0
\$\begingroup\$

I am using a DC to DC converter (Murata MEJ2S0505SC (datasheet)) (5 V to 5 V at 500 mA) for isolation purposes. Now I want to calculate the input decoupling capacitor.

Can anyone please suggest any methodology to calculate it? (Note: It is not mentioned in the data sheet.)

\$\endgroup\$
5
  • \$\begingroup\$ Have you tried simulating it? \$\endgroup\$ – winny May 25 '20 at 11:57
  • \$\begingroup\$ Now i added data sheet,and i did not simulated and i need theritical explanation. \$\endgroup\$ – yugi May 25 '20 at 12:03
  • \$\begingroup\$ The capacitor is usually associated with an inductive element to form a front-end \$LC\$ filter. You dimension these components based on a maximum allowable input ripple current at a certain frequency. Usually a small \$L\$ is adopted as it supports the dc current while a large capacitor can cope with the high-frequency pulses absorbed by the dc-dc module. Have a look at a seminar on filters I taught at an APEC conference in 2017. The example starts slide 80. Don't forget to damp the filter. \$\endgroup\$ – Verbal Kint May 25 '20 at 12:12
  • 1
    \$\begingroup\$ Did you look through their application notes, DCAN-68 "Low power isolated DC-DC application notes"? \$\endgroup\$ – SteveSh May 25 '20 at 12:18
  • 1
    \$\begingroup\$ murata.com/-/media/webrenewal/products/power/appnote/… \$\endgroup\$ – Andy aka May 25 '20 at 12:53
1
\$\begingroup\$

The input filter specifications depend on: (1) your source impedance and (2) dynamic frequency of the DC-DC converter (3) the spectrum of your step load. (4) radiated EMI of source current and conducted noise as a result of (1,2,3)

This is a 2-way filter spec. If you require load regulation noise on the source, the unbalanced load impedance being the input to the DC-DC can generate noise on the source.

Consider the input cap. as a LPF to the source impedance including path inductance, both DM and CM. Ideally, you want 0 Ohms and expect a well-designed DC-DC converter from Murata will have dampened chokes with suitable Q's or L/R ratios.

Design thought experiment

  • Consider the voltage divider relationship of a repetitive pulse with load impedance and the ESR of your input cap
  • define your allowed load ripple voltage error / step load current = Step Zout
  • Consider the impedance of Cin @ f DC-DC = Zc(f)
  • be aware that the C*ESR =Tau product is limited by the type and quality of capacitor material < 10us for low ESR e-cap and > 100us for G.P. e-cap. while Ceramic is lower density much much faster Tau.
  • Now define Z cap in terms of minimum C and maximum ESR @ f for step load I at output ripple Vpp

To meet all the considerations in the beginning; it may end up being a CM choke with two caps in a CLC or Pi filter with a certain differential attenuation Bode Plot that is trivial to design on Falstad's Analog filter site after you define impedance and attenuation @f.

\$\endgroup\$
2
  • \$\begingroup\$ Might also suggest using multiple caps in parallel possibly of different types, due to self-resonance and series resistance - to get filtering over a wider range of frequencies. \$\endgroup\$ – scorpdaddy May 25 '20 at 13:26
  • \$\begingroup\$ Certainly @scorpdaddy,, I imagine the Zcap(C,f,+ESR) of the caps over the load spectrum must be closer to zero than any possible negative incremental resistance For the input Z(V,f) of the DC-DC in the boost startup mode in order to be stable. \$\endgroup\$ – Tony Stewart EE75 May 25 '20 at 19:31

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.