Hello, I have a project with a custom PCB which will use a ATmega328P powered by a single Lithium cell.
I plan to omit any type of LDO, and run the uC at 8MHz (or 1 MHz), so the low voltage (3,6V) should not be a problem.
The PCB will be powered through approx. ~30cm wire from the battery.
What kind of capacitors should be added near the uC? Besides from the standard 100nF bypass capacitor on Vcc.
Is a buffer capacitor (preferably 4,7uF or 10uF low ESR tantalum) needed?
Thanks in advance.
It is recommended of course, it depends a bit on what other parts you intend to power, the battery will be slow to respond to increased current draw, but the fluctuations will hardly affect the MCU.
Nobody uses tantalum anymore.
If you need any bulk capacitance it will depend on what else is on the PCB.
Because you get get 10uF (or higher) ceramic caps that are smaller, cheaper, perform better, and are less likely to catch fire.
Thanks for the explanation, it was my intention to ask @jim-p for further details.
Regarding the other components, nothing high consumers. Two very low current PIR sensors will be powered from the custom board, and two LEDs possibly (max. ~5mA using series resistors).
So a 10uF ceramic should be okay on the Vcc line before uC?
No, ceramics are a bad choice for input capcitors. They are not very good at absorbing input transients such as when you connect and disconnect the battery. A good electrolytic, 33uF or 47uF, with a voltage rating of 10V or more would be good.
Show your schematic and also a link to the PIR sensors. It is likely that these also have a large capacitor across the power rails. Also state if the PIR sensors are to wake the device from sleep mode.
Noted, thanks! Since it is a battery powered project, capacitor leakage current is also important.
I found this 33uF 10V and luckily it states "After 2 minutes' application of rated voltage at 20°C, leakage current is not more than 0.01 CV or 3 (µA), whichever is greater".
I think 3uA will apply to the 33 uF variant.
No schematic yet, I'm just in the data collecting period. 
I'll add it to the topic, as soon the first version is available.
PIR sensor datasheet shows some type of internal capacitance, but it's just a bypass cap I guess. Just the PIR sensor will be used, not a "breakout" board.
Yes, sleep mode will be used and either of the PIR sensors will wake up the uC.
Maybe you need new glasses.
All those yellow smd parts are tantalum.
You can find two under a classic Nano.
Leo..
Are you planning on running your circuit at 10V?
So what do you recommend that @amazed use as the input cap on his board?
No, on a single 3.6V cell. I guess, your question rather relates to the datasheet equation?
Because it always confuses me what voltage should be used in the given "0.01 CV" formula, because it states
"After 2 minutes' application of rated voltage at 20°C, leakage current is not more than 0.01 CV or 3 (µA),"
and the rated voltage is the nominal to me. But it would make more sense to calculate with the 3.6V, because leakage current is more or less the result of applied voltage and the insulation resistance, right?
For that PIR, it looks like you need a high value (say 100k) pull down on the input to the ATmega328p (pin D2 or D3) and wake up on rising edge trigger. You could also use a pin change interrupt if you can't use D2 or D3.
If the project is purely to drive lights, you may also want to include a photo resistor and only power the PIRs during darkness. Here is gets more complex because you will also have to wake the device regularly (maybe every 8 seconds) to sample the ambient light level.
For a 33uF, 10V cap: 0.01CV would be 0.01 x 33u x 10V = 3.3uA
But since you won't be stressing your capacitotor at the rated voltage the leakage should be less.
How much less, who knows.
Thanks for the pull down suggestion.
D2 (INT0) and D3 (INT1) will be used by the way.
The main goal is data collecting, which will be done also in darkness. The LEDs are only for testing/signaling purpose (will be switchable by jumper).
I found two capacitors, should is choose the one with lower impedance? Or it's not gonna make any difference?
UUD1A330MCL1GS
vs.
UCD1A330MCL1GS
The UCD version is said to be lower impedance version, price is the same.
Difference regarding leakage or something else?
Leakage should be the same, the same "0.01CV or 3uA" formula is given in both datasheets.
Low impedance would be better for higher ripple current, right?
You said there is nothing drawing a lot of curent so I don't think ESR will really matter in your case
