Electromagnetic interference (EMI) has a repute for being one probably the most tough features in power-supply designs. I believe that this repute comes, largely, from the truth that most EMI-related challenges should not issues that may be solved by trying on the schematic. This may be irritating, because the schematic serves because the central place the place the engineer goes to know what the circuit does. Certain, you recognize that there are related features within the design that aren’t within the schematic—issues like code.
You additionally know that the schematic doesn’t signify issues like printed wiring board parasitics. Nevertheless, in EMI, parasitics like these can have dominating results in your potential to satisfy the necessities, forcing you to have the mandatory expertise to acknowledge what kinds of parasitics can positively or negatively affect the EMI spectrum. This Energy Tip article will discover how these kinds of parasitics can create an sudden EMI filter resonance in a gallium nitride (GaN)-based onboard charger (OBC) for an electrical automobile (EV).
Determine 1 exhibits a high-level system illustration of the OBC. Its major operate is grid-to-vehicle voltage and present battery charging. A secondary operate is vehicle-to-grid energy circulate in order that the EV can complement renewable power sources which may have a fluctuating peak capability.
Determine 1 The schematic exhibits a high-level system illustration of the onboard charger (OBC). Supply: IEEE
Now, let’s flip our consideration to the EMI concerns contained in the OBC.
Onboard charger’s EMI assessment
EMI consists of differential-mode (DM) and common-mode (CM) noise. For an OBC system, DM noise is principally generated by the enter present of the ability issue correction (PFC), whereas CM noise may result from each the PFC and the conductor-inductor-inductor-inductor-capacitor (CLLLC). Determine 2 exhibits the cooling resolution (chilly plate) for the OBC within the decrease proper nook of the schematic. The chilly plate is crucial in stopping the elements from overheating; nevertheless, its presence introduces parasitic capacitance that impacts the EMI.
Determine 2 Parasitic results inflicting EMI are proven within the decrease proper nook of the schematic. Supply: Texas Devices
As proven in Determine 2, there are parasitic capacitances between the switching nodes to the chilly plate, between the first and secondary grounds to the chilly plate, and between the first and secondary winding of the CLLLC transformer. These parasitic capacitances can generate or have an effect on CM noise-current ranges within the system.
With estimated parasitic capacitances, simulations present that within the worst case, the naked DM noise with solely the two.2-µF enter capacitor (CX1) is about 110 dBµV. Likewise, the naked CM noise with none CM filter is about 115 dBµV at about 350 kHz. Designing a two-stage EMI filter helps attenuate EMI noise beneath the Comité Worldwide Spécial des Perturbations Radioélectriques (CISPR) 32 commonplace.
The common-mode impedance of LCM1 and LCM2 at 350 kHz is about 3 kΩ. Their leakage inductances are about 6.4 µH, that are used for DM noise attenuation. CX1 and CX2 are 2.2-µF movie capacitors for DM noise attenuation, and CY1, CY2, CY3 and CY4 are 4.7-nF ceramic capacitors for CM noise attenuation.
Ideally, with the designed filter, each the naked CM noise and naked DM noise must be attenuated by greater than 65 dBµV, and the EMI noise ought to meet the CISPR 32 commonplace. There are nonetheless some sensible challenges to handle, nevertheless.
EMI filter resonance
EMI filters are full of resonances by design. In truth, it’s these resonances that allow the filter to attenuate noise and allow a system to cross the EMI commonplace. Determine 3 exhibits a typical attenuation curve for an EMI filter. Discover that at frequencies above 100 kHz, the filter does a pleasant job of decreasing amplitude. There are some resonances beneath 100 kHz, nevertheless, that might be fairly problematic in the event that they exist on prime of a switching frequency.
Determine 3 That is how typical EMI filter attenuation seems to be like for an onboard charger. Supply: Texas Devices
Clearly, nobody would deliberately place a resonance on the switching frequency, however interconnect impedances, element parasitics or each can typically push the system to function in an unintentional means.
Determine 4 exhibits a barely modified EMI filter in comparison with Determine 2. The variations are within the elements in crimson. LP1 and LP2 signify the parasitic inductance of the interconnect between the EMI filter and the PFC enter. The presence of LP1 and LP2 required some native capacitance for the PFC present to circulate by way of. Subsequently, shifting CX1 to the enter of the PFC and including CX0 elevated the attenuation of the filter. The 4 components in crimson mix collectively to create a resonance at 240 kHz. On this design, 240 kHz is the conversion-combined switching frequency of the two-phase PFC. This resonance goes to amplify the switching present and subsequently make the EMI worse at this frequency.
Determine 4 The EMI filter is proven with a resonance on the switching frequency. Supply: Texas Devices
Determine 5 exhibits the time-domain waveforms of the AC line present flowing by way of LP1 in magenta, together with the AC enter voltage in blue. Discover that the present has a big 240-kHz sine wave with a 28-A peak-to-peak amplitude. This sine wave is the direct results of the triangular PFC present flowing by way of the unintended amplifier created by the crimson elements in Determine 4.
Determine 5 The time-domain waveforms of the AC line present are proven flowing by way of LP1 in magenta. Supply: Texas Devices
Damping a resonance like this may be difficult, because the obligatory damper will usually require an inductor or capacitor bigger than these used within the circuit. One other potential resolution could be to decrease the inductance of the interconnect in order that the resonance is not on prime of the switching frequency. In idea this sounds good, however virtually talking, that interconnect is there for a motive; so, making it smaller isn’t actually viable.
Another choice is to contemplate the need of retaining each CX0 and CX1. You possibly can’t take away CX1 because the PFC requires some native enter capacitance for the high-frequency present. Nevertheless, CX0 is there to extend the capacitance, with the intent of accelerating the attenuation. Eradicating CX0 improved the EMI by roughly 6 dBµV. The amplitude is diminished by 50% and a big share of the require attenuation (65 dBµV) wanted to cross the usual. That’s a fairly whole lot.
Two design takeaways
The takeaway right here is twofold. The primary is the unique premise: the schematic doesn’t inform the entire story for EMI. On this case, there was an unintended resonance brought on by interconnect inductance that amplified the switching frequency noise. Recognizing the basis reason behind the issue is all the time probably the most crucial step in debugging.
The second takeaway is that typically much less of a usually good factor (filter capacitors) is best. You possibly can normally handle EMI points by including elements, however on this case, the presence of a element makes the problem worse. Subsequently, by eradicating CX1, we have been in a position to scale back the dimensions of the filter, decrease the system value, and enhance the EMI.
Brent McDonald is a system engineer within the Energy Provide Design Providers staff at Texas Devices.
Associated Content material
