# Does decoupling capacitors try to charge the fluctuating source back?

In common examples of decoupling capacitors, the scenario is like this: When the voltage fluctuates below or above its ideal value, capacitor either stores or releases energy to compensate the fluctuation in power so that the load gets a nice, flat voltage.

simulate this circuit – Schematic created using CircuitLab

Now, when I read the circuit form left to right (first looking at the source, then the cap parallel to it, then the load parallel to it) this explanation sounds right because it seems like power is delivered from left to right and capacitor behaves like a buffer in between.

But when I think about it, if cap is at 1V and power source suddenly drops to 0.95 volts, there is a short circuit between 1V and 0.95V so a relatively high current must flow from the cap to the voltage source. This not only tries to charge the source back but also wastes most of the power on that little short between 1V and 0.95V.

So, how should I think so that this circuit will make sense to me?

• Depends on the source Jun 16, 2020 at 2:36
• The source cannot accept a charge unless it is a battery. A typical DC power supply would have diodes to block back-feed, but capacitors in the power supply output can charge from either direction. Current to correct a 50mV drop would be small and short-lived. Wasted power would be tiny if changes in voltage are not rapid and repeated and have a significant change as a percentage of 'normal' voltage.
– user105652
Jun 16, 2020 at 2:45

Generally speaking, it isn't the source that fluctuates, it's the load. One common example is digital circuits that draw short-term spikes of current when their outputs switch. And the best way to think about it is that there is some significant impedance between the source and the decoupling cap, especially at higher frequencies.

simulate this circuit – Schematic created using CircuitLab

So while the DC signal flows from the source to the load, the high-frequency AC noise imposed by the load flows mostly through the capacitor, both charging and discharging it. It "decouples" the noise from flowing through the power supply impedance, shorting it to ground instead. The goal is to keep the green noise loop as short as possible.

The peak noise current is never greater than the DC current, so there is never a net flow "backwards" through the power supply.

• This makes a lot of sense. I was reading The Capacitor Handbook - Cletus J. Kaiser and that mentions fluctuation in source, that confused me but your explanation was helpful. I do still wonder though, what would happen if the source fluctuates? Jun 16, 2020 at 3:59
• Source fluctuations are generally much lower in frequency (often a harmonic of the line frequency), so you add much larger "bulk" capacitors (hundreds or even thousands of uF) close to where the power source is connected to the PCB. Again, working with the power supply's source impedance, these form a low-pass filter to attenuate that kind of noise. Jun 16, 2020 at 4:00
• I can think of one particular scenario where a source can fluctuate: think of a noisy linear stabilizer with a capacitor on its output. The datasheet typically mentions some capacitor value for output filtering, which tends to be "not very big" - say 1 uF ceramic. If you add capacity and lower ESR, the stabilizer's own noise will get amplified as output current noise into the low-ESR capacitive load. Not so well behaved stabilizers can ring/oscillate if memory serves... Even some SMPS bucks, designed for wet elyt, can oscillate if loaded by a solid poly.
– frr
Jun 16, 2020 at 6:30