Merci Alain,

I can't see any change with the decoupling cap and the power is ok. I'll go for some other values and check again.

PBP manual: PULSIN normally waits a maximum of 65535 counts before it determines there is no pulse.
It looks to be the same for RCTIME... but as you say, RCTIME has not the same behaviour as PULSIN.

I think, depending on the IR remote type and protocol, I miss the first bit of the pattern.

Is it possible to make PULSIN scan a pin without time limitation (=until a signal is detected)?