Also! Erik DeBenedictis invented an optimization to S2LAL that can compute the inverses as-needed, rather than keeping both the 0,1 signal pairs around:

I don't quite understand this - the diagrams in Erik's paper are still using the wire pair convention, no? Launching a new stream with an inversion is still useful for sure but I think this is an incorrect description

Hi @Riking28, there are two levels of pairing normally in S2LAL. The active high/low signal pairs are electrically required to control the transmission gates, but they do not correspond to logic 0/1. If one signal pair represents logic 1, then you normally would need another, completely separate signal pair to encode its logical complement (because normally every signal is required to go inactive at a certain time in the clock cycle, so simply swapping the rails in a single pair does not work). But, Erik's note is showing how the signal pair for logic 0 can be generated dynamically from the signal pair for logic 1 (or vice-versa).