“If you’re not making mistakes, then you’re not doing anything.” — John Wooden
Over the years, I have seen several things that make me scratch my head. Many of them are things I have done (there, I admitted it). I would like to look at these situations to (maybe) help not to make the same mistakes. So, to be clear: Do not do these things.
One would think that a thick aluminum chassis with a number of screws would be a good shield, but on one power supply, we were failing an aerospace radiated emissions scan in the 1 to 10 MHz range by several decibels. We chased the problem for several days, if not weeks, before I noticed we were making a mess of the metal by how we handled the chassis when we opened the unit. The aluminum case was covered with fingerprints and getting a bit greasy. So, I took some alcohol and wiped down the surfaces. The emissions dropped 20 dB, and we passed the test. Repeatedly.
This is the concept of good chassis bonds. Metal to metal contact must be in the micro-ohm range to work well for many shields. This is because if a current is flowing in the chassis when it reaches the slightest impedance, a voltage will appear across that impedance, which will be the source of the radiation. Chassis coatings, whether for environmental protection (such as anodizing or conversion coating), or esthetics (e.g. paints), or by contamination, will degrade the contact quality. So even my aluminum chassis that should have had over 100 dB of shielding effectiveness was degraded because of fingerprints.
Some may wonder if more screws would help. In this case, not likely. First, if the screw surfaces were coated, the screw would not make a good bond to the chassis. Also, screws are inherently inductive, and if the surface the screw is threaded into is an insert or captive nut, the impedances increase. If the concern is that the screw spacing must be less than 1/20th of a wavelength apart, the real issue may be missed. The problem looks more like a folded dipole than an aperture. Openings and apertures are when we consider wavelengths and frequencies. However, considering two metal surfaces in contact as an aperture may be misleading.
Overlaps and bonding from metal to metal is a bit of an art as much as sound engineering when designing chassis. Do you overlap seams? Do you use a conductive gasket, or do you need to? How far apart do you need to put the screws? In the words of consultants worldwide: It depends.
I had a client who was evaluating shielded cables for a computer peripheral. One set of cables appeared to have no shielding effectiveness. Upon investigation, it was a shielded cable, with a foil shield, and the drain wire was grounded on both ends. However, the aluminum over mylar foil shield had been installed with the drain wire on the inside of the shield, but the aluminum was on the outside of the shield. No contact was made between the drain wire and the shield—thus no effective shield.
As consultants, we see countless cable shields which are terminated in a pigtail or wire, typically with a “service loop” which is 10 cm long or more. Or they are bonded only at one end—which is an article all on its own on why that does not work—or both.
But even a perfectly shielded enclosure must have wires running from inside to outside. Any conductor that pierces the shield will allow signals on the inside to come out, or outside in—unless there is an excellent filter at the point of penetration. This is typically a capacitor that has low impedance from the conductor to chassis over the frequency range that is needed.
Thus, we place a filter at or very near the connector, which often includes capacitance from line to chassis (or whatever structure and the return path is needed to return the current to its source). And on the schematic, a capacitor of the proper value is shown. When constructed, the capacitor is placed very near the connector pin. So far so good. Then what? It goes to a ground symbol on the schematic, but how is that constructed? Does that “ground symbol” involve a long trace to some corner of the circuit board where a capacitor connects to a pad that casually touches a screw that mounts the circuit board to the chassis? How long, and thus, how inductive, is that path? Remember that inductance is a high-frequency impedance. So, is the path inductance filtering the capacitive filter to keep it from working?
I once worked with a client on a project in which we got the design to pass conducted emissions. The engineer submitted the final drawings for the qualification of the product. Upon qualification testing, the unit failed conducted emissions. In fact, the emissions were about 20 dB worse than they were for the design we saw pass earlier. I asked if he changed anything. No, I was told. So, I asked if he improved anything—well yes, he did. He made one of the filter capacitors 10 times larger. Since I knew we had some large ceramic caps installed, I asked him where he found such large ceramic capacitors. He said, “Oh, they are not ceramic, they are electrolytic.”
The construction and dielectrics used in capacitors are very important. For example, the electrolytic capacitors in question have high equivalent series resistance and inductance, which render them poor filters above 100 kHz or so. While great for bulk storage, these caps do not provide the high-frequency filtering needed for our test. When we replaced the caps with ceramics, all was well with significant margins across the frequency range. That engineer soon afterward quit the industry and went into another line of work. True story. The bottom line, more is not always better.
Not all caps are alike, even among the same style. Maybe ceramic capacitors are the most finicky. Class I ceramics are nice and stable over temperature and bias voltages but are bulkier as a result. Class II and Class III may have wide swings in effective capacitance over temperature ranges and have significantly reduced capacitance with a bias voltage but are physically smaller and less expensive.
Stories About Common Mode Inductors
To reduce common-mode energy from an AC to DC power supply, an engineer wound a common mode inductor. Once installed, we repeated the measurement only to find the emissions went way up. Having seen this before, I asked if he was careful about how he connected the inductor. He was not, not thinking it was important. The inductor was connected with the line and neutral currents adding to the core magnetic field, not cancelling, as should be with a common mode inductor. As a result, the core was going in and out of saturation with each half sinewave, 800 times a second (on a 400 Hz power source). This radical change of impedance turned the inductor into an emissions generator instead of an absorber or filter.
To pass radiated emissions, a common mode inductor was created with a few turns on a nickel-zinc ferrite (permeability about 700). When this worked well, the engineer decided to use a higher permeability core in the design—something with a permeability of about 2,500. However, this material was a manganese-zinc compound, which did not have the needed bandwidth for radiated emissions.
Another engineer heard that the inductance increases as the square of the number of turns. Wonderful—let us double the number of turns and we have four times the inductance. Well, yes and no. Parasitics and core saturation must be accounted for. For example, in the case of a common mode inductor, it can be modeled as shown in Figure 1. Notice the presence of the following components:
- CW which is the inter-winding capacitance. Each turn will capacitively couple to the next. Adding more turns will increase this capacitance
- CP-S which is the capacitance from winding to winding, line to line. Some baluns are wound with turns on opposite sides of the core to minimize this. However, this can lead to increases
- LL/2 which is the leakage inductance. In a common mode inductor, leakage inductance can appear as a series inductance in each lead, but that reduces the mutual inductance from the coupled fields between the two windings
- LM which is the mutual inductance value, the inductance we use for the common-mode rejection
In his talk, Dr. Michael Schutten (reference below) describes how to measure each of these components in a common mode inductor. I suppose if all the parasitics of all the components were modeled, we would have a better idea of what to expect from the performance.
So, adding more turns may increase CW, which will decrease the impedance of the core at higher frequencies—the thing we don’t want to happen. More is not always better.
I have found that ferrites are great for common mode inductors, but often not for differential mode inductors. Ferrite material is susceptible to saturation, which can render the core ineffective. I’ve found that magnetic fields from hidden structures can cause problems. A trace ran next to a DC-DC switching regulator was known to couple the switching frequency into that trace from an internal transformer inside the integrated circuit, causing a conducted emissions failure. I’ve found that when you try to filter common mode currents with differential mode techniques, it does not work well. A line-to-line capacitor may work great on differential mode currents, but it might make the common mode currents more common mode as it were. And, I’ve found the best way to control EMI is to understand where the currents flow, and remember they have to flow in complete loops—from a source and back again. It is how they do that which gets us into trouble. And finally, “ground” is not a hole you can pour circuit noise into and it goes away. Many in the EMC world avoid the use of that term except for safety purposes and prefer to think of reference planes and current return paths.
So maybe Charles Dickens was an EMC consultant, since he wrote in A Tale of Two Cities: “It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness.” I have made some foolish mistakes, been a partner with others’ mistakes, but hopefully learned from them. And hopefully, you can learn from ours.
Below are some lectures, papers, and talks I have heard, which I believe would be useful. These are brilliant people who explain difficult concepts in easy to understand manners, and which I used in the creation of this article. There are numerous other sources of excellent information I can highly recommend.
EMC Fundamentals for Switch-Mode Power Converters, 2020 IEEE Symposium on EMC & SI – Michael Schutten, Ph.D., Cong Li, Ph.D.
Filter layout and mechanical design. Why filters can fail, 2020 IEEE Symposium on EMC & SI – Dr. Arturo Mediano, Senior Member IEEE, University of Zaragoza
Design and Implementation of EMI Filters, EMI Filter Basics, 2020 IEEE Symposium on EMC & SI – John G. Kraemer PE, Collins Aerospace
Conducted Emissions, 2019, 2020 IEEE Symposium on EMC & SI – Lee Hill, SILENT Solutions LLC & GmbH
Filter for EMC, 2016 IEEE Symposium on EMC & SI – Dr. Arturo Mediano, Senior Member IEEE, University of Zaragoza
Shielding, 2015 IEEE Symposium on EMC & SI – Dr. Todd Hubing
Decoupling Capacitor Design on PCBs to Minimize Inductance and Maximize EMI Performance – Bruce Archambeault, et al., April 2015