6
\$\begingroup\$

Is there a general self resonance frequency characteristics for MLCC capacitors? If I make a RC low pass filter, say using 1kohm and 15nF 0603 with fc = 10.6kHz, I guess it won't work as LPF beyond the self resonance frequency of capacitor. I couldn't find the self resonance frequency characteristics for capacitor (C0603C153F3GACTU) in its datasheet. Is there a general estimation? I am interested to know the frequency response of the filter for around 100-200 MHz.

\$\endgroup\$
  • 2
    \$\begingroup\$ Have you tried this: ksim.kemet.com/Ceramic/CeramicCapSelection.aspx it doesn't have 15nF C0G though. \$\endgroup\$ – Dejvid_no1 Oct 5 '15 at 12:30
  • \$\begingroup\$ Thanks. I just checked it. 10nF are the maximum available and they already have less than 100MHz SRF. \$\endgroup\$ – zud Oct 5 '15 at 13:35
  • \$\begingroup\$ @Dejvid_no1: thanks, that Kemet tool is an eye-opener. Notice the choice to "combine impedances". Things that have surprised me: 1) a particular capacitor model, I guess it was 0603 220nF X7R, has two resonant minima, spaced rather wide apart... 2) if I combine several C0G capacitors, say in a range between 10 pF and 22 nF, I get several parallel-resonant maxima ! well above 10 times the nearby "expected" impedance value. \$\endgroup\$ – frr Jan 9 '18 at 13:10
  • 1
    \$\begingroup\$ For the record: the problem of parallel resonance between several different-size parallel-combined MLCC's (also called anti-resonance) is described in fair detail, along with possible countermeasures, in the Murata C39E appnote. murata.com/~/media/webrenewal/support/library/catalog/products/… I've googled for a few hours now and this is the best paper on the topic that I have found. \$\endgroup\$ – frr Jan 9 '18 at 20:10
  • \$\begingroup\$ To summarize my impressions based on the Kemet KSIM, if I should trust the tool: C0G achieves some impressive ESR values in tiny capacities at high resonant frequencies, but for the purposes of power blocking, the other side of the coin is high Q, and it becomes a downside when combining multiple capacitors of different values, as the high Q boosts anti-resonance. For power blocking, you should look at X7R at the lowest available voltage and preferably low capacity, and maybe combine several pieces (equal or unequal capacity). Keep the voltage at twice your actual operating voltage. \$\endgroup\$ – frr Jan 10 '18 at 7:26
5
\$\begingroup\$

MLCC devices in surface mount packages all have a self-inductance (as indeed does anything carrying current).

The typical values for some common case sizes:

0402 : about 700pH

0603 : about 900pH

0805 : about 1.1nF

1206 : about 1.4nH

I am also using reverse geometry devices, specifically 0204 and 0306 with self inductances of about 280pH and 350pH respectively.

These are approximations as the effective inductance is determined by the specifics of the actual sizes of the internal plates (which varies across manufacturers).

You can calculate the SRF using the standard series resonance formula and this should be within about 5% (because the self resonance variation is the square root of the LC variation).

[Update: Added calculated SRF and impedance plot]

The part above would be self-resonant at about 43MHz, and here is the plot of Z vs. F:

Plot of Z vs. F for 15nF 0603 MLCC

HTH

\$\endgroup\$
3
\$\begingroup\$

I'd say that capacitor was unsuitable for your requirements. The Kemet spec says: -

Extremely low ESR and ESL

But fails to state what they are and of course you need the ESL to calculate the SRF.

Kemet also say this: -

Preferred capacitance solution at line frequencies and into the MHz range

How many MHz they mean is unclear but if you expect performance at 100-200 MHz you might be disappointed.

I've just taken a look at some data sheets from AVX and they do quote SRF (in graphical form) but the problem with these is that for anything higher than about 1nF, the SRF is going to be about 100 MHz - my advice is parallel up a 15nF with a 100pF - the 100pF will have a much higher SRF and the combination of both is likely to yield the result you need but not without a bit of digging around.

\$\endgroup\$

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.