2 comments
Most linear and switched voltage regulators have upper limits for the allowed load capacitance, they become unstable (oscillating) above the limit. Replacing 100nF for 3.5uF everywhere is very ill-advised. Better to stick with 100nF, and have a distributed (total n*100nF) capacitance of few uF on the whole load line, for stability and low frequency decoupling, no more.
In audio or especially musical instrument applications there may be some "analog" to this.
You need clean DC power for analog or digital circuits, so you have to start with a power supply having adequate specs. And if it's good you can kind of take it for granted from there.
But that's just getting started.
The circuit you will be powering will usually have a schematic that is best represented separately from the power supply. What you really need to look at on most drawings is the digital logic or analog device interconnections without all the extraneous wiring that will exist on the final circuit to carry the supply voltage from the V+ & V- to every single component that needs it.
In an ideal world on a non-powered breadboard you connect the components according to the schematic just like it is in the drawing without any power lines yet. You may or may not be only about halfway done. Then from the power supply module you run the power lines to the key points on the breadboard that need it.
The power supply is so clean, and the world so ideal, that no further decoupling capacitors are needed beyond that contained within the power supply already.
But that's a lot of wire like spaghetti all over the place. And you may be handling high frequencies, or digital signals having similar slopes on the scope, even if they are not sine waves.
Depending on the current and impedance, some of these stray wires can broadcast these "signals" a few inches like the wires were radio antennae. And sometimes these signals or interference can be picked up by other traces acting as receiving antennae and behave as unwanted input to the analog or digital circuit. You can only imagine the inaudible feedback that could develop beyond your hearing range under some conditions, and the compromises it can make in the audio spectrum anyway. Or you could use the scope.
Keeping the interconnecting traces short and carefully laid out cuts back on the receiving antenna effect for the heart of the circuit. Adding a "local" capacitor between + & - on every PCB that is connected to the power supply module still helps even though the supply itself exceeds specs and has enough storage, filtration, and decoupling caps inside to do its job well. Its job is over but yours starts where it leaves off.
If the current fluctuation on the daughter boards is not much you probably won't need any significant local storage caps there, only tiny caps for filtration like the 100nf range. This is not a random value, but it is a completely generic one, simply the most suitable component most of the time when chosing between 10nf, 100nf, and 1000nf. Definitely not a calculated capacitance almost every time you see it.
So that beyond a certain point (depending on other working circuit parameters, you would have to measure that or at least calculate it) all the higher frequencies, digital signals beyond a certain slope, or interference spikes narrow enough to be beyond that range, which appear on the power lines for any reason, will all be shunted directly between + & - hopefully without allowing the working circuit or sensitive component to be influenced by the undesirable signals at all.
In the prototyping phase, sensitive components or difficult PCB layouts have the weak points identified and decoupling judiciously applied, after the original design schematic did not include them at all. On a sensitive component the closer the decoupler cap is to the semiconductor, the better it can protect from interference appearing on the trace leading right up to the component.
No logic is supposed to change whether many or few caps are added, but this can require troubleshooting that was not fully anticipated during the original design phase.
A trend developed where often every chip has a 100nf and nobody bats an eye.
With analog, you hear everything within range, and decoupling can effect the sound for better or worse.
Beyond that and more subtle, though quite perceptible, can be the effect on performance.
But it's just a few capacitors sprinkled in, look at the equations, that's not supposed to change anything meaningful . . .
It's the ideal world, at it again.
You know, the planet where star grounding and separate analog and digital grounds or planes aren't needed either.