Conducted immunity test, coupling path?

[KlausKragelund]

Hi

I have a product that has failed conducted immunity tests at 30+ MHz
The product tested it in metal enclosure, and using M12 shield cables (so 360 degrees shield towards the product).

The cable that has failed has differential output, and in the test setup, a unit (AE) is measuring this signal, also with differential input.

The test is conducted by using a 50ohm generator, a 6dB attenuator and a 100 ohms series resistor that feeds current into the shield of the cable (with a clip on the cable, insulation removed). See setup figure attached.
I have reproduced the setup in my own lab, and measured that I can indeed measure a differential voltage on the AE unit.

The signal pair in the cable would ideally be 100% symmetrical, so no voltage should occur, but in practice it seems a signal pair would normally have about 5% mismatch.
I have tried with a cable with double shield, but it does not seem to make a big change.

I have also tried to measure the shield resistance over frequency, but are not seeing big changes in between the cables. Seems that most of the current injected into the cable is due to CM capacitive coupling from the cable to the surroundings.
I have also tried to measure just the cable coupling, from current into the shield to a voltage on the cable pair with a differential probe. Sort of like a poor mans transfer impedance test.

Anyone here have experience with this test and how the current is coupled into the signal pairs?

[selcuk]

The definition in chapter 12.3.4 (Common-to-Differential-Mode Conversion) of the below book is similar to your description:

https://link.springer.com/chapter/10.1007/978-3-031-14186-7_12

I didn't have a similar issue before, but what are the protection components against common mode and differential mode noises at the termination of the cable on your board?

[mtwieg]

Hi

I have a product that has failed conducted immunity tests at 30+ MHz
The product tested it in metal enclosure, and using M12 shield cables (so 360 degrees shield towards the product).

The cable that has failed has differential output, and in the test setup, a unit (AE) is measuring this signal, also with differential input.

The test is conducted by using a 50ohm generator, a 6dB attenuator and a 100 ohms series resistor that feeds current into the shield of the cable (with a clip on the cable, insulation removed). See setup figure attached.
I have reproduced the setup in my own lab, and measured that I can indeed measure a differential voltage on the AE unit.

If so then that means the "Decoupling device" fitted between the direct injection point and the AE apparently isn't doing a good job.

Is that the only issue, or is there some other specific reason you "failed" the test. Keep in mind the pass/fail criteria for these tests are usually tailored for the specific EUT, so we have no way of knowing what "fail" means. But it's usually based on what happens to the EUT, not the AE.

[KlausKragelund]

The definition in chapter 12.3.4 (Common-to-Differential-Mode Conversion) of the below book is similar to your description:

https://link.springer.com/chapter/10.1007/978-3-031-14186-7_12

I didn't have a similar issue before, but what are the protection components against common mode and differential mode noises at the termination of the cable on your board?

During the test, the EUT failed, due to too large deviations seen on the signal at the AE (recording unit). I repeated the test, even just removing the wires from the AE, just lets them hanging. And I can measure a voltage on the pairs of the cable when left both floating and connected to the AE.

So to make it even simpler, I am testing only the cable, injecting a current into the shield and measuring the voltage on the signal pairs. I see a significant signal, so driving the shield with 7Vrms at 25MHz, I see about 100mVrms on the signal pairs. So some coupling from the shield to the pairs is happening. Capacitive and/or magenetic coupling

[KlausKragelund]

Hi

I have a product that has failed conducted immunity tests at 30+ MHz
The product tested it in metal enclosure, and using M12 shield cables (so 360 degrees shield towards the product).

The cable that has failed has differential output, and in the test setup, a unit (AE) is measuring this signal, also with differential input.

The test is conducted by using a 50ohm generator, a 6dB attenuator and a 100 ohms series resistor that feeds current into the shield of the cable (with a clip on the cable, insulation removed). See setup figure attached.
I have reproduced the setup in my own lab, and measured that I can indeed measure a differential voltage on the AE unit.

If so then that means the "Decoupling device" fitted between the direct injection point and the AE apparently isn't doing a good job.

Is that the only issue, or is there some other specific reason you "failed" the test. Keep in mind the pass/fail criteria for these tests are usually tailored for the specific EUT, so we have no way of knowing what "fail" means. But it's usually based on what happens to the EUT, not the AE.

The test failed since the signal at the AE end deviated by 1% as defined in our spec for the max allowed deviation.

In the standard the decoupling device is defined as in the attached picture. It should create a defined impedance towards the AE. Still, since the resistance injected ti 150ohms, that decoupling impedance needs to be high to matter, and fi you wind as described, a big inductor can have parasitic capacitance, so it might not be all that good.

I am suspecting that we are in fact just testing the cables and the immunity of the AE instead of the EUT. The separate tests of cables alone seems to support that claim

[mtwieg]

Thanks for the clarification. I'm assuming this error signal you're seeing is linearly related to the injected interference signal (i.e. same frequency, and has proportional amplitude).

Is it possible that the differential voltage you're seeing is truly there, and not just a measurement artifact (for example, poor CMRR on a differential probe)?

I would try repeating the test while doing the injection at different sites:
1. At the GND/shield pin of the connector on the EUT side
2. The EUT enclosure (which I'm guessing is connected to circuit GND)

I've not worked with circular cable assemblies very much, but it wouldn't surprise me if the connection of the termination of the shield at the connectors is the issue. For example, in a lot of cheap USP cables the braid/foil does not actually connect to the connector shell, instead the connection is made with a single drain wire. This means that for a short length, all the "shield" currents are carried by just another wire, which may couple to the other signal wires in bad ways.

[tszaboo]

I am suspecting that we are in fact just testing the cables and the immunity of the AE instead of the EUT. The separate tests of cables alone seems to support that claim

You very well might be doing that. I've failed EMC testing for immunity at 100MHz for a 100mbit Ethernet device.
Ask yourself, if you are selling the product including the cable. And if the failure is representative to real world conditions, eg, that's how they connect your EUT in the real world.
And for these common mode noise problems, I also found that capacitor tolerances are super important. You have two capacitors on the two data/signal/ADC lines, and they are 25% then at high frequency the CMRR goes down because you filter one side better than the other

[KlausKragelund]

I am suspecting that we are in fact just testing the cables and the immunity of the AE instead of the EUT. The separate tests of cables alone seems to support that claim

You very well might be doing that. I've failed EMC testing for immunity at 100MHz for a 100mbit Ethernet device.
Ask yourself, if you are selling the product including the cable. And if the failure is representative to real world conditions, eg, that's how they connect your EUT in the real world.
And for these common mode noise problems, I also found that capacitor tolerances are super important. You have two capacitors on the two data/signal/ADC lines, and they are 25% then at high frequency the CMRR goes down because you filter one side better than the other

I have seen something similar. A typical differential input stage has capacitors to limit bandwidth. But if the 2 caps are not equal, then it shows up as lower CMRR at higher frequencies

[KlausKragelund]

I was at the testing lab friday, to debug the system.
The test was run at 20Vrms, and at 150kHz, the voltage at the DUT was 35Vpeak (I measured with a scope to check the operation during test).

The immunity test injects a signal onto the shield about 15cm away from the DUT. (by scraping off the insulation of the cable)

I get an error below 1MHz, large signal on the differential pair, which seems to overdrive the output amplifier of the DUT, so that the signal gets DC offset.
The DUT has shielded cables, and shielded enclosure, so just 15cm of shield coupled to the inner wires created large voltage, about 15Vpp on the signal pair.

I then went to the lab at home, used a Bode 100 to measure the gain from the shield to the signal wire, only the cable, no DUT. I did that by injecting onto the shield 15cm away with the shield as reference. Then measured on the output side, the voltage on a single wire developed across the entire length of the cable.

I also did a sanity check, shorting the injection point, just to be sure that the setup did not couple from wire input to output. All good.

So the results show that the coupling from cable shield to inner wire is large, all the way from below 100kHz to quite high frequency, with a nasty resonance at 2.7MHz (actually one point that was shown to be an issue in the fiorst conducted immunity test)

Has any of you guys done similar test before, I was quite taken back that the coupling is so high?

Setup and measurement is attached

[mtwieg]

I was at the testing lab friday, to debug the system.
The test was run at 20Vrms, and at 150kHz, the voltage at the DUT was 35Vpeak (I measured with a scope to check the operation during test).

The immunity test injects a signal onto the shield about 15cm away from the DUT. (by scraping off the insulation of the cable)

I get an error below 1MHz, large signal on the differential pair, which seems to overdrive the output amplifier of the DUT, so that the signal gets DC offset.

Yes, if the injected signal takes the driving amplifier in the DUT out of its linear range, then developing a differential voltage isn't surprising.

The DUT has shielded cables, and shielded enclosure, so just 15cm of shield coupled to the inner wires created large voltage, about 15Vpp on the signal pair.

I then went to the lab at home, used a Bode 100 to measure the gain from the shield to the signal wire, only the cable, no DUT. I did that by injecting onto the shield 15cm away with the shield as reference. Then measured on the output side, the voltage on a single wire developed across the entire length of the cable.

I also did a sanity check, shorting the injection point, just to be sure that the setup did not couple from wire input to output. All good.

So the results show that the coupling from cable shield to inner wire is large, all the way from below 100kHz to quite high frequency, with a nasty resonance at 2.7MHz (actually one point that was shown to be an issue in the fiorst conducted immunity test)

Has any of you guys done similar test before, I was quite taken back that the coupling is so high?

Setup and measurement is attached

The low frequency (<1MHz) response seems straight out of a textbook, nothing surprising there (usually the corner frequency is <5kHz but I'm guessing it appears higher in your case due to the output impedance of the bode100). I'm wondering what you expected it to look like. Not sure about the other modes at >1MHz. How long is the cable?

But I'm not sure this measurement tells you anything useful about your immunity issue. First of all, the method of injection is very different from the actual immunity test. And so what if voltage if a voltage is induced between two ends of the cable? So long as it's induced equally on all conductors (including the shield which is connected to your DUT's circuit GND), then the DUT's driving amplifier should not "see" it.

Could you clarify the path by which the DUT's circuit GND is connected to the reference GND plane (the GND of the interference source)? That's fundamental to diagnosing the issue.

[KlausKragelund]

I was at the testing lab friday, to debug the system.
The test was run at 20Vrms, and at 150kHz, the voltage at the DUT was 35Vpeak (I measured with a scope to check the operation during test).

The immunity test injects a signal onto the shield about 15cm away from the DUT. (by scraping off the insulation of the cable)

I get an error below 1MHz, large signal on the differential pair, which seems to overdrive the output amplifier of the DUT, so that the signal gets DC offset.

Yes, if the injected signal takes the driving amplifier in the DUT out of its linear range, then developing a differential voltage isn't surprising.

Yes, it also shows up in simulations. When the injected current is higher than the current capability of the opamp, it saturates in an unsymmetric way, so the resulting DC is shifted due to lowpass filtering.

Replacing the opamp with a higher current type should solve the issue, or adding a NPN/PNP follower stage to boost the current capability

The DUT has shielded cables, and shielded enclosure, so just 15cm of shield coupled to the inner wires created large voltage, about 15Vpp on the signal pair.


I then went to the lab at home, used a Bode 100 to measure the gain from the shield to the signal wire, only the cable, no DUT. I did that by injecting onto the shield 15cm away with the shield as reference. Then measured on the output side, the voltage on a single wire developed across the entire length of the cable.

I also did a sanity check, shorting the injection point, just to be sure that the setup did not couple from wire input to output. All good.

So the results show that the coupling from cable shield to inner wire is large, all the way from below 100kHz to quite high frequency, with a nasty resonance at 2.7MHz (actually one point that was shown to be an issue in the fiorst conducted immunity test)

Has any of you guys done similar test before, I was quite taken back that the coupling is so high?

Setup and measurement is attached

The low frequency (<1MHz) response seems straight out of a textbook, nothing surprising there (usually the corner frequency is <5kHz but I'm guessing it appears higher in your case due to the output impedance of the bode100). I'm wondering what you expected it to look like. Not sure about the other modes at >1MHz. How long is the cable?

But I'm not sure this measurement tells you anything useful about your immunity issue. First of all, the method of injection is very different from the actual immunity test. And so what if voltage if a voltage is induced between two ends of the cable? So long as it's induced equally on all conductors (including the shield which is connected to your DUT's circuit GND), then the DUT's driving amplifier should not "see" it.

Could you clarify the path by which the DUT's circuit GND is connected to the reference GND plane (the GND of the interference source)? That's fundamental to diagnosing the issue.

The cable is 4m long. During the test, the injected current is only on the first 15cm of the cable. I did not expect so high coupling, but guess it's more or less just a single turn transformer, with some high capacitance coupling also. The coupling only happens on those 15cm though.

The reason I used this test, was to dig into how the current on the shield actually produced a voltage on the differential pair towards the AE, and how it couples current into the EUT/DUT.

I see two ways:

1. The current onto the shield generates a voltage over the length of the wire of the differential pair. That produces a voltage at the far end (at the AE), which is LP filtered and should still keep the DC level. But, since a differential pair normally has 5% mismatch, some of the developed voltage is actually converted into differential voltage instead of common mode.

2. The current on the shield couples to the internal wires, and a part of that current is pushed into the output of the amplifier in the DUT. As written above, the amplifier saturates and developes DC voltage shift. I would have expected the current into the output of the opamp to be small, since the AE side is high impedance, so the voltage would be developed at the AE side instead of the DUT side, and the current into the opamp would be small

It seems that there is a large current into the opamp. Will be masuring that next.

As for the DUT, the shield of the DUT is connected to GND of the DUT, with parrallel connection of a resistor, tranzorb and a lot of small ceramic capacitors. Also the DUT output has a 1nF capacitor to shield, so in case of a transient voltage on the signal wires, the current is diverted into the shield instead of into the opamp.

If I increase the size of the DUT capacitance to shield, then the simulations shows improvement, but I cannot increase it much due to the system will then have a too low bandwidth wrt the actual signal.

Attachment shows the output stage. TL082 can only supply 26mA. So a BCP53/BCP56 stage could boost that.
Attachment also shows how the injected signal is put onto the shield towards the DUT

[mtwieg]

The cable is 4m long.

Then I don't think transmission line behavior of the cable can explain the modes <10MHz... must be related to some external capacitance somewhere... not sure it's worth investigating though.

During the test, the injected current is only on the first 15cm of the cable. I did not expect so high coupling, but guess it's more or less just a single turn transformer, with some high capacitance coupling also.

Correct, in that simple setup it's just a transformer, and also the resistance of the shield and source which forms that cutoff around 20kHz.

The coupling only happens on those 15cm though.

The reason I used this test, was to dig into how the current on the shield actually produced a voltage on the differential pair towards the AE, and how it couples current into the EUT/DUT.

I see two ways:

1. The current onto the shield generates a voltage over the length of the wire of the differential pair. That produces a voltage at the far end (at the AE), which is LP filtered and should still keep the DC level. But, since a differential pair normally has 5% mismatch, some of the developed voltage is actually converted into differential voltage instead of common mode.

2. The current on the shield couples to the internal wires, and a part of that current is pushed into the output of the amplifier in the DUT. As written above, the amplifier saturates and developes DC voltage shift. I would have expected the current into the output of the opamp to be small, since the AE side is high impedance, so the voltage would be developed at the AE side instead of the DUT side, and the current into the opamp would be small

I think these are both plausible. But I still question whether doing the injection in the manner you described above (reply #8) is wise, as it's very different from how it's done in the actual immunity test. It's possible that mitigations for one setup won't work for the other...

As for the DUT, the shield of the DUT is connected to GND of the DUT, with parrallel connection of a resistor, tranzorb and a lot of small ceramic capacitors.

Having trouble parsing this... the shield of the EUT is connected directly to circuit GND, then what are the other components in parallel with...?

Also the DUT output has a 1nF capacitor to shield

You mean the DUT shield or the cable shield? Not sure if the cable shield connects directly to the DUT enclosure shield.

Can you show these connections on the schematic?

so in case of a transient voltage on the signal wires, the current is diverted into the shield instead of into the opamp.

Not sure what sort of event could cause such a transient within the DUT, but this also means shield transients/EMI may also transfer to the signal wires. Seems counterproductive, especially if you happen to be having a conducted immunity issue...

If I increase the size of the DUT capacitance to shield, then the simulations shows improvement

That's counterintuitive. I'm guessing you're adding equal amounts of capacitance to both of the drivers. Perhaps doing so causes the coupling to "match" better between the drivers, causing the differential perturbation to decrease. But the common more perturbation is likely increasing a lot more... I'd still expect this to cause problems.

[khs]

I was at the testing lab friday, to debug the system.
The test was run at 20Vrms, and at 150kHz, the voltage at the DUT was 35Vpeak (I measured with a scope to check the operation during test).

The immunity test injects a signal onto the shield about 15cm away from the DUT. (by scraping off the insulation of the cable)

I get an error below 1MHz, large signal on the differential pair, which seems to overdrive the output amplifier of the DUT, so that the signal gets DC offset.

Yes, if the injected signal takes the driving amplifier in the DUT out of its linear range, then developing a differential voltage isn't surprising.

Yes, it also shows up in simulations. When the injected current is higher than the current capability of the opamp, it saturates in an unsymmetric way, so the resulting DC is shifted due to lowpass filtering.

Replacing the opamp with a higher current type should solve the issue, or adding a NPN/PNP follower stage to boost the current capability

The DUT has shielded cables, and shielded enclosure, so just 15cm of shield coupled to the inner wires created large voltage, about 15Vpp on the signal pair.


I then went to the lab at home, used a Bode 100 to measure the gain from the shield to the signal wire, only the cable, no DUT. I did that by injecting onto the shield 15cm away with the shield as reference. Then measured on the output side, the voltage on a single wire developed across the entire length of the cable.

I also did a sanity check, shorting the injection point, just to be sure that the setup did not couple from wire input to output. All good.

So the results show that the coupling from cable shield to inner wire is large, all the way from below 100kHz to quite high frequency, with a nasty resonance at 2.7MHz (actually one point that was shown to be an issue in the fiorst conducted immunity test)

Has any of you guys done similar test before, I was quite taken back that the coupling is so high?

Setup and measurement is attached

The low frequency (<1MHz) response seems straight out of a textbook, nothing surprising there (usually the corner frequency is <5kHz but I'm guessing it appears higher in your case due to the output impedance of the bode100). I'm wondering what you expected it to look like. Not sure about the other modes at >1MHz. How long is the cable?

But I'm not sure this measurement tells you anything useful about your immunity issue. First of all, the method of injection is very different from the actual immunity test. And so what if voltage if a voltage is induced between two ends of the cable? So long as it's induced equally on all conductors (including the shield which is connected to your DUT's circuit GND), then the DUT's driving amplifier should not "see" it.

Could you clarify the path by which the DUT's circuit GND is connected to the reference GND plane (the GND of the interference source)? That's fundamental to diagnosing the issue.

The cable is 4m long. During the test, the injected current is only on the first 15cm of the cable. I did not expect so high coupling, but guess it's more or less just a single turn transformer, with some high capacitance coupling also. The coupling only happens on those 15cm though.

The reason I used this test, was to dig into how the current on the shield actually produced a voltage on the differential pair towards the AE, and how it couples current into the EUT/DUT.

I see two ways:

1. The current onto the shield generates a voltage over the length of the wire of the differential pair. That produces a voltage at the far end (at the AE), which is LP filtered and should still keep the DC level. But, since a differential pair normally has 5% mismatch, some of the developed voltage is actually converted into differential voltage instead of common mode.

2. The current on the shield couples to the internal wires, and a part of that current is pushed into the output of the amplifier in the DUT. As written above, the amplifier saturates and developes DC voltage shift. I would have expected the current into the output of the opamp to be small, since the AE side is high impedance, so the voltage would be developed at the AE side instead of the DUT side, and the current into the opamp would be small

It seems that there is a large current into the opamp. Will be masuring that next.

As for the DUT, the shield of the DUT is connected to GND of the DUT, with parrallel connection of a resistor, tranzorb and a lot of small ceramic capacitors. Also the DUT output has a 1nF capacitor to shield, so in case of a transient voltage on the signal wires, the current is diverted into the shield instead of into the opamp.

If I increase the size of the DUT capacitance to shield, then the simulations shows improvement, but I cannot increase it much due to the system will then have a too low bandwidth wrt the actual signal.

Attachment shows the output stage. TL082 can only supply 26mA. So a BCP53/BCP56 stage could boost that.
Attachment also shows how the injected signal is put onto the shield towards the DUT

You may reduce the voltages your clipping diodes are connected, because they may work due to the voltage drop of the diodes,
so some current flows always via your opamp..