Steve:
IR Receivers have a nasty habbit of flopping up and down on their output due to random light in a room. Particularly if fluorescent lights are in use. They are also really good at picking up the bounce off the walls in another room nearby where a real IR transmitter may be in use.

It is for these reasons that it is so important to check the protocol quite closely in a decoder including the ID part of the code. In some codes they send part of the code in true & inverted format and it is wise to check that too.

RC/5 code from Philips is so great becasue it is biphase encoded (same as manchester). Biphase can be self clocking if you make the decoding algorithm robust enough. I notice that most of the algorithms for bi-phase decoding of RC/5 that you find on the internet use fixed sampling at 1/4 or 3/4 of the bit cell time. These algorithms just are not as reliable as they could be -- especially for longer transmission distances with the multi-path bounce and presence of the fluorescent lights.

I have one such robust algorithm that I made that runs in the TPU of a 68332 processor for a video conferencing system. I think I'll take a look at porting that to 8051 and posting it on my page here. A good project for another weekend.
Mike Karas