What is the reason you find 100nF || 47uF in circuits?

As a newbie in electronics here comes a basic question of mine:

Lately I often heard or saw something about parallel-capacitor-wiring which i cant quite understand.
Ofc, using capacitors in front of an input does help the voltage be more constant. no doupt here.
I know that

C_12 = C_1+C_2

in a parallel wiring of capacitors. But why would someone do this with 100n || 47u?
47.1u or 47u? I'd think that this is not much difference and let the 100n be.

An example is the scheme of the ArduinoMEGA2560. Here, at page 1 in the upper left, you can see the
"NCP1117ST50T3G". On its input-side you have what i described before. One of them seems to be an electrolyt-capacitor since it has a "+" at its symbol. The other one doesn't. Is this a reason for the wiring? A current the other way will not hurt the electrolyt-capacitor? (does that idea even make sense?) Or do i get everything wrong and "PC1" is no capacitor? (i doupt it)

ALL IN ALL: Why do i need the 100n-capacitor?

I used to wonder this, and someone explained it last year.

It has something to do with the series resistance or leakage resistance of the capacitor. If you assume an "ideal" capacitor it would be, as you say, rather pointless. But in the real world, capacitors do not behave as "ideal" capacitors. Therefore you use two different types of capacitor to meet two of the two different goals of having the capacitor there ( such as voltage smoothing and avoiding high-frequency noise ).

And yes, electrolytic capacitors DO have a correct polarity and will be degraded by reversed voltage, so the polarity does matter.

Thanks for your answers guys ! (Wow, this frequency hurts the eyes :smiley:

michinyon:
[...]and avoiding high-frequency noise

And by that you mean that the 100n will result in an higher resistance at higher frequencies and therefor the circuit wont be so sensitive to these frequencies? I'm not sure here

JkbS:
in a parallel wiring of capacitors. But why would someone do this with 100n || 47u?
47.1u or 47u? I'd think that this is not much difference and let the 100n be.

It's not about capacity, it's about response time. A small ceramic will respond very quickly giving the slower electrolytic time to respond with its larger capacity.

Grumpy Mike has a page about this, but we are talking about decoupling here - decoupling
means keeping the power and ground rails from changing voltage when the load current
changes. This is easy at very low frequencies as the power supply is designed to do this
and the wiring has a very low impedance.

However at the speed logic gates switch (nanoseconds), the wiring has a much much
larger impedance (hundreds of ohms), due to the stray inductance
involved in being a wire. Without decoupling many strange failure modes can occur
and in pattern sensitive ways - the circuit becomes unreliable and unpredictable basically
because the power and ground pins are jumping around in voltage at nano-second
timescales.

Power and ground planes have a lower impedance than wires, but its still in the ten+ ohms
range, and without a 4 or 6 layer PCB you won't have both power and ground planes.

Wider PCB traces have less inductance, this is why power and ground traces are
wider than signal traces (the DC resistance isn't normally the issue with small logic chips).

This means that every power pin on a chip needs to be directly connected to a
low-impedance capacitor to ground with a short wide trace. At these high frequencies
this means ceramic multi-layer chip capacitors ideally as these have minimum inductance.

Smaller capacitance values tend to have less series-inductance so a 100nF device is
almost always used (these are good to GHz), but to gain better (lower impedance)
at lower frequencies (10MHz range) adding a 10uF or similar is often done. Lead
series inductance is less important at these lower frequencies so such larger caps
don't need to be right next to the chip, and can easily be shared between nearby chips.

When adding decoupling caps to a breadboard use the shortest route between power
and ground pins you can (straddling the cap over the chip is a good way).

It isn't ESR so much, but inductance. A large aluminum electrolytic capacitor is made by winding two thin strips of aluminum foil into a cylinder. It tends to have a rather high self-inductance. So above the resonant frequency of the capacitance and its self-inductance, the impedance goes up with frequency.

Ceramic caps have much lower self-inductance, this combined with a much lower capacitance means a much higher resonant frequency, and so much higher frequency before the impedance becomes primarily inductive. There is also the inductance of the wire leads and PCB traces. Which is why you are supposed to put those 100nF capacitors as close to the IC power pins and ground as possible.

That's a great graph Polymorph. Goes to show that you never can tell ... when you read some of these "basic" threads, you might still learn something. 8)

I didn't draw it, I just found it. There are a lot of graphs that show the frequency response overlapping, but don't show them summed. Misleading unless they are summed.

I figured, based on the caption. Still, it illustrated the explanation very well. Sometimes I regret that particularly lucid lectures get lost in the archives. Keyword search is never a complete replacement for stumbling across good material by happenstance.

Yes, that goes for the internet in general.

michinyon:
And yes, electrolytic capacitors DO have a correct polarity and will be degraded by reversed voltage, so the polarity does matter.

not ALL electrolytic caps have polarity....

If you're going to be pedantic, even bipolar electrolytics have polarity, they're just stuffed two-to-a-can so they can't pass DC either way.