Measurements and Limits of Conducted EMI: How Conducted EMI Is Measured

For measuring EMI, we need to use an ISN ('Impedance Stabilization Network'). In off-line power supplies, this becomes a LISN (Line Impedance Stabilization Network) - also called an AMN (Artificial Mains Network). See Fgr. 2 for a simplified schematic. Note that the LISN, as recommended for CISPR-22 compliance, is detailed in CISPR 16.

The purpose of the LISN is multifold:

++ It’s a source of clean ac power to the power supply.

++ It provides data to the measurement receiver/spectrum analyzer.

++ It provides a stable, balanced impedance (as seen by the noise signals emanating from the power supply).

++ Most importantly, it makes the measurements repeatable anywhere in the world.

Note that from the viewpoint of the noise generators in the power supply, it’s the LISN that forms their load.

Let us assume that the values of L and C used in the LISN are chosen so that the following statements hold true, unequivocally:

++ The inductance L is low enough not to impede (ac) line current (50/60 Hz) - but high enough to be considered 'open' over the frequency range of interest (150 kHz to 30 MHz).

++ The capacitance C is low enough not to pass the ac (line) voltage - but high enough to appear as a 'short' over the frequency range of interest.

Note that Fgr. 2 doesn't really represent the LISN per se, but is really the equivalent schematic required for calculating the level of noise picked up by the receiver. So in fact, the impedance of the cable + receiver has already been factored in - into the values of the components shown (namely the 50 ohm resistors). We know that a typical coaxial cable going to a measuring instrument (analyzer/receiver or oscilloscope etc.) presents a 50 ohm impedance to a high-frequency signal (because of transmission line effects). So when the receiver is measuring the noise, say between L and E, the LISN actually uses a relay/switch to place a real resistor of 50 ohm across the opposite pair, that is, between N and E. So in that case, the "50 ohm" shown across L-E is in reality just the impedance of the cable going to the receiver.

In this way, as we toggle the switch to measure either VL or VN, the lines are kept balanced at all times. Note that the choice of 50 ohm also simulates the impedance of typical mains wiring to high-frequency signals. But in any case, the procedure makes the measurement virtually

"blind" to the actual impedance of the mains wiring - making it repeatable in any location.

To know what the measured voltages VL and VN are, we now look at Fgr. 3. The voltage due to the common mode component is 25 Ohm multiplied by the current flowing in the earth connection (i.e. 50 ohm times the current in each leg). The voltage due to the differential mode component is 100 Ohm times the differential mode current. Therefore the LISN provides the following load impedances to the noise generators (in the absence of any input filter)

++ The CM load impedance is 25 Ohm.

++ The DM load impedance is 100 Ohm.

As we flick the switch on the front panel of the LISN, we will measure the following noise voltages:

VL = 25 × Icm + 50 × Idm or VN = 25 × Icm - 50 × Idm

Fgr. 4: The LISN MATE

Fgr. 5: A CM and DM Separator

Both the VL scan and the VN scan obviously need to comply individually with the limits.

But how different can the VL and VN scans be? In fact, the above two equations have inspired a rather misleading statement often found in related literature - "if the noise emission is predominantly DM, the VL and VN scans will look almost the same. The scans also look identical if the noise is predominantly CM. And if the VL and VN scans look very different, that implies that both CM and DM emissions are present." However, in the case of an off-line power supply, this statement is clearly not true. Because, that would imply that somehow the emissions on the L and N lines are different. However, we know that in any typical off-line power supply (with an input bridge rectifier), the L and N lines are essentially symmetrical - both from the viewpoint of the operating current and therefore the noise spectrum. So every successive ac half-cycle, the operating current, and the noise distribution get transposed from one line to the other. True, at any given moment, the noise on L will be quite different from that on N, but when averaged over several ac cycles (as any spectrum analyzer would do), equality (symmetry) is restored. Any remnant differences between the VL and VN scans can be traced back to some undocumented asymmetries between the two halves of the test circuit, or some severe radiation source impinging asymmetrically on the cables or traces close to the inlet of the power supply.

We observe that the standards don’t require us to measure the CM and DM components individually, but rather a certain sum as described in the preceding equations. However, there are times when engineers do want to see both the CM and DM components separately - for troubleshooting and/or diagnostic purposes. So various people have come up with clever ideas to separate the CM and DM components. Some of these are mentioned below.

++ A device called the 'LISN MATE' is rare now. It was invented by an engineer named Nave. It provides about 50 dB attenuation for the DM component, but the CM component comes right through (slightly attenuated - by about 4 dB). The schematic for this is shown in Fgr. 4.

++ The transformer-based device shown in Fgr. 5 exploits the fact that common mode voltages cannot cause transformer action - because transformer action requires that a differential voltage be applied, so as to produce current in the windings, and thereby causes the flux to swing within the core. Unlike the LISN MATE, in this case both CM and DM noise components are outputted. This device used to be available from AEMC in France, at www.aemc.fr.

++ Both methods above unfortunately require modifications to the standard LISN - because they invoke a certain simultaneous math between the VL and VN components. However, a LISN normally provides either VL or VN at any given moment - not both (at the same time, as required here). We can modify the traditional LISN, but that is not only tricky to do, but also hazardous because of the high voltages involved. Therefore, a completely different approach is simply to buy a LISN explicitly designed for the purpose of providing separate CM and DM noise scans (besides providing the necessary "summed up" scan for achieving compliance).

An example of such a LISN is the ESA2000 from Laplace Instruments at www.laplaceinstruments.com.

Table 1: Conducted emission limits

Fgr. 6: CISPR 22 Limits Plotted Out