Today we stand at another tipping point in the professional audio/video industry. The transition from analog-based AV distribution to digital AV distribution was a little bumpy, but is now complete. I had the privilege of overseeing and helping drive that transition from my previous position as Technology Manager of DigitalMedia for Crestron. The analog sunset helped create a new juggernaut in video distribution, but the transition we face today is much more significant, and may shake up the existing order again.
The transition from traditional point-to-point style matrix switches (or “circuit switches”) to Ethernet switches (“packet switches,” more generally) is inevitable. In fact, commercial AV is a bit of a holdout. The effects will be massive, and the technology is upon us now. But first, a little background.
Picture the image of a 1950s telephone operator moving and connecting patch cords on a telephone board. When you pick up your handset to call your friend, you first speak to the operator to ask to be connected to your party. The operator does so by connecting a patch cord between the terminals from your phone and the terminals to your friend’s phone. The operator is building a directly connected circuit between the two of you. This is what a circuit switch is. Eventually, of course, machinery was developed to replace the human operators, but the principle remained in place for decades. Circuit switches are (relatively) easy to build and easy to control and, therefore, every new switching technology starts out on a circuit switch. So, when television signals came onto the scene, and early broadcasters and producers needed to switch and distribute AV signals, analog circuit switches were a natural choice; they became ubiquitous.
Today’s HDMI matrix switches (or switchers, as they’re also commonly called) work the same way: Just replace the patch cords and analog switching machines with high-speed digital transistors. The core of these matrix switches consists of expensive, power-hungry chips called “crosspoints” or “crossbar switches,” which blindly connect inputs to outputs, as instructed by software.
Circuit switches have some advantages. Their architectural simplicity means that they can dedicate their full bandwidth to moving data from point A to point B without overhead. They also do not introduce any timing uncertainty of their own. (Although most HDMI matrix switch systems do not offer synchronized outputs, this is for reasons other than the matrix switch itself.)
However, circuit switches also have significant disadvantages. Because of the need to dedicate pins and wires to every single I/O port individually, the crossbar chips are huge and very expensive, running into the hundreds of dollars for a single chip. Further, although the chips are architecturally simple, the circuit boards required to support them are complex and expensive to design, especially as bandwidth and I/O counts increase. The switches created are inflexible and do not scale well. An input port is always an input port, period, and an output is always an output. If you connect a single cable between two matrix switches, you can run only one data stream in one direction between them. If you want two signals to get to the same destination (picture-in-picture, anyone?), you will have to run two wires.
Moving different data types independently (for example, routing audio separately from video) requires dedicated hardware for each data type, all of which adds not only cost, but size and power consumption. Finally, the infrastructure costs to support these systems are huge, because each video path in a system must be represented by an individual cable, and this infrastructure is single purpose, dedicated strictly to AV data.
The alternative to circuit switching is called packet switching. Packet switching works by chopping up the (digital) data stream into—you guessed it—packets, and adding a bit of information to each packet is called a header, which describes the intended destination for the packet. The switch at the heart of a packet switching network is necessarily more complex than a circuit switch. It maintains a connection to all of its ins and outs simultaneously (unlike the circuit switch, which connects and disconnects ports all the time), and inspects the header of each incoming packet. Based on the destination information contained there, the switch decides which output port (or ports) to use to send the packet onward. Of course, the most popular technology for packet switching is called Ethernet, and it is far and away the most successful and widely used communication standard in the world.
Ethernet was invented more than 40 years ago. Let that sink in. Richard Nixon was President of the United States. There were no cellphones or personal computers. The cassette tape and the VCR hadn’t been invented yet. You may have used a few other technologies with a 40-plus-year lifespan today, such as fire…and the wheel. How many others can you name? I bet none of them are electronic communication systems.
Packet switching took over the phone network in the latter decades of the 20th century. The internet took over all our lives in the first decade of the 21st. Today, we see Ethernet quickly reshaping the AV industry outside of the commercial AV world. Think about it: In the consumer space, nearly all video content is delivered via Ethernet. Not only are the internet-based streaming services like Netflix and YouTube reliant on it, but your cable and satellite TV use IP (the Internet Protocol that rides on top of Ethernet) to deliver your traditional programming to your set-top box.
In the broadcast industry, the replacement of SDI matrix switches with high-bandwidth Ethernet systems has been taking place for a few years. (Side note: Why don’t we just co-opt their technology? Because broadcast manufacturers don’t contend with HDCP, funny PC resolutions or the cost concerns of commercial AV customers.)
The advantages of packet switching to the AV market are incredible. Because Ethernet is inherently bi-directional, each switch port can serve as input or output, or both. High-bandwidth trunk ports on switches mean that multiple switches can be cascaded efficiently, to allow systems to easily scale to greater and greater sizes, without having to replace huge chunks of switching infrastructure.
And since we are talking about Ethernet, the AV industry gets to ride the coattails of that monster-sized industry, which means a faster pace of technology advancement and ever-falling prices. Plus, every enterprise customer already has a managed Ethernet infrastructure; thus, instead of having to purchase a completely new infrastructure with a unique set of management headaches and costs, they can expand what they already have, and we can show them how to apply it to their AV needs.
This sounds great, and consumer and broadcast AV have already started reaping the benefits. So what is taking commercial AV so long?
Three issues have held back commercial AV from being able to transition to Ethernet, but each of these issues has been resolved in just the past couple of years.
The obvious limitation is bandwidth. In all prior generations of video and Ethernet, the data network simply didn’t have the horsepower to handle the demands of commercial AV video…at least not without high compression and high latency. A decade ago, one- gigabit networking was common, but 720p or 1080i video demanded at least 1.4Gbps. Today, 4K video requires less than the 10Gbps available from mass- market Ethernet gear, and the trend is here to stay. Ethernet products of 25G, 40G and 100G are already available, and Ethernet will continue to outpace the introduction of new video standards.
Another problem until recently has been cost. Since 2007, 10-gigabit Ethernet has been available to serve the high-end datacenter market, but at a cost of up to $1200 per fiberoptic port. Since that time, the price of 10G switching and optics has fallen, and 10GBaseT over copper has been developed. Today, one can buy 10-gigabit switches from Amazon for around $100 per port, whether copper or fiber. Compare that to what the “big boys” in commercial AV charge for HDBaseT switches: An 8×8 retails for more than $1000 per port! And what if you need nine inputs? Those 10G switches on Amazon come in 8-, 12-, 16-, 24- and 52-port varieties. Ethernet switches with more than 1000 ports are available from Cisco, Arista and others. Which ports are inputs and which are outputs? That’s up to you.
The final challenge for moving uncompressed AV signals over Ethernet is the complex technology required to repurpose Ethernet technology for AV applications. Uncompressed AV signals are inherently synchronous with extremely large amounts of data having to arrive in a particular order and “just in time.” Watching frames (or pixels) out of order does not work. Ethernet networking is asynchronous, and while high bandwidth, it is designed for mostly disconnected pieces of smaller data sets that are not impacted by microseconds or even milliseconds of latency. Nothing in the standard Ethernet/IP/TCP delivery stack guarantees on-time arrival of data. This limitation had to be overcome before high-quality, zero-frame-latency video could be delivered via Ethernet.
Today, technologies exist that can overcome this challenge. Time Sensitive Networking (TSN, formerly AVB) provides “hooks” in the Ethernet architecture to address this challenge by requiring each packet in the network to carry detailed timing information with it, although few commercial AV products have been designed to carry video using TSN (it is more popular for audio, which faces the same challenge). TSN imposes a requirement that the Ethernet switches themselves must offer support for AVB’s timing mechanism, which limits widespread adoption.
A solution to enable the synchronization of video signals across asynchronous Ethernet networks, without the need for any specialty capability from the Ethernet network itself, is currently being used by many OEMs to deliver an end-to-end system for the switching of uncompressed 4K AV signals over 10G Ethernet.
With real products available and now shipping from several companies, the direction this is all headed is clear. In five years, we will not use circuit switches in commercial AV. They will have been completely displaced by the use of Ethernet infrastructure to deliver all content, whether the underlying technology is zero compression and zero latency, or highly compressed to fit into mobile links and handheld devices (more on that distinction another time).
A new focus on Ethernet-based switching means some great things for our industry. Smaller manufacturers, unable to sink tens of millions of dollars into creating a full lineup of matrix switches, can instead leverage billions of dollars of industry investment in Ethernet hardware. With focus on bringing unique and thoughtful capabilities to a handful of endpoint devices (the boxes that convert HDMI, etc., into Ethernet), these companies can compete on equal footing with the matrix dinosaurs. This means a whole new level of creativity and innovation, since manufacturers can focus on ease of use and good software rather than the complexities of building switch gear (which Cisco will always be better at anyway).
For integrators, consultants and designers, this is a great opportunity to study up on this new technology and gain a market advantage. Those who embraced digital distribution and all its hardships in 2009 and 2010 are now established, successful, leaders in the field. The same will be true of those who learn to design and integrate Ethernet networks to support high-bandwidth video distribution.
We in the AV industry have an opportunity to stop creating and pushing new standards for distribution and switching of digital data. Let’s take advantage of the biggest and most widely deployed communication and data distribution standard in the world: Ethernet. The IT industry is a trillion-dollar global industry. The more we can leverage its R&D and ideas, the faster our own industry’s state of the art can advance.
Why reinvent the wheel?