Feed thru vs common mode choke

[likephysics]

I am looking for some comparison of feed thru caps, ferrite beads and common mode chokes.
when to use which etc.
Feed thru basically shunts noise to ground. But if the circuit ground to Earth is not so exactly low impedance, some of the noise will end up in the circuit.
Isn't a common mode choke better in this case.

 

[the_emi_guy]

You need to diagram out your circuit. Diagram needs to include impedances, even if just estimates, at the frequencies you are concerned about.

It is difficult to make any kind of general comment about whether a feedthrough cap or ferrite would solve an issue.

 

[likephysics]

The power to my board comes from an adapter: 100-240VAC, 5V DC output.
From this 5V, I have a buck with 3.3V output with 1MHz switching frequency.
I don't want this 1MHz or it's harmonics going back into the adapter and the adapter wire acting as an antenna.

At present, I have feed thru caps at the input of my buck converter.
The SMT feed thru caps are connected to circuit ground, not frame ground.
There's a 1Meg resistor with 0.1uF cap in parallel connected from circuit ground to frame ground.

So any common mode noise has to pass thru circuit ground and then to frame ground.
But if I use a common mode choke, I don't have to be concerned about ground path.

 

[the_emi_guy]

What are your regulatory requirements? Class A, Class B, Tempest?

 

[likephysics]

Class B.

 

[the_emi_guy]

You have 3 nodes of interest: 5VDC input, circuit gnd, and chassis.

I am assuming your circuit ground is a ground plane (GND). If this is not the case then this discussion will get more involved.

First manage noise between 5VDC and GND. In other words, at the point where 5VDC leaves your ground plane (at the power connector) you do not want any noise between 5VDC and GND. Thus, any common mode noise on your DC power cable will be due to noise between GND and chassis - which is a separate problem entirely.

You can achieve this by a cap-ferrite-cap pi filter to GND on the 5VDC line. First capacitor provides path for noise current to return to the supply via GND (this is where the noise wants to go, noise currents want flow in loop back to source). Ferrite and second cap provides low pass filter reducing RF noise on 5VDC at connector.

In other word your first line of defense is to prevent noise currents from leaving your PCB and encountering the chassis. You do not want to include the chassis in series with the noise current that wants to flow from 5VDC to GND.

You will want to look at the switch node of the power supply. Zoom in on the rising and falling edges and look for VHF/UHF ringing (ignore slow KHz ringing). Some power supplies are worse than others, depending on how "snappy" the MOSFETs are. You can often reduce noise by slowing down the switching by inserting a few ohms in series with the output of the gate charge pump power supply (Vboost).

Also, be aware that 5VDC, prior to your filter, will be noisy. If it is a plane then watch for sneak paths that allow it to couple to other signals that go off board unfiltered.

In other words think about confining the noise to the PCB first before worrying about what to do the the shield (separate topic).