SMPTE 2014: Envisioning the IP Media Facility

HOLLYWOOD— Why bother retrofitting a perfectly functional serial digital interface facility with one predicated on Internet protocol networking?

“Ethernet is bidirectional, which SDI is not,” said Al Kovalick, consultant and industry cloud evangelist. Ethernet also carries bigger payloads farther.

The nature of the two architectures requires new lossless network strategies, he said. SDI is circuit-switched while IP is packet-switched

“In world of SDI, how do you achieve lossless? Through circuit-switched connectivity,” he said.

With IP, Kovalick predicted two classes of push user-data protocol service. UDP is a core component of the IP suite. There will be an economy class at a fixed rate and a “best effort, like TCP,” or transmission control protocol, also a core component of the IP suite.

With TCP, data essence would be pulled as needed, he said. Receiver buffering will smooth out irregular delivery, but there will be latency issues.

Interoperability and bridging are key to transitioning from SDI to IP.

“You have to have interoperability. Hybrid facilities will be the norm. Legacy and IP will live together,” he said.

Within the IP-based facility, pushed transport loads comprise separate flows per essence type, so no SDI wrapping. Everything is an IP packet… the SMPTE rasters, audio, metadata, forward error correction, control, etc.

“What does it take to move something over Ethernet? You have to identify what you’re sending, and have a time stamp. If you have a time stamp, you can make up for a lot of sins in the network. You can do a lot of things with just ID and time stamps,” he said.

All devices and flows use the same system clock and timestamps to align.

IP’s bidirectionality enables push and pull transport. With an SDI source, the pushed essence is streamed over lossless connectivity. With an IP source, there can be pushed UDP essence, for example, and pulled essence, TCP or something else.

“With regard to pull, think of the ‘client-server’ and HTTP over lossy networks,” he said. The configuration is of value for unmanaged QoS transport, e.g., public Internet and clouds, he added.

“TCP comes in when you have a network that you don’t control,” he said. “You will see TCP pulled-frame-at-a-time system, and it will probably be unmanaged networks.”

Kovalick said IP enables bundling, both native and virtual, something SDI can’t do.

He gave an example of native a bundle of audio and video, each coming out of a couple of cameras The signals can be bundled and moved together or split up.

“Say you want the audio out of a particular bundle,” he said. “With SDI, it has to be de-embedded. With IP, you switch it out.”

For frame-accurate splicing, Kovalick said there were three general methods. Source-timed, where control comes in at the source site and changes the UDP port number.

“When I see port number change from 500 to 501, I know the path has to be changed,” he said. “It’s good for multicast groups. The second thing is switched-timed. If a packet could be switched on the VDI line in SDI, we’d be done. It is possible. Fox has worked on this with an [field-programmable gate array]-assisted switch.”

Kovalick concluded that the “future is leveraging IP and the Internet… for infinite workflows. That’s not an exaggeration.”