How a stand-alone active EMI filter IC shrinks common-mode filter size

How a stand-alone active EMI filter IC shrinks common-mode filter size
Other Parts Discussed in Post: TPSF12C3-Q1, TPSF12C1-Q1

Automotive on-board chargers and server power supplies are highly constrained system environments where power density is a primary metric. It’s important to reduce the volume of the electromagnetic interference (EMI) filter components so that the solution can fit into demanding form factors.

Common-mode (CM) filters for these and other high-density applications often limit the total Y-capacitance – related to touch-current safety requirements – and thus require large-sized CM chokes to achieve a target corner frequency or filter attenuation characteristic. The result is a compromised passive filter design with bulky, heavy and expensive CM chokes that dominate the overall filter size.

With advances in passive components lagging behind high-speed power semiconductor devices as well as circuit topologies, the volume of the passive filter is one of the limiting factors for increasing power density. Practical filter implementations can occupy as much as 30% of the total volume of a power solution, as shown in Figure 1.

 

Figure 1: A conventional single-phase passive EMI filter in a 3.3-kW totem-pole power factor correction reference design


Reduce system size, weight and cost 

 Learn more about our power-supply filter ICs.

Active EMI filter (AEF) circuits enable more compact filter solutions for next-generation power-management systems. Space-constrained applications can use active power-supply filter integrated circuits (ICs) to reduce the size of magnetic components and the overall filter. Additional benefits of an AEF include lower component power losses for better thermal management and higher reliability, reduced coupling between components within a confined space, easier mechanical and packaging design, and lower costs.

Figures 2 and 3 are schematics of single-phase and three-phase filter circuits, respectively, where an active solution replaces a traditional passive design. The single-phase TPSF12C1TPSF12C1-Q1 and three-phase TPSF12C3TPSF12C3-Q1 AEF ICs, positioned between the CM chokes, provide a lower-impedance shunt path for CM currents. As illustrated, the active solution has CM chokes LCM1 and LCM2 with much lower inductance relative to the same components in the passive filter.

 

Figure 2: Single-phase passive EMI filter (top) and corresponding AEF circuit with lower CM choke inductances (bottom)

  

Figure 3: Three-phase passive EMI filter (top) and corresponding AEF circuit with lower CM choke inductances (bottom)

AEFs

With Y-rated sense and injection capacitors connected to the AC lines, the circuits aim to reduce the total filter volume yet maintain low values of the line-frequency leakage current to chassis ground. This is possible by using an active circuit that shapes the frequency response of the injection capacitor – effectively increasing its value for high frequencies. In turn, the amplified injection capacitance over the frequency range of interest for EMI mitigation will lower CM choke inductances relative to the values of a passive filter with comparable attenuation.

The circuit advantages using an AEF are:

  • A simpler filter structure with a wide operating frequency range and high stability margins (calculated using the common-mode AEF quickstart calculator tool).
  • A reduced CM choke size with lower volume, weight and cost. This also enables much lower copper losses and better high-frequency attenuation performance from reduced choke self-parasitics.
  • No additional magnetic components – the AEF circuit only uses Y-rated sense and injection capacitors, with no impact to peak touch current during a fault condition.
  • Enhanced safety using a low-voltage AEF IC referenced to chassis ground.
  • A stand-alone IC implementation that offers flexibility in terms of placement near the filter components.
  • Immunity to line voltage surges to help meet International Electrotechnical Commission 61000-4-5.

The X-capacitor(s) placed between the two CM chokes in Figures 2 and 3 provide a low-impedance path between the power lines from a CM standpoint, typically up to low-megahertz frequencies. This allows current injection onto one power line, usually neutral, using only one injection capacitor. If the three-phase filter is a three-wire system without a neutral, the SENSE4 pin of the TPSF12C3-Q1 ties to ground and the injection capacitor couples through a starpoint connection of the X-capacitors.

Practical AEF implementation

Figure 4 shows a practical AEF implementation suitable for the converter in Figure 1. Using the TPSF12C1-Q1 single-phase AEF IC achieves CM noise attenuation.

  

Figure 4: Single-phase filter evaluation board with the AEF rated at 10 A

Figure 5 shows EMI results with the AEF disabled and enabled.

  

Figure 5: EN 55032 Class B EMI results with the AEF disabled and enabled

As evident in Figure 5, an AEF provides up to 30 dB of CM noise attenuation in the low-frequency range (100 kHz to 3 MHz), which enables a filter using two 2-mH nanocrystalline chokes to achieve CM attenuation performance equivalent to a passive filter design with two 12-mH chokes. To make a fair comparison, these chokes come from the same component family (made by Würth Elektronik) with a similar core material. Table 1 captures the applicable CM-choke parameters for the passive and active designs, and Figure 6 highlights the volume, footprint, weight and cost savings.

  

Filter design

CM choke part number

Qty

LCM1, LCM2 

(mH)

RDCR (mΩ)

Size 

(L × W × H, mm)

Total mass (g)

Total power loss 

(W) at 10 A, 25°C

Passive

7448051012

2

12

15

23 × 34 × 33

72

6.0

Active

7448031002

2

2

6

17 × 23 × 25

20

2.4

Table 1: CM choke parameters for the passive and active filter solutions

Figure 6: Footprint, volume, weight and cost reductions enabled by an AEF (a); choke size comparison (b)

 

The AEF in this example achieves a 60% total copper loss reduction at 10 A (neglecting the winding resistance increase from the temperature rise), which implies lower component operating temperatures and improved reliability.

Conclusion

It’s challenging to achieve a compact and efficient design for the EMI filter stage in high-density switching regulators, particularly for automotive and industrial applications where solution size and cost are such priorities. Practical results from an active filter solution to suppress the measured CM noise signature indicate a significant volumetric reduction of the CM choke components when benchmarked against an equivalent passive-only filter design.

Additional resources:

Anonymous
Parents
  • Hi Timothy,

    I just have a question on GND for the AEF controller. Is the AEF controller's GND connected to the primary side GND or chassis (metal case)?

    See below snapshot for further illustration.

    Thanks,

    Martin

  • Hi Martin,

    Thanks for the question on this active filter implementation. The AEF circuit is referenced to chassis GND as shown in Figure 9-2 of the AEF white paper. This is effectively the same as the Y-cap grounding in a passive filter implementation. Chassis GND is usually not the same as primary-side GND in a circuit with AC rectified input. Figure 6-1 of the white paper is a simplified equivalent circuit -- here it is updated to show just the LISN impedance to GND (effectively 2 x 50Ω in parallel for common mode), as this avoids the ambiguity with the common-mode source voltage.

    Regards,

    Timothy Hegarty

  • Hi Timothy,

    Thanks for your clarifications. I am clear now with GND connection of AEF.

    I also have a concern on leakage current measurement on AEF circuit. As a safety requirement, the leakage current/touch current is measured between Line to chassis or Neutral to chassis. As shown below, there are new paths between grid L/N to the chassis via CINJ, Csen1 and Csense2. Since the total impedance of the AEF varies to the frequency. We should have a calculated comparison of the leakage current between the conventional passive filter and AEF to convince the AEF may have sufficient margin for leakage current.

    Moreover, if the AEF is connected to the chassis GND, the clearance between the AEF's bias supply, grid L/N and primary GND should be very careful due to safety requirement (>8mm). In practical application, the bias supply is usually provided by the auxiliary power supply. It is very difficult to keep sufficient clearance between the AEF supply and the primary GND due to safety requirement.

    Thanks,

    Martin

  • Hi Martin,

    The sense and inject caps of the AEF circuit are effectively Y-caps that replace the existing Y-caps positioned between the chokes in a conventional passive filter. If the total value of these sense and inject caps is maintained less than or equal to the original Y caps in the passive design, then there is no increase in leakage/touch current amplitude.

    The TPSF12C1 and TPSF12C3 family of AEF ICs “amplify” the effective Y-capacitance between the power lines and ground, mainly in the frequency range above ~50 kHz. Since the impedance profile below 1 kHz is not affected by the AEF circuit, the influence of the AEF on the safety requirements for leakage/touch current is very limited. As a result, the AEF circuit can attenuate CM noise without compromising safety. More specifically, AEF will not impact the leakage/touch current value observed at line frequency. Note the measurement circuit for peak touch current (for example, that specified in IEC 60990) has a 1kHz RC filter that removes the higher frequency content, so it's really the line-frequency component that is measured.

    In terms of PCB layout, our filter board designs use 150mils (3.8mm) spacing rule from both Line and Neutral to chassis GND. The required spacing can change depending on whether the routing is on internal or external PCB layers and also the specific isolation requirements for the application (e.g., 8mm as you mentioned). Routing of the VDD supply lines from an aux supply to the IC should adhere to the required spacing rules as needed.

    - Timothy Hegarty

  • Hi Timothy,

    Thanks for your clear clarifications. Agree that the Leakage/touch current would be less than or equal to the original design of passive filter.

    I also have concern on lightning surge current passing through the AEF IC and TVS. As we know, the surge current would be very big (20~30Apk) and would cause hard failure of the TVS and IC.

    Do you have any recommendation for surge protection?

    Best regards,

    Martin

  • Hi Timothy,

    I made some modification on AEF IC connection. The modifications aim to move the AEF IC's bias supply to the grid side L/N instead of connecting to the PE GND/chassis so that the clearance between the AEF IC's bias supply and the AC grid L/N is not critical as original connection.

    I am not sure if the modification can work?

    Best regards,

    Martin

  • Martin, the surge test schematic does not show the MOVs, which are essential for clamping the input surge transient. With MOVs installed, the SENSE pins have internal clamps (adequate for the 680pF sense caps), and the external TVS as shown protects the INJ pin.

    In terms of the alternate schematic, sense caps are required to connect to the L and N power lines to detect the CM voltage disturbance (effectively the sum of the two voltages relative to chassis GND) and also to reject the line frequency voltage and DM disturbance (the sensed voltage difference). Note that this AEF circuit uses a voltage sense, current inject (VSCI) topology - see white paper SNVAFJ9 for more detail.

Comment
  • Martin, the surge test schematic does not show the MOVs, which are essential for clamping the input surge transient. With MOVs installed, the SENSE pins have internal clamps (adequate for the 680pF sense caps), and the external TVS as shown protects the INJ pin.

    In terms of the alternate schematic, sense caps are required to connect to the L and N power lines to detect the CM voltage disturbance (effectively the sum of the two voltages relative to chassis GND) and also to reject the line frequency voltage and DM disturbance (the sensed voltage difference). Note that this AEF circuit uses a voltage sense, current inject (VSCI) topology - see white paper SNVAFJ9 for more detail.

Children
No Data