bypass capacitors supply power pulses to the components that they bypass.
this both allows the component to function normally since the pulses no longer get dropped across the power bus impedance and protects the rest of the system from the power pulses that the component requires.
You often add another bypass capacitor since the different capacitors work for different frequencies. You usually need at least two in critical high frequency systems, say a 0.01uF and a 100pF. If the system sees low frequency pulses as well as would be the case if you were taking a microcontroler in and out of power down you should add a 1-10uF as well.
The lowest values need to be at every pair of power pins as close as possible to the pins with as short traces as possible. larger values do not need to be as numerous or as close the the components as smaller ones.
Are we talking about the same capacitors?
I was talking about capacitors that are connected in parallel with the input pin, not in series, so how do they protect the PSU from the pulses?
Bypass caps just love to eat transitions on a DC voltage path. They gobble em right up. Once they smell a DC change they put on their dinner coat. Bypass caps keep a DC voltage nice and smooth. Integrated circuits on the other hand hate transitions on their supply line. It makes them seasick and queasy, they throwup in the form of oscillations, eventually they get a fever, burn up and die. Seems like a symbiotic relationship...
Bypass caps just love to eat transitions on a DC voltage path. They gobble em right up. Once they smell a DC change they put on their dinner coat. Bypass caps keep a DC voltage nice and smooth. Integrated circuits on the other hand hate transitions on their supply line. It makes them seasick and queasy, they throwup in the form of oscillations, eventually they get a fever, burn up and die. Seems like a symbiotic relationship...
Ahhh, I see your question now, and this is a good one, let me see if I can give a well deserving response.
So we are taught that caps in parallel add, so one would think that putting a 100pf cap in parallel with a 1uf cap would serve no purpose. But this is not the case.
Going back to my earlier analogy, caps like to eat, but 100pf caps only like fast small glitches, large caps being big and clunky as they are are somewhat slower, they can't catch those fast glitches, but in one fell swoop they can swallow one fat slow glitch. So little caps fast glitches and noise, big fat caps for big glitches and slow noise. End result nice quiet DC line.
Somewhat a childish explanation and I do not mean to insult your intelligence, but sometimes simple explanations are best.
-------------------edit---------------------------
For most designers, they use decouple caps like salt and pepper, just sprinkle them on a board and no longer analyze if they are needed or not. Why? because they are cheap and can't hurt and in most cases prevent possible problems from creeping up
Going back to my earlier analogy, caps like to eat, but 100pf caps only like fast small glitches, large caps being big and clunky as they are are somewhat slower, they can't catch those fast glitches, but in one fell swoop they can swallow one fat slow glitch. So little caps fast glitches and noise, big fat caps for big glitches and slow noise. End result nice quiet DC line.