Filter module


I need some calculation for my thesis. A found this picture on the internet. My project is almost the same. It's a pi filter for a 3 phase BLDC. The peak current is 60A. It runs from 12V car battery. I need to design a filter which can handle the sudden current demand, while the voltage ripple is not too large on either side.

How should I choose the values and calculate the voltage ripple? Thank you for any help!

  • \$\begingroup\$ You are talking about 12V battery but in schematic there is 48V. What is correct? And why do you think you need a PI type filter for stabilizing the voltage? \$\endgroup\$ Sep 22, 2018 at 14:03
  • \$\begingroup\$ This is just a picture from the internet. I have to design a transmission control units filter. This is why 12V. At my workplace they used pi filter. Do you think that other filter type would be better? \$\endgroup\$
    – Tardief
    Sep 22, 2018 at 14:14
  • \$\begingroup\$ Do you know U=LdI/dt and I=CdU/dt. That’s pretty much all you need to know but it could help to calculate for example corner frequency. You select the values to satisfy your voltage (and current) ripple requirements. In essence you need to define what ”not too large” is. \$\endgroup\$
    – winny
    Sep 22, 2018 at 14:49
  • \$\begingroup\$ This question is similar to, and may be better answered, here: Input filter LC for switching power supplies \$\endgroup\$ Mar 31 at 6:15

1 Answer 1


I just read, this is not your design but an example from the web ( without a citation)
- Do you have any test results for Motor step load regulation at DC feed and at the battery and motor current?
- Do you have any RLC parameters for EVERY physical part and path?
- It all matters if not done right.

Do you have any good experience with Sim Tools.? I like Falstad's because it is realtime for tweak values and easy to draw. Shift or ^ and mouse wheel, right click and menu have many hidden features to tweak values and draw standard filters or make your own as I partially did to follow your schematic, obviously with fake values. The parts are ideal, so you must create the model with discrete RLC values for each wire, cap etc.

Painful but this is the reality. If you do not have an expensive system like COMSOL (which has a steep learning curve and expensive) then this is one free way to do it after you research the values for the model. enter image description here

It helps if you have experience on ESR*C values like 1us is low ESR and 100us is general purpose and then ceramic is much lower and those with W/L geometry<1 are microwave caps. open wire is 10nH/m approx, paired wire xx pF/m.

I did NOT model the ground return path which also has an RL component model. So I assume you have tried to make this negligible like short connections to chassis, Litz wire and 50uOhm terminal fasteners for example.

I suggest you spend a few days learning a good Sim like Falstad's and try to get a network analyzer to get the transfer function on your physical model to match your mathematical model to see why it has so much ripple, noise and EMI, ESR losses, crosstalk, interference on sensors etc.

A pal of mine recently built an 800VDC laminated busar with expensive big caps so I did a quick and dirty model for the 75kHz full bridge without final values of the busbar and cap where ESL was critical for 10kW with very high dI/dt. enter image description here

You can see the effects Q and resonance from the component value and ratios can lead to PWM frequency issues if done wrong... leading to IGBT shoot-thru or SOA thermal issues.

Final note: The motor BEMF has a spectral shape and not a flat sweep so the net ripple spectrum will be determined by the RPM, reluctance model and motor/generator spectral density.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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