Ethernet in broadcast

I remember getting a job at a parts distributor back in 1988. The new owner decided that the only way to keep track of inventory was to put in one of these newfangled computers. When this was announced, half of the sales force resigned or retired. They had been used to doing inventory by hand for more than 50 years, and the fear of this “new technology” was palpable.

The fear of the unknown is one of the main brakes on implementing any new technology. So when someone says, “We're moving to Ethernet,” expect a lot of people to retire. Plus, Ethernet is taking over the known world, so changing jobs and moving to a different industry probably won't help. Obviously, the data world has embraced Ethernet; the factory floor is Ethernet; theater lighting is going to Ethernet; and homes are moving in the same direction. The best bet, then, is to double down on understanding the technology.


Why is Ethernet use so ubiquitous today? Ethernet was invented by Bob Metcalf and David Boggs in 1973 when they worked for Xerox. It turns out Xerox didn't want to be the next big computer company and gave Ethernet lock, stock and barrel to the IEEE. Thus, Ethernet became a “free” standard. Because free is a good deal, the Ethernet solution became hard to resist.

But you're reading a broadcast magazine. Are broadcaster's going to use Ethernet? Well, they've already moved in that direction, in case you haven't noticed. Certainly, office management, traffic, billing and similar applications rely on Ethernet, and you almost can't find a machine that doesn't have an RJ-45 Ethernet jack on the back somewhere. That jack, at first, was for machine control and diagnostics. You could even network many of these machines together and have a control network. Put that on the Internet with a password or two, and a manufacturer could look at its machine, update software and fix bugs. But these were control functions; none of this was on-air.

Now, we're talking about content, including audio and video. A video server is first and foremost a server. That data going in and out is data. The fact that it is video might require a certain data rate and bandwidth, but it's just a different kind of data.

Video on a twisted pair

Does that mean you can just put video on your existing Ethernet network? That depends on what kind of video you are talking about: compressed or uncompressed? Some boxes out there will take your HD video and turn it into a 150kb/s stream that can be carried on the network. While the picture at the other end might look pretty good, it would not be broadcast-quality. What's needed for broadcast quality is an uncompressed network, and today's broadcasters need a network that can handle not only HD (with a data rate 1.5Gb/s), but also 1080p/60, also called 3G or 3G-SDI (3Gb/s ).

Another good reason not to combine your office e-mail with on-air content is reliability. You don't want someone downloading a huge file to slow or stop your video. Fortunately, there's no limit to the number of networks you can install. All mission-critical applications use dedicated networks, and they can be configured with redundant rings, so network failures are extremely rare.

Of course, at the mere mention of a separate dedicated network, your IT person will be at your office door protesting. Data networks are their territory, not yours. But, you argue, you are talking about content, i.e., on-air content. This data isn't just some spreadsheet or Word document. This is an argument that continued in many facilities until the two jobs became one.

What kind of Ethernet network can carry 3Gb/s? Certainly not the 10Mb/s of yesterday or a 100Mb/s network, made popular by Cat 5 data cable. Not even a 1Gb/s network could handle HD or 3G. Sure, these networks could carry compressed versions, and those compressed versions are easy to send anywhere or to record and play back. You can already get servers for the newsroom that ingest, edit and output compressed formats on 1Gb/s networks. But uncompressed HD and 3G require something more, the next level of data networking: 10Gb/s networking.

Hollywood clearly saw this coming, so SMPTE set up a committee, DC-28, to research this. One of its conclusions was the need for a 10Gb/s network. The technology initially required a fiber infrastructure, but it's now possible to handle 10Gb/s on twisted-pair copper. This is called 10GBASE-T, which simply means 10Gb/s networking on twisted pairs.

The copper version requires four data pairs, and each pair carries 500MHz. All four pairs send and receive data at the same time in duplex mode, so there's some pretty amazing number crunching going on to get those 10Gb/s on only 2GHz (total) of the four pairs. This is accomplished by complex coding called pulse-amplitude modulation (PAM)-16, which means that there are 16 levels from zero to one. This provides a data rate of 10Gb/s that can theoretically carry six HD bit streams, or three 3Gb/s bit streams. The only thing missing is the box to convert from uncompressed HD or 3G to Ethernet.

This need for speed means that your video network is going to be a dedicated, broadcast-only network, and a redundant ring or two might be a good idea to increase reliability. The real problem is the installation of the network.

Physical connections

Back in the day of 10Mb/s networks (10BASE-T), almost anything would work as long as you got the right wire order. By the time the industry reached 100Mb/s (100BASE-T), the cable and connectors became more critical. At 1000BASE-T (1GBASE-T), the physical connectivity demands became much greater.

At a gigabit, Cat 5e cable was measured differently because all four pairs were working, not just two pairs as before. Cat 5e cable was the same as Cat 5 cable (100MHz per pair) but measured differently for this higher data rate. Cat 6 extended the bandwidth to 250MHz per pair to reduce the complexity of the boxes. We're now at 500MHz per pair with 10GBASE-T.

But as bandwidth goes up, so do the demands placed on the physical path. For instance, connectors, which were a nonissue at 10BASE-T, became important at 100BASE-T and critical at 1000BASE-T. You could easily put a male plug on a 10BASE-T network (Cat 3), but you had to be a bit more careful with 100BASE-T (Cat 5). By the time you get to 1000BASE-T (Cat 6), some odd things often happened with connectors. Many of those simple connectors would no longer function correctly. Male RJ-45 plugs had up to a 70 percent failure rate when used for Cat 6 GigE.

Why is that? There are a number of reasons, including simple things like the untwisting of the twisted pairs. At high frequencies, the impedance of the pair is a critical value. The cable has a native impedance of 100Ω.When the pairs are untwisted and separated, the impedance changes: It's no longer 100Ω. Signals going down those pairs see an impedance mismatch, something other than 100Ω. And as RF engineers understand, mismatched impedances produce reflections. The result is that signals reflect back to the source and even radiate from the cable at that point, and crosstalk between pairs in the connector increases, perhaps causing the network to fail.

This is why, in the networking world, installers can put punching wires on the back of a jack to make it work, but not on a male plug. Instead, factory-made patch cords must be used. But broadcast engineers don't install jacks and patch cords to route audio and video. If your audio and video were on Ethernet, a broadcaster might simply wire point-to-point.

It's no wonder that the worst data installations I have ever seen were in broadcast stations: These engineers aren't data installers; they're broadcast installers. These people simply did not understand the issues behind Ethernet and proper installation techniques.

Once a network's speed exceeds 10GBASE-T, it's time to think about a different solution. One could buy a handheld tester to test everything, but that may cost almost $10,000. One possible solution is to let a manufacturer build your network before it is installed, called preterm. The engineer provides the manufacturer a CAD and tells the vendor how many and low long the cables should be.

Cables can then be built in the factory and often packaged in groups of six, bundled in mesh with a pulling eye on each end. This solution uses a patch panel with feed-through, female-to-female jacks. All you have to do is pull in the cables and plug them in. The entire network is then prebuilt and pretested, with verified test results. The labor savings for installation and testing can be dramatic. When employing this tactic, don't let a vendor tell you that a 10Gb/s network “has to be shielded.” That just adds the potential for ground loops. There are plenty of unshielded, ungrounded installations that work just fine.

Don't fear this plug-and-play network. It doesn't mean you're going to be out of a job for lack of things to do. The cable in a network installation accounts for about 3 percent of the total cost. You'll still have the other 97 percent to keep you busy and gainfully employed.

Steve Lampen is multimedia technology manager and product line manager for entertainment products at Belden.