In live production, humans are required to make real-time editorial decisions based on multiple, simultaneous streams of audio and video. If latencies between these streams become noticeable, the production can be affected. As such live production systems must be designed so as to keep latencies within tolerable levels. As broadcasters move to IP-based systems for live production, there are new potential sources of latency which must be understood and designed for.
In this paper we identify specific latency requirements for live production, model the sources of latency in an IP-based system, and outline a systems approach for achieving latencies that are as good, or better, than the typical latencies found in SDI-based systems.
Requirements are Driven by Human Factors
In many broadcast applications (delivery of program streams, for example) delays on the order of multiple seconds are routinely experienced and go completely unnoticed by the party receiving those streams. However, in live production, the latency requirements are much more stringent. This is because humans are observing and interacting with multiple video and audio signals from the same event in real time. Delays between signals, if large enough, can be noticeable and distract production crews. This in turn can lead to errors during a live broadcast. In order to reduce the potential for on-air errors, live production systems must be designed such that latencies are below human perceptual tolerances.
As human perceptual thresholds vary depending on the type of latency, it is important to consider each type of latency separately. For example, we illustrate here three types of latency which need to be managed in a live production system.
- Tactile to Visual Latency
- Audio to Visual Latency
- Visual to Visual Latency
As a goal then, in order that a production crew working a live event perceives no delay and an "instant response" to button pushes, latencies need to be kept within specific values. These maximum latency targets are independent of video format and of the transport technology used in the system. Rather, they are based entirely on the ability of the human brain to detect differences in the arrival times of various stimuli.
The fundamental latency requirements for live production are invariant with changing network transmission speeds and video frame rates. They are based on human factors. By carefully designing for each variable in the model, practical live production systems can be built using IP technology.