A minimum of 10-µF tantalum capacitor is required at the output to improve the transient response and stability.
and
An input capacitor is recommended. A 10-µF tantalum on the input is a suitable input capacitor for almost all applications.
and the example circuit shows a 100 µF on the output, 10 µF on the input.
The Arduino Nano schematic shows… a 1 µF cap on the LM1117 output, and no capacitor on the input.
Hm.
Any insights? For Nano based circuits, should 10 µF caps be added on the 12 V rail (if that’s what’s powering the Nano) and 5 V output? (Of course if there are other ICs in the circuit their bypass needs would have to be satisfied too.)
The PSRR is given in the datasheet, so if one were to define typical noise levels and loads on the in and outputs, the required capacitance can be calculated.
Now for a more practical consideration. It’s already common practice to add 10uF across the power rails close to the power connector. Unless a circuit would involve ridiculously large currents or noise components (a valve, perhaps?), adding a second 10uF doesn’t seem necessary to me. Rather, I’d add the standard 100nF close to the power pin for some high frequency rejection.
The 5v output then. 1uF close to the regular is already a nice step. One shouldn’t be using the onboard regular for large loads anyway. Probably the most effective option would to be add 100nF accros the input of any components being powered from the 5v rail.
Personally I power the Arduino externally whenever the 5v rail powers little more than the Arduino itself. In that case the 7805 or whatever regulator used needs its prescribed input and output caps, obviously. By the same logic as above , I add 100nF to the 5v input pin of the Arduino, and any other components on the 5v rail.
Maybe it’s got something to do with keeping it suitable to run on USB power. For some reason I don’t honestly understand, the USB standard limits how much capacitance you can put across the power pins. I’ve put 4400uF on those rails and nothing bad seemed to happen from it, but those guys have gotta worry about the specs a bit more.
The “common practice” 10 µF from the 12V rail to ground at the supply input was what I was wondering about and what prompted me to look into the schematics in the first place — not a second one at the VIN pin.
As for the 5V output, TI’s “A minimum of 10-µF tantalum capacitor is required” seems a very strong statement. They say “recommended” on the input but “required” on the output. If USB power isn’t used, is there a performance downside to putting 10 µF at the 5V pin?
TI’s example does not add a 100 nF on input or output.
In the Nano schematic there are a 4.7 µF and a 100 nF from VUSB to ground. And for some reason, there are two 1 µFs and a 4.7 µF in parallel from +5V to ground next to the ATMEGA328P (maybe they’re in more separated physical locations on the board? But they do show the parallel 1 µF next to the LM1117.) No bypass cap is shown near the FT232RL.
Maybe it’s just a matter of the datasheet author trying to cover TI’s tuchus for as wide a range of applications as possible. There’s no way to anticipate all of them, but 800mA is a generous enough current that it could conceivably supply quite a few chips that power on and off from time to time. So they’re specifying enough capacitance on the rails that it can respond to those changes in load.
Whereas the Nano has just the one mcu, and Arduino were desigining it to be their cheap and small option.
