As the number of inputs and outputs grows the total count of crosspoints grows.
Though complexity is high with virtual crosspoints, the relative cost is less.
What constitutes routing today is not what it once was. Electronic switches were developed to expand the capability of facilities, improve quality and reduce reliance on mechanical patching systems. It became more and more difficult to support large systems of distribution amplifiers and patch panels.
In some ways television routing is less complex than the telephone switchboards of decades ago. NTSC and SDI systems are inherently unidirectional. POTS service is inherently bidirectional. A video routing switcher only needs to open a circuit from one input to one output, unlike a POTS or computer network connection that must manage bidirectional service. Video topology is different. A device with an input and output, for instance a VTR, requires two circuits. Both are switched independently, but the topology of video facilities today supports only point-to-point, always-open connections.
One solution to fostering the growth of video facilities about 25 years ago, was to provide an electronic patching system where fewer distribution amplifiers were necessary because every input was available to all outputs simultaneously. A routing switcher could reduce complexity and improve performance. But as the number of inputs and outputs grows the total count of crosspoints grows. A 20×20 system has 400 crosspoints, while a 100×100 system has 10,000 crosspoints. The cost of the system was directly related to the number of potential paths. In the above system the crosspoints account for 98 percent of the circuitry.
In some newer digital systems the switching is done in a time domain switch where the crosspoints are “virtual,” and though the complexity is high the cost is less. (See Figure 1.) This is done by using time domain multiplexing of the signal within the switcher. Some manufacturers have designed such switchers. In a small switch the cost of the multiplexing equipment will swamp any savings, but in large systems the cost can be significantly reduced and new capabilities provided. For instance, consider a large central switch that actually consists of several smaller distributed time domain switchers. Inputs and outputs could be wired to nearby connection frames, but the actual time domain multiplexed signals can be connected to multiple frames, creating a fabric where any output can access any input without regard to the location of the input. This reduces the length of I/O cables, and thus the cost of implementation, without resorting to tielines between discrete routers.
What about bandwidth? Today services exist in video facilities that vary in required bandwidth from under 100Kb/s to 1.485Gb/s, a difference of four orders of magnitude. The complexity and range of signals to be switched seems to expand at least annually.
It is no longer always practical to have individual routers, treated by the control system as a “level” for each signal type. It also wastes resources to expand the router in a flat system where every signal is available to every output even if the logical connection is invalid. In a system with AES and SMPTE 292M, there is no intrinsic reason for using a single fabric for switching both. However, signal types can be combined in one flat matrix — especially where the structure uses time domain crosspoints — to save implementation cost in the end, particularly in large and complex systems where many signal types are present.
It is common today for video routers to accommodate signals from below 10Mb to 1.5Gb. This combines at least SMPTE 292 with conventional SDI signals. By doing so, a television station facing the implementation of a digital plant and the addition of DTV can save by not buying two routing frames. Such routers often offer reclocking at only one rate. The ability of SMPTE 259M to travel reliably about 1000 feet without signal restitution, and the relatively small number of devices the signal passes through without processing and reserialization into SMPTE 259M, makes reclocking for SDI signals less critical than it is for HD-SDI where the signal can only traverse a few hundred feet of cable before reclocking is important. As a result some manufacturers have chosen to make their routers reclock at only 1.485Gb.
A few years ago the cost of the line drivers, serializers and de-serializers for SMPTE 292M was very high. Fortunately these costs have dropped significantly. The difference in cost between a wideband router and an SDI router can be as little as 10 percent. It seems prudent to give serious consideration to a wideband system today. The insurance it gives is well worth the marginal difference in cost. In the future the distinction will reverse itself and the common implementation will be wideband.
The future will hold the convergence of many factors that may fundamentally change the way circuits are set up and managed in video facilities. For instance, DTV can only deliver unidirectional service without additional back-channel capability. The digital RF broadcast service mandates no return circuit to support interactivity. However, deliver the same ATSC stream via a network connection like fiber to the home and it may well become valuable for the originating system to be aware of the status and capabilities of the receiver.
Inside a video facility many devices today are being equipped with monitoring and status reporting capabilities that require bidirectional topology. If you “route” a signal to a monitoring station it would be valuable to send status and control information at the same time, using the minimum number of levels for the routing system. In fact the concept of “levels” could become unnecessary if the video circuit is bidirectional and carries the program services (video and related audio), metadata, status and monitoring, and control.
This is a change at the very core of video facility design. It is not difficult to envision a television plant as a fabric where devices provide input and accept output in an unconventional view. Consider a production switcher that is viewed as only a processing engine. It could have access to a multiplexed fabric of virtual crosspoints without having any of those circuits itself. It could process entirely in the time domain, and deliver the result back to the fabric in its assigned time slot. Individual functions within a production switcher could be available to the whole system without tying up the whole production switcher. For instance, master control could access a keyer for a weather crawl without having a discrete keyer of its own. But such holistic approaches to facilities require seamless bidirectional topology not only for signal interconnection, but also for control, status and monitoring.
The range of signals requiring point-to-point connections is steadily increasing as the impact of computer power in video and audio equipment grows. One major manufacturer announced at NAB this year that in the future every device they build will be designed with Ethernet addresses and connections. Wouldn't it be easier to put an ATM connection that carried everything on the device? What if such a system were built where access to video would not require that bandwidth be held in reserve to audio, and where the device being “served” could simply ask for status when it wants it, audio when it needs it or perhaps only the luminance video channel if a luminance key is the objective? In such a system the multilevel router of today would become redundant. Only the switched fabric of the “video network” would be needed, and each device would only need access to a homogenous single network connection. Scopes would simply request the data they need to display the intended signal. Digital audio mixers could request the signals in digital form as they are needed instead of having tons of unused input and output connections.
Routing has been at the core of television facility design for more than a quarter of a century. Today the topology issues, which include issues of how to grow complex systems at affordable cost require rethinking of how routing is viewed in the global context.
John Luff is vice president of business development for AZCAR.