For all slides, click here.
Our industry is about to join the rest of the world and use network architectures to switch and distribute signals. It sounds like a simple task, but AV-over-IT presents us with decisions and problems we never had to consider before….
At long last, the message has gotten through. It’s only taken a decade and a half, but our AV industry is finally moving AV signal management to networks—specifically, to those that use Transport Control Protocol/Internet Protocol (TCP/IP), fast switches, firewalls, Virtual LANs (VLANs), subnets and a host of other features we’ve never previously had to consider.
Even so, judging by the enormous turnout for my AV-over-IT class at InfoComm 2017 (and the previous 10 years of classes), there are still plenty of consultants, dealers, rack builders and end users who are still sitting on the fence, or who are just dipping their toes in the water.
For those folks, I offer a word of caution: The tail wags the dog now—that is, the “tail” of consumer electronics is largely shaping our world. Smartphones, tablets, the Internet of Things, the cloud, YouTube, Wi-Fi, Facebook, Instagram, drones, robots…all of them originated in the IT-centric consumer world and, now, they’re firmly entrenched in ours. There’s no turning back now….
Perhaps nothing has been more affected by this trend than the management of video and audio signals. Flash back to the turn of the century, when the dominant connections were composite and component analog video, the 15-pin D-Sub VGA computer video connector, the ubiquitous RCA plug, nine-pin RS232 jacks and “mini” phone plugs (not to mention different configurations of BNC connectors).
Life was simpler then, although somewhat confusing with all those incompatible signal and cable formats. But the analog signals we transported through those connections were easy to move around—low-resolution formats with slow pixel clock and data rates that didn’t pose much of a bandwidth challenge.
As digital audio and video came to market and as high-definition video was born, those clock and data rates began a slow, but steady, uphill climb. New connectors were developed to handle denser and faster video and audio, multiplexing them into a single plug. Copy protection and self-configuration through metadata—curiosities and headaches at the start—became an integral part of every interfacing system.
Early versions of display interfaces were upgraded to handle more data, 3D and 4K/Ultra HD. As each version of copy protection was hacked, a new one would replace it. (And quickly be hacked, too!) Display screens keep getting bigger and bigger, requiring more pixels. Two incompatible display interface standards (HDMI and DisplayPort) jockey for market share, while simultaneously pushing their speeds upward, and color bit depth and frame rates are increasing steadily.
Today, we’ve created Frankenstein monsters of bulky, heavy and power-hungry 16×16, 32×32 and even 64×64 matrix switchers to keep up. It’s a mess! Category wire signal extenders are intermixed with bulky HDMI and DisplayPort connectors. We’re even using format converters to extend the transition-minimized differential signals from HDMI to digital formats, hoping to push them several hundred feet.
Reset The Game
Clearly, that trend is not the right path to pursue anymore—not when we can switch and distribute AV signals using software and hardware that are less bulky and expensive. While our industry was busy building monster HDMI products, the IT industry was quietly bringing out faster, compact switches; upgrading network capacity from 100Mbps through 1Gbps to 10Gbps, with 40Gb hardware now starting to ship.
In fact, you can buy a 1Gb “managed” network switch with 12 ports (10 copper ports and two optical) for less than $200 today. And that switch is fast enough to handle lightly compressed Full HD (1920×1080) and Wide UXGA (1920×1200) video, plus multichannel audio and control. Need more speed? A 10Gb managed switch with 12 to 16 copper/optical ports for transporting 4K/UHD video would cost you just $2,500 to $5,000.
OK, so we know we can build an AV network that’s fast enough to transport just about any signal we need to move around. Get rid of those bulky hardware switches. Toss out those TMDS-to-packet signal extenders. Standardize on Cat6 and Cat7 copper wire and optical fiber for interconnects. Toss in a few video/audio encoders and decoders, and we’re all set…right?
Not so fast. The move to AV-over-IT introduces a new concept…one we probably hadn’t thought about before. It’s known as quality of service (QoS), and it’s a combination of several factors, among them image and audio resolution, levels of compression, signal latency, network speeds, bit error rate (BER) and traffic density.
To complicate matters further, you might be considering a decision to place your AV gear on an existing network. And that can be a very sticky wicket, given that the IT administrator(s) of that network must focus on avoiding disruptions to their own QoS benchmarks. They might not be all that welcoming to a bunch of unfamiliar hardware popping up on their local area networks—hardware they have no real operational control over, and that might serve as an unwitting doorway for infiltrating viruses, malware, spyware and ransomware.
Although a detailed discussion of network architectures and protocols is beyond the scope of this article, we can get you started in the AV-over-IT game with a review of two important game pieces: video codecs and network switches.
Video Codecs: What You Need To Know
Because bandwidth is usually fixed and expensive to upgrade, we will use video compression to achieve better efficiencies through our AV-over-IT system. And we have several options to do that.
The first is to use a video codec with high compression ratios for maximum efficiency, such as Motion Picture Experts Group (MPEG) compression. MPEG is a so-called “lossless” compression system that takes advantage of redundancies from one frame of video to the next. It uses a process known as discrete cosine transform (DCT) to sample and reduce image pixels to mathematical bits, known as blocks.
MPEG-2 has been around for more than 20 years, and it’s widely used by broadcasters, cable TV systems and satellite broadcasters. However, a more efficient and practical choice would be MPEG-4 Advanced Video Coding (AVC) H.264, first introduced in the early 2000s and 50 percent more efficient than MPEG-2 in compressing a given signal.
An H.264 video encoder analyzes and converts strings of video frames in sequence, looking at both the start and the end of the sequence to see what has changed both in time (temporal motion) and in object position (spatial motion). Then, it makes “inter-frame” compression decisions, in effect copying and pasting areas of the video frame that haven’t changed to successive frames while, simultaneously, refreshing and updating areas that have changed.
Each sequence of encoded video frames is bookended by a pair of uncompressed intracoded (I) video frames, along with predictive (P) and bi-directional interpolated (B) frames, and it’s known as a Group of Pictures (GOP). The distance from one I-frame to the next—the GOP length—can range from 15 frames (a common length for broadcasts) to 90 frames (a common value for streaming video).
The process of analyzing continuous GOPs means the encoder must effectively look into the future—that is, analyze video before it is compressed and transmitted to you. Because time travel hasn’t been perfected yet, a delay is added in the coding and compressing process. Consequently, there is latency introduced between the live event and when you see it appear onscreen.
By sampling and compressing video with H.264, we can achieve tremendous efficiencies. Using inter-frame compression, we can reduce bandwidth by 20:1, 30:1, 50:1 and even 70:1 ratios. But, to ensure all the video packets get there (and in the right order) before reassembly, we must add forward error correction (FEC) bits—and that increases latency.
A new codec format, known as High-Efficiency Video Coding (HEVC) H.265, promises a bit-rate reduction of 50 percent over H.264. Unlike H.264, H.265 uses coding transform blocks (CTBs) and transform units (TUs) to sample, digitize and compress video in larger chunks.
It’s ideal for non-real-time transmission of 4K/Ultra HD using bit rates that would only be fast enough for Full HD with the older MPEG-2 standard. However, at present, H.265 encoder hardware is expensive, with basic encoders starting at $5,000. Unlike previous versions of MPEG, H.265, at present, is a software-intensive codec, which accounts for its higher cost. Some companies have introduced lower-cost H.265 decoder chips for set-top boxes and IPTV streaming as of this writing.
To get around the latency issue with MPEG-based codecs, we have another option: the Motion-Joint Photographic Experts Group (M-JPEG) codec. M-JPEG also uses the DCT process to encode video, but it cannot perform inter-frame analysis and compression, just “intra-frame” compression. Consequently, there are no P and B frames and no Group of Pictures in an M-JPEG stream—just a steady series of I-frames.
Although not nearly as efficient as MPEG, latency is extremely low with M-JPEG encoding, making it a better choice to connect live video feeds, such as image magnification of a concert. Note that, with both M-JPEG and MPEG formats, you can set the encoder to adjust dynamically the amount of compression required, based on “best image quality” and/or “network conditions” parameters.
There are numerous M-JPEG encoders coming to the AV market, and several manufacturers tout their low latency numbers as a practical way to replace HDMI-based switching. Both H.264/H.265 and M-JPEG encoders and decoders can operate quite happily on the same network, in case you need to mix and match.
Network/Switch Speed Requirements
Now that we know how to compress video streams down to manageable levels, we need to choose the appropriate switch to transport our AV signals. As a best practice, always select what’s known as a “managed” switch for your AV-over-IT installations. That will provide you with a full suite of controls and software settings you’ll need when setting up an AV/IT network. You will also want to select an OSI Layer 3 switch, although there are now Layer 2+ switches available that will work nicely.
For streaming H.264 and H.265 video, a 1Gb switch (1GigE) is usually adequate. The resulting bit rates using either codec will be in the range of 20Mbps to 200Mbps with Full HD and 4K video sources. Full HD video compressed with M-JPEG will also work through a 1GigE switch. If you want to use lighter compression (less than 6:1) and plan on transporting 4K or higher-resolution signals, then you’ll need a 10Gb (10GigE) switch.
Here’s how you make the call: You want to stream a 3840x2160p/30 signal with 4:4:4 (RGB) video and eight-bit color. The clock rate for this signal is 297MHz (4400×2250 pixels with blanking (x) 30Hz = 297MHz). Multiply that rate by three (for RGB color), and then again by 10 (eight bits plus two bits ANSI overhead), and the result is 297 (x) 3 (x) 10 = 8.91Gbps, which is the uncompressed data rate through an HDMI or DisplayPort connection.
A data rate of 8.91Gbps would pass through a 10GigE switch without any compression, but it would gobble up most of the available bandwidth. If we apply 4:1 low-latency M-JPEG compression, we’d still need a 10GigE switch to pass the resulting 2.2Gbps signal. Even 6:1 compression wouldn’t drop the data rate below 1Gbps—we’d need a compression ratio of about 10:1 to get under the bar. (Double the refresh rate to 60Hz, and we’d need 20:1 compression to get 4K video through a 1GigE switch.)
If we use the maximum recommended compression for M-JPEG (20:1), we’re now in the ballpark of 445Mbps for our 4K video stream. But that will introduce image artifacts with certain types of content and, if you’re connecting to a large 4K display, that’s a big “no-no.” That’s why 4K and Ultra HD content transmitted through IT networks requires at least a 10GigE switch.
Of course, if latency isn’t an issue, we could compress our 4K video stream even more by using the H.265 codec, winding up with a bit rate well below 50Mbps, which would easily pass through a 1GigE switch. For non-real-time content delivery, that is a practical approach.
Which codec to use? How much compression should I apply? Those questions are part of your QoS decision-making. If you are connecting multiple large HD or Ultra HD display screens in real time, then light compression and high bit rates will be the order of the day. But, if you are feeding multiple HDTVs in a sports bar—where everyone sits some distance from each screen, and they aren’t paying a lot of attention to each screen anyway—then you might be able to get away with 1280x720p resolution and lots of compression, because latency isn’t an issue.
For AV-IT applications, your managed switch must support multicast modes. In a multicast, a server creates copies of the stream and sends them to each connected user. That means all the ports in your managed switch should be set to multicast mode to pass the streams from your encoder (or encoders) to decoders. Make sure Multicast Forwarding and Forward Unknown Multicast are both enabled for all switch ports you use.
H.264 and H.265 video encoders will use an internet protocol known as Real-Time Streaming Protocol (RTSP) to preserve the packet order through the network. That protocol will often be paired with User Datagram Protocol (UDP), commonly employed for streaming media. And, of course, the AV packets will have the usual IP and TCP headers to get through the network—just as with all other internet traffic.
A separate protocol, Integrated Group Media Protocol (IGMP), must also be supported in your managed switch, along with IGMP Snooping, IGMP Querying and IGMP Snooping Fast Leave for each switch port you want to connect. IGMP Snooping lets your switch “listen in” on all IGMP multicast traffic (that would be our AV stuff) traveling between encoders and decoders. A switch equipped with IGMP Snooping can then pass multicast traffic only to those switch ports that require it, preserving bandwidth on the network.
The IEEE standard for one frame of Ethernet data packets is 1,500B, which is adequate for H.264 compression. But, if we want to use lighter video compression (M-JPEG, JPEG2000) with higher data rates, we’ll need to set our switch port settings to pass Jumbo Frames, which can be as large as 9,000B—six times as large as standard Ethernet frames. (9,216 is a typical maximum value for switch settings.)
Another extremely useful feature in a managed switch is Dynamic Host Control Protocol (DHCP) addressing. Any device connected to a network must have an internet address, and it can be assigned manually (static addressing) or automatically (DHCP). With DHCP turned on for each switch port, IP addressing for all connected equipment will be automatic. (You will have to log into the control menu for each encoder and decoder you use to set their IP addressing to DHCP, too.)
If you are IT-savvy, then you can set all IP addresses manually. But, if you connect to another network with a DHCP server, make sure the block of addresses you use doesn’t conflict with those that the other network’s server assigns.
For 24/7 video connections, you might also have to disable any “green,” energy-saving switch settings that would close ports due to inactivity. Most new switches come with that feature. It’s fine for the occasional bursts of data in TCP/IP traffic, but it’s impractical for continuous streaming of video. Finally, look for Power over Ethernet (PoE) functionality. That lets your switch provide the “juice” to operate your encoders and decoders, and it eliminates a bunch of wall transformers.
As you can see, setting up an AV/IT network is a bit different from installing an HDMI matrix switch. But you don’t need as much hardware, and your cabling requirements are simplified considerably. Your signals can travel longer distances (especially with optical fiber), and network connections can be expanded with less hassle. You just need more IP addresses and switch connections.
Wow…that was a quick summary of AV-over-IT! In future articles and columns, we’ll drill down a bit deeper on network configuration and operation, so stay tuned….