Connector Emission Filtering for CE mark

[Pack34]

One of my coworkers is out running a product through some pre-compliance testing and I heard that there will most likely be an issue with an interface connector that we provide for customer use. On the connector, we have a a power line, a couple simple digital inputs and outputs, and an analog output. Nothing here is by any means high frequency.

I'm assuming that the issue is that the connector and associated cable are acting like an antenna, which is causing the issue.

What type of filtering would be needed for something like this? Just a simple bead and/or resistor in series with all the pins on the connector?

Would removing the ground plane under the connector have any real impact on a low-frequency connector?

[T3sl4co1l]

Well, very very broadly: yes, a possible risk.

You didn't mention:
- If the connector is attached in standard use (i.e., what they'll be testing)
- What the signal level, bandwidth, repeat rate, filtering, etc. is on the lines
- Internal grounding/shielding, consistency of grounding / ground loop voltages/currents, other internal sources (logic, switching noise?)
- If the connector does radiate, then what does it 'push' against, i.e., what other boards or cables or enclosure might carry the displacement current

Removing ground plane is almost always a terrible idea, but I'd have to see the entirety of the design (system level interconnection; mechanical; electrical (schematics and PCB)) to provide much practical input...

Tim

[Pack34]

Maximum frequency is about 1kHz
Voltage level is 3v3
Connector should be terminated on a frequency generator to simulate customer signals.
The ground planing is odd (I didn't design this). The connector is on a daughter board that connects to the main unit internally through a cable. The analog and digital circuits have isolated grounds on this connector board but the grounds are combined when they connect to the main unit.

There is an on-board switcher but the resulting rail that goes to the output connector goes though an LDO.

For the ground plane under the connector, I know that this is standard practice for Ethernet connectors but wasn't sure if it would have any effect in this case.

[T3sl4co1l]

Maximum frequency is about 1kHz

Are you certain of this?

You mentioned digital.  What's the rise/fall time of those signals?

Maximum frequency is determined by edge rate, not clock rate!

Tim

[Neilm]

Basic question - what test? I am assuming from what you have said it is radiated emissions that are the problem.

Removing the ground plane will have an effect - it will make it worse.

If doing pre-compliance shows issues, then a good way to start is to clamp ferrites on cables. This will show if it is that cable that is acting as the antenna. If so, then look  at the frequency that is the problem. Does this relate to the rise / fall times of signals (or harmonics there of), and SMPSs in the unit. I have seen 1kHz signals radiate at over 100MHz on a 2 layer board as the signal path went all over the PCB.

[qno]

I think it is probably a susceptibility problem.
What is the test level? 3 or 10 V/m or maybe 30 V/m??

One or more of the wires is probably acting as an antenna.
Limit the bandwidth of the interface add 10 nF to the inputs if it does not influence the bandwidth of the interface.

You can always ask the test house for advice.

[Neilm]

I think it is probably a susceptibility problem.

It says emission in the topic subject - it does not say radiated or conducted

[Pack34]

I think it is probably a susceptibility problem.

It says emission in the topic subject - it does not say radiated or conducted

Radiated

[Neilm]

Well, here is where the fun begins.

It is now a matter of investigation, skill, experience and a bit of luck.

What frequency is the problem? Look at the frequency and look at the plots. If you see a repeating pattern of peaks that has an interval of a clock frequency, then that is most likely the problem.  It is not just clocks, but switch mode power supplies as well as any edges being transmitted over the cables.

A few points - I have seen noise transmitted up 0V lines, so don't assume that because the signal does not go up a cable it can not be the problem.

Get a spectrum analyser and some probes and sniff around your circuits. You should be able to see the offending signal (be aware that you may also see near field effects that do not transmit in the far field) Even cheap kits should allow you to narrow in on the part of the circuit that is responsible.

Neil

[Jeroen3]

A easy way, though not calibrated, to test low frequency emission on your board is to pull the ground strap off your probe, clip on the hook-thingy, and scan the board without touching it. The hook thingy is a very efficient antenna.
With the scope on mV's per div. If you're picking up 10's of volts, be afraid...

This way you can at least see if the bodges you are attempting are of any help without calling the guy at the test facility to spent his costly hours performing futile bodges. It won't work in the high Mhz ranges. But if you're emitting there it is no longer the connectors fault.