Switching-Mode Power Supply (SMPS) and EMI -- Measurements and Limits of Conducted EMI: Differential Mode and Common Mode Noise

• Differential Mode and Common Mode Noise
• How Conducted EMI Is Measured
• The Conducted Emission Limits
• Quasi-peak, Average, and Peak Measurements

Here we take up the concepts of 'common mode' and 'differential mode' noise. Also, regulatory conducted emission limits and related measuring techniques are discussed.

Differential Mode and Common Mode Noise

Initially, we are going to stick to more conventional descriptions of these parameters. But gradually, we will start discussing certain nuances/differences that can arise in applying the concepts to the area of power conversion.

Fgr. 1: Differential and Common Mode Noise

Conducted emissions fall into two basic categories:

++ Differential mode (DM), also called symmetric mode or normal mode

++ Common mode (CM), also called asymmetric mode or ground leakage mode.

Looking at Fgr. 1, 'L' stands for Live (or "Line" or "Phase"), 'N' for Neutral, and 'E' is the "Safety Ground" or simply, 'earth.' 'EUT' stands for Equipment Under Test. Note that the earth is shown represented by the IEC symbol for Protective Earth (ground with a circle around it) and is occasionally labeled 'PE' in literature. The DM noise generator is across the L and N pair. It tries to push/pull a current Idm through these two wires. No current flows through the earth connection on account of this noise source.

Note: There is nothing special about the DM noise current direction indicated in Fgr. 1. It can well be the other way around - that is, going in through either L or N, and coming out of the other. In off-line power supplies, we will see that in fact, the direction reverses every ac half-cycle.

Note: The designer may realize that the basic ac input operating current of the power supply is also differential in that sense - since it flows in through one of the L or N wires and leaves by the other.

However, the Idm shown in Fgr. 1 does not include this component. That is because the operating current, though differential, is firstly not considered to be "noise." Further, its frequency is low (twice the line frequency - 100 or 120 Hz, thus being virtually dc). Even its harmonics are well below the range of standard conducted EMI limit curves (150 kHz to 30 MHz). However, it must not be forgotten that the operating current will dc-bias the noise filter choke, and can thereby adversely affect the performance of any EMI filter (and also of the current probes being used to gather data). So, though we can ignore the ac (line) component in that sense, it still has an indirect effect on the high-frequency input filter.

In Fgr. 1, the CM noise source is shown connected at one end to earth. On the other side, it goes through equal impedances to each of the L and N lines. It will therefore also drive equal noise currents into the L and N wires - and in the same direction. However, note that that assumes equal line impedances too. We realize that if the impedances are unbalanced, we will get an "asymmetrical common mode" current distribution (in the L and N wires). And that is, in fact, a common scenario in actual power supplies. Also note that this "asymmetrical common mode" is equivalent to a mixture of true-CM mixed with some DM (we will demonstrate that a little later).

Note: Since CM noise is itself often called "asymmetric," it’s preferable to call this type of operating mode "non-symmetric" rather than "asymmetric."

Engineers often instinctively tend to disregard common mode noise present on the output of their converters. So a typical output noise and ripple measurement is always deliberately differential in nature - we spend a long time trying to get the oscilloscope probe positioned correctly on the output terminals (with minimum length of probe ground-wire), simply to avoid picking up common mode noise. But suppose the converter is providing power to an actual subsystem (not a resistive "dummy load"). Looking into the input of this subsystem, we will rarely (if ever) see equal (balanced) impedances (i.e. from each of its input terminals to the earth ground). So what happens is that any "common mode" noise existing previously on the output rails of the converter becomes a differential input voltage ripple (of high frequency) for the subsystem. In other words, common mode noise gets converted into differential mode noise, if the line impedances are unequal. Therefore, it’s no surprise that if an unbalanced filter (e.g. a choke present in only one of the two lines) is present at the input of the subsystem, it will only make matters worse. And further, no amount of CMRR (common mode rejection ratio) in the subsystem will help either. True, the subsystem will usually contain a front-end line-to-line input capacitor that will help decouple some of this incoming DM noise. But eventually, the subsystem can start misbehaving. Therefore, reducing common mode noise at the point of creation is always the highest priority.

Thereafter, equalizing the line impedances becomes important. The latter can often be achieved by placing balanced filters at the input of the subsystem (e.g. with two inductors, one on each input line).

Note also that since by the very nature of their creation, common mode currents usually have a much higher high-frequency content than differential mode currents, they also have the capacity to cause severe radiation (besides also causing inductive and capacitive coupling to nearby components and circuits). In fact, it’s often said that the rule-of-thumb is that a mere 5 µA of common mode current ina1m length of wire can cause FCC Class B radiation limits to be violated. For FCC Class A limits this number goes up to 15 µA. Note also that the shortest standard ac power cord is1min length.

To avoid confusion, we should note that the net common mode current going through the Earth is called "Icm" in our case (Icm/2 in each line). However, in related literature, this is often called "2Icm"(Icm in each line).

Note: There is nothing special about showing the CM noise current in Fgr. 1 as coming out of the equipment (through both the L and N wires). It could well be in the reverse direction. And like DM noise, it too could be sloshing back and forth, depending on what part of the incoming ac half-cycle we are on, at a given moment.

Note: We will see that in an actual power supply, differential mode noise is initiated by a swinging (pulsating) current - but the DM noise generator is itself closer to a voltage source. On the other hand, current mode noise is initiated by a swinging voltage, but the CM noise generator itself behaves more like a current source. That is actually what makes common mode noise so much more "stubborn" - like any current source, it demands a path to flow through. And since its path can include the chassis, the enclosure can itself become a large high-frequency antenna.

Let us now do some simple math to split the measured "non-symmetric" currents in the L and N lines into true-CM and DM components. To avoid algebraic errors, we first establish a convention for what is a 'positive direction' for the current flow. Let us assume that in Fgr. 1, the direction from right to left is a positive direction, and left to right is negative.

We also keep in mind that a current 'I' flowing in one direction on any wire is equivalent to '-I' in the other direction (on the same wire).

Suppose we measure 2 µA going from right to left in one wire (say, the L wire). Then we measure 5 µA going from left to right in the other wire (N). We want to estimate the CM and DM components from these two measurements only.

We have by definition (from Fgr. 1)

Solving these simultaneous equations...

This means we have a current of 3 µA flowing from right to left in E (common mode component). And we have 3.5 µA (differential mode component) flowing from right to left in L, and left to right in N (differential mode component).

Suppose we measure a current of 2 µA flowing from right to left in the L wire, and no current in the N wire. Estimate the CM and DM components from these two measurements.

We similarly get ...

This means we have a current of 3 µA flowing from right to left in E (common mode component). And we have 3.5 µA (differential mode component) flowing from right to left in L, and left to right in N (differential mode component).

Suppose we measure a current of 2 µA flowing from right to left in the L wire, and no current in the N wire. Estimate the CM and DM components from these two measurements.

We similarly get ...

So a non-symmetric mode can be considered part asymmetric (CM) and part symmetric (DM).

Fgr. 2: Simplified Schematic of LISN -- (N) MEASURE

a) IL TO GROUND, (VL )

b) IN TO GROUND, (VN )

c) If different take higher of the two

Fgr. 3: Impedance Presented to the CM and DM Noise Generators by the LISN: DM Load Impedance = 100

CM Load Impedance = 25