[Introduction]Inductor-capacitor filters are often added to the input and output of switch-mode power converters to reduce reflected ripple current and output noise while meeting EMC radiation and susceptibility limits. Converter manufacturers sometimes specify recommended filter inductance values, but parts with the same nominal performance from different component suppliers can have very different performance over the entire frequency range, resulting in poor results and increased conduction and radiation interference. This article will explore different variations in inductor performance.

Most power converters today, including all isolated DC/DC converters, are in “switch mode” where the external DC voltage is “chopped” at high frequencies and then converted to AC to supply the internal isolation transformer. The AC output of the transformer is adjusted and rectified by the duty ratio control and then returned to DC. The whole process has high efficiency and low loss. The downside is that switching produces high-frequency ripple on the input and output, as well as conducted and radiated noise spikes, which can interfere with other equipment. There is a trend for power converters to operate at higher frequencies and faster slew rates to improve efficiency, but the resulting noise spectrum is much wider.

LC filter reduces output noise

Any commercial power converter has basic filtering to reduce ripple and noise to a typical peak-to-peak value of about 1% of the DC output. Usually this is acceptable, but if the application is sensitive a lower level is required and the simple solution is to add an external LC filter (Figure 1).

Figure 1: External LC filter reduces output ripple and noise from switch-mode power supplies

Theoretically, the impedance of the inductor is zero at DC, and the impedance of the capacitor is infinite, so the required DC is not affected. But as the frequency increases, the inductor impedance ZL increases while the capacitor impedance ZC decreases, resulting in an increasingly larger “voltage divider” effect. The filter corner frequency can be selected to reduce converter switching frequency ripple, but noise spikes are composed of frequency spectrum up to tens of MHz, so the attenuation of noise spikes is more difficult to predict.

The reason is that at certain frequencies, when the values of ZL and ZC are equal, the LC network “resonates” and the noise is amplified rather than reduced, although this effect is damped by the load resistance. Above resonance, there is still some noise attenuation, but other parasitic effects start to appear. For example, the self-capacitance of an inductor creates another resonance at a much higher frequency. This capacitor also makes it easier for noise to bypass the inductor. At high frequencies, due to the “skin effect”, the core loss of the inductor increases, the AC resistance of the inductor line also increases, the impedance of the capacitor becomes smaller than the equivalent series resistance (ESR), and the capacitor begins to play the role of resistance. . The equivalent series inductance (ESL) of the capacitor also has a high frequency effect. If the equivalent circuit of the LC filter included these parasitic components, it would be closer to Figure 2.

Figure 2: External filter with parasitic components

Inductor parasitics alter noise attenuation performance

For the influence of the magnetic core loss involving frequency in the circuit, using LLOSS 1, 2 and RLOSS 1, 2 is a simple way of modeling; different LLOSS values will produce different impedances, so that different resistance components RLOSS 1 and 2 have effects at different frequencies. Adding more LLOSS/RLOSS networks can make the model more accurate, but it is difficult to calculate the component values with reference to the inductor data sheet, so if you want a complete inductor and core model you can get the values empirically. Figure 3 shows simulation plots of filter attenuation with and without LLOSS/RLOSS networks, including assumed values of L and C and their parasitic components, showing that core loss has a large effect on high frequency attenuation of noise, at At 10MHz there is a difference of 20 dB. Unfortunately, core losses do not appear in typical inductor data sheets and can vary widely.

Figure 3: Comparison of LC filter attenuation with and without core loss

Input EMC filter component selection

When selecting a commercial inductor for an EMC filter at the DC/DC converter input (Figure 4), the inductor manufacturer’s data sheet usually only provides information on the inductance, DC resistance, and individual resonant frequency. Although this can attenuate the reflected input ripple by a certain amount, it is difficult to predict noise spikes and spectral attenuation without data to define the spurious components. From the analysis of the output filter, it can be seen that high frequency effects such as core loss can again have a great impact on noise attenuation. It is understandable that inductor manufacturers will not provide relevant information because there are so many variables. For example, core losses depend not only on the magnitude and shape of the AC component of the waveform, but also on frequency, DC current bias, and temperature.

Figure 4: Typical DC/DC converter input EMC filter

This makes it difficult to select the ideal inductor, which in the worst case can result in conducted and radiated noise levels exceeding operating or statutory regulations. The answer may not even be known until the final product undergoes independent EMC testing, but making any changes at this point would be extremely costly.

If suitable test equipment is available, such as an antenna and an EMC chamber, samples of inductors with the same title ratings from different suppliers can be tried in the circuit to check actual results. Large inductance values may seem good but will lower the resonant frequency, and smaller components may have higher DC resistance and thus voltage drop, depending on the converter’s load and power consumption. Large inductances also have larger self-capacitance, thus reducing high frequency attenuation. A final consideration is that large inductances themselves can cause voltage spikes when subjected to load current steps. A small inductor with a large capacitor is also an option, but if the electrolytic type is used due to cost and size, it will not have good high frequency characteristics. Other types, like ceramic, are good at high frequencies, but are expensive and bulky for high capacitance values.

The correct combination of L and C will be affected by factors such as cost, size and performance. When choosing an inductor, the market has a confusing choice of types. This includes ferrite, powdered iron cores, some special options such as polycrystalline cores, then drum, toroid and E-type cores, as well as through-hole or SMD mounting options that also affect performance. Buyers will also find devices that appear to have the same inductance specification and current rating, but at very different prices.

Each inductor offered is suitable for a specific application. Ferrite has the lowest losses but the material is more expensive than iron powder. Iron powder is also more resistant to overcurrent and retains inductance better than ferrite. “Toroidal” or toroidal cores have less magnetic field leakage, but are more difficult to wind and terminate than drum or “spool” cores. So design, production, EMC, procurement and process engineers all need to be involved in selecting the best solution.

Preferred converter manufacturer-proven solutions

As a manufacturer of AC/DC and DC/DC converters, RECOM understands the difficulty of choosing the right inductor and therefore offers a range of cost-effective inductors and recommended capacitors that match most power converters that pass their Conducted and radiated noise testing in EMC chambers. Customers can take advantage of a proven one-stop noise reduction solution, saving time, money, and potentially faster time-to-market.

references

[1] RECOM: www.recom-power.com

[2] Modelling ferrite core losses: http://ridleyengineering.com/design-center-ridley-engineering/39-magnetics/185-a03-modeling-ferrite-core-losses.html

The Links: **C070VW02-V0** **MCC%20200-16IO1** **BEST-IGBT**