accapacitor
I have seen here and there schematics with electrolytic capacitors put on AC. And this sounds weird to me.
Electrolytic capacitors have a polarity, right? If we invert the polarities on DC, bad things happen. As far as I understand, AC inverts polarity every now and then (commonly 50Hz). Why can we put such capacitors on AC without damaging them?
Example:
from demonstration here: http://youtu.be/qdXbnhb1bVo?t=5m57s
isn't it true that the capacitor will stop as many Volts it has written on it
No.
It's the opposite of what you want. You want the protection device to pass low voltages and block or attenuate high voltages.
Capacitors are not designed to be used as fuses, therefore you cannot really use one as a substitute for a voltage or current limiting device to protect IC pins.
The voltage rating on a capacitor can't really be relied on in that way. I wouldn't be surprised if a specific 60V capacitor that passes a 60V AC signal also passes a 120V AC signal. The rating only means the manufacturer won't guarantee the cap works above 60 V, they certainly don't guarantee that it fail at exactly 61 V (say).
What you are looking for is a "voltage clamp" or perhaps a polyfuse or similar.
Arduinos also have this problem of unprotected IO. There are variants that include better protection which seem to me to be a useful example
From here
"Every I/O pin is protected by a 5.1V zener diode and 220 ohm 30mA PTC (resettable fuse). The equivalent circuit is shown in this figure"
You'd have to adjust component values for a 3.3 V microcontroller or SoC such as the Rpi.
See also Pi GPIO protection - using (sacrificial?) buffers with ESD protection on a daughterboard.
For a detailed explanation on this and other related topics, though probably much more than you may be interested in, you can check the excellent "Input and output capacitor selection" Application Report, from Texas Instruments,
http://www.ti.com/lit/an/slta055/slta055.pdf
You can also check some general advice for troubleshooting circuits with linear regulators of the 78xx family,
Linear regulator (L7805CV) outputting 5.8V
Answering specifically your concerns:
You MUST NOT use BIG ceramic capacitors on input or output. It will do more harm than good due to the very low ESR (Equivalent Series Resistance) of these capacitors. On the contrary, small ceramic capacitors (<1uF) can be added to increase the transient response at high frequencies (>1MHz). This can be of huge importance in high speed designs. However, you should better use those 100nF (0.1uF) ceramic capacitors for decoupling directly the power pins of the high speed IC devices.
Keep the 100nF ceramic output capacitor in your prototype. However, it will probably make no difference, as your Raspberry Pi will have some its own high speed decoupling capacitors.
The LM1084 is available in 3.3V, 5.0V, 12V and adjustable versions. Which one are you using? I am assuming you have the 5.0V fixed version. If so, the recommended capacitors for most of the applications, according to the datasheet (pages 8-9), are:
1) Bulk output (load) capacitor: recommended 10uF (Tantalum) or 47uF (Aluminum electrolytic). You are not allowed to use BIG ceramic capacitors at the output of the LM1084. However, the datasheet explicitly states that "Output capacitance can be increased indefinitely to improve transient response and stability."
So, you are perfectly OK with a bulk 220uF electrolytic capacitor at the output. In fact, the added capacitance will improve your circuit under "large" and/or fast changes in the load current (like those generated by a USB hub or a Raspberry Pi board).
2) Bulk input (decoupling) capacitor: recommended 10uF (Tantalum) or 47uF (Aluminum electrolytic). Apparently, you are not allowed to use BIG ceramic capacitors at the input of the LM1084. And I say apparently because the input capacitor ESR has very litte or none impact in the stability of a regulator.
My advice here is twofold:
Keep the 470nF ceramic input capacitor in your prototype. It will help remove input transients and noise. It is low valued (<1uF) and at the input, so it won't create stability issues.
Add a 47uF or bigger Aluminum electrolytic capacitor. If you have a spare one, another 220uF like the one at the output will serve good. If you don't add this bulk capacitor, you regulator may oscillate under heavy loads and/or get damaged.
My personal, ethical choice: I try to avoid Tantalum capacitors, whenever possible, due to the enviromental concerns and conflicts it is creating worldwide, and specially in Congo and other African countries,
Best Answer
"Can" and "should" are two things. Should you do this? No: this use is outside the specified operating parameters of ordinary electrolytic capacitors. You seem to understand this already. Can you do it? Yes, as the video demonstrates. To understand why requires some understanding of what's inside the capacitor.
A capacitor is two conductors (usually plates) separated by an insulator. The more surface area, and the closer together they are, the higher the capacitance. Electrolytic capacitors have a thin film rolled up in the can. This film is covered in a thin oxide layer, and the thinness of this layer is what gives electrolytic capacitors their high capacitance relative to their size.
This oxide layer is created by the chemistry of the materials in the capacitor, and the polarity of the voltage applied to each side of the film. A voltage applied in the correct direction builds and maintains the oxide layer. If the polarity is reversed, the oxide layer dissolves.
If the oxide layer dissolves, you no longer have an insulator between the two plates of the capacitor. Instead of two plates separated by an insulator, you have two plates separated by a conductor. Instead of a device that blocks DC, you have a device that conducts it. Basically, you have a wire in a can.
Usually, when you encounter this failure mode, a large current flows, rapidly heating the internals of the capacitor. The expanding fluid and gas ruptures the safety vent or the can explodes.
Why then, does the capacitor in this example not explode?
The reverse polarity voltage is never applied for very long, and never without a correct polarity voltage applied soon after to repair whatever damage was done.
The oxide layer doesn't dissolve instantly when a reverse voltage is applied; it takes time. The time depends on the voltage applied, the size of the capacitor, the chemistry, etc, but half a cycle of 50 Hz AC is probably not long enough to cause serious damage. When the other half of the cycle comes around, the oxide layer is restored.
Any fault current is significantly limited by the series resistors.
With those resistors in series, the power available to heat the capacitor is small. There simply isn't enough power available to catastrophically destroy the capacitor because most of the available energy goes into the resistors. Perhaps you just warm the capacitor slightly. When the voltage reverses direction, the oxide layer can reform.
Probably you still damage the capacitor eventually, to some extent, but it is operational enough for the demonstration.