Handling Your Internal Cables

by William D. Kimmel, P.E.

and Daryl D. Gerke, P.E.

Kimmel Gerke Associates, Ltd.

We’ve written a lot about external cables, including shielding and terminations, but haven’t done much with internal cabling. Surprisingly often, the internal cable is a key factor in EMC, both for emissions and immunity. While the best time to address this is during the design phase, often the problem can be solved when troubleshooting at the test lab. Let’s take a look at the problem and some possible solutions.

The Problem

Two common problems occur with internal cabling, ground impedance and adjacent element coupling.

The first enemy is ground impedance, which occurs in any interconnect situation, even between motherboard and daughterboard, where the interconnect path is quite short. No wonder that you have a ground impedance problem when the cable run lengths can easily be a third of a meter or longer. Figure 1a shows a number of signal lines sharing a few ground lines. All of the lines have impedance, but the most significant is the ground impedance – any common mode current traveling along the cable will result in ground bounce. Interference is usually common mode, but common mode rejection has its limits, with the end result being common mode to differential mode conversion. Similarly, driving high speed signals down a cable creates a common mode situation on the ground path. In extreme cases, the ground bounce due to signal returns will be enough to create a data error.

A second aspect (also Figure 1a) is the loop area created by the location of the signal and its nearest ground wire, the area being proportional to the spacing between signal and ground. The antenna efficiency is proportional to the loop area for both emissions and RFI.

In addition to intra-cable effects, you have coupling effects between cable and an adjacent cable, to nearby structural members and to the enclosure seams. Figure 2 shows this effect.

Design Your Cable

There are two things you can do to make your cable perform well – fight for as many ground pins as you can get and distribute them wisely throughout the cable. You need to make this decision early in the design phase, before the pinouts are set. Sadly, no one wants to waste pins on ground – clearly, one ground wire is enough, isn’t it?

From an impedance standpoint, your ground bounce will be inversely proportional to the number of ground wires in the cable – this applies to bounce produced by signal return and also to that supplied by an external transient source. It also applies to emissions caused by common mode voltage generation. So do yourself a favor, use plenty of ground pins.

How many pins are needed? This involves many variables, such as length, data speeds, and amount of noise tolerable. But for common data buses running in the MHz range, you would do well to budget one ground for every five data bus signal lines. High speed lines, say those upwards of 100MHz, need one return per signal line.

Once you have settled on the number of ground pins, use them wisely – don’t bunch them all together. As a minimum, distribute the ground lines roughly evenly along the cable (Figure 1b. But better, yet, make sure your critical lines (e.g., high speed clocks) are adjacent to a ground pin. If you are piping DC voltage down the cable, treat them as though they were a ground line, and decouple them at both ends.

If you are stuck with the pinouts, your next best option is to add a microstrip line to the cable. This is a ground strap molded alongside the signal wires. In order to be effective, the ground strap needs to be connected to the ground wires in the cable at both ends. Implemented well, this approach can be very effective in controlling ground noise.

Position Your Cable

Whether or not you have any control over the cable design itself, you can still accomplish a lot by controlling the cable routing. There are two primary effects of interest – coupling to adjacent cables and electronic members and coupling to the enclosure skin. The most notable electronic members are the active electronic elements – computer clocks and switching power supplies make dandy emission sources; reset lines and edge-triggered devices make dandy receivers. If you route your cable immediately above these components, you are really asking for trouble.

Away from the board, you will be interested in cable-to-cable coupling. The best control method is distance, whether you are talking electric or magnetic field coupling, but in close quarters orientation is a significant factor.

Antenna effects can also be reduced by routing critical cables alongside a ground plane. Particularly, when cables are routed helter-skelter inside an enclosure, crosstalk and EMI effects are unpredictable, and will vary from box to box. We shudder to think of how a personal computer ever passes an emission test – every time you open the box and move the cables a little bit, you have a whole new emission profile.

A word of caution – when routing cables against a ground surface, make sure the surface is continuous underneath the cable. Cables routed immediately adjacent to gaps in the ground create very effective slot antennas. This is particularly problematic when the ground surface is the skin of the enclosure.

As a matter of practice, we generally discourage using the enclosure skin as a ground plane for cable routing, but often that is the only plane you have. Where the shield is well-built, this is of little consequence, but most commercial boxes have mediocre shields. The best you can do, then, is to avoid the seams, particularly the cover.

One thing about equipment with a number of internal cables – you probably don’t have the room to place all cables in desirable locations. Fortunately, you will often find that one or two cables are the most prone to emit or receive energy, so you may be able to resolve the problem by directing your attention to those specific cables. You may have to experiment a bit, however, as it is not always obvious which cables are the problem.

As a last resort, you can always use an internal shield over selected cables. Usually, the shield is best grounded to enclosure ground. Multipoint grounding is preferred for high frequency RF and transients, but for the short runs inside the enclosure, a single point ground may be adequate. Avoid pigtail grounds – clamp the shield to ground for best results.

Summary

You can avoid a lot of emission and immunity problems if you take a little care up front in the design – get enough ground wires and spread them out over the entire cable. But even if you encounter a problem on the test floor after the design has been pretty much cast, you can still accomplish a lot by repositioning or shielding internal cables. Just to note, this is not a replacement for good external cable shielding and terminations.