Public Enemy #1: Pigtail GroundsNo matter what the EMI problem, we usually find ground impedance at the root. Nowhere is this more conspicuous than pigtail grounds, usually associated with a cable shield termination. When on an EMI troubleshooting consultation for radiated emissions and immunity and for ESD, our first look is at the cable connectors — often the problem is there, and detectable by a visual inspection. The basic problem lies in the impedance of the termination of the cable shield. Let's start by looking at the impedance of a pigtail ground. Ground ImpedanceThe basic issue is ground impedance, particularly at the frequency range of interest. At power line frequencies, impedance is primarily resistive, so conductor cross section and path length are the driving factors. Above audio frequencies, however, inductance becomes the major issue, and we need to look at impedance. At radio frequencies, inductance of even a short length of wire (or pigtail) becomes the limiting factor.  Figure 1. Equivalent Circuit of Pigtail Cable Termination Figure 1 shows the phenomena. All cables have high capacitance between the shield and the signal or power conductors, typically 25 to 50pF per meter of cable length. Combine this with the inductance of the termination, and you have a nice coupling path between the shield and internal conductors, complete with a resonance to emphasize coupling at a particular frequency range. The purpose of the ground is to provide a low impedance path to shunt interference currents, so the impedance of this path must be low. Our rule of thumb is the inductance of a wire is about 20nH/inch or 8nH/cm of length. While this rule has met with resistance in some camps, no one has come up with another reasonable rule-of-thumb — but if you have one, go ahead and use it, remembering that no matter how much you argue, there will be inductance and you will have to deal with it. Using our rule of thumb, the impedance of one cm of wire will be: Z = 2*pi*f*L, where pi is 3.14, f is frequency in Hz and L is inductance in Henry For L = 8nH, and f = 100MHZ, Z = 5ohm, hardly a short circuit. So how low do we need to get the ground impedance? Well, this varies widely with the need, but our observation is that if your ground impedance is over an ohm or so, its effectiveness is questionable. The impedance of a cable end will vary depending on standing wave conditions, ranging from a fraction of an ohm to perhaps 100ohms. So even providing a ground with one ohm impedance may be very inadequate. Achieving a Low Impedance GroundHaving established that we need to have a low impedance ground, how to we go about getting it? We start with the ideal low impedance ground — an infinite plane. By the way, this applies to any ground, not just a cable termination.  Figure 2. Circuit Spreads Out in Ground Strap Obviously, we can't get an infinite plane, but we don't have to — we are only concerned with that portion of the ground path that carries tangible amounts of current. Figure 2a shows the current spreading out between two contact points on a plane. We can cut away a portion of the plane that isn't carrying current without raising the impedance of the path and, in fact, can afford to squeeze the current a little bit as shown in figure 2b without seriously increasing the impedance of the path. Our rule of thumb is that the length of the path should be no longer than five times the width, or a 5:1 ratio. This puts us into the range where the ground starts to work reasonably well — it may not be good enough, either, but it is a good place to start. A few comments are in order. First, the path needs to be significantly shorter than a wavelength for this strap to be effective. The ground impedance reaches a maximum at a quarter wavelength, so you need to be well below that, say lambda/20. Next, the question of whether we can fold the strap in two to minimize the width — sorry, this doesn't work, it just gives you a narrow strap twice as thick, and thickness is not important — it's a question of length-to-width ratio. Next, there is the question of whether a braid is better than a solid strap. Actually, there is essentially no difference in inductance between the two choices — use whichever one you want. A braid maintains greater flexibility, especially where you also need cross section to carry power level currents. But if current capacity is not an issue, the solid strap works just fine. Cable Shield Terminations Figure 3. Cable Terminations As we said earlier, we need to keep the cable shield termination impedance low -- Figure 3 shows the possibilities. The common practice of running the cable shield ground wire through a connector pin to a termination point inside the box is totally unacceptable — this will inevitably result in a ground path of three or four cm, perhaps more, resulting in perhaps. Sadly, foil cable shielding encourages this practice — the drain wire is all too handy. You are doing a little better if you run the termination to a screw at the connector — you may be able to reduce the path length to a cm or two, and even make the strap wider at the same time. This may be good enough, depending on the application — but keep the strap as short and fat as possible. If you are in a position to clamp the cable shield directly to the housing at the entry point, eliminating the pigtail entirely — this is about the best you can do without a circumferential connector termination. In practice, there are many cases where even a short strap won't be good enough. No matter how good a contact you have, you will still have magnetic field coupling between the ground connection and the wires in the cable. The only way to avoid this is to do a circumferential termination — the cable shield is grounded to the connector shell around the entire perimeter. This not only reduces the ground impedance to essentially zero, but it also keeps the magnetic field completely outside the cable. This is the best practice. But you need to be sure the termination is circumferential. Military style connectors are made to facilitate this connection and, even then, may be incorrectly assembled. But commercial style cables are mostly done incorrectly. The only effective way to make a circumferential connection, is to make a compression fitting — certainly feasible with a braided shield, but difficult to do with a foil shield. But the fact is, many commercial cable shields are terminated to a screw post inside the connector — you may think you have a circumferential wrap, but you really don't. Good cables are available, but you won't find them at your friendly computer store. We haven't even touched on the sad fact that many commercial shields are grounded at one end only — the almighty single point ground. The fact is a single point grounded shielded data cable is never appropriate. Period. If you buy a cable at your computer store, check the package. If it doesn't say it is grounded at both ends (and this is most of them), you have a single point useless ground. SummaryLow impedance grounds are absolutely essential. At high frequencies, inductance becomes a major limitation and can be minimized by keeping the ground path short and fat. This is impossible if you use the drain wire for a ground path. Circumferential cable shield terminations are the only fully effective answer, but a short pigtail ground may be adequate for moderate requirements. Finally, never trust a store-bought shielded cable. | advertisement |
