
Power Tip 40: Common-mode currents and EMI in non-isolated power supplies
(Editor’s note: to see a linked list of entries #1 to #37 in the Power Tips series, click here; to see a linked list of all entries from #1 to the latest one, click here.)
(Additional editor’s note: if you are interested in EMC or EMI/RFI, be sure to check out our EMC Basics series, here.)
Have you dismissed common-mode currents in a non-isolated power supply as a potential electromagnetic interference (EMI) source?
In high-voltage supplies, such as one you might find in an LED light bulb, you may find that you can’t. On inspection, it really is no different than an isolated supply. There will be stray capacitance to ground from switching nodes that will generate common-mode currents.
Figure 1 is a schematic of a LED power supply showing the parasitic capacitance that is the main cause of common-mode current in this buck regulator. It is the capacitance to earth from the switch node. It is surprising how small this capacitance can be and still create a problem.
Figure 1: Even just 100 fF of capacitance from the switch node
can create an EMI issue
(click here to see enlarged image).
The CISPR Class B (for residential equipment) conducted emissions limit allows a 46 dBμV (200μV) signal into a 50Ω source impedance at 1 MHz. This translates into only 4 μA of allowable current. If the converter switches with a 200 Vpk-pk square wave on the drain of Q2 at 100 kHz, the fundamental will be around 120 volts peak. Since the harmonics decrease in proportion to frequency, there will be about 9 Vrms at 1 MHz.
That can be used to calculate an allowable capacitance to ground of around 0.1pF, or 100 fF (or a 2-MΩ impedance at 1 MHz) which is an entirely feasible amount of capacitance from this node. There is also capacitance from the remainder of the circuitry to earth that provides a path for the common-mode currents to return. (This is notated as C_Stray2 in Figure 1.)
In an LED light application, there is no chassis connection: only hot and neutral are available, so common-mode EMI filtering is a problem. That is because the circuit is high impedance. It can be represented by a 9 Vrms voltage source in series with a 2 MΩ capacitive reactance as shown in Figure 2, and there is no realistic way to add impedance to reduce the current.

Figure 2: Even 100 fF can cause you to exceed EMI limits
(click here to see enlarged image).
To reduce the emissions at 1 MHz, you need to reduce the voltage or reduce the stray capacitance. Two ways to reduce the voltage is with dithering or rise time control. Dithering varies the operating frequency of a power supply to spread out the spectrum.
For a discussion on dithering, look at Power Tip 8 (February 2009). Rise-time control slows the switching speed in the power supply, to limit the high-frequency spectrum and is better suited for EMI problems above 10 MHz. Reducing the stray capacitance from the switching node can be as simple as minimizing the etch area or it may involve shielding. Capacitance from this node to one of the rectified supply lines does not create common-mode current, so you can bury the trace in a multilayer printed wiring board (PWB) and reduce much of the unwanted capacitance.
However, you can not completely eliminate it because there is still capacitance remaining from the drain of the FET and inductor. Figure 2 provides a graph and steps you through the calculation of the EMI spectrum.
The first step is to calculate the spectrum of the voltage waveform (red). This is accomplished by calculating the Fourier series of the drain-voltage waveform, or more simply by calculating the fundamental component and approximating the envelope as one divided by the harmonic number and the fundamental.
A further adjustment is made at high frequency [1/(π×rise time)], as shown above 7 MHz. The next step involves dividing this voltage by the reactance of the stray capacitance. Interestingly, the low-frequency emissions are flat with frequency until you cross the pole that is set by rise time.
Finally, the CISPR Class B limits are also plotted. With only 0.1 pF of stray capacitance and a high-voltage input, emissions are close to the limits.
EMI problems can also exist at higher frequencies due circuit resonances and radiated emissions caused by resonances of input cabling. Common-mode filtering can help these issues because there is a reasonable amount of capacitance in C_Stray2.
For instance, if it were 20 pF, its impedance would be less than 2 kΩ at 5 MHz. Common-mode inductors of sufficient impedance can be added between the circuit and the 50 Ω test resistor to reduce measured emissions. This is also true at higher frequencies.
To summarize, with high-voltage, non-isolated power supplies, common-mode currents can cause EMI emissions to exceed standard limits. In two-wire designs (no chassis connection), they are particularly difficult to handle because of the high impedances involved.
The best way to approach this kind of challenge is to minimize the stray capacitance and to dither the switching frequency. At higher frequencies, where the impedance of the distributed capacitance from the remainder of the circuit becomes small, common-mode inductors can reduce both radiated and conducted emissions.
Please join us next month when we will discuss power supplies for DDR memory.
For more information about this and other power solutions, visit: www.ti.com/power-ca.
About the author

