It seems to be usual practice to put in decoupling capacitors of several different values, for example 0.1uF and 1uF (or sometimes 10uF, or 0.01uF depending on the device etc). I'm wondering if there is anything that can replace a 0.1uF/1uF combo with a single device? In particular, would a low inductance 1uF (reversed geometry, LICC, or 2T-LGA) be a suitable replacement?

I have looked at the graphs that show what happens in terms of impedance when combining multiple values, but I haven't seen a similar graph for LICC that I can compare directly.

EDIT An example of the parts that I'm looking at is:

Compare these with totally ordinary 0.1uF 0402 capacitor: if I'm reading the datasheets right, the LICC devices have (far) lower impedance at all frequencies 0.1-100MHz. Hmm...

  • 1
    \$\begingroup\$ dont mix up esr with parasitic leakage inductance. You need to find capacitors that show graphs of impedance versus frequency and there are plenty that do show this info. Those that don't are not worthy of consideration. \$\endgroup\$ – Andy aka Oct 13 '13 at 21:17
  • 1
    \$\begingroup\$ Can you give a link for the datasheet of the part you are considering using? \$\endgroup\$ – The Photon Oct 13 '13 at 23:00
  • \$\begingroup\$ @ThePhoton: Linked a couple of datasheets above. Let me know what you think? \$\endgroup\$ – Alex I Oct 14 '13 at 2:18
  • \$\begingroup\$ @Andyaka: The datasheets often show ESR vs frequency and Z vs frequency in the same graph. I think Z is the more useful figure of merit, correct? \$\endgroup\$ – Alex I Oct 14 '13 at 2:22
  • \$\begingroup\$ They are both useful in different ways. \$\endgroup\$ – Andy aka Oct 14 '13 at 7:17

I only looked at one of your proposed parts, the Murata LLL153C80G105ME21. I compared it with a same-value part in a larger package (GRM21BR71E105KA99#, 0805 size), the key improvement is in the available voltage rating. The 0204 part is rated for 4 V, while the 0805 part is rated for 25 V.

Even if your application only applies 4 V to the cap, take note of the capacitance change with applied voltage charts. The value of the 0204 part will be reduced to a bit above 30% of nominal (e.g. 0.3 uF instead of 1 uF) with 4 V applied. The 0805 part will still be at 95% of its nominal value with 4 V applied, and only loses about 45% of its value at 25 V applied.

So the smaller part can be used if you can accept its reduced temperature range, but its value will be reduced to just a bit more than the 0.1 uF value that has been typically recommended for use as the near-chip bypass capacitor over the past decade or so. If you really want 1.0 uF of bypassing, you'll still have add some larger parts in parallel with the suggested 0204 part.

On the other hand, if you can live with the low WV rating and you use this part in place of the "traditional" 0.1 uF 0402 part (in parallel with additional larger-value caps), you will gain a 3 - 4x increase in effective capacitance, so that is a substantial improvement.

Also, in a high-reliability application, you may want to use a package at least one size up from the minimum needed for the capacitor value and WV you are using. The smallest available size is pushing the limits of what the manufacturers can do, and can have reliability issues.

  • \$\begingroup\$ You make a great point about capacitance over voltage, I had not noticed this effect before. I am working from a reference design for a microcontroller that has a 1uF and 0.1uF, both standard 0402 4V X7R parts. The supply is 1.8V, so they would already be derated significantly - they have similar capacitance over voltage curves. This is what I am trying to match/beat. I think a single 1uF LICC 4V 0204 would do that, even though it would be similarly derated. Does that sound like a workable plan? \$\endgroup\$ – Alex I Oct 14 '13 at 7:22

No, because the two capacitors serve different purposes. The larger capacitor provides power during large bursts of demand, and the smaller capacitor filters high frequency noise from/to the supply. A single capacitor cannot perform both tasks effectively.

  • \$\begingroup\$ Thank you for the explanation. I do wonder about it though. If replacing a 0.1/1uF combo with a 1uF clearly there is enough total storage for large bursts of demand, since the total is about the same. So the only potential issue is filtering. The smaller capacitor filters high frequency noise because it has a lower impedance at high frequencies, because it's self- resonant frequency is higher, correct? In that case, would not any capacitor with a similarly low impedance at the frequencies of interest do just as well? \$\endgroup\$ – Alex I Oct 14 '13 at 3:27
  • \$\begingroup\$ Unfortunately my analog is a little weak, so I don't have an answer for that. \$\endgroup\$ – Ignacio Vazquez-Abrams Oct 14 '13 at 3:33
  • 1
    \$\begingroup\$ The small decoupling capacitors are supposed to go very near devices to catch things like spikes when the device switches IO pins etc., the big power supply smoothing caps go nearer the main supply. My boss is an RF/Analogue guy and he complains bitterly about data sheets etc. that just scatter an assortment of caps all over the board in the hope it will work, as it can actually lead to the various different components forming oscillators as their ESR's change with the various frequencies and harmonics present. Blame laid firmly at the feet of "digital guys"... \$\endgroup\$ – John U Oct 14 '13 at 8:28

If you can accomplish the same filtering and supply requirements with a single capacitor, use one. Generally, it is not the case, as stated above. However, a multiple capacitor solution has several disadvantages: 1. higher cost (2 components, 2 placements). 2. larger board space. 3. If using two capacitors, the frequency response plot has a zero between the two poles, which might cause some problems.

  • \$\begingroup\$ Could you explain about the zero between poles? Do you mean there is a frequency at which the two capacitor combo has high impedance due to anti-resonance? \$\endgroup\$ – Alex I Oct 14 '13 at 7:41

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.