SDI Is About to Give Way to Ethernet
Serial digital interface (SDI) needs to be put out of its misery, but somebody has to show up with a stick big enough to beat it to death. And, lo, along comes Mellanox, swaggering in with its 100 gigabit Ethernet routing chips.
Moving to Ethernet in production facilities will certainly contribute to making the entire video industry more efficient, but together with other innovations, it can be a requisite for a significant transformation in how broadcast video is delivered and consumed.
Video production and broadcast facilities are among the last holdouts still using the serial digital interface. Why? SDI-based video transport systems have been working fine. Changing workflows is hard, and harder still if it has be done on the fly. The affordable Ethernet alternatives weren't that much better than SDI. The most desirable (which is to say, fastest) Ethernet alternatives have been too expensive.
All of that has changed, Mellanox Technologies Ltd. (Nasdaq: MLNX) marketing vice president Kevin Deierling explains to Light Reading. Mellanox, which specializes in high performance Ethernet routers, adapters and cables/connectors, believes that upgrading video production from SDI to Ethernet now represents an excellent opportunity.
There are several trends that together constitute a death warrant for SDI in video production.
The transition to all-IP: Video production is a last bastion for SDI. Almost everything else in the video chain, including storage and distribution, has already gone all-IP. (There are notable exceptions: the last mile of cable networks, for example.) It is inevitable that SDI will be entirely replaced by Ethernet; the only question is when?
The advent of ultra HD: HD-SDI transport runs at about 1.5 Gbits/s. Production facilities edit uncompressed video. The transport rates of uncompressed video vary depending on several factors (including bit rate and frame rate). Uncompressed HD can run to a few hundred megabits per second; SDI has no problem there. But an uncompressed 4K UHD stream needs 7-to-12 Gbits/s. An 8K UHD stream is about 24 Gbits/s or so. As video production facilities start to handle UHD, SDI will be immediately inadequate, as will GigE. 10 GigE is barely adequate for one UHD stream, and production facilities typically handle several at a time.
Data centers are driving Ethernet costs down: 40 GigE was once too expensive for video production facilities, but costs have plummeted as a result of truly enormous demand from data centers, driving economies of scale. Data centers are now driving down prices for 100 GigE too. 100 GigE ports started out at $100,000 each; now one is $1,000, Deierling notes.
Mellanox claims a large percentage of the 40 GigE market, and it expects to keep its lead at 100 GigE. It believes it can dominate the SDI replacement market. Deierling says the company has yet to size the market, but it could run into the tens of millions of dollars.
here on Light Reading.
The transition from SDI to 25, 40, 50 or 100 GigE will not be just a matter of swapping out routers and connectors, however. Workflows will be entirely different, and new models, practices and frameworks will be necessary to make sure it all works correctly.
To that end, the European Broadcasting Union (EBU), Society of Motion Picture and Television Engineers (SMPTE) and the Video Services Forum (VSF) have joined to sponsor the Joint Task Force on Networked Media (JT-NM). The organization's charter is to help to drive development of a packet-based network infrastructure for the professional media industry by bringing together manufacturers, broadcasters and industry organizations (standards bodies and trade associations) with the objective to create, store, transfer and stream professional media. Mellanox is a member. The JT-NM published its first version of a reference architecture just a couple of months ago; find it here.
The British Broadcasting Corp. (BBC) is one of the broadcasters trying to get out in front of all of this. It has set up an IP Studio, which is experimenting with 100 GigE transport. "In order to send many uncompressed UHD video streams over a single 100Gbps interface, we must use specialized techniques to avoid inefficiency in traditional network software, take zero-copy approaches from the application layer right down to the network card, and take advantage of hardware features to control stream rates in order to avoid overflow in network equipment packet buffers," the company reported.
The Beeb is also working with the 2550100 Alliance (25, 50 and 100 -- get it?), which says it is "designed to pick-up where the standards development and promotion of the Ethernet Alliance and 25G Consortium leave-off," working on specific solutions at some of the new Ethernet technologies and rates that are still being developed and standardized.
Broadcasters are aiming not just at being able to handle UHD video, but shifting to all-IP as a prerequisite for a transformation of how broadcast TV is delivered and experienced. The basic idea is disassemble the elements of a video stream (video, audio, captioning, etc.) so that they can be variably reconstructed in each viewer's device.
For example, all viewers would get the weather at the same time. The difference is that where most viewers are likely to opt for the typical weather report, a hearing-impaired viewer might opt for video of the same report presented by someone signing. Someone speaking a different language might be able to select an alternate stream of a weatherperson speaking in that language. Anyone might choose to get a captioned crawl; perhaps that crawl might be in a different language.
Each element would be an object, defined by meta data and assembled to provide the combination requested.
Deierling also explained how object-oriented (OO) video production might help spur the transition to 8K video. Under the most common TV-viewing circumstances, most people will not notice an appreciable difference between 4K and 8K video. What some broadcasters are considering is the possibility of using those techniques to allow people to zoom in on, say, a specific athlete during a play, or a specific participant in performance -- all without any loss in resolution.
— Brian Santo, Senior Editor, Components, T&M, Light Reading