More on syncing MLB GameDay audio to DirecTV video
For today’s Giants-Phillies NLCS game 6, I’m using the same setup as described in my earlier post. It’s working moderately well and is of course miles better than the torture that is listening to Joe Buck and Tim McCarver.
The strangest thing is how the delay between GameDay audio and DirecTV seems to vary minute by minute. It’ll be steady at around nine seconds, then drop to seven and go up to twelve, all without any pauses in the audio stream. This means I have to keep my VNC window (all the audio processing software is on my old Dell Windows laptop) open and adjust the buffer in Radio Delay every so often. I can see several reasons for the delay between radio and TV and therefore between radio or GameDay and DirecTV, but I can’t quite grasp how the delay can vary in realtime without the stream dropping or pausing.
It seems to me that the overall delay is the result of the differing paths the audio and video take to reach me. If I was listening to KNBR on a standard AM radio, I’d be hearing the play-by-play in near-realtime; the signal goes from the ballpark, through KNBR’s onsite processing/uplink, to KNBR’s studio, to Sutro Tower and then to my radio. Most of that happens via broadcast, i.e. speed-of-light, and is point-to-point with no satellite bounce. Add another couple of steps for retransmission to MLB’s studio so it can go out over GameDay audio, and of course the streaming audio is affected by Internet speeds.
The video path is very different and a lot longer. It goes from the ballpark, through Fox Sports’ onsite processing/uplink (“the truck”) to Fox Sports’ central studio or processing center–I don’t know where that is, or if they use KTVU’s facilities–to KTVU’s studio in Oakland, to DirecTV’s facility in Denver and then via direct-broadcast satellite to me. There are at least two satellite round-trips there, possibly three depending on how the signal is sent from the Fox Sports truck to their central facility. The signal is encoded, decoded and re-encoded an unknown number of times during the whole process, and that takes time as well.
Most of the above is speculation based on some knowledge of how these things tend to work and a bit of research. I’d love to know any better and/or corrected details.

