Clock and data recovery (CDR)
Clock-and-data recovery is how a SerDes receiver figures out the beat of the music when nobody sent a metronome. There is no separate clock wire — the timing is hidden in the data's own transitions — so the CDR watches the incoming edges and regenerates a clock locked to them, then uses it to sample each bit at the optimal instant: dead-centre in the eye, as far as possible from the blurry edges. It's like tapping your foot to a song: you infer the tempo purely from where the beats land.
Most CDRs are a phase-locked loop with a twist: a bang-bang (Alexander) phase detector samples the data at the edge as well as the centre and votes 'sampling too early' or 'too late', nudging a controlled oscillator until the centre sample sits in the eye's middle. The CDR's loop bandwidth is a careful trade-off — wide enough to track the transmitter's slow wander and spread-spectrum clocking, narrow enough to reject high-frequency jitter. If the CDR loses lock, the link goes blind and the bit-error rate explodes.
CDR is why line coding (like 8b/10b or scrambling) matters: the data must guarantee frequent transitions, because a long run of identical bits gives the CDR no edges to lock onto and it drifts off-beat.