The new scheme is different, but quite simple: we divide whatever input clock we get until we reach the "base" rate of 44.1/48k. We distribute this around the interfaces as "loop sync." Each interface then taps off the loop sync and recreates an ultra-low jitter local clock. The local clock is then scaled internally according to the session sample rate. This scheme gives excellent jitter performance throughout the system, and means completely solid internal timing. This is really needed, too, since the higher sample rates are much less forgiving, timing-wise.