV never trips an RCD, touch wood. nor does a Leaf AFAIK).
This testing is time consuming as the trip may occur every time or every 10th time allowing a minute per test and the volt may be getting annoyed with us.

The longer time event triggers the scope as well so there may be a second event of usual doublet that we don't see. You're right, the two may be independent. Without further investigation it does all seem random.

Perhaps the turn on phase within the 20ms sine wave is important ? i.e. no zero crossing point switching used at turn on ?

This was the first time we had gone as far as to visualise the earth current with oscilloscope, but next step may be to view the mains voltage waveform on one trace and the spike on the other to see if there is a time relationship ?

Even if some phase relationship shows up, the question is what can be done about it ?

[acmotor]

postal leap frog there.
Common thought train re turn on phase ?

[acmotor]

Just thinking back to filter theory...
The 3 caps arrangements were common mode, bypass and Y.
Bypass caps (to earth) would cause the largest turn on spike in earth current given that the A,N are unbalanced wrt earth anyway.
Common mode caps would cause no earth current and Y caps (common mode + cap the earth would cause 1/4 the earth current spike (assuming same value caps) of bypass filter caps.

[weber]

... it might relate to battery current rushing in to charge an output capacitor on the charger.

But why would that current flow in the earth lead?

When I wrote that I had no idea of exactly how it might relate. I just thought (and still think) that the non-J1772 contactor that acmotor mentioned hearing was likely to be switching the DC side, since J1772 is already switching the AC side. But since you so rightly ask for a mechanism, I have come up with one that also involves your class-Y capacitors. Good thinking re the mains filter.

For other readers: the class-Y capacitors are the capacitors in the mains filter that connect between Active and Earth and between Neutral and Earth, even when they are not connected in a Y configuration as in acmotor's diagram.

The mechanism would be common-mode induction into the active and neutral (but not earth) on the output side of the mains filter. However the lack of any relationship to the time spent off, as reported by acmotor, makes this now seem unlikely.

With the Y capacitor at the input to the charger, their current would depend both on what value the mains is at when the contactor comes on, and also what residual charge the capacitor is at; this would depend on what part of the cycle the contactor switched off last time, and how long since last time. I doubt you could predict anything from that, unless you could measure the capacitor voltage directly.

Mains filters often have bleeder resistors for their caps, so people can't get zapped from appliance plugs. They also often have a common mode choke upstream of the class-Y caps which would prevent this problem if it was due to upstream switching, but not if it was due to induction coming back from downstream.

acmotor, good idea about sampling the mains phase at the same time.

And yes, what to do about it? It might well be possible to get rid of the problem in a safe manner without needing to fully understand how it is occurring.

It certainly can't hurt to try adding a common mode choke on the mains side to see if it makes the problem go away. But if that doesn't work, it probably requires tracing the circuit of the actual mains filter in the charger, and possibly modifying it by adding a common mode choke between the filter and the rest of the charger, or (and this might well be controversial) reducing the size of the Y caps. They could be disconnected temporarily (and/or temporarily increased in value) to prove that this is the problem.

Search for "comon mode" in this page http://www.marcspages.co.uk/pq/3342.htm

[weber]

Typical mains filter:

[acmotor]

OK, that is normal filter / funtion.
Now putting what may be our turn on spike current path....



Will the common mode inductor help ? Phasing of the windings may need to induce opposite current in the neutral line to reduce core balance (RCD) unbalance at the mains end since the RCD doesn't measure earth current, measures A/N unbalance ?
   

[acmotor]

Just a thought too.... I guess the Active and Neutral should be swapped before the EVSE (as they are interchangeable unless something is wrong) as a test to see that the cap to ground (bypass cap) is present and the same on both AC lines. i.e. the volt actually has a symetrical filter.

CyA in the above diagram is perhaps the issue since CyB across N-E has basically no potential across it. A check that CyB exists and is the same value as CyA is what I'm thinking to check. Most likely is but you never know. This is assuming that at least some of the bypass caps are downstream of the contactor that pulls in in the volt once J1772 handshakes are completed.

Hey J, I hope you are following this. We'll need your volt again.

[acmotor]

Also, the RCD trips don't occur (at least rare I think J said) with one particular EVSE, the 'fbox'.
We need to open that fbox EVSE and see if it has any additional (to other EVSEs) filtering components in it. They may be a clue to a fix.

[coulomb]

Will the common mode inductor help ?

Yes, I believe so.

Phasing of the windings may need to induce opposite current in the neutral line ...

Yes, phasing is critical. As shown, common mode current (that flows from active to neutral and nowhere else) induces flux in the direction of the black arrows when active is positive with respect to neutral. With equal turns and equal current, the fluxes cancel and the inductor is effectively a pair of short circuits.

For the current flowing into the Cy A capacitor, the current flows to earth, and not through neutral, so your red current causes the red flux, and there is nothing to balance that red flux. So the red flux causes a voltage spike in the neutral winding, causing current to increase through the Cy B capacitor or other paths (e.g. the right hand Cx capacitor, or the charger load) until it balances the current through the Cy A capacitor. Paths other than through Cy B will affect the current through the top part of the circuit, but it doesn't matter too much since the common mode inductor will keep adjusting to balance the currents. These other currents may delay Cy A charging.

The net effect is, the RCD should see more balanced currents, and not trip. I think it's important for the CyB capacitor to exist and have about the same value as the Cy A capacitor. But I don't think it's worth checking for this, except that swapping active and neutral would be easier than inserting the common mode inductor. I've never seen the Cy capacitors not appear in pairs. I suppose the split phase system in the USA has a lot to do with that.

[ Edit: actually, what may happen is that the inductor presents significant inductance to asymmetrical currents, so it looks more like an open circuit to fast changing asymmetrical currents. This would cause Cy A to charge much more slowly than otherwise, which will still have the effect of reducing the tendency for the RCD to trip. ]

[coulomb]

Also, the RCD trips don't occur (at least rare I think J said) with one particular EVSE...

My guess is that that particular EVSE has a common mode inductor as we have been discussing. But it would be good to confirm this.

If this one has such an inductor but trips are still happening rarely, then the solution would seem to be to increase the value of the inductor (e.g. more turns, and possibly a larger core).

[acmotor]

OK, thanks guys. Will check all this out and try a common mode inductor.
Will report back soon.

[coulomb]

Actually, it occurs to me that an inductor like this could possibly defeat the purpose of having an RCD at all. If a human manages to touch the active inside the car's charger somehow and chassis at the same time, 50 Hz current will flow from active to earth (through the human), and in this case, we don't want the common mode inductor to even out the currents and prevent the RCD from tripping.

I think all will be well since the Cy capacitors are small and the inductor should only have the Cy B capacitor to even out the earth currents through. [ Edit: also, the inductance will be quite small, so the impedance at 50 Hz would be quite low.] But if you install one of these and it seems to work, please make sure the RCD can do its intended job by causing a say 30 mA current to earth from active (through say a ~ 6k8 10W resistor with suitable voltage rating) and make sure it still trips the RCD. It might make sense to wire this resistor before inserting the common mode inductor, to make sure it works and causes sufficient imbalance current to positively trip the RCD.

[acmotor]

Yes, valid point. You have probably answered yourself though as the common mode inductor won't do much at 50Hz. (and may not even fix the volt problem yet )

We have RCD test unit that measures time and trip current and will be running it over the equipment as you quite rightly suggest. We did that to start with to confirm the RCD functions to standard as the Holden folk like to blame the RCDs not the volt.

[g4qber]

Tonight I used the 10A mitsu evse on the volt
Has the evse connected to the jaycar 15a to 10a adaptor
Set volt to 10 amps

Came back to find that rcd/cb had tripped

Put on the clever watts power meter and found that the volt negotiated
10.49A
Thus the jaycar rcd/cb was doing its job and shutting down the charge.

The volt seems to like to draw maximum and over current

The jaycar rcd also tripped a few times on startup.
At least I didn't have to go outside to reset the main rcd.

Finally set the volt to 6A mode
And once again volt goes overboard with 6.33A

Looks like volt may not follow standards
Or
Is only compatible with the evse that Holden supplies

Read elsewhere for what happens when the 2010 imiev uses the Holden evse.

Bottom line is that it is best to install a dedicated 15A circuit for charging
Even Brian m's leaf draws 18A with e-station's 16A fbox. I had to do a double take.

Another weird thing with the volt which could be related to the over amp draw is that on a 16a fbox the volt negotiates a 10A charge.

This is why I've upgraded to e-station's 32A unit. The volt negotiates a 15A charge.
Interestingly the 16A card fbox negotiates a 15A charge.

[coulomb]

Put on the clever watts power meter and found that the volt negotiated
10.49A
... Looks like volt may not follow standards

If you are using the Watts Clever Energy Watts Monitor, it is specified only to 2400 W (10.0 A RMS). It should still be pretty accurate at 10.5 A, but it may not be. Note that its stated accuracy of 0.5% means that a real 10.0 A may read as high as 10.50 A, so really the 10.49 you have measured is within specification.

Certainly it should be able to measure 6 A to to its stated accuracy of 0.5% (which is 0.3 A at 6 A, so a measurement of 6.33 A may actually be as low as 6.03 A).

Even Brian m's leaf draws 18A with e-station's 16A fbox.

Is this also measured with the Watts Clever 10 A meter? If so, it is well outside of its specifications. Whatever iron is in there may well be saturating, for example. Or the shunt could be overheating, increasing the resistance, and hence increasing the voltage drop and therefore the indicated current.

[g4qber]

Thx for the info

I shall try this again tonight with the clipsal 15a power meter. I used the clipsal at Brian m s place as I didn't have the watts clever unit

The WC unit seems to be able to measure up to 16 amps.

I had a previous WC model whose kwh could not be reset. So I used cents per kwh as a work around

on the imiev says 13.18A on WC but 12.688A on clipsal
3558W vs 3148.8W

[g4qber]

CW          Clipsal.                          Difference
10.60 -   10.212.     Mitsu evse.      0.388
10.34 -    9.955.      Holden evse     0.385
10.52 -   10.052.      Mitsu.               0.468
10.50 -   10.126.       Mitsu.              0.374

[g4qber]

elektrobay bollards

Volt engine fault light shows up after charging at
A) subiaco
http://www.recargo.com/sites/11382
B) rac dtec
http://www.recargo.com/sites/9158

This fault light eventually goes away after a while as long as you don't charge at these bollards

[marty11]

Bottom line 15 amp is best practise. Agreed.

Also, for your clipper creek LC 25.. many have suggested to oversize to 40 amp when using a clipper creek on the volt.

(Have read a few post over on GM Volt forum about this same issue. )

Have you tried out the clipper creek on an outlet rated higher than 25 amps yet?

[g4qber]

i might need acmotors help to find a 25 amp circuit with/out RCD.

the interesting thing is that all the EVSEs ive tested work with the LEAF and imiev, but not the volt.
so it seems to point to the (my) volt having the issue.

I was thinking of posting the ECS-20 to you for testing on your volt, but I might still need it here to compare with the 30mA unit that Will is building

[marty11]

ok, you could send it to me, and I will test / send it back.

Maybe A/c or someone might know of a place that you can test the clipper on though.

My standard Holden EVSE has not tripped at 10 amp anywhere, but I have access to recently built motel rooms with trigger happy RCD units, so I might try and charge from one of those rooms, and see if it will trip there. Two places I regularly charge at have older wiring so no RCD to trip.

Also kinda curious whether a 20 amp circuit breaker on a dedicated line would be of any use for BEV's like the Imiev or leaf. (So they could use 15 amp without any probs hopefully)

As a business owner, I might try and provide better than the normal 10 amp, without going to too much expense.

Also interested in the most inexpensive option for places along the highway , other than a weatherproof 10 amp outdoor outlet for ev users to charge from. Would be handy to be informed a little, so I can try to convince a few highway businesses to consider providing a service to EV users.

Either way , if you want to post it to me to test on my Volt, PM me and I will give you postal address.