FCC Conducted Emissions Failure

GoogleGroups
1

We are planning to use the Beagle Bone black in a commercial product, We have designed a custom cape and put the whole thing inside an aluminum housing. The system has been sent off for EMC testing but unfortunately the testing lab has failed conducted emission. There where a lot of peeks but the following frequencies failed

180MHz
203MHz
228MHz
250MHz
288MHz
312MHz
372MHz
384MHz
396MHz
420Mhz
468MHz
480MHz
875MHz
1000MHz

To try and isolate the issues we’ve acquired a spectrum analyzer and a set of near field probes. From my testing with the near field probes It seems that most of the failed frequencies are coming from the beagle bone black.

I’ve noticed the 1000MHz and 250MHz peeks (which where the biggest) seemed to go when the Ethernet cable was is removed. Many of the other peeks seemed to match up with the peeks on the beagle bone black FCC report. (http://elinux.org/Beagleboard:BeagleBoneBlack#Regulatory_Compliance_Documents) The once that don’t also seem to go when the beagle bone is powered off.
The system sits inside an extruded aluminum housing and the end plates are actually PCBs with single plane of copper. So there is lot I can still do to tighten up the enclosure. Although based on the report I would have expected the BBB to pass outside an enclosure.

Has any one else had similar experience with the beagle bone black and EMC problems? And how did you solve them?

2

When doing this, you need a good power supply with a choke on it and a good Ethernet cable, one that is shielded costs more money than your typical cable. I have no idea what your cape is doing, Having it in a case can help, but the real issues are related to the cables that come into and out of the enclosure.

Gerald

3

Does your case have a removable cover?
One that runs in a sliding slots on both sides?
Is the extruded aluminum anodized?

If so, then it is approximately worthless for RF shielding.

Aluminum anodizing is an insulator. Which means that the cover is not well grounded to the rest of the case.

Research “slot antennas.” A slot in a plane of metal is just as good a radiator as a metal dipole in open space.

What is the resonant frequency of the metal box when viewed as a resonant cavity?

It will actually increase the radiated spurs of the basic product (the BBB and anything else with clocks) at the resonant frequency(s) of the case.

As an experiment, drive some metal screws into the junction between the case and the cover, to get better grounding.

Grind the anodizing off the case at any point that you need a metallic ground to the aluminum case.

If you are using the slots inside the case to hold circuit cards, don’t assume that you are getting any grounding through the slotted ridges, because they are anodized.

As Gerald suggests, get some clamp-on ferrites or ferrite beads (appropriate for the frequency range of concern) and start putting them on every cable entering or exiting the case.

Use shielded cables, and ground the shields at the point of exit/entry.

Good luck.

— Graham

4

Our test lab graciously retested (just for positive peeks - not quasi peeks ) with the device outside the housing. You are correct some of the peeks where a few dB worse outside the housing although overall the system still failed.
The 250MHz and 1000Mz did disappear when the shielded Ethernet cable was used too.

At this stage I’m looking at moving to a beagle bone green. I did some rough pre compliance tests with a H probe ~ 14mm above the board. The green had the same peeks but about 15 to 20 dB lower than the black. You can see the results in the document here http://bit.ly/2abNcVa . I need to do some more investigation but the green is looking a lot better at this early stage.

5

David:

Since you have provided no specifics about your product or the enclosure, I can only guess what you are trying to do, and how.

So, some general thoughts:

If your product is small enough, Hammond makes a line of “submersible” extruded aluminum cases, where all four sides of the box are part of the same extrusion, and therefore no side slots. If your product is small enough to fit in one of these, don’t use their rubber end gaskets, and put your PCB end covers right on the metal, so you have metal to metal all around the end plates. This combination can actually provide some EMI shielding.

If you can live with 10 Mb Ethernet, lock your Ethernet port down at 10 Mbps, which will significantly reduce the higher (multiples of 125 MHz) spurs.

Get an ferrite clamp-on cable EMI KIT from Laird, or Fair-Rite, or Wurth. Even if you don’t end up putting these on the final product (although you might have to) you can use them to identify which wires coming out of the case are carrying which spurs. Ask your EMI lab for a recommendation, and the names of local reps that carry them.

Get rid of (or at least suspect) the USB power supplies. Many no-name USB power supplies do net meet FCC/CE (no matter what the label says), and therefore any product they hook to, will not pass. Only power your product with known good, clean power.

My guess is that the BeagleBone Green might be a little quieter, but will likely not get you across the goal line.

Observe good grounding and shielding hygiene on all wires/cables entering or leaving the box. Keep all signal wires and their ground returns twisted together. Never split a signal and it’s ground return. Use shielded cable where ever possible, and ground the shield at the entrance/exit point.

Once again, good luck.

— Graham

6

We need to do something similar and I'm trying to find a way to reduce the eth0 / wlan0 speed to 10 Mb, but it looks as though ethtool will not work in Debian on the Beaglebone Black. Is this a shortcoming of drivers or the hardware?

7

Never tried reducing the eth0 speed.

We where able to reduce the USB speed by making code changes to the linux kernel. (that was fun) But in the end soldering the USB connectors and providing a better contact of the Ethernet port to the chassis fixed most issues.

Also the beagle bone green was a lot quieter than the black.