Question and Answer Session for 'EMC Troubleshooting Tools and Techniques' Webinar

Answer: I did have to evaluate a medical product with a very small PC board. For emissions, I’d test like any other board. However, for identifying emission sources, I’d use a very small loop probe. Beehive Electronics, Aaronia, Langer and Rhode & Schwarz all make near field probes with very small loops that will yield good resolution. You can also make your own by fashioning a small loop at the end of a piece of small-size coaxial cable. You’ll likely require a broadband preamplifier to boost the very low signals from the smaller probes. For radiated immunity, I feed a small RF field into one of these small loops and run the field around the PC board, looking for areas of sensitivity.

Question: Our cables are folded and going in different directions in the instrument. Will they have both vertical and horizontal components of emissions, and is the emission frequency value dependent on the total length of cable, or is it only the length of cable that is horizontal (or vertical)?

Answer: Good question. I’m assuming these are all cables attached to the product and spread out around the measurement table. If the cables are oriented horizontally, they’ll generally product horizontal polarized emissions; if vertical, then vertical emissions. If they droop over the edge of the table, they’ll generally produce a combination of horizontal and vertical emissions. Folding (serpentining) the cables tends to cancel the field in that section, but may produce a minor amount of vertical emission. It really depends on how tight the folding is. Of course, the objective is to eliminate the common mode currents on all your cables so cable emissions don’t enter in to the equation!

Question: For immunity, I thought an analog circuit was more susceptible than a digital circuit?

Answer: Generally, you’re correct. However, I’ve seen cases for radiated (and especially ESD and other impulse type signals) that will reset the microprocessor IC or cause glitching in digital signals.

Question: We typically use a 3m chamber to test our IVD instruments, as seems typical in the US and Europe. We were questioned by Japan as to why we do not use a 10 m chamber, since limits are in terms of 10 m in all of the standards, not 3m. How do we justify this?? Should final reports always be at a 10 m test site, not a 3 m chamber?

Answer: While a 10m distance seems to have been adopted by the standards committees as a “standard”, not everyone can afford a 10m chamber. Therefore, 3m data should be accepted. The issue for 3m arises mainly at the lower frequencies (say, under 200 MHz) where the chances of reflection errors might be greater. I always suggest that margins from the limit be increased when using 3m chambers, if at all possible. Also, the 4m antenna height can’t normally be achieved in a 3m chamber, so you’re not fully comparing the higher “take-off angles” of fields emanating from the product under test between the two measurement distances. Finally, depending on the definition you use (typically 1/6^th wavelength), you’re approaching the near field at the lower frequencies. In other words, the assumption of “plane waves” is not quite true and so the distance factor from the product under test will also depend slightly on factors of 1/r^2 and 1/r^3 terms, rather than the simple 1/r term for plane waves. One thing you might consider is to take a comb generator with some standardized (electrically short) antenna and compare between your 3m chamber and a commercial 10m chamber. This will provide you a sort of calibration curve between the two measurement distances versus frequency. This should help you determine a “safe” margin when using your 3m chamber and may even provide the justification required by other country regulatory agencies.

Question: How do you troubleshoot frequencies above 1GHz?

Answer: Assuming you’re asking what tools and techniques I use for troubleshooting radiated emissions above 1 GHz. Basically, I use the same tools and troubleshooting techniques. For very high (above 6 GHz) emissions, I use the smaller PC board log-periodic antennas and higher frequency broadband preamplifiers. These higher frequency antennas are available from www.ka5vjb.com.

Question: What can you do if you notice the emissions noise is actually on the shield of the cable?

Answer: That’s typically where you find these noise currents (also known as common mode currents) traveling along the outside of your shield. Bonding the shield well to the shielded enclosure or adding on-board common mode chokes or installing an external common mode choke around the product end of the I/O cable are typical solutions. Even better is to design your PC board and system designs properly to avoid or minimize these common mode noise sources in the first place.

Question: How do you validate the E-Field power level when using test antennas in a test bench area? For pre-compliance testing, how do you correlate your measured signal to compliance levels with homemade antennas?

Answer: When testing with a sense antenna 1m away from the product under test, it’s possible to get “quick and dirty” answers by adjusting the limits upward by 20 dB from the 10m limits. Understand, this is only a very rough ballpark limit, as you’ll be well into the near field for the lower frequencies. See the discussion above relating to 3m versus 10m testing for more detail on some other issues. It’s much safer to set up a calibrated EMI antenna 3m from the product and use 3m limits for more accurate results. The advantage of setting up an antenna close to the product is that you can get immediate results (did the harmonic go up, down or no change?) during the series of experiments and potential fixes. This technique will really boost your efficiency. It’s also very helpful to have calibrated test data in hand showing where harmonics are close to (or over) the limit. Realize, though, that a 10 dB decrease at 1m does not necessarily equate to a 10 dB decrease at 3m or 10m. However, any decrease will be a good thing. You’ll just need to optimize the emissions as best you can on the bench and then have the product retested in a decent chamber on occasion in order to check progress.

Question: When do we use an E-field versus a B-field probe for EMC testing?

Answer: H-field (or B-field) probes are best used to detect H-fields – typically from wires, cables or traces where you might have a high di/dt (currents in loops). E-field probes are best used to detect the E-fields from enclosure slots or gaps. They are also useful for detecting high dv/dt, such as in switching power supplies. Understand, though, that when measuring in the near field the electromagnetic wave will have components of both E- and H-fields. It mainly depends on which field might be dominant (currents or voltages).

Question: I am confused about the purpose of the pigtail. What is the purpose or the point and why is it useful? What’s the recommended method in terminating cable shields?

Answer: First, let me state that “pigtail connections are bad news” and can generate high levels of common mode currents, which tend to radiate and cause your product to fail. The use of pigtail connections is a “cheap and dirty” way to connect cable shields to their connector shield terminations. More expensive connectors have provisions for connecting or clamping the cable shield in a 360-degree bond, which is ideal. The longer the pigtail used, the worse you can expect emissions. If you’re forced to use poorly designed connectors, it’s recommended to use two (or more) short pigtails to the connector shield, one on each side of (or surrounding) the internal wires. This will tend to cancel the resulting fields.

Question: I have some really sensitive pressure sensors and am picking a lot of noise from radiated frequencies on the immunity tests. I can see the 2Hz frequency on the product, as if it were the signal I am trying to measure! Even when shielding them with EMI plastic and/or metallic shields. These shields are not connected to ground nor earth, because doing that worsens the results. Maybe I am doing something wrong?

Answer:  If your sensor is designed to pick up 2 Hz signals, it should be relatively easy to add simple low-pass filtering at the input in order to remove higher-frequency noise. You’ll want to ensure the sensor electronics is completely shielded for additional protection. Be sure the low-pass filter is located right at the sensor input and just inside (or at) the shield boundary.

Question: With high frequency these days, how can you have ventilation holes in your shield at 1/20 wavelength?

Answer: The 1/20^th wavelength of a slot or seam is merely a rule of thumb, which approximates about 20 dB of shielding effectiveness at that frequency. When designing ventilation holes or slots, you need to consider the highest frequency used (or harmonic produced) by your product. Then take 1/20^th wavelength at that highest frequency and calculate the maximum hole or slot size from that. For example, for 1 GHz, 1/20^th wavelength is about ½ inch. If you need to go higher in frequency, then it’s generally best to use patterns of small round holes – maybe ¼ inch, or smaller, in diameter.

Question: Where could I obtain the details of the PC antennas?

Answer: They are available from www.wa5vjb.com.

Question: Are not inductive and capacitive coupling just special cases of radiated emissions?  If so, then there are still only two types of coupling?

Answer: I suppose you could say that. However, what we’re really talking about is electromagnetic fields and how they propagate and couple together. Conductive coupling is noise (usually low frequency) traveling along a pair of wires (normally due to a common resistive path), which form a loop. Breaking the connection stops the interference (think so-called “ground loops”). Radiated coupling is what you would imagine – EM waves propagating through space from one point to another, whose coupling factor reduces by a factor of 1/r. The couplings you mention – inductive and capacitive – are really near field effects, which are reduced rapidly with separation (near field terms of 1/r^2 and 1/r^3). Inductive coupling requires two loops coupled together, with the field of the one inducing noise into the other (think transformer). An example would include one cable inducing noise into an adjacent cable. Capacitive coupling requires two “plates” with the noisy plate inducing a voltage change in the other due to dispersion effects (think capacitor). A typical example might be the heat sink from a SMPS inducing voltage changes in a nearby cable or chassis. Capacitive coupling is generally a high dv/dt phenomenon, while inductive coupling is a high di/dt phenomenon.

Question: How do we minimize/eliminate radiated emissions for circuits with multiple ground returns?

Answer: I’m not sure whether the question is framed around designing a single PC board or whether you’re talking about an entire system. Assuming we’re speaking about a PC board design, the first question you need to ask yourself is, “why is it necessary to have multiple ground returns?” If the answer is, “well, we have digital, analog and, say, motor controller circuitry and we need to keep them isolated from each other”. Well, that’s fine. In that case, I recommend using a single signal/power return plane in your product for all signal technologies. The trick is to keep noisy motor controller or digital signal/clock currents from penetrating into sensitive analog circuitry. You can do this by segmenting your PC board into separate areas and ensuring that no signal, data, address, or clock signals cross over the analog section and that the potentially noisy motor control signals avoid crossing over either the digital or analog areas. Dr. Todd Hubing has a lot more to say about proper PC board design on his web site: http://www.cvel.clemson.edu/emc/index.html.