Electrical – Emi Filter Design with Tracopower isolated power supply in railway application

A "typical" two-stage power supply EMI filter looks like this:

^{simulate this circuit – Schematic created using CircuitLab}

The inductors attenuate common-mode noise (and also offer some differential attentuation via their leakage inductances). C1 and C4 shunt differential-mode noise. If there's an 'earth' point available you will also see two more capacitors between CM1 and C4, connected from each rail to the earthing point. These provide more common-mode attenuation.

Sometimes you will see a single stage, sometimes two. As long as your filter is attenuating the necessary noise components and are OK thermally, I don't see a pi filter being 'wrong' necessarily. (For lower frequency conversion, it's easier to get the attenuation via the inductors - just add more turns / more core etc.)

Assumptions:

You understand what the proposal meant by
"Recommended external EMI filter for EN55022 class A:"

the spec A good PSU with similar proposed filter installed

Imagine this to be 20~30 dB higher in the UHF band if not installed unless shielded in an enclosed like a proper Faraday Shield.

It means this proposal is necessary to meet many user applications to get below the Class A Industrial threshold for unintended radiations. Class B Consumer level is 10 dB harder. (lower)

My questions is about C6 and C7: What effect do these capacitors have?

• neglecting C7 since you have this shorted out with a common ground (assuming it is a square low inductance connection )
  • What does C6 do?

C6 bypasses the radiation of noise in the PSU in the unit acting as an unintended radiator. By careful low impedance ground planes under the part it shunts the RF that is radiating like an antenna between input and output.

An example of 1nF 100V 1812 MLCC for C6 180x120 mil (4.5x3mm) (not to scale 8-} but shown for dramatic reduction in UHF unwanted emanations of common mode and differential input output noise.

A 1nF has no effect below 10MHz and very little below 100MHz but "may have" (depending on s-parameters) low Z above this f to radiated noise if the load is 2.5 Ohms on the output.

So the transient switch noise internal to the regulator can radiate due to ~1~2ns rise times of some switches. (assumption and presumably why they propose this)

So C6 attenuates with the L1 (rising X(f) on input and Load Caps on the output. Output caps help to attenuate load regulated and source regulated noise but not the differential noise between input and output from the regulator itself. This is a common proposed filter and relies on your microwave circuit layout experience to include a spot for this during testing with ground planes around it and 1kV insulation rules applied to IN-OUT gaps.