[BC] DTV Audio Levels

[Robert Orban]

At 06:49 AM 6/16/2009, RichardBJohnson at comcast.net wrote:
>don't think it would take more than a few milliseconds to process 
>the audio. There shouldn't be an resynchronization problems because 
>synchronized audio was delivered with the video. Synchronization 
>problems come into play when one has multiple paths for both the 
>video and the audio. The TV station already has synchronized audio 
>available from the network. Extracting digital audio from the 
>composite signal is a trivial operation, with the maximum latency 
>determined by the dictionary length in the compression. This is all 
>"realtime" as far as audio is concerned, consuming a few tens of 
>audio frequency cycles, maximum. Actual signal processing is 
>instantaneous when related to slow varying information like audio.

Our processors for digital channels, including digital TV, typically 
delay the audio by 25 ms and can be padded to one frame. The 25 ms of 
delay results from a cascade of the phase-linear crossover in the 
multiband compressor and low-IM look-ahead limiting. We believe that 
this system's extremely clean handling of speech justifies the delay, 
speech being particularly important in sound for picture applications.

To compensate for the one frame of additional delay, we recommend 
that the audio makeup delay elsewhere in the system be reduced by one 
frame. Because the MP2 video codec in DTV has far more coding delay 
than the AC3 audio codec, reducing the audio make-up delay by one 
frame is almost always practical.

The upcoming HD-SDI interface card for our 8585 surround processor 
includes video delay to ensure that the audio and video remain in 
sync when passed through the unit.

Bob Orban