EMC questions answered (part 7)

-April 14, 2015

Several great questions on troubleshooting radiated emissions were asked during my recent webinar, Troubleshooting Radiated Emissions, hosted by UBM/Techweb and sponsored by Rohde & Schwarz last April 2015. Unfortunately, we ran out of time to answer them all, so here they are with answers.

If you missed the webinar, you may GO HERE to download a copy of the slides and listen to the webinar "on-demand".

Questions on PC Boards:

Q: Do high speed traces crossing gaps in the return plane create differential-mode or common-mode currents?

A: A high speed trace is a differential current. Differential currents flow in loops. The (signal) wave propagates from source to load and then back to the source. When a differential signal crosses a gap in the return plane, this  creates a conversion to common mode (CM) currents in addition to the differential mode current. These CM currents also return to the source, but can also radiate out I/O cables, generating emissions, before finally returning to the source.

Q: Why did the customer put the floating [copper gap] below the memory IC?

A: You're referencing that picture of the PC board with a large gap under the memory IC. I really don't know. I find that many circuit board designers have not been trained in EMC design, as it's not generally taught in the colleges and universities. One of the first things I check (and you should, too!) is to examine all the return planes, identifying gaps and slots.

Q: I'd like to try that measurement on my board, where exactly did you attach your clip-lead? [Also] in this page, you connect a wire to GND plane and just probe the wire?

A: In reference to the "trace over a gap" experiment, the clip lead may be attached anywhere on the return plane. There are common mode currents running all over the PC board, just looking for a wire from which to radiate! Placing a current probe around that wire allows you to observe the high frequency harmonic currents.

Q: Are high speed signals that pass through via's not recommended? If so, how would the radiation from the via compare to the high speed signal crossing a gap in the return path? I assume that it is not nearly as bad?

A: The issue really is high speed signals referencing more than one return plane. Vias introduce an impedance "bump", which can cause reflections. Running a high speed signal through too many vias will impact the signal integrity. When a high speed trace penetrates from one reference plane to another (assuming both plane are the same potential, eg., "ground"), you need to design in a return path - ideally adjacent to the signal via. For example, the addition of one, or more, adjacent connecting vias from one plane to the other would work to allow the return current to flow back to the source in the shortest path. If the two planes were different potentials (power and return, for example), then you'd need one, or more, "stitching" capacitors installed adjacent to the high speed signal via. Running a high speed trace over a gap is much worse than running through a via, because even with differing potential planes and no defined path, the parasitic capacitance between planes helps somewhat to create a return path.

Questions on Cables:

Q: Should the chassis/cable shield be connected to the PCB ground? Should they be connected with an inductor? or a 0 ohm? [Also] Do you happen to have a picture of how a cable shield is grounded onto the chasis? [Also] How do you suggest grounding I/O connectors with plastic enclosures? Q2 In the case of plastic enslosures, what is the best way to handle I/O grounding?

A: For shielded enclosures, the cable shield should ideally be connected via a "360-degree" bond to the enclosure, right AT the I/O connector. This is best done by using shielded connectors. In addition, it's important to bond the I/O connectors on the PC board to chassis or product enclosure through short standoffs. The idea is to make the voltage difference between the PC board signal/power return plane, connector ground shells, and chassis/enclosure all the same potential (zero!).

For products without shielded enclosures, cable shields should generally be connected to the signal return plane. In order to prevent (or reduce) the amount of common mode current flowing out the outside of the cable shield, you'll need to design in common mode filtering on the board and located nearest the I/O connector. You should never use an inductor or ferrite choke in series with a cable shield bond to chassis or the signal return plane.

Q: What are the best methods for non-shielded cables isolated from the metal housing?

A: Non-shielded cables (for example, power), which penetrate a shielded enclosure need to have common mode filtering installed on the PC board at the penetration point or nearest the power connector.

Q: Is that a square wave driving the pigtail?

A: For the various experiments, I'm using a harmonic comb generator from Applied Electromagnetic Technology (AET) that essentially is a 1.8 or 10 MHz pulse generator. The pulse width is about 2-3 ns. I've written several articles on using these, the latest of which may be found here.

Q: if I connect [an] HDMI cable signal twisted pair ground (drain wire) with outer braid, what will happen?

A: There should be no effect whether the drain wire is connected to the shield, or not. Of course, the actual cable shield should be well-bonded to the metal connector shell (in multiple places). I might add that many of the cheaper HDMI cables have poor shield-to-connector ground shell bonding. If you're using a cheap HDMI cable for compliance testing, you may be "shooting yourself in the foot".

Q: You'll notice that the Arduino is in a plastic case. How do you solve the issue of grounding the case of the Ethernet jack (e.g.) to a non-existent metal enclosure?

A: This is similar to the general question above regarding connecting the I/O cable shields to the signal return plane. However, for most good Ethernet connectors, there is a built-in common mode filter (and sometimes ESD transient protection). Most Ethernet cables are just two unshielded sets of twisted pairs, so it's very important to ensure some form of common mode filtering exists in your design. In the case of the Arduino, I suspect the Ethernet connector is the cheaper unfiltered type, so I would expect a large amount of emissions from the cable. That's why I use it for my EMC demos!

Q: A 1.3m cable has a resonance frequency around 90 MHz that leads to a increase of the noise at this frequency. it could be a disadvantage during tests at those frequencies. How can we minimize this frequency?

A: Boy, that is the important point I was trying to get across and I wish I'd had more time to explain this in more depth! Because most I/O cables are about 1m long, they will all have this resonance at about 90 MHz (and higher-order resonances (180, 270, etc.), which will tend to cause peaking of the harmonic energy in these frequency bands. If common mode currents are allowed onto the cable shield, you'll observe these peak emissions during the radiated emissions testing. The important point is to reduce, or eliminate, these common mode currents through propoer filtering.

Questions on Probing:

Q: What tool is most useful for measuring high frequency currents in cables?

A: As I mentioned in the webinar, one of my most-used EMI troubleshooting tools is the RF current probe. By knowing the amount of RF common mode currents flowing on an I/O cable, one can calculate the expected E-field at a given distance, d. This is a very powerful technique for assessing the EMI profile of your product prior to compliance testing. EMI fixes during troubleshooting will be observed in real time as the current levels decrease at the various frequencies of concern.

Q: The current probe examples appear to measure total current – conductor, return, and shield – which should sum to zero. Please clarify.

A: Ahh, excellent question. I answered this one live, but will repeat it for the other's benefit. Assuming a good quality shield (a tight weave with very small openings), currents propagating along the signal conductor will go from source to load, but the return current will return on the INSIDE of the shield and should not be measureable when looking at the outside of the shield. This is the result of skin effect, where the inside and outside of a cable shield appear to be two separate and isolated conductors. Therefore, all we're measuring is the common mode current flowing on the OUTSIDE of the shield, which is all we care about as far as radiated emissions. Cheaper cables with looser weaves or cables (especially HDMI cables) with pigtail connections to the connector shell, will allow interior signals or common mode currents, respectively, on the outside of the shield.

Q: What kind of current probe (brand? model? specs? cost?) do you use on the clip lead wire?

A: I was using the Fischer Custom Communications F-33-1 current probe in those pictures. However, there are many vendors of quality current probes - Rohde & Schwarz, ETS-Lindgren, Com-Power, Pearson, Solar, and many others. Select one based on the frequency desired (generally, 1 to 100 or 200 MHz is the best range for most common mode currents). You'd have to contact these vendors for pricing, but generally, they run $1,000, and up.

Miscellaneous Questions:

Q: Could you give some examples of a few types of common mode filtering that are cost effective?

A: For I/O ports, the surface-mount common mode chokes are effective. It's also a good idea to include transient suppression devices, as well, for ESD and EFT protection. As a last resort, a clamp-on ferrite choke works sometimes, but they are generally more expensive than the surface-mount parts. As far as the definition of "cost effective", what's it worth to have your product pass compliance testing first time and start shipping? As always, there's a tradeoff between taking the time for a proper EMC design and skipping this step, resulting in weeks of troubleshooting and repeated testing.

Q: Multiple turns through a ferrite core increases the impedance. If you can't do that, use multiple beads. Sometimes it takes six or more before you suddenly get the effect you are looking for.

A: Good input. There's a couple important points underlying your suggestion. For a ferrite choke to work, its impedance needs to be much greater than the wire, cable, or trace its connected to. Also be aware that high bias currents (power supply or power distribution applications) can dramatically reduce the impedance of the smaller ferrites, because of saturation. Applying multiple turns to a ferrite choke, while increasing the impedance, can also degrade the high frequency effectiveness through parasitic capacitance between the turns. Try to keep the turns spread apart as much as possible.

Q: Why use a tool instead of fingers for foil & tape?

A: You must never have cut your yet fingers on the edges copper tape!

Q: If the product has 4-5 switching power supplies, how to identify the high frequency return current paths for each or entire product?

A: This is an important "systems" question. It's important to keep each secondary voltage loop area as small as possible. Likewise, for the primary loops. It's important to run the voltage return wires/traces in close proximity to the source wires/traces all the way from power supply to PC board. Using twisted pairs of power and power return will accomplish this. Generally, but not always, the common (return) connections of all your supplies all connect together at a common power return plane on the board. If these are all "on-board" supplies, the same principle holds true. Always ensure that current return paths are well defined, whether it be from power supplies or high frequency digital or analog sources. As for the whole product, you might refer to Henry Ott's, Electromagnetic Compatibility Engineering.

Q: Does Rohde & Schwarz offer EMC Test Services?

A: No, they don't...they're primarily a test & measurement instrument company. However, if it's an instrument or measurmeent application, they have several good field engineers that can help. For EMC compliance testing, there are a number of good test houses. Check the directory section at www.incompliancemag.com or www.interferencetechnology.com.

Q: what do you do when the ambient noise floor is high?

A: I'm not sure whether you are referring to the measurement "noise floor" of a spectrum analyzer, or the ambient RF signals that interfere with your observing product emissions outside of a shielded chamber. Let's assume the latter. Ambient RF signals from broadcast radio/television and other mobile communications systems can certainly affect your ability to discern the actual harmonic emissions from your product. Knowing the exact frequencies of the top harmonics is important. Moving the location of your troubleshooting test area to a basement can help a lot. Reducing the span and/or resolution bandwidth of the spectrum analyzer can also help exclude interfering signals. I also try to troubleshoot with the receiving antenna about 1m away from the product under test. This serves to boost the harmonics so they're better observed. These techniques are all described in EMI Troubleshooting Cookbook for Product Designers, as well as some of my blog postings and magazine articles.

Thanks for attending and if you have any other questions, don't hesitate to email me at ken@emc-seminars.com.

For more questions from our past webinar participants, check out these other blog postings:

EMC questions answered (part 1) - Cables and EMC mitigation

EMC questions answered (part 2) - PC boards and EMC mitigation

EMC questions answered (part 3) - pre-compliance testing for radiated emissions

EMC questions answered (part 4) - EMC measurements and product design

EMC questions answered (part 5) - general questions

EMC questions answered (part 6) - general questions

Loading comments...

Write a Comment

To comment please Log In