Estimated reading time: 5 minutes

Current Flow on Traces, Part 2: Common Mode and Mode Shift

If you ask someone what is the difference between differential mode and common mode (on a differential trace pair), you might get an answer like this:

Differential mode is when the current goes down on one trace and returns back on the other. Common mode is when current flows in the same direction on both traces.

This answer is sort of correct, but there is a dangerous misconception conveyed here. If you have read my article “What is Current And Why Do We Care?” (Note 1), you will (hopefully) remember that current flows in a loop. The definition of common mode, above, says nothing about the return current. But there must be a return current (there always is), and it is flowing somewhere. The problem with many common-mode currents is that we don’t know where they are flowing, and if they are flowing in a large enough loop, they present a very possible EMI problem.

There is another definition that I really like, and it’s only partly tongue-in-cheek. It goes like this:

Differential mode is when the signals go where we expect them to, based on the traces and the schematic. Common mode is all the other signals! (Note 2)

Differential vs. Common Mode

Let’s take a short step back and look at differential and common mode signals by themselves. If we have a differential trace pair, and the signal on the return trace is exactly equal and opposite to the signal on the forward trace, there is no common-mode component. If they are not exactly equal and opposite, there will be a common-mode component to the currents on the trace. Carrying that idea one step further, any combination of currents on a differential trace pair can be broken into differential and common mode components (See Figure 1.)

Figure 1. Differential-mode and common-mode components on a differential trace pair.

Id (Figure 1) represents the differential-mode component. Ic represents the common-mode component. Id + Ic is the current on the top trace. Ic – Id represents the current on the bottom trace. The return current for the common-mode component is represented by 2xIc flowing on an undefined path. Suppose we hypothesize the following values:

Current on top trace = 20 maCurrent on bottom trace = -20 ma

It follows that Id = 20 ma and Ic = 0.0 ma. This is a case where the currents on the two traces are exactly equal and opposite. The differential-mode component is 20 ma (on each trace) and the common-mode component is zero (on each trace.)

Now let’s let the currents be unequal on the traces. Let:

Current on top trace = 20 maCurrent on bottom trace = -19 ma

In this case, the differential-mode component, Id, becomes 19.5 ma and the common-mode component, Ic, becomes 0.5 ma. The common-mode return current is 2x0.5 = 1.0 ma flowing on some other (probably undefined) path. It is this return current loop that poses an EMI danger.

Let’s look graphically at what might happen along a trace. Consider the two square waves illustrated in Figure 2. Let the red wave be the positive driver output and let the black wave be the negative driver output. The blue line is the algebraic sum of the two waveforms. Since the two waveforms are exactly equal and opposite (i.e., 180 degrees out of phase) the algebraic sum is zero. There is no common-mode component.

Figure 2. Two waveforms exactly equal and opposite. The algebraic sum is zero.

Now let the positive waveform lag the negative waveform by a few degrees. This very slight difference between the signals could be caused by a slight offset in the driver, or perhaps a slight difference in trace lengths. Figure 3 illustrates the results.

Figure 3. A slight delay in the positive signal produces a significant common mode component.

As a result of the slight timing difference between the two signals (here the black waveform switches slightly before the red one does), the signal representing the algebraic sum (the blue waveform) is quite pronounced. It is the blue waveform that represents the common-mode component and since its return path is not controlled it may be causing a very dangerous EMI problem. (See Note 3.)

How Common-Mode Components Develop

So the question is, why would the currents on the two traces be different in the first place? Well, there can be several reasons:

First, even the best differential drivers can have a small offset between their positive and negative outputs. If there is a slight difference in timing, a series of common-mode pulses would then be generated during the times each signal was switching. If there were a slight difference in amplitude, then a slight waveform similar to one of the square waves would be generated as a common-mode component.

Second, even if there is no offset at the driver, a difference in trace length may cause an offset to develop. If a differential pair turns a corner, the inside trace follows a shorter path than the outside trace.  As a result, the timing changes and a series of common-mode pulses would be generated at each switching point. Such a situation is called mode shiftor sometimesmode conversion. Mode shift occurs any time the signal relationship on the two traces changes, causing a change in the algebraic sum signal component (the common mode component.) It is often desirable to add a little trace length to the inside trace to equalize the path lengths between the inside and outside traces.

Third, if there is a change in propagation speed between the traces, one waveform will slow down with respect to the other. One example of a difference in propagation speed is known as the weave effect (Figure 4). Propagation speed is inversely proportional to the square root of the relative dielectric coefficient of the material surrounding the trace (see Note 5). The two traces illustrated in Figure 4 are routed in different environments as far as the material goes, so their propagation speed will be different. This can also result in mode shift, and a changing common mode component.

Figure 4. The weave effect can cause different propagation speed on the two traces.(Image courtesy of Bert Simonovich's Design Notes)

Conclusion

The terms “mode shift” or “mode conversion” relate to anything that has resulted in a change in the algebraic sum of two signals. A change in the algebraic sum results in a change in the common-mode component. Any increase in the common-mode component may be troublesome from an EMI standpoint. Therefore, in our board layouts, we want to (try to) avoid anything that would contribute to mode shift.

Notes

1. See “What is Current And Why Do We Care?”
2. This quote was given to me by a person who attributed it to Paul, Clayton, "A Comparison of the Contributions of Common-Mode and Differential-Mode Currents in Radiated Emissions," IEEE Transactions on Electromagnetic Compatibility, Vol. 31, No.2, May 1989.
3. The example in Figure 3 might be criticized as being exaggerated.  After all, many waveforms don’t switch with as fast a rise time as suggested here. Nevertheless, the point is made that it does not take much of an offset to generate a sharply defined common mode component in digital circuits.
4. Picture in Figure 4 taken from “Bert Simonovich’s Design Notes”
5. See my article “Propagation Speed in Microstrip: Slower Than We Think”