Signal management: Essential choices

signal man

Hurrairah bin Sohail examines the shift of video signal transmission to IP networks and discovers the trade-offs between latency, compression and bandwidth that different solutions offer.

Justin Kennington, president of SDVoE Alliance, says: “Codec performance can be thought of like a triangle, with quality, latency and bandwidth at the three points but you can only pick two. So, Netflix provides good quality with good bandwidth consumption but it can afford high latency. HDBaseT has good latency and good quality but has infinite bandwidth requirements. “The point with HDBaseT, as is true of any point to point signal transmission be it HDMI, DVI, SDI or anything else, is that you have to dedicate one whole cable to your signal. So actual bandwidth or datarate doesn’t mean too much. You can’t do anything with that cable but one signal. Whereas with any kind of IP solution, you get to mix and match the kind of data, subject to the datarate or bandwidth. You can fit 50 20Mb signals on a 1Gb link; or you can fit two 500Mb signals on that link.”

Delving into the characteristics of codec performance, requirements around latency are based on human physiology and what the human eye can perceive. The other two characteristics can be thought of as variable. Kennington says: “The performance requirements around quality are going higher and higher – 8K, HDR, 120Hz etc. But the requirement around bandwidth consumption is that it generally be kept low and this is relaxing and will always continue relaxing. As bandwidth gets cheaper, the ‘keep it low’ requirement will change.”

Blink and you’ll miss it

The topic du jour when it comes to video signal transmission is latency. With a lot of marketing terms being bandied Kennington offers some clarity: “Latency is a universal constant. Even when you are interacting face to face in the same room, the speed of sound and the speed of light means that there is latency. This is also true for signals. It takes time for the electrons to travel from one point to another. In pure scientific terms, there is no such thing as ‘zero latency’. That being said, there are ways to ensure that latency is kept as low as possible.”

Joel Mulpeter, product manager at Crestron, says: “Our marketing has always been ‘zero perceptible latency’. Typically this latency number is between 20ms to 32ms. In the past, 50ms used to be the benchmark and perceptible latency can change based on factors such as age.”

Creston’s DigitalMedia NVX product is marketed as having no latency penalty, which translates to actual end-to-end latency of 30ms, as Rob Carter, technology manager for DigitalMedia at Crestron, explains: “When we talk about no latency we are referring to our NVX patent-pending technology that combines codec and scaling latencies. As a result, NVX has the same latency performance as our HDBaseT scaling products which we have been selling for years before anyone was talking about latency.”

All latency is not equal as Kennington adds: “There are really big classes of ‘meaningful latency’ in AV applications. I’d classify them as high latency, more than about 150ms – 10 frames of video; medium latency between the two to ten frame class; and low latency in the sub-millisecond range. Nothing much exists in the range between 1ms and 30ms.”

Depending on the application, latency requirements can change. Bob Michaels, CEO of ZeeVee, says: “When discussing latency it is extremely important to consider the application. A video signal that is half a second behind might not be the end of the world, but what if that delay was present at a broadcaster’s desk reporting from inside a sports stadium? What level of customer experience does that provide? What if the delay is present in a mission critical control room?”

Kennington from SDVoE says: “The range of tolerable latency depends on the application. A surgeon using an endoscope during surgery can’t have latency and he also needs high image quality. For such an application latency of 10ms would be the upper limit.”

Finding balance

Video signal transmission changed with the introduction of 4K resolution content and the convergence of AV and IT. 4K60 4:4:4 content represents approximately 12Gb of raw, uncompressed data. Traditional AV signal transmission infrastructure does not have the bandwidth to handle this amount of data and neither does IP infrastructure. Currently, the only way to transmit 4K60 4:4:4 content is with compression. With compression in play the topics of latency and quality gain increased importance.

Chris Fitzsimmons, product manager at Biamp, says: “The assumption is that compression is a dirty word. But with demonstrations, this perception has been overturned.”

Compression can be heavy or light. Both take into account the bandwidth available, with IP networks increasingly being chosen as the infrastructure to transmit video. Heavy compression makes it possible to transmit 4K video over 1Gb networks while light compression transmits 4K over 10Gb networks.

Samuel Recine, director of sales, Americas and Asia Pacific at Matrox, believes that image quality of very low bitrate streams is good enough for most corporate applications: “Crestron and other brands manufacture exceptionally balanced products using JPEG 2000 compression. Other companies, like Matrox have achieved 4Kp60 4:4:4 quality and latency levels using H.264 that are incredibly good at bitrates even well below the 1 GB/s requirements of JPEG 2000. The majority of AV-over-IP applications can tolerate a small amount of latency in exchange for massive networks, a high degree of interoperability, ease of network set-up and maintenance and scalability. This is the crux of the AV-over-IP market. No codec competes with H.264 for the balance of quality, latency, bitrate and interoperability with millions of other devices.”

Carter from Crestron says: “Choosing the right compression technology is critical, since it has a direct impact on the end user experience. JPEG 2000 can offer amazing quality with latency low enough to fit within our scaler latencies, which is why we selected it. Unfortunately not all JPEG 2000 implementations are created equal, so there are other network video products on the market that use it but still have quality and latency problems. VC2 is another codec out there but it sacrifices quality for latency.”

Kennington from SDVoE Alliance says: “High compression of the type required to put 4K video onto a 1Gb network will always demand compromises in both latency and image quality. These technologies need to use a whole frame of video to make a reasonably good image quality in the low bandwidth environment. Just the act of waiting on a whole frame (16ms at 60Hz) takes time and in all cases the system delay is at least two frames.”

Fitzsimmons from Biamp says: “10Gb solutions for video transmission are a hedge against quality. In our experience, 4K signals over 1Gb can be very good. The quality is better than ‘good enough’.”

10Gb vs 1Gb

The choice of compression is inextricably linked with the choice of network. Laurent Masia, senior PLM for smart and managed switches at Netgear, however believes that 10Gb is catching up: “The adoption of 10Gb is happening worldwide. It was ratified by the IEEE and became a standard in 2003, which is important in the world of IT because it ensures inter-operability. 10Gb switches were very expensive in the beginning but the price has been coming down rapidly. At Netgear, we shipped our first 10Gb switch in 2009 and we are already shipping our third generation of 10Gb switch products.

“There are two drivers for the adoption of 10Gb infrastructure. The first is servers and virtualisation, which is having your servers and your storage separate. Rapidly, 1Gb was the bottleneck for virtualisation and 10Gb was the solution for replacing fibre channels. 10Gb is very popular for servers and data centres today.

The second driver was something that happened 10 to 15 years ago, namely the arrival of Gigabit to the desktop. Computers, laptops and all PCs today ship with 1Gb ports. To connect all these products to the network, you need a larger core.” While data centres and applications where virtualisation is important might be shifting to 10Gb infrastructure, the majority of the existing networks which could be brought into service for video signal transmission are 1Gb.

Mulpeter from Crestron says: “The core for a network can be 10Gb, 40Gb or even 100Gb depending on its size. But following the star or tree topology for networks, from the core the network steps down and ends up as 1Gb at the edges.”

Fitzsimmons from Biamp says: “Right now, with regards to networks, it is definitely 1Gb around the edges. In our experience, this isn’t changing and even new installations are still maintaining 1Gb around the edges.”

According to Masia, the fight is not for existing networks. He details: “For most SMB installations, there will be a separate network or VLAN for video signal transmission. The majority of networks at the edges are Gigabit networks and if you just insert video into these networks there will be problems. Seeing that most installations for video signal transmissions will be dedicated, the question is not whether you have a 10Gb or 1Gb installation. The real question is whether you want 4K60 4:4:4 video with sub-millisecond latency or whether you just want 1080p.”

The battle lines between the proponents of a 10Gb solution and a 1Gb solution are well drawn. Kennington from SDVoE Alliance says: “For the history of AV we have been installing new infrastructure to support new AV installations. Why would that stop? Nobody says ‘you can’t have a matrix switch – you don’t already have HDBaseT infrastructure!’. No, they say ‘we’re going to have to add some new infrastructure so you can have a matrix switch’. Same thing here – you build extra capacity onto your existing network to support new AV applications. You should be expanding with 10Gb today.”

Mulpeter from Crestron says: “The difference been 1Gb and 10Gb solutions for video streaming is that 1Gb is scalable. You can scale a 1Gb solution to a larger number of endpoints. With 10Gb decoder and receiver you are passing 10GB of data. If the transmitter and receiver are on separate switches I need to pass this data not only from box to box but from switch to switch. If your switch uplink ports are 40Gb you would only be able to send four streams across. But if you were using a 1Gb video transmission solution you would be able to manage 40 streams.”

It is also important to understand the central technical problem of sending video over Ethernet, namely video is synchronous data that needs to be sent over an asynchronous network. Mulpeter from Crestron explains: “When you are browsing a website and there is missing data the network just sends it to you again and you don’t notice it. However, if you miss data on video you cannot add it back in at a later time and this is definitely noticeable. The challenge for us to make sure this does not happen. The technology for ensuring synchronous data is properly sent over asynchronous networks, such as multicast, is well developed and has been used for years.” AVB has been present on the scene with the exact promise of making this video transmission a reality. At present, Biamp’s TesiraLux can accomplish video signal transmission over the network with AVB.

Fitzsimmons from Biamp says: “TesiraLux is a frame by frame compression engine. The determining factor for latency is the frame rate and the resolution of the content does not have as big of an impact. We can state that Lux has a higher compute power than its competitors. Transmitting 4K resolution or 1080p is not different for Lux. It is more than capable of processing 4K images in similar times.”

Regarding AVB’s strengths, Fitzsimmons says: “AVB has several advantages. Its method of being ‘deterministic’ is better than the alternatives in that it provides greater flexibility. It offers better bandwidth management. AVB also had immediate advantages when setting up the network. It is more robust. Our competitors use IGMP snooping to try and achieve the same results but it is much harder to get the same performance with that route.”

But the lack of AVB switches in the market has hindered its adoption. Fitzsimmons elaborates: “AVB struggles in two regards. Firstly, it is an open standard. That means that there is no one to build the hardware for you. There is no manual and there are no specifications to follow. Research and development is required to produce AVB hardware. Secondly, the adoption rate for AVB products is dependent on the adoption of AVB network infrastructure. AVB might be technically the best standard but that does not translate into it being the best in every scenario. If the AVB switches are not available in the market, then it cannot be specified by IT departments.”

While the debate between compression and bandwidth continues, what is agreed upon is that the landscape of video signal transmission has changed as converged networks are emerging. Michaels from ZeeVee says: “Most companies may not have an extensive AV department. But they will certainly have an IT department or IT personnel. IT is our new customer and we need to realise that while a management team within company X may have been our client for AV products and services, we have an entirely new ‘customer’ within company X to satisfy and that is the IT folks.”

At the same time, the AV industry is waiting to see how the market evolves. Mulpeter from Crestron says: “The big question is whether resolutions will continue going up. 4K60 is what we are doing right now and it is a resolution that many projectors and displays can do. But is 8K going to come onto the scene? Is 8K even something that is desirable?”

Fitzsimmons from Biamp says: “Multi Gigabit or N Gigabit technology is an exciting step forward. Being able to transmit more than 1Gb on Cat6 cables would open up a lot of avenues for video signal transmission.”