11
\$\begingroup\$

I was wondering if someone could explain why larger package capacitors (1210) are supposed to have more ESL and ESR than a smaller - say 0603 package?

I would imagine the larger package is still essentially many 0603 equivalents in parallel for a multilayer ceramic. Say we are comparing a 0.1-1uF 0603 to a ~10uF 1210 package, wouldn't the 10uF be more effective for decoupling? Why are smaller packages recommended for decoupling when larger packages "seem" better in my mind.

Many thanks!

\$\endgroup\$
6
\$\begingroup\$

Generally-speaking, larger capacitor packages increase the current loop through the part, so the inductance (ESL) is greater. Similarly, the extra material means the resistance (ESR) is higher. When you put the ESL and capacitance together for decoupling applications, you get an LC tank circuit with a resonant frequency that decreases with increasing inductance and capacitance. The ESR in this circuit represents the minimum impedance at resonance.

When decoupling, you normally want to get below a certain impedance over the operating frequency range of the device in question. To achieve this you need multiple LC circuits covering different parts of that frequency spectrum. This is why you need a range of different capacitor sizes.

In order to achieve your desired ESR, you may also need several capacitors in parallel rather than one, as the ESR goes in parallel as well and will therefore be lower.

As a last note, also bear in mind that the escape pattern you use (position and number of vias & traces) from decoupling caps can dramatically affect the decoupling performance as well, because they add to the inductance. When you get below 0201 caps you can find that the overall inductance actually increases with smaller cap size because of this.

More information here:

Additional information

\$\endgroup\$
  • \$\begingroup\$ Additional information link is dead (looping redirect). \$\endgroup\$ – user4718 Mar 1 '16 at 0:25
4
\$\begingroup\$

Smaller packages have different resonance points than larger packages. Larger packages also have higher lead inductances(you'll need to think about through hole packages).

Smaller packages are always better for high speed since they reduce the length a signal has to travel. As you know, for high speed design, the longer the length the more problems one has. This is why FGPA's can operate so fast even with many paths because the paths are all crammed on such a small area.

There is a good analysis of smd package sizes online somewhere(I don't have the link but saw it ones). It talks about why one should use both large and small sizes for bypassing which has to do with resonance. Decoupling is a different story though. It all depends on what kind of signal you want to decouple.

Smaller generally is better simply because it allows one to reduce the signal path. This is always good. It is not always the case the smaller is better though(you end up with other problems such as crosstalk).

Note that when you parallel things you may reduce some factors you will also increase others. If you parallel resistors you might reduce their resistance but you increase their capacitance. It may be a higher capacitance than if you just used one resistor with the combined resistance in the first place.

When dealing with decoupling capacitors another factor is leakage. This paralleling capacitors increase the leakage. This is usually pretty bad for decoupling because you are not as decoupled as you may want.

\$\endgroup\$
  • 1
    \$\begingroup\$ There is a good analysis of smd package sizes online somewhere(I don't have the link but saw it ones). It talks about why one should use both large and small sizes for bypassing which has to do with resonance -- I'm very interested in this if anybody finds it. That's exactly the question I need an answer to... \$\endgroup\$ – user4718 Mar 1 '16 at 0:26

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.