EMC Testing Setup For Automotive Products

[Emanuel]

There is something that I can't understand, perhaps someone will be kind enough to explain.
The usual EMC test setup for conducted and radiated emissions require a table with a conductive plate, LISN and a power supply usually placed outside the test chamber.
The DUT housing, if has a metal enclosure, is attached to the table with a grounding strap. The power cable from the DUT to the LISN, if shielded is also grounded near the LISN.
So far so good, but when I say grounded it is in fact earthed.
This is not what happens in a car, where the chassis practically is the battery -.
The customer requires the power to be supplied by 2 wires and the metal enclosure not connected to the -, but when installed in the vehicle the enclosure will be connected to the car's chassis.
I know that an EMC test is supposed to be made in a setup identical to the real installation, but the standard test setup in this case is not reflecting the final use.
The reason I am asking is that I have a strange problem - While testing the DUT without earthing it the conducted emissions are not ideal but below the limits. But the moment I connect the grounding strap they go up way above the limits.   

[xavier60]

Maybe something that is capacitively coupling HF current to the housing like a switching transistor using the housing as a heat sink.

[Emanuel]

No, nothing like that, everything is SMT and placed on one side of the PCB only.

[jonpaul]

please post photos of test set-up, Q PCB, schematic or blocked diagramme
Testing EMI compliance of,véhicule PSU and,electronics is very well documented
Customer should specify the applicable EMI regulations or standard.

Eg Automotive Avaionics, Marine.
Régulations different for régions, USA, CA, EU, UK.

Bon journée

Jon

[Emanuel]

Well, the question here was about the test setup in principle, as I tried to explain.
The standard setup is with the enclosure and cable shield earthed, while in the real life, in a car, they are connected to the battery -, same as the circuit's ground.
About the specific test I am making and the strange results I think that I will create a separate topic.

[JPortici]

If you are affiliated with a vehichle manufacturer, they all have a very thick document in which all tests to perform are described, ask for said text and follow the guidelines. guidelines say earth the thing? you do that and don't worry about what really happens in the car.
Otherwise you have a certain degree of freedom anyway in how to determine if the test has passed or not. You could go the easy way and specify that the enclosure must be left floating, problem solved.

[Niklas]

There are pictures of the different setups (remote or local grounding) in the CISPR25 standard. Both LISNs should be placed on the reference metal plate that is also under the DUT.
One thing to test is if the emission pattern changes if you connect the PCBA 0V to the metal housing through capacitors. This will break up the DC path, but still give an AC path. 100nF with at least 2x the battery voltage in rating, but you might need more if the customer has another requirement. Sometimes they ask for 500V insulation test voltage between the housing and the supply pins. Could also place a 100k resistor in parallel with one of the capacitors to bleed out electrical charges.

[T3sl4co1l]

Tying the enclosure to the plane is equivalent to mounting it to the chassis, so the setup sounds reasonable.

Sounds like common mode with respect to the enclosure: differential would be present regardless of connection, or one line drops off when grounded but the other remains.  If a CM/DM network is available / can be made, that may give a better idea of which frequencies are coming through this way.

Tim

[Emanuel]

If you are affiliated with a vehichle manufacturer, they all have a very thick document in which all tests to perform are described, ask for said text and follow the guidelines. guidelines say earth the thing? you do that and don't worry about what really happens in the car.
Otherwise you have a certain degree of freedom anyway in how to determine if the test has passed or not. You could go the easy way and specify that the enclosure must be left floating, problem solved.

Well, in this case it is a military vehicle, required to pass MIL-STD-461. They always use a shielded cable and always connect the housing to the vehicle's chassis with a grounding strap.

[Emanuel]

There are pictures of the different setups (remote or local grounding) in the CISPR25 standard. Both LISNs should be placed on the reference metal plate that is also under the DUT.

The required standard is MIL-STD-461, for the conducted emissions test it resembles CISPR22 but stops at 30MHz

One thing to test is if the emission pattern changes if you connect the PCBA 0V to the metal housing through capacitors. This will break up the DC path, but still give an AC path. 100nF with at least 2x the battery voltage in rating, but you might need more if the customer has another requirement. Sometimes they ask for 500V insulation test voltage between the housing and the supply pins. Could also place a 100k resistor in parallel with one of the capacitors to bleed out electrical charges.

The customer recommends an RC snubber between the 0V and the metal housing, R=1ohm C=680nF. But if there is another setup with different values or in parallel I will be glad to hear.

[Emanuel]

Tying the enclosure to the plane is equivalent to mounting it to the chassis, so the setup sounds reasonable.

Sounds like common mode with respect to the enclosure: differential would be present regardless of connection, or one line drops off when grounded but the other remains.  If a CM/DM network is available / can be made, that may give a better idea of which frequencies are coming through this way.

Tim

I suspect that the problem comes from a different place.
The PCB is an aluminum substrate type, single layer. I don't have a bottom layer as ground plane, and the aluminum substrate is connected to the housing.
Maybe this explains why the filter is not performing as well as expected. But I made a lot of projects using this setup and all passed the tests. In fact I have a similar product but with a different buck converter which works perfectly.
Is it possible that the aluminum substrate causes this EMI increase as a result of (maybe) current passing through it?

[xavier60]

Because metal clad PCBs usually have a very thin dielectric layer compared to a typical fiberglass PCB, every track and pad has a much higher capacitance to the metal substrate. So, every track that carries fast voltage signals will be causing significant common mode EMI currents back out of the power input, assuming that there is adequate bypass capacitance between the input power rails. 
At a minimum, Y capacitance is needed and possibly a Common Mode Choke.
https://www.etechnog.com/2021/08/what-is-y-capacitor-applications-uses.html That's a very annoying webpage, try this one, https://www.emcbayswater.com.au/blog/emc-testing/conducted-emi-emissions-typical-problems-common-solutions/

[xavier60]

This better explains the function of Y caps, page 12.
http://www.interferencetechnology.com/wp-content/uploads/2015/10/EMC-Introduction-to-Clean-and-Safe-Power-Final-3.27.2015.pdf

[Emanuel]

This better explains the function of Y caps, page 12.
http://www.interferencetechnology.com/wp-content/uploads/2015/10/EMC-Introduction-to-Clean-and-Safe-Power-Final-3.27.2015.pdf

Thanks, I will try that. As I mentioned, in the specifications they recommended an RC snubber to be placed between the 0V and chassis, with 1ohm and 680nF.
In the links they are talking about offline circuits and I know that the Y caps are used in these applications. The question is if in a DC powered device they will have the same effect.
Yes, I am considering also a common mode choke, have to find room for it.

[T3sl4co1l]

Do you have easy contact to the metal core (vias)?  If nothing else, even if you have to bolt a strap to ground, that's workable.

Simply treat the enclosure as ground plane.  There will be some (AC) voltage drop along your circuit ground paths; these can be minimized with respect to the outside by single-point grounding (i.e. to the chassis) at the last point in the filter.

If nothing else, all the noise within, can be kept there by filtering each line (+V and GND) with respect to chassis.  This loses the advantage of a CMC (the single inductors will be somewhat bigger), but will definitely work.

If there are any other connections to the circuit, you'll have to consider the common mode between connectors.

Tim

[Emanuel]

Here is an update:
As xavier60 recommended I placed 1uF Y caps after the PI filter and the result is surprising, all the harmonics peaks were minimized.
Right now everything is just below the CE102 limits without the 20dB attenuator, with the attenuator looks just perfect, a clean line with a single peak at the buck frequency of 200kHz which in any case is located on the limit slope where the level is higher than the flat portion.
I tried also with an SRF0905-102Y CM choke but no serious improvement was obtained.
Now to the hard part, radiated emissions RE102, but this is not something I can check in my lab.

[Emanuel]

Do you have easy contact to the metal core (vias)?  If nothing else, even if you have to bolt a strap to ground, that's workable.

No vias, it is a single layer PCB

Simply treat the enclosure as ground plane.  There will be some (AC) voltage drop along your circuit ground paths; these can be minimized with respect to the outside by single-point grounding (i.e. to the chassis) at the last point in the filter.

This circuit is working on 12 or 24 VDC, I don't understand where will be the AC voltage drop.
The thing is that during the test the lab is attaching a grounding strap to the housing but they are grounding also the cable shield. So it is a 2 points earthing, but I assume that the test table surface will create an easier path so nothing will pass between the 2 grounding points.

[T3sl4co1l]

Trace impedance. At AC, conductors are not ideal. Very wide conductors -- planes -- are the best you can get, but even there, some voltage drops between points.  (Defining a "point" itself becomes problematic because fields, but as a hand-wave, it's okay enough.)  Inductance is on the order of 1nH/mm trace length.

Tim

[xavier60]

No, nothing like that, everything is SMT and placed on one side of the PCB only.

In case it is not realized, having a switching transistor on metal core PCB is very much like a switching transistor attached to the case via a thermal washer. In both cases, HF current flows capacitively between the transistor's tab and the case, assuming that the MCPCB's substrate is connected to the case. 

[Emanuel]

In case it is not realized, having a switching transistor on metal core PCB is very much like a switching transistor attached to the case via a thermal washer. In both cases, HF current flows capacitively between the transistor's tab and the case, assuming that the MCPCB's substrate is connected to the case.
[/quote]

You are right, as I am using MCPCB's a lot (we are making LED lamps) I had such problems in the past.
In this case I am using an LM4164 buck converter with integral switch. This IC has a SOIC8 package with grounded thermal pad on the bottom and very far from what is considered HF, the oscillator is set at 200kHz.
It may be that the problems will appear at the radiated emissions test.
Here is the PCB layout

[xavier60]

I manufacture LED lighting for railroad use. I have settled on the HV9910 controller. I have managed to avoid MCPCB  even when I transitioned from gull-wing LEDs to SMD LEDs.
But I still include Y caps and CMC.

[Emanuel]

I manufacture LED lighting for railroad use. I have settled on the HV9910 controller. I have managed to avoid MCPCB  even when I transitioned from gull-wing LEDs to SMD LEDs.
But I still include Y caps and CMC chokes.

wow.... I liked the HV9910, I used it in a lot of projects years ago.
But it was very problematic with radiated emissions.

[T3sl4co1l]

The fundamental is around 200kHz but the harmonics extend up to 100MHz or more (for typical regs; I don't know LM4164 specifically).

You have basically ~pF of capacitance from that node to chassis, pushing your circuit up by as much.  If there's 100pF GND-chassis capacitance, that attenuates it up to 40dB maybe, but it's coming from 10s of V and you need ~mV.  The LISN has ballpark 50 ohms to chassis (per line, and give or take how the impedance works out at the far end of a cable), so the equivalent circuit is a capacitive divider into a resistor load.  The Thevenin equivalent source is ~100pF at 100s mV, into 50 ohms; below cutoff (Fc = 1 / (2 pi R C), ~32MHz), it's asymptotic 20dB/dec, and 200kHz is 2 decades below cutoff so you get another ~40dB there.

It should be close, actually.  But it's all proportional to switch node amplitude and capacitance.  Easily solved with a larger capacitor (some ~nF) from GND to chassis, and even moreso with a CMC or individual chokes inline.  The CMC must be used with capacitance to chassis, not to circuit GND.

The spectrum will be ~flat up to some upper cutoff, because square wave harmonics go as 1/N, balancing the cap divider's response.  In the span from ~30MHz to that cutoff, amplitude may be up or down -- how much depends on ringing at the switch node or other points, and finally the rise/fall time of the regulator.

Tim

[Emanuel]

I manufacture LED lighting for railroad use. I have settled on the HV9910 controller. I have managed to avoid MCPCB  even when I transitioned from gull-wing LEDs to SMD LEDs.
But I still include Y caps and CMC chokes.

I tried to send you a PM

[Emanuel]

Easily solved with a larger capacitor (some ~nF) from GND to chassis, and even moreso with a CMC or individual chokes inline.  The CMC must be used with capacitance to chassis, not to circuit GND.

Speaking about the CMC, the standard I need to pass tests conducted emissions between 10kHz-10MHz. I am using a CMC with high impedance at low frequencies.
What about the radiated emissions coming out through the power cable? Shielded cables are not always used and the recent version G of the MIL-STD-461 requires testing without shielded cables.
So we protect everything with a grounded metal enclosure but the problem remains with EMI from the wires.
An additional CMC for high frequency will solve the problem?