2023 03 02 HW EMC A

Silence.
Good morning. Good morning,
Mr. Frank, is your connection okay now?
Yes. No. Yes.
I had to rejoin me because my microphone is not right working, but I see that's fine.
Yulia, hi. Hi. Good morning. Morning. Morning.
As I said, Yulia, we can skip your topics for this week, if you want.
Yeah, join only for one topic. Please.
We can start with your topic and prioritize this instead of the EMC topic.
Yeah, this is my topic.
I have to join too, Rob, but start your explanation.
Okay. Okay.
I think we need Julia's support.
Okay.
Please, then.
So, then, if it's okay for everyone, I will take the screen and share the presentation I've created.
Yes, please. Please. So, let's see.
Let's start with the introduction here.
We had first only tests 1 and 2 performed with 150 millimeters position for BCI,
and we have reported we can pass the test.
Then in the turn of the in-rush current modification,
We went to further testing and we decided also,
okay, it would be good to check the further positions.
And then we were surprised and not amused that we have failures at 450 and 750 millimeters position.
So there we failed repeatedly, not every time.
And we have then also re-performed this last position,
several 50 millimeter with another hardware without modification,
and we found, okay, this one has a failure as well.
So what happens is then the question.
I've taken some screens from typical tests.
I don't want to share all the approximately 100 pages of reporting.
So, typical test, this one was at 150 mm, terminal 30 and CAN in the loop, not terminal 40.
We see some disturbances of the cell voltages, we see some disturbances of the current,
but the circuit breaker status, which is supervised here, is OK.
The only thing that toggles is the redundant power supply.
So, this is the behavior we have expected, that is fine, but this is 150 mm.
Current deviation typically less than 200 mA and 10-20 mV on the cell voltages at a certain resonance frequency,
which is around or between 360 and 380 MHz.
So, the next slide has the 750 mm position, I concentrate on this one,
because we have done then most of the test activities with the clamps at 750 mm.
Also, terminal 30 and CAN in the loop, the voltage disturbance is a bit higher than before.
The current is also, it's more easy to see, it's higher than before,
but it's still within our set, within the limits of the test plan.
But here we see at 376 MHz the circuit breaker opens.
And the next, that's not good, and the next test with terminal 40 in the clamp, in the injection clamp,
we see a really high deviation of the cell voltage.
Could be that at some time the communication is lost to the ASIC and then regained.
So, we might need to check in detail.
The current deviation is above 500 mA, so we would call this a failure from the current deviation.
And also the circuit breaker opens.
And here there's obviously also a mechanism that it tries to close again and then opens again.
So, I'm not sure what exactly happens.
This is why we have Julia in this round.
And we have the same frequency, basically.
One question, Robert, sorry for interrupting.
Also, on this slide, as well as on the previous slides, I see EMC, BCI, C1 introduction.
Is it still C1 or is it actually the modified C1 with the C2 sample changes?
This one is the DUT10, so this is a modified one.
But we can still call it C1 because when we check now the unmodified sample for comparison,
we see in general a similar behavior.
So, there might be, we thought in the beginning the modified one is a bit more stable,
but I come to that later.
Okay.
It seems not to be true.
So, basically, this behavior happens for both C1 unmodified and C1 modified, but it is not fully reproducible.
This comes then to more details in the test results.
I don't show any further curves.
So, if you want to go back to the logs and plots, we need to go to the beginning and because it looks quite similar each time.
So, what we did, we first went back to an old hardware version.
We still have one DUT alive with the C0 plus revision D.
We only built 10 of that, which has an old software version.
And with this one, we can pass the test at 750 millimeters.
We did not do further testing because it seems there is some issue with that sample.
Yeah, but it did not affect this test result, but let's not stress it.
Maybe we need it again for something else.
So, this one then was at a stop, but we could test both terminals at 750 millimeters and C deviation,
which is at the limit for current, but no circuit breaker opening.
Then we tried a software swap.
We have unfortunately some compatibility topic that the old software does not run on the C1.
So, we could not test the software version one-to-one.
And we went only from the 724 to the 730 and did a lot of tests to check if there's an influence.
In the end, I would say between those two software versions, there's no real difference in the behavior.
Okay, so it leaves us just with the hardware.
Yeah, not really.
We could not swap to software six because this could not be installed on the C1,
and we didn't want to mess around too much with the old sample and try a new software on that.
Maybe, yeah, we might decide together with Julia if we find a way to do that.
But so far, we were limited here.
Then we played with the hardware a lot with both DUTs, and we found during the testing,
we had varying results.
Sometimes we can pass, sometimes we fail, and the C1 DUT10 seemed a bit more stable.
But in the end, it was not really reproducible.
The same DUT would, after exchanging, would fail where it passed before.
So then somehow we got the idea, maybe it's related to the mounting.
Because each time we change a sample or we flash, we open the battery,
and if we exchange the DUT, we need to loosen all the screws,
the terminals, connections, and the internal wiring harnesses.
I think I already understand what's there.
And then we thought, okay, our DUT10, usually it should have a fixation for the internal cell harness.
For this, it was broken, too much handling.
We didn't really pay a lot of attention because, well, it's a development sample, who cares?
In the end, we have to care about that because if you then put it in the proper position
by this distance holder, it's, yeah, plus minus one millimeter, correct position, we reproducibly fail.
And yeah, for the DUT11, this wiring harness fixation is okay.
We put it there, we fail.
So there is obviously an influence here on how we place the harness.
And we have here a fixation point, which cannot be really seen.
We have one here on the BMS itself.
It's a point, so we have some rotation freedom.
And we also have another drawback that is not really shown here.
Here, we have the opening of the connector to the internal wiring.
And because we are working a lot with the same battery and the same harness,
we need to pull this one up.
And yeah, after several tries, the cable is not straight anymore here.
So it almost touches the busbar, it comes very close to it.
But I did not find an influence there, we tested.
So maybe it's important, maybe not.
I could not recreate an influence here, if we force a distance or not.
Anyways, we then thought, okay, we have an influence with this harness, what can we do?
And the first is, let's make it reproducible.
Try this way to force the wiring harness into the proper position.
We fail.
This is what we say here, reproducible fail.
Then we put the wiring very close to the housing here.
We have passes, but we are close, so it's not very good.
The circuit breakers stayed close for these tests.
We'd made three tests.
Let's talk about reproducibility and some repeated testing.
But in one of the tests, the current was too high, the current deviation.
But anyways, the behavior is better, more stable.
And reproducible.
And if we do not open the battery, reproducible.
So then the next way what we do in EMC is we try a shield on this harness.
I guess it's not really, how to say, not something we would like to put into series,
but we made the copper foil around this harness and in the first attempt, if we magnify this a bit,
we see it has a direct contact copper shielding to the battery housing.
To the housing, okay. Yeah.
And with this setup, we have passed the test for both terminals, even in over-testing condition.
Usually our injection current is 200 milliamps according to BMW norm,
which is much higher than all the other norms I know.
And but this way we have passed with 250.
So we know, we did this to, to, to be sure we are stable at 200.
The shield is only connected at that point.
It's not connected to ground or KL41 or something like this. Exactly.
It's not connected to the wiring, only to the housing here.
And yeah, we had some discussion if we should try it later on,
but I don't think it is really a good idea to connect it to the wiring, to our local ground,
because I think this is where exactly our noise couples in.
But one question here.
So where the harness different with the C0 plus where we passed the test?
No, we use the same battery, the same harness. Okay.
So, so what I can say is shielding makes the measurement values more stable.
I have a table later on where we can see it.
But that's not the root cause then.
And the circuit breaker, we still need to further detail what's happening.
At least for our measurement values, cell voltages and current,
we have a real improvement if we make a shielding and suppress the noise here.
Let me try a remark.
So this housing is from my point of view, not connected to any grounds.
It's more or less floating.
This housing is electrically connected to the L-shape, but the L-shape is floating.
We have, in this case, not a glue, but we have still an isolation area
with an isolating tape between our power PCB and the L-shape.
Are you going to try to find out the root cause?
Yeah, we did further tests.
Okay.
Yeah, we did not stop at this point, but for us, it was important to see,
okay, we have a lot of noise coupling in through this harness.
Maybe let me continue the second test.
We have a shield without connection to a housing.
This is not successful.
So this shielded harness, but without electrical connection to the battery housing,
we fail, even at the nominal test current.
This means that basically the shielding gives us like a path, which basically leads to,
let's say, the pulse towards ground and dissipates basically over the housing.
Yes, yeah, exactly.
And the second attempt was then we put the harness in the proper position and have a copper tape to the battery housing.
And this works again, as well as the first try.
So both two positions connect with copper tape and we can pass.
So we need to suppress the noise, for example, by putting it away to the housing. Good.
We did not stop here.
Then since we found we have some topic here, especially how we place the cable, not even the shielding.
Before we go to hardware options, we tried what happens with the communication harness.
Because each time we close the battery, the cover is not transparent and we cannot see
exactly where's the position.
And since we want to test fast, of course, what we did is, yeah, we mounted it without this orange protection cap.
Then the communication harness lies usually between the BAT plus terminal and here's would be the terminal 40.
Okay. Interesting. Yeah.
And this one is then after checking internally, also with all mechanical colleagues, is not the intended position.
And it's also not a good location.
When we put it between BAT minus, BAT plus, yeah, we had even a pass with terminal 30.
So the noise that comes in this harness, maybe it does not couple as good anymore to the cell harness.
If we put it here above the pole protection as it should be in Sirius, then we have a similar behavior.
It seems better on terminal 30, but still it's not 100% reproducible.
Anyways, since this is the Sirius design with pole protection mounted, we did all the further tests in this setup.
We still have a margin where this harness is in the end and I cannot really check it.
This is the moment before closing.
And also in this design, somehow it is difficult to close, but maybe we come to that later.
I think there's also a change at EV side for this harness, so we can come to this later.
Then let's see what's happening inside the battery.
We see here is the connector, we have the wiring internally to the cells along this track and also here,
and there are some NTCs, I think it's here and here, there's an NTC position.
And this is just where the terminal 40 or the BAT plus busbar goes above the battery stack.
So it really goes here, then goes here to the battery plus pole.
So it's a long way, you will say.
Yeah, it's a long way and some chance physically for noise over coupling.
What's the distance between the busbar and in the height, let's say the 60 centimeters?
No, no, no, in the height. Not now.
The height maybe... 64 centimeters?
Not that much, no less. Maximum is centimeter.
So the busbar is very close above this cover and there's not much distance here.
Of course, it's understandable because you want to use all the space you have effectively.
Yeah, sure.
But for sure, we have some potential coupling here, radiated from the busbar to the inner wiring.
And with this in mind, we went back to the hardware and thought about what can we do.
In the schematics, I don't have it fully open right now, but what we have is six NTCs inside the cell stack.
These NTCs share a common ground and this is the ground where the cell ASIC is on.
And yeah, we saw if we suppress the noise from the whole cell harness, we have an improvement.
So put a ferrite into the local NTC ground.
It's only the ground here at NTCs.
It does not affect the cell ASIC here.
So basically, just an additional ferrite on the ground path.
Yes, and since we could not do it in the layout because there is no pad and whatever, we had to put it in the harness.
So now we have one modified harness with a ferrite.
This is not alone effective enough.
So with this test, we had fails at both terminals.
Then we saw, OK, we have ESDH protection capacitors for all the return paths of the NTCs.
We have removed those because if we have a coupling into the harness,
we have six wires for the NTCs and the capacitor comes without resistor in front, so it can couple directly into the ground.
With both measures, we still failed at terminal 40, but the terminal 30 was passed and the noise level goes down a bit.
And then we went back to an original harness without ferrite.
And here you see it was the opposite way.
So again, we are at a point where we say the behavior is still not 100% reproducible because
we are at the limit what our design can somehow withstand, usually even above.
So anyways, the circuit breaker behavior was not reproducible, but the measurement values,
the cell voltages and currents are improved by any of these measures.
I have a table later on where we can check this.
Usually, this is the topic where Julia comes into play.