In this lecture, we will look into more details related to measurement of conducted electromagnetic interference, and the use of the line impedance stabilization network or LISN. We'll talk about differential-mode and common-mode EMI, and we will illustrate the discussion using a simulation example based on a boost DC-DC converter. Let's look at the LISN first, what is shown here is one-half of a complete LISN. On the left-hand side, a LISN is connected to a power source, on the right-hand side, it's connected to the equipment under test, and in between, LISN operates as a filter. You can see that any AC current that is going into the equipment under test, would mostly be close through this capacitive coupling to the output port of the LISN, whereas only a little bit of current would be flowing through the series inductance. So most of the AC noise, generated by the equipment under test, is diverted to the output port of the LISN. At the output port, we connect a spectrum analyzer to investigate the frequency domain characteristics of the noise generated. A spectrum analyzer standard input impedance is 50 ohms, so the voltage across that 50 ohms will be the voltage that is going to be evaluated in frequency domain to see where the equipment under test is meeting standards or not. The purpose of the LISN is to standardize the test environment for EMI testing. Without the LISN, the results would heavily depend on the characteristics of the particular power source at the site where the equipment is connected, and it would be an unfair measurement, so to make the measurement fair and uniform, we insert the LISN for the purpose again, of standardizing the test setup. In the switch.lib library that we use for simulations, we have included a netlist for this half LISN circuit, and the netlist is simply showing elements that are shown on the circuit diagram. We should immediately know that this is only one-half of a complete LISN circuit that's connected between the line and the chassis ground. A complete LISN, as you will see in a moment, consists of two of those circuits connected back-to-back. Here, is a simplified diagram of a complete test setup for the case when the equipment under test is a DC-DC converter. Let's suppose on the left-hand side, we have a DC input voltage source, between the input voltage source and the DC-DC converter, we insert a LISN. You see that LISN consisting of two-half circuits connected together in the middle, to a point that is then connected to a common or chassis ground across the test setup. On the input side, a DC-DC converter would typically include some type of filtering, let's suppose we have just a decoupling capacitor on the input side. Now, when you look at this diagram here, it appears as if the entire circuit is just floating above the chassis ground, and in fact, that is indeed the case. But what will, in fact matter, is the fact that the source and a DC-DC converter and the interconnects between them, all must have some amount of parasitic coupling to the chassis ground that is capacitive in nature. To show that explicitly, we would visualize that we have capacitors here connected between the power lines to a chassis ground, and similarly on a DC-DC converter side, there's going to be some amount of capacitance connecting the power lines to the chassis ground. Then from the internals of the DC-DC converters, we'll have likely multiple, parasitic capacitance connecting various nodes to the chassis ground. Output ports of the LISN, there would be two of those, one of those is going to be connected to a spectrum analyzer to evaluate the amount of noise present at this node. The other node, which is symmetric, around the chassis ground, is going to be terminated by a standard 50 ohm impedance. The spectrum analyzer can be connected to either of the two nodes, and in fact, standards require meeting the constraints or the limits imposed by the standard at either of the two LISN signal output nodes. Now is a good opportunity to look into two types of noise that equipment under test or a DC-DC converter would generate. One is called Differential-Mode-EMI, and is perhaps the closest one to the way we think about ripples generated by a DC-DC converter. On the input side, the input current will consist of a DC component plus some amount of AC component that represents the residual ripple visible on the input side. That AC component of the input current is the differential mode electromagnetic interference, and that AC components as shown on the diagram, is going to be mostly diverted to the output port of the LISN, and would be then evaluated by a spectrum analyzer. Notice the direction of differential mode current flowing through the LISN into the DC-DC converter and through the return path of the DC-DC converter. So there's plus and minus terminals for the DC-DC converter, and the differential mode current is closed, closing through those plus and minus terminals and back through the LISN device. At a two ports, the magnitudes of the voltage is generated by the differential mode AC current are going to be the same, but opposite in sign. Now there is another component to the electromagnetic interference that is called Common-Mode EMI. The Common-Mode EMI originates from these parasitic capacitive couplings that we talked about a little bit earlier. Here we are going to focus on parasitic capacitive coupling from the DC-DC converter itself to the chassis ground. A typical physical reason for that type of coupling would be, for example, from a switching node of the power converter, which may be a drain of a power MOSFET to the heat sink, and the heat sink being connected to the chassis ground is then providing that conductive path. If inside a DC-DC converter we have a switching node with high pulsating voltage waveforms with fast transitions. Those fast transitions are going to induce currents flowing through these parasitic capacitive coupling components to the chassis ground and will create what is called a common mode current. The distinguishing feature of the common mode electromagnetic interference, is that from the equipment under test point of view, that current flows in the same direction on both terminals of the equipment or DC-DC converter. So if you have a first transition right here, let's focus on this falling transition of the switch node waveform. That falling transition is going to create a current spike through the parasitic coupling capacitance. That spike is going to flow through the chassis ground, then back through the LISN, the midpoint of the LISN, and then through the AC coupling capacitors inside the LISN in the same direction, in both sides. So in this case, again, we will evaluate the amount of noise with the spectrum analyzer, and you can notice that the two output ports have exactly the same amount of common mode noise. These two are the same and the same sign. Here as an example, we have the EN 55022 class B limits. Class B limits are shown here in two different ways the spectrum analyzer can be set up, and one is called the quasi-peak, the other one is called the average. The measurement of the two output signal nodes, or the LISN, must both meet the limits that are prescribed by the standards in both modes, in both quasi-peak and average model operation of the spectrum analyzer. For the class B limits, the standard prescribes operation between 150 kilohertz and 30 megahertz. This is particularly challenging for DC-DC converters with often have very significant frequency components generated in that frequency range. So we will try to evaluate conducted EMI on a boost DC-DC converter example. Here's the boost DC-DC converter. You have the input port, inductor, you have a power MOSFET, power diode, and the output filter capacitor load connected across the output filter capacitor. In this example, the main power MOSFET is driven in an open-loop manner by a pulsating waveform through a driver with a duty cycle of about 0.75. The input voltage on the left hand side, is given as a DC value of 50 volts with a 0.75 duty cycle, the output voltage is about 200 volts, and the output power is about 400 watts. Converter operates at 200 kilohertz switching frequency. Now, let's show here how the circuit is actually connected in order to evaluate the conducted EMI. So on the input side, we show the LISN, which is really two of these LISN1 blocks connected back to back with the middle point connected to the chassis ground. So this node right here, is going to represent the chassis ground. On the other hand, when we simulate DC-DC converters, we use this ground symbol on the negative side of the converter. In this case, we're careful not to do that because we want to represent the case where there is potentially common node coupling between the nodes inside the converter and the chassis ground. Therefore, we need to really connect the converter between the input voltage and the N terminal, the N terminal being the bottom part of the bottom LISN. So the input source has L and N terminals with respect to the chassis ground. The output of the LISN has the input and the end terminals that are connected to the DC-DC converter, and we will then model the presence of a network analyzer at the output ports of the LISN as 250 ohm resistance as shown right here. So if you simulate the circuit, you expect to see here switching waveforms, pulsating waveform at the switching node and a triangular wave ripple in the current, and indeed, that's what we see. So here is inductor current waveform, a familiar triangular current waveform. We can easily compute the amplitude of the ripple in that current waveform, which goes from approximately 6.5 amps to about 14 amps. The average value of that current, is really the input current seen at the input of the DC-DC converter right here. So this input current is a filtered version of the inductor current because we have this input filter capacitor connected between the input node and the end node of the converter. Now, that input current does look almost DC, but if you look closely, does have some glitches and it has some residual ripple. Those glitches in residual ripple are going to be responsible for whether this example is meeting or not the limits prescribed by the standard. We also want to pay attention to the switch node voltage. Here is the switch node voltage pulsating waveform between approximately zero and approximately 200 volts at the output. Of course, the output voltage is a filtered value of about 200 volts. Now, what we're really interested in this example, is the amount of conducted EMI. We're going to look at that in three different ways. A complete value of the noise at the output of the LISN. We will also separate out the differential mode and common mode noise by definition. Here is how those waveforms actually look like. See here is that the bottom, is your differential mode noise, and you can see that the differential mode noise is not equal to zero. In fact, the peak value right here is around 200 millivolts, the peak value here is about minus 300 millivolts. The waveform is AC, but it does contain a significant amount of switching ripple. Compared to the standards that are prescribing values that are in the low millivolts or sub-millivolt range. We are clearly well above those limits. The RMS value of the differential mode noise is about 0.176 volts, so it's pretty significant. The common mode noise, as we discussed earlier is typically consisting of the spikes of current associated with sharp transitions in the switch node voltage waveform and parasitic coupling to the chassis ground. How does the parasitic coupling actually happens? Well, in the circuit model, we have included a coupling capacitor of just 20 Picofarads from the switch node voltage to the chassis ground right here. So we have modeled the entire parasitic coupling from the converter to a chassis ground with a single relatively small value capacitor connected between the switching node to the chassis ground. Again, that capacitor would really represent coupling from let's say, the drain of the power MOSFET to the heat sink. The value of just 20 Picofarads it looks pretty insignificant, but the problem is that at this point we have sharp transitions in the voltage. We have high value of dv per dt and thereby that high value of dv per dt is going to generate a significant amount of common node noise. As shown here, that common mode noise indeed consists of this large spikes, the spike of about 30 volts with negative spikes being a little bit smaller in magnitude because the opposite transition has a little bit lower dv per dt. The common mode RMS value is 0.5 volts, that's huge. Of course, when you add these up, you get a significant amount of total value of noise. If you wanted to look at these waveforms in frequency domain, spice offers an FFT function. In the View menu, you can change the view to FFT and look at the spectrum of the noise observed. So here is the spectrum of the differential mode EMI compared to the limits imposed by this particular standard example. You can see that we are not meeting the limits. Our fundamental ripple component, that's in 200 kilohertz and then harmonics of the ripple are well above the limit imposed by the standard. If you look at the common mode EMI frequency domain, things get even worse. Just this 20 Picofarad of parasitic capacitive coupling is causing huge amount of common mode noise that's observed here in red. Then in green, we have a total overall EMI that would be measured by the spectrum analyzer. We see that we're really not meeting even close the requirements of the standard. So we would need to add significantly more filtering to be compliant with this particular standard that we're looking at. Here's just an example of a circuit configuration of an EMI filter. You will see that filter at first looks fairly complicated, but it's actually much better to look at that filter in terms of how it behaves with respect to differential mode interference and how it behaves with respect to common mode interference. So with respect to differential mode interference, we have effect of the filter components that are connected between the two terminals of the input of the DC-DC converter. Those components include a filtering capacitor. That filtering capacitor is going to include the CF values shown right here plus a series combination of these two capacitances that are referred to as a common mode components. You will see they play the major role in a common mode filtering. So CF plus half value of CCM is really the main filter capacitor that we have on the input side of this DC-DC converter. In parallel with that, we have what is called a damping network. You will learn why that damping network is necessary in order to mitigate the effects of the input filter on the frequency responses, and closed loop operation of the DC-DC converter itself. In terms of the differential mode filtering, the next main component in the L_C filter shown here is L_f and that L_f is in series, so you will have the DC current of course, going through L_f and is going to be delivered to the DC-DC converter, but L_f will serve the purpose of filtering the differential-mode AC components of the input current. In series with that, is what is called the leakage inductance of the common mode choke. This common mode choke consists of two coupled inductors with dots shown right here. So for differential-mode purposes, when you look at the current that is flowing in from left to right and then coming back from right to left, those two inductors that are coupled, are coupled with dots that are opposite, and so all that is left is just the leakage inductance between the two coupled inductors. These coupled inductors in a common mode chokes are usually coupled with very high coupling coefficient, so the leakage inductance is relatively small and the filter is going to be dominated for differential-mode purposes by the series inductance. Then finally, there is an additional capacitor on the input side, C_IN, for in the differential-mode part equivalent circuit model of the complete input filter. For common-mode purposes, the common-mode current is going to see the components with respect to the chassis ground. Those components with respect to the chasis ground with remember, the common mode current is having the same direction in both terminals of a DC-DC converter, are going to be dominated by two components. One is the parallel combination of the two C_CM capacitors from the either one or the two nodes to the chassis ground, and then the series combination of the two coupled inductors with dots now being in the same direction for the common-mode current. So this is going to be a common-mode representation of the complete filter. To conclude, we have differential-mode EMI determined by the converter input current ripple that we're familiar with, which depends on converter configuration, converter parameters, and of course, switching frequency. We should note immediately that converters with pulsating input current will generally generate larger differential-mode EMI and will require a large amount of filtering to meet the differential-mode EMI requirements. We will in this course, study filter damping that is necessary to mitigate the impact on converter frequency responses and closed loop regulation performance. In addition, common-mode EMI is present and it's determined by currents through parasitic capacitances between high dv predicting nodes of the circuit, and the chasis ground. Common-mode EMI is a difficult problem because it depends on how large dv per dt is. It depends on very detailed issues related to components for example, intertwining capacitances are important. Component parasitics in general are much more important for the common-mode EMI generation. Finally, the common-mode EMI depends strongly on circuit and mechanical layout. This is why we say that electrical simulation models as we have done in this simple example, can be used to gain some insights into common-mode noise and maybe to compare common mode performance with respect to different filtering or different values of dv per dt and so on. But generally speaking, you cannot rely on simulation to tell you whether your DC-DC converter is going to meet requirements or not. A measurement is really the judge and we cannot really assume that simulation model can be so detailed to capture all the parasitic effects that are involved in common-mode generation. Finally, the way we approach designing input filter is to split the problem into two simpler problems. One is the problem of designing the differential-mode filter, and the other one is the problem of designing common-mode filter. Much of this class is going to be about differential-mode filtering. The other problem of common-mode filtering has to be looked at separately. What is coming next is going to be focused on differential-mode filtering, and the damping necessary to mitigate the impact of the input filter on the performance of the converter itself.