Cable Crosstalk

By William D. Kimmel, P.E.
and Daryl D. Gerke, P.E.
Kimmel Gerke Associates, Ltd.

We don’t write much about crosstalk – not because crosstalk isn’t important, but because it is so difficult to quantify. The simple models using lumped circuits have been shown to be so inaccurate that we refuse to use them. If you really want to do crosstalk calculations, get a computer and the latest software (that is one place where EMI software works quite well).

But we find that most cable crosstalk problems don’t need detailed analysis – you just need an idea of whether crosstalk might occur, and how to handle it. The good news about cables is that they can usually be changed without modifying the equipment itself – this is really important when a problem is uncovered in the field.

There are two key differences between cable and circuit board crosstalk. The first is that while crosstalk in the circuit board is often a problem, there are still many circuit boards built that simple don’t experience a crosstalk problem. With cables, however, crosstalk is almost always a consideration.

The second is that while capacitive crosstalk tends to dominate in circuit boards, inductive crosstalk dominates in cables. This brings to light a major consideration when dealing with cable crosstalk – most ground treatments involve a single point ground, which works well for low frequency capacitive crosstalk, but is not appropriate for inductive crosstalk.

So when you buy cables, you will often find the need for shielding. The good news is that there is no shortage of shielded cables. The bad news is that there is little guidance on how to handle the shields, leaving us to uncover a lot of mishandled shields – when you buy a shielded cable in the local computer store, you don’t always know how they are terminated.

So let’s take a look at a number of scenarios in shielded cables:

The Single Shield

If you have a homogeneous cable with a single shield, you don’t have much intra-cable crosstalk protection – your concern is simply to keep inside energy in and outside energy out. But do you ground at one end or both? And what’s this talk about pigtail terminations?

The tried and (sometimes) true rule of single point ground is appropriate only for low frequency applications, including audio frequencies. This approach originated with telephones, and continued with stereo sound systems and sensitive instrumentation, where the principal concern was keeping 50-60Hz hum out of the input wires. With high impedance input circuits, electric field coupling is the key concern, so a single point ground is adequate, and even preferred. A multiple point ground will create a ground loop – a potential problem. And at audio frequencies, the pigtail isn’t a concern – the inductance of the pigtail is of no consequence.

The single point ground principle breaks down when you get to higher frequencies, where the cable starts to look like a distributed element. At that time a single point grounded shield looks more like an antenna than a shield. So the shield needs to be grounded at both ends when the wavelength of the highest problem frequency approaches 1/20 wavelength. Also, the cable termination becomes a significant factor – at 100MHz, a one inch pigtail has about 20nH of inductance, and about 12 ohms of impedance. So we need to eliminate the pigtail, preferably with a circumferential termination such as you would get with a compression fitting.

The multipoint ground also works well for low frequency magnetic fields as well, an increasingly important consideration. If you are dealing with magnetic fields, you need to terminate both ends to provide a return path to cancel the magnetic field. A single point ground does not provide the return path so cancellation cannot occur. Well, such a need doesn’t show up very often, does it?

It turns out the answer is, increasingly, “yes.” The reason lies in the increasing use of motor drives. The old days of DC or sine wave AC motors are coming to a close, what with the modern motor drives used for promoting high efficiency and precise speed and position control. But this means you are driving square waves down the cable. And with substantial voltages and currents.

Now, consider a typical three phase drive. Typically, you switch one phase at a time, so you always have a return path: there is at least one phase sourcing current and one phase sinking current. But what happens at the transitions – a glitch. Big time, and that returns on a fourth path, typically a ground wire, better yet, a shield grounded at both ends. Looking at the cable from a distance, the currents cancel fully, and the associated magnetic fields cancel almost completely (complete cancellation, unless you have spatial concentricity, only happens with coaxially spaced conductors). You can’t get that condition except for special cases (obviously including coax), but you can get closer with a circumferential shield than you can with a simple adjacent ground wire. You’ll get even better cancellation by twisting the wires, then shielding, but that isn’t practical for most power level applications. So, for magnetic field shielding in motor drives, ground both ends of the shield.

Note that the shield needs to surround the entire set of power conductors. If there are shields around the individual power conductors, they can be single point grounded to block electric fields, but this won’t block magnetic fields. Yes, you can block those magnetic fields by grounding that shield at both ends, but you’ll end up carrying a lot more current on the shield than you would like.

Multiple Shields

You may be faced with driving a variety of power and signal wires between the source and the load. We have already mentioned the big noise generator of the bunch, the motor drive. But you may be running just about any kind of signal between the two boxes, including low level analog, serial data, encoder data or even video.

The best solution to this situation is to have separate cables for each kind of signal or power wire. A little spacing does wonders in minimizing crosstalk. But if you have decided you need to put everything in one cable, now what? Now, you have crosstalk issues big time, and you are going to need several sets of shields within the cable.

You may well be in the position of having multiple shields surrounding your conductors. Now you have the best of both worlds. You single point ground to block electric fields and two point ground to block magnetic fields. Generally, you ground the inner shield at one end, and the outer shield at both ends (see the figure). Given a good termination on the outer shield, it will block high frequencies as well as low frequency magnetic fields. The ground voltage produced by the ground loop is intercepted by the inner shield.

Summary

Cable shielding needs to be handled differently for different needs.

Low frequency high impedance circuit shields are best handled by a single point ground (usually, you can ground at either end, but there are cases where the ground works better at one end than at the other). The termination can be pigtailed without adverse consequences.

Low frequency low impedance circuit shields (such as motor drives) are best handled by grounding at both ends. The termination can usually be pigtailed, but we advise against long, indirect pigtails – they may carry currents and associated magnetic fields.

High frequency circuit shields (longer than 1/20 wavelength) are best handled by grounding at both ends. Pigtails are to be avoided – circumferential termination is best.

Multiple shields can be used to achieve the best of both worlds.

If you are working in the field, don’t be afraid to try some alternate grounding techniques – sometimes conventional logic doesn’t apply.