Common Mode Noise

by William D. Kimmel, P.E.

and Daryl D. Gerke, P.E.

Kimmel Gerke Associates, Ltd.

Common mode noise is at the root of both emissions and susceptibility currents. EMI begins and ends at a circuit, but coupling paths are primarily common mode. Common mode interference is much more of a problem than differential mode interference, as it is much harder to get rid of, so your best line of defense is to design so as to minimize the generation of common mode in the first place. The good news is that common mode is entirely parasitic – you don’t need it for your equipment to function.

Let’s take a look at how common mode is generated and what to do about it.

Common Mode Generation

First, we need to define common mode, as opposed to differential mode. Figure 1 shows a cable, with signal path and associated return paths plus a current component that is common to all the wires in the cable. The current goes down the cable and the return is through an undocumented stray path, most commonly via a stray capacitive path to ground. So how does this originate?

The most prevalent case is due to ground impedance. Figure 2 shows a signal path, with currents following the normal round trip. This is the “normal” or “differential” mode. Now, all conductive paths have some impedance – there is no such thing as a zero impedance path, even though the text books keep on making that assumption. In practice, there will be some resistance at any frequency, and as the frequency increases, inductive impedance becomes a major factor. As the figure shows, the primary current path is through the desired return path, but a small portion of the currents divert and go the other direction. This will be a very small percentage in the case of a planar return path, but even one part in one thousand is enough to create a problem.

Unless the current is blocked or diverted, it continues on out the cable, to radiate from there. This, then, is the common mode problem.

The second case is due to field coupling, both capacitive and inductive. Any electric or magnetic fields generated in the course of normal operation will couple to surrounding metallic elements. While the currents may be differential mode, the coupling paths are mostly common mode.

Figure 3 shows electric field coupling from a noisy chip to a nearby connector. The connector pins intercept the currents and carry them out to the outside world.

We have a similar condition with external interference, which is largely common mode. A perfectly balanced circuit will, in principle, ignore common mode currents. In practice the cancellation is limited to the balance of the circuits and by the dynamic range of the input circuit. Even the best circuit has some imbalance – assault your circuit with a thousand volts of common mode, as you would with one of the transient tests, and a 1% imbalance will be a significant problem. And while differential amplifiers are pretty good at audio frequencies, their CMRR degrades at higher frequencies.

What To Do

For emissions, the first line of defense is to minimize the generation of common mode currents. On the circuit board, the key is Ohm’s law: E = I*Z. On the circuit board, you need to reduce the ground voltage by reducing the current in the path or by reducing the impedance in the path.

Note that we are talking about the impedance in the ground path – that which is connected to the outside world. If you are in a position to steer the offending currents along another path, you can reduce the current in the problem path. This, in fact, is the principle in single point grounding – you provide an alternate path for the currents, keeping them away from the problem path. Unfortunately, this approach is difficult to effect at higher frequencies – it is difficult to adequately avoid the stray paths. And at still higher frequencies, antenna effects start to become significant.

So, for practical purposes, your usual choice is to reduce the ground impedance. There are basically two feasible approaches. First is to keep the offending signal path as short as possible, so that the return path is as short as possible. The second is to make sure the return path is continuous – any gaps in the plane will significantly raise the effective impedance, not to mention creating a local antenna.

For emissions, the big problem is the periodic waves, most notably the clock and high speed buses, so concentrate on those signal traces.

For immunity, the problem will mostly originate from off board and enter the board via a cable. Here, since the current is largely common mode, it is basically ground noise. As long as the ground impedance is low enough, the common mode will have little effect on the ground plane, but will ultimately show up downstream wherever the ground impedance is evident. Again, the slots on the circuit board pose a potential problem, and the problem is not limited to any particular category – upset any signal, and you have a likely problem. If the ground impedance is satisfactorily low, then the next potential problem is to any connectors attached downstream of the noise. While a ground plane impedance is usually adequately low, ground impedance through the connector is always much higher, high enough that any transient currents will bounce the ground across the connector.

Ideally, we could reduce the impedance across the connector by assigning more ground pins. While this is a good idea, it suffers from two basic limitations. First, pins are always at a premium, so you are going to have to fight to get more (What? Waste all those pins on a ground that doesn’t do anything?). But no matter how many pins you have, the impedance across the connector will be much higher than that on the ground plane, so that is the place to concentrate your efforts – if you can increase the number of ground pins from two pins to four pins, you will reduce the impedance to half, and every bit helps.

More likely to be feasible is to implement a common mode choke at the cable input or at any other associated connector boundary. A CM choke is basically a 1:1 transformer, with as many conductors as needed to accommodate the cable. All conductors pass through the CM choke, including the ground wires. The idea is to insert as much series impedance as you can, to reduce the common mode currents. Note that differential mode currents are essentially unimpeded in a CM choke, so signal degradation is usually not a concern.

Of course, if you have a filter connector and a metal enclosure, you can divert the currents from the signal line to the case, before the current can get to the circuit boards.

The other common mode generator is field coupling. While both electric and magnetic field coupling does occur, electric field coupling dominates on the circuit board. The path, such as shown in Figure 3, can be intercepted with a Faraday shield interposed between the noise source and the metallic receiving element. This is commonly accomplished using a chip shield, but is sometimes handled by building a fence between the chip and the connector.

Field coupling paths are much worse in power supplies and converters, where the components are stacked up and placed in close proximity. The key is to identify the noisiest node and arrange the attached components so as to minimize coupling to elements that go to the outside world. As magnetic elements are more commonly found in power supplies, you need to look for magnetic field coupling paths around the transformers and inductors. You also need to keep loop areas small in the main power paths.

Summary

Common mode currents are difficult to get rid of – the best approach is to minimize generating them in the first path. Your key lines of defense are to keep the ground impedance low, paying particular attention to the return path of the critical circuits. For field coupling paths, look for the key noise generators and lay out the board so as to minimize stray capacitance to those circuits.