http://en.wikipedia.org/wiki/GPS_signals#Demodulation_and_decoding http://en.wikipedia.org/wiki/GPS_signals#Demodulation_and_de..., currently paragraph 4 descri

http://en.wikipedia.org/wiki/GPS_signals#Demodulation_and_decoding http://en.wikipedia.org/wiki/GPS_signals#Demodulation_and_de..., currently paragraph 4 describes pulling out the data. Basically, even though the signal is below the noise floor, part of the signal is the output of a pseudo-random number generator with known properties, so by setting up your own PRN, and then trying different (time) alignments, the correct alignment will stick out, despite the noise (which all averages out). Once you have the time alignment, you can read the rest of the signal, and then you can tell how far you are from that satellite ((it's clock - your clock) / c). Repeat for other satellites and you're done. Can't find any good references, but that's what our information theory lecturer said! Also, it explains why a hot start is quicker than a cold start (if you know where you are, and the accurate time, you know what alignments to try to start looking for the satellite signals.)