1
\$\begingroup\$

I am trying to select common mode chokes for a board that I am designing. The point of doubt is common mode and differential mode noise filtering and the parameters/values of the components selected.

Now, the system consist of the following - battery connected to my PCB via wires(Vcc and GND) of 20V or so, Buck regulator and the rest of the system.
Now, having long wires from the batteries I understand that there can be common mode noise that can creep into my system. So I am having a common mode choke (XFMR1). L1 and C1 will filter out differential noise.

I understand how these 3 work. Now, what purpose does C1 serve ?

Suppose my buck regulator supplies 4Amps,this would mean I will have to select a common mode choke greater than 4Amps rating,right ? Now, the common mode choke has the property of having a high impedance at specific frequencies. What frequency will my choke be of ? As it is coming from the battery and since the battery is connected to many other boards, it is difficult to zero in on a certain frequency,right ? How do I start about it please ?

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
  • \$\begingroup\$ You have designed it so why are YOU fitting C1 - who told you or what data sheet informed you it was a good idea. You also say that you understand how they work. I'm confused as to what you need to know. \$\endgroup\$ – Andy aka Nov 23 '15 at 11:46
  • \$\begingroup\$ My understanding is that common mode noise can be generated and can creep up the wires that connect it to the battery and hence the rest of my system. What I am confused about is that I will need to use it on the ground plane/line as well. Will that not affect my ground stability \$\endgroup\$ – Board-Man Nov 23 '15 at 11:50
  • \$\begingroup\$ @Andyaka. I am refering this document murata.com/~/media/webrenewal/products/emc/emifil/knowhow/…, page 4. \$\endgroup\$ – Board-Man Nov 23 '15 at 13:20
2
\$\begingroup\$

C1 is your first line of defense against differential noise coupled onto your wires, and your last line of defense against your device sending signals out into the world on those wires. Twisting the two wires tightly will help with differential noise as well.

If your buck supplies 4 A, it does not necessarily draw 4 A (as a linear regulator would!). If you have a specification on its efficiency, and you have good knowledge of its operating point, you might get away with less. But 4 A should certainly be enough, so that is a good 1st point. If you can't get one you like rated at 4 A, then sharpen your pencil (so to speak) and figure out exactly how much current will be needed.

Regarding the frequency for the common mode choke, the best option is to build it without one and measure the noise. Then select one that is best suited for where noise is the worst. If that is not feasible, then just try to get it below your circuit operating frequencies.

Do you have a separate chassis / earth ground, or is your negative coming in grounded?

\$\endgroup\$
  • \$\begingroup\$ Wow. Thanks. Yes I have a chassis ground. What I will do is connect my PGND or DGND(digital and power ground) to the chassis ground in the digital section via screws(metal). I read that they must be shorted at the digital section and not at the analogue section. The reason I am not able to understand. \$\endgroup\$ – Board-Man Nov 23 '15 at 13:17
  • \$\begingroup\$ But regarding the value of the C1, I must observe the supply lines and then connect the capacitor. I must understand the frequency of the noise (differential) and then connect a capacitor with very low impedance at that frequency,right? How much is low impedance? I mean what is the min and max that classes as a low impedance ? \$\endgroup\$ – Board-Man Nov 23 '15 at 13:19
  • \$\begingroup\$ I would aim for 5 ohms of reactance, maximum, and hopefully less than 1 ohm. You don't want to necessarily go really really high in capacitance (i.e a calculated reactance of microohms) because that often means that you are passed the point of the self-resonant frequency, and the Z=1/jwC equation is no longer valid. So, if your noise is 1 MHz, find a cap with a SRF of at least 1 MHz (generally greater, especially if the noise is wideband). Don't use a 10 uF thinking you'll get 0.016 ohms. use a 0.1 uF or even a couple of 0.01 uF caps in parallel. \$\endgroup\$ – MikeP Nov 23 '15 at 13:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.