Common-mode current in non-isolated power supplies

Have you ever tried to eliminate the common-mode currents of non-isolated power supplies, which can be a source of electromagnetic interference (EMI)? In some high-voltage power supplies, such as those used in LED light bulbs, you may find that you can't eliminate them. Upon closer inspection, non-isolated power supplies are actually no different from isolated power supplies. The switch node ground parasitic capacitance generates common-mode currents.

Figure 1 is a schematic of an LED power supply that shows the main cause of common mode current in this buck regulator. The cause is the switch node ground capacitance. It is surprising that such a small amount of capacitance can still cause problems. The CISPR Class B (for residential equipment) radiated regulations allow a 46 dBuV (200 uV) signal at 1 MHz into a 50 source impedance. This means that only 4 uA of current is allowed. If the converter switches a 200 Vpk-pk square wave at 100 kHz into the drain of Q2, the reference voltage is about 120 volts peak. Since harmonics drop in proportion to frequency, this would be about 9 Vrms at 1MHz. We can use this to calculate the allowed capacitance and get about 0.1pF, or 100 fF (equivalent to 2 megohm impedance at 1 MHz), which is a perfectly possible capacitance for this node. In addition, there is capacitance to the rest of the circuit ground, which provides a return path for the common mode current, shown in Figure 1 as C_Stray2.

Figure 1: Switch node capacitance of only 100 fF creates EMI problems

In LED lamp applications, there is no substrate connection, only heat and insulation, so common-mode EMI filtering becomes an issue. This is because the circuit is high impedance. It can be represented by a 9 Vrms voltage source in series with a 2 megohm capacitor (as shown in Figure 2), and there is no way to increase the impedance to reduce the current. To reduce the radiation at 1MHz, you need to reduce the voltage or reduce the parasitic capacitance. There are two ways to reduce the voltage: dither tuning or rise time control. Dither tuning expands the frequency spectrum by changing the operating frequency of the power supply.

Figure 2 100 fF can cause EMI limits to be exceeded

To discuss jitter tuning, first read Power Tip 8 (February 2009). Rise time control limits the high frequency spectrum by slowing the switching speed of the power supply, and is best suited to address EMI above 10MHz. Reducing the parasitic capacitance at the switch node is easy, just minimize the etch area or use shielding material. The capacitance from this node to the rectified power line will not form common-mode currents, so you can bury the traces in a multilayer printed wiring board (PWB), thereby reducing a lot of the unwanted capacitance. However, you can't completely eliminate it because there is still capacitance remaining from the FET drain and inductor. Figure 2 shows a graph that walks you through the steps to calculate the EMI spectrum. The first step is to calculate the spectrum of the voltage waveform (red). This can be done by calculating the Fourier series of the drain voltage waveform, or simply calculating the fundamental component and then approximating the envelope (1 divided by the harmonic number and the fundamental component). Further adjustments are made at high frequencies (1/(pi * rise time)), as shown above 7MHz. Next, divide this voltage by the reactance of the parasitic capacitance. Interestingly, the low frequency emissions are flat and steady until the frequency crosses the pole set by the rise time. Finally, the CISPR Class B regulations are plotted. With only 0.1 pF of parasitic capacitance and a high voltage input, the emissions are close to the regulations.

EMI issues also exist at higher frequencies due to circuit resonances and radiation caused by input line transmission resonances. Common-mode filtering can help solve these problems because there is a large amount of capacitance in C_Stray2. For example, if the capacitance is 20 pF, its impedance is less than 2 K-Ohms at 5MHz. We can add a high enough impedance common-mode inductor between the circuit and the 50 Ohm test resistor to reduce the measured radiation. The same is true at higher frequencies.

In summary, when using high voltage, non-isolated power supplies, common-mode currents can cause EMI radiation to exceed standard regulations. In some two-wire designs (no substrate connection), this problem is particularly difficult to solve because there are many high impedances involved. The best way to solve this problem is to minimize parasitic capacitance and implement high-frequency pulsation for the switching frequency. At higher frequencies, the impedance of the scattered capacitance of the rest of the circuit becomes smaller, so common-mode inductance can reduce both radiated and conducted emissions.