The evolution of production switchers

It’s funny how our business reinvents itself on a regular basis. Even funnier still is how commodity IT technology has been in the forefront of switcher developments of late.

For decades, production switchers were rigid in their architecture. Each M/E bank fed into an internal router but cascaded together in a fixed structure. M/E 1 could not have M/E 3 as a source, though the reverse was possible. There is a simple reason for this: Analog switchers could accommodate only a fixed amount of delay, and cascading effects busses together lengthened the latency of the signal.

To accommodate this, fixed delays were inserted where appropriate in a switcher so that when cascaded, things still would work. In analog color composite systems, this was necessary to keep the color subcarrier in phase.

A game changer

In the late 1970s, Grass Valley introduced a switcher (Model 300) that, although still analog, broke that paradigm by using delay lines of various lengths to allow “infinite re-entry.” This was a game-changing innovation. It was complex and expensive, but it sustained the 300 as the premier switcher for many years. For years, there were “300 emulation modes” in some switchers to allow TDs who are reticent to change to do things in familiar ways, long after digital switchers killed the market for analog production switchers altogether.

Analog switchers were expensive for many reasons, but it is important to remember that they were truly a system. Replacing a single module might require extensive setup of the entire switcher for (especially) levels and phase. This meant that initial checkout and fine tuning in the factory was labor-intensive and required considerable time while electronics “cooked” together. Digital switchers can be produced one board at a time and don’t need to be checked and tuned as a system to anything like the precision necessary in the analog world.

Feature creep in digital switchers is often a matter of writing new code, though in some instances that code must be run in FPGA chips, which have extraordinary capabilities. Enabling features not purchased initially can be as simple as inserting a code to authorize your use, since it is not practical to deliver many versions of software with many combinations of features to users all over the world.

Today’s capabilities

“Infinite re-entry” was revolutionary in its day, but today switchers come with a dazzling array of capabilities. A modern 1 M/E system can do most of what a 3 M/E Model 300 would have been capable of, is more stable, contains multiple digital effects channels, and may be capable of both HD and SD.

Take keying, for instance. A 300 had two keyers, but a modern switcher may have four keyers per M/E, or in at least one case five keyers per M/E. In older analog systems, chroma key was an option that often cost considerably more and could be assigned to one M/E at a time. Now digital switchers often have more than one on each M/E.

But nowhere is the difference more striking than when considering digital video effects. When Vital Industries introduced the Squeezoom in 1977, it cost $200,000 and only did 2D manipulations on the X and Y axis without perspective. It was a “wirewrap wonder,” a bit flakey to keep working and fundamentally changed our industry. A couple of years later, AMPEX introduced the ADO, and Grass Valley added digital effects in association with NEC that filled an entire rack. Compare that to a switcher today that might have eight channels per M/E.

More horsepower

But good news always creates a shadow. Increases in production capability have required manufacturers to find ever more capable processing hardware solutions to enable the steady advance of features that require significantly more horsepower than their predecessors. This is where commodity IT hardware becomes part of our solution.

Take a look at any computer game today, and you will see incredible rendering capability evidenced. A big part of that is in software, of course, but rendering complex images in real time is done in increasingly sophisticated graphics engines. The compute power behind those card level solutions that every teenager wants to have are not integral to video switchers. Just look at your smartphone and rotate the screen. That real-time image processing is far from trivial, and the availability of hardware and software solutions to do many of the manipulations we do in professional switchers is critical to the ability of manufacturers to build systems at prices we are willing to pay.

The next step

My column has been about technology in transition for more than a decade. I have to admit that I only partly saw the logical extension of the evolution of switchers until recently.

What if there was a simple box with tons of processing capability in a platform best described as a deep layering engine? Think of a conventional M/E as what it really is, a highly structured set of layers with transitions possible in each layer.

Taking this to the logical extension, what if that same layering capability could be described not as nested or infinite re-entry of descrete M/Es, but rather as a set of layers that could even be described completely in human readable text (XML) so that new instructions for switcher “configuration” could be authored in many different applications. Transitions can be made to appear global or affect a single element in a single layer of a deep composite of many elements.

Make that processing stack deep enough, with a boat load of memory, present sufficient inputs to it, and you effectively have a very complex, very powerful switcher that can be redefined any way necessary. Control panels truly become a user interface.

Though this may seem a bit futuristic, it is quite possible with current processing technology that arises in part from video processing developed for the gaming industry. Such a system would be inherently reprogrammable to do many different tasks, including digital effects, titling, compositing and essentially all the processes we are used to seeing in a production or master control switcher.

When musing about this with a colleague, it was suggested to me that a system like this might cost a fraction of a large production switcher and could allow a training facility to be built. TDs who have no experience with the latest model might sit down at perhaps a touch screen control panel, which could be configured to look just like any production switcher, but more importantly could be reprogrammed to act like others at will.

Even more powerful, though, is changing the function of the switcher completely to suit the immediate need. The processing frame could be a master control switcher one minute and a production switcher the next. Integrating graphics and commercials into a sports broadcast might be as simple as giving control of the master control electronics to a production control room with a different user interface running, perhaps at the same time that master control sees its panel just like usual. Lower cost in total, more utility and future extensibility — all in a compact package that harkens to its roots as a computer game console.

John Luff is a television technology consultant.