3
\$\begingroup\$

In classic decoupling designs, it is common to see a large (10 µF or bigger) electrolytic capacitor in parallel with a smaller ceramic capacitor (in the range of 0.1 µF). The larger electrolytic capacitor can deliver more current over a longer period of time, but suffers from high ESR, thus a ceramic capacitor was used to handle higher-frequency current spikes.

However, it's 2016; today our circuits run on 1.8 V instead of 18 V. We can easily get ceramic caps as large as 100 µF in the voltage ratings we need.

Is there any reason we still need the classic 0.1 µF + 10 µF design? Why not just go with a single, 10 µF ceramic cap, which has similar ESR as the 0.1 µF?

I ask this because I've seen a large number of reference designs that have switched their polarized electrolytics over to ceramics, but still use the multi-value parallel approach. In one particular reference design, I even remember seeing a 22 µF, 0.1 µF, and 0.01 µF — all ceramic — run in parallel.

\$\endgroup\$
  • \$\begingroup\$ No, old habits die slowly. Go with ceramic. \$\endgroup\$ – winny Jul 11 '16 at 16:09
  • 1
    \$\begingroup\$ Beware the hazards only using ceramic capacitors: cds.linear.com/docs/en/application-note/an88f.pdf \$\endgroup\$ – user2943160 Jul 11 '16 at 16:34
  • 2
    \$\begingroup\$ Another problem with ceramic capacitors to look out for: Ceramics often have reduced capacitance with increase in circuit voltage, especially for smaller packages. I have read that the decrease in capacitance can be as much as 80% and this is not always evident from the datasheet. See: Ceramic Capacitors FAQ See also: Why 47uF capacitor drops to 37 uF, 30uF or lower \$\endgroup\$ – Tut Jul 11 '16 at 17:25
4
\$\begingroup\$

The approach is still relevant today. Why it is so is explained quite well by Dave Jones in this video (EEVBlog #859).

The reason why paralleling bypass capacitors is still relevant is because the frequency response (i.e. actual impedance) of the single capacitors is different due to different parasitics. To obtain a setup that will effectively bypass a large range of frequencies you need to combine capacitors with different nominal capacitance (and/or different technology and/or construction).

This all has little to do with the specific technology of electrolytic caps. In modern digital electronics the power pins of the digital ICs draw sudden pulses of current due to internal circuitry switching. These pulses have large bandwidth and need a bypassing scheme that attenuates the voltage drops these pulses induce in the power rails.

In other words, if you want to keep the voltage at the rails constant, you need to attenuate a very large range of frequencies, and this is not attainable (in general) with a single bypass capacitor, especially in digital circuitry with high clock rates.

To give further insight: think of a single capacitor as a tank circuit, where the capacitor may resonate with its parasitic inductance. Assuming a simple model where a capacitor has a single resonant frequency, paralleling different caps, with different resonant frequencies, allows us to "combine" the different frequencies responses so as to get a flatter overall response, hence a larger bandwitdh.

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

Some benefits of the old way:

  • The low value 0.1 uF is available in small packages like 0402 or even 0201. These packages have lower parasitic inductance, allowing them to decouple higher-frequency transients. This can be particularly important if your project needs to be qualified for radiated emissions.

  • A tantalum or aluminum electrolytic likely has higher ESR than the same-value ceramic part. This can be beneficial in actually absorbing power from the circuit rather than simply reflecting transient signals elsewhere. Of course whether higher ESR is beneficial or not will depend greatly on the nature of the decoupled load and the board layout. But it is important to keep in mind, for example if you're updating an older design and are considering replacing electrolytics with ceramics.

\$\endgroup\$
  • 2
    \$\begingroup\$ It is certainly possible to degrade (say) a '10uF' ceramic cap (which might actually be 5-8uF under bias) to simulate a 4.7uF tantalum by adding a small (fraction of an ohm to a few ohms) resistor in series. This can be useful for eliminating the ringing you (and @user2943160 in a comment) refer to (or stabilizing some older voltage regulators). \$\endgroup\$ – Spehro Pefhany Jul 11 '16 at 16:38
  • 1
    \$\begingroup\$ @SpehroPefhany, that's true, and you might even be able to do it using a constricted trace rather than an smt resistor, if you're willing to bird-dog your vendor to make sure they don't mess up the trace dimensions. But either way you increase the ESL, and you still don't get a single capacitor to cover all your decoupling needs like OP wants. \$\endgroup\$ – The Photon Jul 11 '16 at 16:44
  • 1
    \$\begingroup\$ Yes, good point. There still might be advantage in replacing 1-2 ceramic caps and a tantalum with 2 ceramic caps and a resistor (reliability, size, cost). \$\endgroup\$ – Spehro Pefhany Jul 11 '16 at 16:52

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.