3
\$\begingroup\$

I am working on a pcb that has to have a CPU with capacitors between its gnd and vcc pins, other components like sensors also have to have capacitors connected to them for whatever reason indicated in the datasheet, and these capacitors are also connected between gnd and vcc, but capacitors between the 2 pins will impact the overall capacitances on the entire board.

For example, if I need a 10pF capacitor between vcc-gnd pins of a CPU, and another one between vcc-gnd pins of a sensor, the equivalent capacitance will become 20pF, and if these kind of connections are too frequent, I will end up with a huge equivalent capacitance that may not work well for those components.

\$\endgroup\$
6
  • 2
    \$\begingroup\$ Why do you think adding up a few or a few hundred of 10pF capacitors would not work well? \$\endgroup\$
    – Justme
    Dec 4, 2023 at 19:30
  • \$\begingroup\$ This is only an example, suppose you have hundreds of 1uF capacitors, then your sensor that required 100pF as decoupling cap will endup with hundreds uF of equivalent capacitance, which does not respect datasheet specs. \$\endgroup\$
    – maths soso
    Dec 4, 2023 at 20:38
  • \$\begingroup\$ Capacitance on a power network stabilizes voltage and reduces noise. Noise is bad, so usually you would not think of your system having low noise and highly stable power as a problem. \$\endgroup\$ Dec 5, 2023 at 1:41
  • \$\begingroup\$ @mathssoso Bear in mind that anything less than about 10 nF is extremely unlikely to ever be used as a decoupling capacitor. Capacitors in the pF range are usually used for other purposes. \$\endgroup\$
    – Hearth
    Dec 5, 2023 at 4:50
  • \$\begingroup\$ "I am working on a pcb that has to have a CPU with capacitors between its gnd and vcc pins". Isn't that true of all CPUs, and indeed most chips in general? \$\endgroup\$ Dec 5, 2023 at 10:13

1 Answer 1

11
\$\begingroup\$

Yes, but.

Decoupling capacitors work best when they're as close as possible to the pins of the IC (some ICs even have internal decoupling capacitors placed directly on top of the silicon die itself). A capacitor providing decoupling to one chip will not necessarily be adequate decoupling for another chip a few inches away, even if it's nominally enough capacitance for them both.

At low frequencies (e.g. when measured with a multimeter), two capacitors associated with different chips will indeed measure as the sum of their capacitances. But at higher frequencies the impedance of the connecting traces (which is most of the reason these capacitors are used!) will become significant.

Outside of the outputs of amplifiers, it's unusual for too much capacitance to cause a problem, especially if it has significant series impedance (due to being halfway across the board). Note that linear regulators are amplifiers and can exhibit this type of instability, though modern ones are typically more tolerant of large output capacitances. Switching converters, as long as they use some form of soft-start, are unlikely to go unstable, though the cheapest ones might have issues starting up into a 10 mF load.

Decoupling capacitors halfway across the board are not generally something you will need to worry about for any chips other than your PMICs, and then only in fairly extreme cases.


Also, you're extremely unlikely to be using 10 pF capacitors for decoupling. It's usually 100 nF for most chips, a few to a few dozen μF for regulators, and probably somewhere in the 1 to 100 nF range for complex CPUs and FPGAs (times a lot, for the dozens of power supply pins that need decoupling).

\$\endgroup\$
8
  • \$\begingroup\$ There can be such as too much capavitance on a power supply. They might become unstable, or there may be a limit for maximum capacitance due to some specification which devices must pass, like USB. \$\endgroup\$
    – Justme
    Dec 4, 2023 at 19:33
  • 1
    \$\begingroup\$ @Justme Yes, but I would consider USB a somewhat unusual case. As for power supplies becoming unstable, that's also less common with modern regulators, but you're right that it is something to consider. \$\endgroup\$
    – Hearth
    Dec 4, 2023 at 19:37
  • \$\begingroup\$ More generally to say -- regulators are amplifiers, so are subject to the same stability considerations; whether they're stable with any load, or not, can be proven by design or testing. There can be energy limits, whether due to startup current or power, switching capacity (like for USB), or etc. Often, one or another consideration dominates in a given application, but it may be worth mentioning all for completeness. \$\endgroup\$ Dec 4, 2023 at 19:45
  • \$\begingroup\$ Why is USB unusual? People use it daily for memory sticks to charging laptops at 100W, so specifications and standard compliance is important. Also, put too many millifarads to a simple buck regulator and it likely fails to start, no matter how modern it is, unless soft start is utilized. But hardly adding hundreds of 10pF caps is near the limit though. \$\endgroup\$
    – Justme
    Dec 4, 2023 at 19:48
  • \$\begingroup\$ @Justme It's unusual to me, perhaps; I rarely use anything USB in my circuit designs, even just for power. \$\endgroup\$
    – Hearth
    Dec 4, 2023 at 19:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.