A Pox on Common Mode

by William D. Kimmel, P.E.

and Daryl D. Gerke, P.E.

Kimmel Gerke Associates, Ltd.

We’ve written about common mode (CM) problems in previous articles, but the continuing trend toward plastic enclosures prompts us to revisit this problem – common mode currents are hard to eliminate without a shielded enclosure. So let’s take a look at how common mode currents originate, why they are such a problem, and what you can do about them.

Common Mode Generation

“Common Mode” means that interference currents are common to all associated conductors, which could be all wires or conductors in a signal or power cable or in the circuit board. This is in opposition to “Differential Mode” (DM) currents that follow the expected round trip loop. Common mode currents are not needed for the function of the circuit – they are entirely parasitic. They are generated as a result of shared current paths. Basic network theory doesn’t allow for common mode (or CM) unless you add in stray capacitive current paths to complete the current path. Figure 1 shows the two most common CM generators.

Figure 1a shows the effect of ground impedance – as seen, most of the signal current follows the intended current loop, but some of the current follows an alternate path out the cable, ultimately returning through a stray capacitive path. Yes, that is only a small fraction, but it is enough – emissions due to common mode currents on the cable can easily be 1000 times greater than those of differential mode currents.

Second is common capacitive coupling paths, commonly to the connector, as shown. Currents from noisy ICs couple to nearby metallic members, in this case, the connector. The currents continue out the connector to the cable, with the same effect as above. Interestingly, this path bypasses filters which may be placed on the circuit board.

Magnetic field coupling will also generate common mode, but is a secondary problem on the circuit board.

What Are the Fixes?

Of the two cases, common mode originating from ground impedance is hardest to fix – common mode currents need to be shunted to a lower impedance path or blocked by inserting a high impedance in the path. Common mode currents can be shunted effectively to the enclosure, if it is metallic, either by connecting ground directly to the enclosure as shown in Figure 1, or with a capacitor shunt if a direct connection is not possible. Note that filtering to circuit ground does nothing – the current is already on ground. But if we don’t have a metallic enclosure, this option is not available and, sadly, the remaining options are far less effective. Let’s look at the options:

1. Insert series impedance in-line. This would usually be in the nature of a common mode choke, most often a ferrite. This option is almost always usable, but it has limited effectiveness – you can’t control the impedance of the cable, and you are limited in how much impedance you can realistically get in-line. You might hope for a 10dB reduction at most. The CM choke needs to be inserted between all the coupling paths and the outside world, including any metallic member that may carry stray currents. Unfortunately, these paths often exist right up to the connector, so that often means putting the ferrite on the cable, a choice most prefer to avoid. One place where the clamp-on ferrite may be acceptable is on the charging cable to portable electronics devices – the ferrite is only needed when charging, not when carrying the device.

2. Reduce the return path impedance for critical signals, primarily clocks, buses and video. (If you are working power circuits, the switching power paths would be the ones to watch.) Nominally, this means the ground path, or the power plane impedance. In a plane, the impedance will be as low as you can get, provided the return path is continuous – your immediate choice is to keep the return path length as short as possible. Your best step is to keep the signal path immediately adjacent to the return plane from source to destination. The plane must be continuous – if the signal crosses a slot in the plane, the impedance goes up. This will also occur if your signal switches reference planes, as it might if your signal passes through a via. The return path continuity must be maintained from die to die – the preferred path into the chip may be through the voltage pin rather than the ground pin. You need to make sure the voltage/ground planes are well decoupled near the signal pins.

3. Keep the critical signal currents as low as possible. If you have a choice (as you might with an ASIC), use low current drivers, just enough to drive the load. Otherwise, insert series impedance immediately at the driver to limit the current exiting the chip – currents that don’t go to the load don’t have to return on the ground plane.

4. Provide an alternate common mode return path. As mentioned above, the metallic enclosure is very effective at shunting CM currents, but if you don’t have one, an alternate metallic return path can be inserted, albeit not as effectively as with a metallic enclosure. Basically, the metallic enclosure provides a low impedance return path for common mode currents. An alternate is to lay a plane immediately below the circuit board to provide this CM path – this could be a metallized coating just at the base of the enclosure, a coat of paint, or a sheet of copper tape. The layer should be placed as close to the circuit board as possible to keep the loop areas small. You can even use a layer of the circuit board. In order for this to work, you need to allocate one layer for the common mode return currents, and is not to be used as a signal return current path, otherwise you will generate common mode currents on this plane. This plane would be stitched to the signal ground plane by placing vias at frequent intervals.

The second source of common mode generation is that of capacitive coupling to the connector. This source is more readily handled than the above case. First choice is to maximize the distance between the noisy chips and the connectors going to the outside world (or any metallic member, for that matter). Place the noisy chips as far from the connectors as feasible – coupling capacitance typically falls off with the square of the distance, so even a little spacing is helpful. Microprocessors are usually at the top of the noise list, but memory chips and video controllers are also potential problems.

If you can’t get enough space, you can use chip shields (Figure 2). The shield is placed over the offending chip and grounded around the entire perimeter of the chip – all four corners as a minimum, more if possible. This shield intercepts the currents and returns them to the chip, rather than letting them couple to the connector. It also intercepts any stray coupling to filter components that may have been placed nearby. Chip shields are available in a variety of sizes and shapes. Heat sinks will also work, provided you ground them all around the perimeter.

We have cited the connector as being the culprit, but any metallic member of significant dimensions will serve as an antenna, so be alert for other chunks of nearby metal.

Summary

Common mode currents are usually much more problematic than differential mode currents – even small currents will create significant emissions. Once generated, common mode currents are hard to eliminate, especially if you don’t have a metallic enclosure. Your best approach is to minimize generation of common mode currents at the circuit board level by reducing return current path impedances and minimizing field coupling using selective shields.