Genlock: keeping the signal usable when the network is not.

Two sources have to stay in step. A feed has to survive a network it was not built for. When that network degrades, something has to decide whether the picture degrades with it or drops out.

Broadcast solved this with dedicated hardware and a single reference that everything else followed. A software pipeline solves it with SRT, RIST, NDI, and a commodity internet connection, which is a different problem with the same exterior. Genlock reproduces that problem in software, with Linux network namespaces standing in for the network, and measures it.

• Where does each contribution protocol stop delivering a usable feed, and what does recovery cost in latency?
• What does sync drift look like without a shared clock, and at what point does it matter for a given production?
• When a feed loses its reference, does the system hold its last good state or start guessing? Guessing is worse, and it is more common than expected.
• AV control systems from Crestron, Extron, Biamp, and Q-SYS each expose a different interface. How much of a shared, read-only view of device state can be built across them?
• Does a model reading raw PTP jitter and offset series directly flag anything the threshold-based fault correlator misses?

SMPTE ST 2110, PTP, and NMOS appear constantly in this research, and not because anyone here is about to deploy broadcast infrastructure in a garage. They are where the vocabulary came from. Someone already spent years defining what "in sync" means to the nanosecond, and that work is worth borrowing.

The hardware stays in the facility. The timing model runs in software.

Genlock gets the signal there and keeps the timed cues on the right frame. MuxNode covers what it costs to encode and move it, and what the file still carries afterwards.