Graphics Technology for Remote Broadcasting

Ten years ago, a remote-production truck would have included a character generator (CG), usually with a dedicated operator, and maybe a still store, usually operated by the technical director (TD).

Ten years ago, a remote-production truck would have included a character generator (CG), usually with a dedicated operator, and maybe a still store, usually operated by the technical director (TD). In 1994, that started to change radically when Fox wrestled NFL away from CBS. In a few years, the graphics effort grew to include impressive graphic displays and animated moves – with sound effects.

A production compartment in a remote truck. Many broadcasters use more than one trailer to cover remote events -- one housing production and another with graphics equipment. Photo credit Gerling & Associates.

Ten years ago, most of us would have never guessed that many CGs today would have audio channels. In that time, the graphics component of a remote production for at least two of the networks has expanded to require its own 53-foot trailer. At the start of each sports season, the mobile-unit vendors review their contracts and re-configure their production and support trucks for the upcoming season. Network and freelance crews, plus talent, often use the same set of trailers each week. The mobile-unit vendor will configure those units based on the unit’s job (A-game unit, B-game unit, etc.). Often, a stand-alone production truck will be re-configured to become the graphics trailer in a particular unit (typically, three trailers comprise a unit).

In the case of ESPN’s Sunday Night Football, the truck vendor, prior to the start of the season, gathers the staff assigned to the graphics-support units at the field shop, along with personnel from the network. Gear supplied by the truck vendor and rental houses are unpacked and installed. Monitors, router heads and communication gear are placed and wired in. This can be one or two items, or, in the case of ESPN, over 100 cases of equipment. Once the gear is wired and in place, the two trucks are interconnected, all the connections, timing and other signal-integration concerns are verified.

ESPN requires a three-trailer unit. The A trailer has production, VTRs and camera controls, along with routing, distribution, and communications infrastructure. The ESPN B trailer carries audio effects (a Calrec C2 console), graphics (two Chyron Duets and an Aprisa), four super-slow-motion systems (two LDK/EVS and two Sony BVP9500/MAV). Other equipment in this unit include robo-cam operating positions and a Grass Valley Profile, which is used to edit feature pieces. The B trailer is a 53-foot expando. The ESPN C trailer carries the Sportvision first-and-ten yellow-line system, the tech manager’s office, and the two Chapman sideline camera cranes, along with all the extra equipment, such as lighting gear.

The Fox A game has A and B units only. The B unit has an iNFiNiT! and an SGI O2, the first-and-ten line equipment, and carryall space for the sideline vehicles. Super slo mos and Profiles are operated from the main truck. The CBS A game is similar to Fox’s. Now let’s look at the systems required for many network graphics packages.

Graphics systems

The current workhorse CG for remotes is Chyron’s iNFiNiT!, which was first shipped in 1989. Just about every remote unit on the road has one. Common configurations would find a 68060 processor with Transform II board and a three-channel mix buffer, a 230 MB Zip, and/or 1 GB Jazz drive. It is used for most sports productions. The notable exceptions are ESPN’s Sunday Night Football and ABC’s Monday Night Football, which use Chyron’s newest CG and still store, the Duet and Aprisa. The iNFiNiT! is used to produce most of the lower-third graphics. It uses proprietary hardware, operating system and application software. Although not used in truck environments, Chyron did begin offering a Windows client (WiNFiNiT!) in 1997 that runs on a separate PC and controls the iNFiNiT! via Ethernet.

Most of the network productions call for an SGI Onyx2, or SGI’s scaled-down O2 visual workstation running VizRT (formally Peak Everest and RT-SET) Discreet, or Sportvision software. These are real-time animation applications that run on IRIX, which is SGI’s implementation of Unix. SGI developed and helped standardize a common set of application-programming interfaces (APIs) known as Open Graphics Library (OpenGL). These APIs are, in essence, applications that sit between the real-time applications that are OpenGL aware and the OS. Calls to an OpenGL API result in calls to an IRIX API. This allows for a common way of creating and manipulating graphics independent of the underlying OS and hardware. The applications OS and hardware just described are used to produce the “Fox Box,” which is the little upper-corner score, down and yards graphic. No matter which network uses this type of graphic, its nickname has become the “Fox Box.” Fox has actually reformatted the box to look like a ribbon across the top of the screen. The Onyx2/O2 is used for all those snappy little real-time animations where flags, time outs and first downs fly in and out. It also is used for scoreboards that fly together and then apart. It takes a lot of processing horsepower to perform these complex graphics in real time. The SGI hardware is based on 64-bit MIPS RISC processors. In the graphics display part of the O2 system, an integrated 32-bit, double-buffered (one buffer to air while the other buffer is filling with the next frame of video), OpenGL graphics engine renders very high-quality images for display at up to 1280x1024 pixels at 75 Hz.

The tape compartment in the Supershooter 18 - the Fox A unit. Photos by Nicholas Traub Photography, Pittsburgh, copyright 2002.

The Onyx graphics engine consists of three sub-components. The first is a Geometry Engine, which has a model of the graphic in 2D or 3D. An object modeled in 3D can be made to spin, tumble or appear to be stationary as the viewpoint moves. Objects, which can be text, pictures or banners, are stored and manipulated separately. They are brought together, or composited, only when needed for output. This allows graphics to be constructed only as templates. The template is a framework to hold pointers. These pointers point to one or more Oracle-type databases that contain the actual objects. Thus the player line-up graphic, whose format doesn’t change game to game as the players do, stays essentially the same. The template for that graphic might hold only pointers such as quarterback and halfback, which point to a database record that contains additional pointers to such fields as QB_picture, QB_name, QB-stats, etc. This second set of pointers would be the variables that change every week, or even at the last minute. The operator would sit at a PC or a laptop, running a thin client that would be used to input changes such as distance or down, and then send the command via LAN to the SGI system. The system would then perform the transitions from old info to new.

In addition, data streams such as game clock, which are often wrapped in TCP/IP, can also arrive into the same box. Thus the graphics box would have sockets open to the operator’s PC and the device providing the clock data via a couple of sockets. Sockets are a method for communication between a client program (such as the operator’s control application) and a server program (the graphics applications running on the graphics box) in a network. A socket is defined as “the endpoint in a connection.” Most OSs today have API calls that support sockets. Calls to these APIs prompt the OS to open threads (separate processes that wrap outgoing data or unwrap incoming data in the necessary protocol to send the data to the appropriate recipient). Often, the wrapping is a few layers thick. The most common way to initially wrap the data is Transport Control Protocol (TCP), which is a service that lets the receiver put the data back in order (because some packets might take longer paths than others from point A to B), and makes the recipient return a receipt verifying that the data arrived intact. The next layer would be Internet Protocol (IP). It provides the addressing that allows routers on the network to push the data towards the receiver (small networks would not need routers). The third wrap or layer would be the information necessary to traverse the local Ethernet, either to a router for travel onto another LAN or out to the Internet, or, if the receiving device is on the same LAN, straight to the device.

The second SGI graphics component is the Raster Manager, which takes the calculations from the 3D geometry engine and creates a 2D raster. This block does all the pixel operations, color and transparency blending, texture mapping and multi-sample anti-aliasing. It must perform these operations at real-time rates.

The Raster Manager hands its output to the third graphics component, the Display Generator, which converts the digital data into the desired analog or digital format. The internal bandwidth of the Onyx2/02 is greater than 6 GBytes/s, so the box can handle SD or HD video.

The blur between graphics and storage continues to grow. Pinnacle Systems is the main graphics package being used by NBC at the Olympics, by CBS in the U.S. Open Tennis and their NFL A game, and by TNT at the Goodwill games. Pinnacle products use Microsoft NT as their OS. Pinnacle has a family of products that treats graphics as a collection of objects. A composite graphic is made up of objects. These objects are stored separately but related to other objects that come together to form a composite or complete video signal only when desired. An object, which could be a block of text, a picture (animated or still), or a background, can be modified individually.

Templates are created that bring the required objects together to create a composite output. What is interesting about this approach is that the objects, such as the network’s “bug,” a banner, or an animated background, can be created on one graphics machine and be made available to other machines. These objects are shared and transferred between machines via an Ethernet LAN. Two machines simply open a socket connection between each other.

CGs are not the only devices that deal in objects. Pinnacle’s DVEs, still stores and even their servers deal in these objects. This means that stills or video clips can be modified remotely by changing a centrally controlled object. In this way, the whole still or clip doesn’t need to be replaced, just the changed object. This concept allows objects to be created and edited on laptops, perhaps by a graphics person sitting at the airport waiting for a flight. But, more likely, the changes will be made on desktops back at the network and downloaded over the Internet by the truck crew once on site.

Chyron, which introduced the broadcast character generator, is still a major player in Graphics. ESPN and ABC use the Chyron Duet – the replacement for the iNFiNiT! – for NFL games. The Duet (and its family of surrounding products like the Aprisa) also treats composite graphics as an associated group of objects. The Chyron application running on the Duet is Lyric. Lyric is a text, animation and graphics authoring software and contains features such as the animation tool and clip playback device used for creating Monday Night Football’s promos, lineups and animated 3D bar charts. The Duet can even have a digital disk recorder installed in it. This allows for a single-channel clip player in the CG with its storage via a SCSI drive.

While the user-interface application on a Duet rides on top of NT, the Chyron’s graphic engine uses VXWorks, which is a real-time operating system. NT passes the OpenGL-requested tasks off to VXWorks. OpenGL is a part of Lyric.

An EVS slow-motion control in a production truck. Photo courtesy of EVS.

ESPN and ABC also use Chyron Aprisa for the NFL. The Duet controls the Aprisa 300 systems, which provide Sunday/Monday Night Football directors the flexibility to include animated clips for the Aprisa. During the game, the stand-alone Aprisa 300 systems are used to play back network-promotion material as well as full-screen, rendered graphic elements. The Aprisa can also run a scaled-back version of Lyric, which allows editing of objects residing on it. The Aprisa comes in a number of variations. The 100 is like the still stores we have known. The difference is an integrated content- and asset-management database that can manage content across the other Chyron still stores, CGs and servers. The 200 adds a DDR to the 100’s capabilities. The 250 is a 200 that has two channels, which can be any combination of clip player, still store or one of each. The 300 adds internal transition effects.

Another item in the graphics racks is the Chyron Codi Telestrator, with a 15” flat-panel touch screen located up in the announce booth that lets the color announcers (John Madden being the most well known) doodle on the screen. Companies such as Sportvision and PVI are now offering systems that use chroma inclusion/exclusion techniques developed for their first-down line to key the drawing in such a way that it appears to be under the players and officials.

Some B games and local college games use a Chyron Maxine, which produces a basic Fox Box without animations.

The four common still stores found on trucks:

Accom (formally Abekas) A-42 still store, the senior member of this family, has been around for nearly 18 years.

Pixel Power Collage still store/CG

Leitch Stillfile

Leitch DSF 3121 four-channel Stillfile

In games where the game clock must still be shot with a camera, older Abekas A-53s are still occasionally used for clock positioning. Many trucks also have a Leitch Logo Compose for bug insertion. Grass Valley Profiles are used for game bumpers and transitions into and out of replays. Two channels are required for video – one provides a key signal and the other the video fill. Two channel of audio are also used for the audio that now accompanies these transitions. The graphics are usually built back at the network by their graphics departments.

Whereas game statistics used to be collected, compiled and interpreted manually, today a lot of the stats you see on screen and stated by the announcers are created automatically. One company that does this is SportsMEDIA Technology Corp. They perform data integration and distribution services and provide a central place to keep stats. What this means is that raw data generated during a game, such as scores as determined by the official scorer and official time-clock information, are fed into a proprietary database. However, the score is often entered manually to allow for timely updating of the score because most automatic systems won’t display a score until it is official. Other game activity is also entered manually into that database. But SportsMEDIA’s database can be used to automatically fill template pages created on a character generator. The CG is set to automatically request data updates as records in the database change. In this way, the CG will automatically update its display as the official score, time and other information changes.

One of the most interesting graphics overlays today is the first-and-ten line. Two companies provide these systems, and they are indeed systems. CBS uses systems from PVI, while Fox and ESPN use systems from Sportvision. The Sportvision system comprises four SGI O2s. The system gathers very accurate camera pan/tilt-head and zoom-lens telemetry, which is generated by a Sportvision modified Vinten head. The collected data is sent via a microphone channel down the camera triax. From the camera’s CCU, it is sent to a Sportvision computer called the gather PC. The gather PC sends the compiled pan/tilt-head data to a computer called the FIO. This is one busy processor. It contains a virtual map of the playing field, and makes the geometry calculations to determine where the first-down line is in 3D space. This computer also accepts data from another PC called the matte PC. The matte PC operator decides what colors in the video are to be keyed over, and what colors can’t be keyed over. The operator will have to make subtle changes as field, sun and uniform conditions change. In the case of an indoor game, this position is often a set-once-and-forget proposition. The FIO computer must do one other computational operation: It must recognize which camera is on-air. It does this by comparing video fed into the system vs. the video from cameras equipped with the modified heads (usually the three-up cameras at both 20-yard lines and the 50-yard line) to determine which camera’s telemetry to use. This approach was taken instead of using tallies because the Sportvision equipment is installed downstream from the switcher, and, because of convoluted video paths from sources to switcher, it was felt that cameras couldn’t always be guaranteed a tally signal.

The output of the FIO computer is data, which indicate which pixels to write to depict the first-down line. The render PC processes this data, and its output is the key and fill signals for the line. These signals go to a keyer. This process takes 22 frames in NTSC (or SMPTE 259) time. To compensate for this, the dirty video out of the switcher is delayed via a Sierra Video Digilinx and then fed as the background signal to the first-down-line keyer. There is a second operator, the lead operator for the system, who turns the key on and off based on director input. A third operator is upstairs in the booth. He communicates to the lead operator where the first-down marker should be placed.

In the case of PVI’s system, the line is added upstream from the switcher. PVI’s system also collects pan/tilt-head telemetry and sends it to a box they call the Elvis. This is a proprietary box based on VME-bus technology. The PVI approach is heavily hardware based. This results in only seven frames of processing delay. A second PC running a user-interface application controls the Elvis. PVI’s system requires three people to operate also.

When it comes to sheer size of a remote production, nothing compares to the Olympics. The Winter Olympics is no exception. In some ways, the Olympics is like an NFL Sunday, but continues every day for nearly two weeks. Just like NFL games, all the various venues need to be transported back to a broadcast operation center (BOC), processe and sent on to the required destinations. The host broadcaster must make every event available to various broadcasters around the globe. In addition, many broadcasters add their own capability to the mix. NBC, the U.S. broadcaster, adds a significant amount of infrastructure, much of it for their graphics packages.

This system is centered around a central server that is a repository of stills and clips created by Quantel Henrys, HALs and Editboxes, plus Apple G4 Macs running Adobe After Effects. The Apple G4 Macs are used to create special effects and elements that are composited in a Henry. NBC Olympics and Proximity Corp. developed this central-server concept. Thus, the initial creation of graphics occurs in Quantel and Apple equipment. The newly developed central server makes the various graphics content available to CGs and still stores at the various venues. NBC is using Pinnacle Fx Deko IIs with ClipDeko internal clip players and Pinnacle Thunder XL four-channel clip/still stores.

The IDS Corp. will provide scoring and event timing data for automatic display. Chyron Codis will input and display event timing data, while results data will be fed into each Fx Deko II at the venues covered. As with football, graphics templates or beds were built by the network – in this case, NBC – in New York.

The two new major items at this year’s Olympics are Proximity’s “Xenoclip” clip translator and the “2 ME Box.” The Xenoclip system allows NBC to translate clips produced on Quantel gear and put them in a data format that is immediately accessible to all character-generator and still/clip-store gear – not just stills anymore, but the ability to unify all clips into one format accessible to all. The “2 ME Box,” produced in collaboration with Pinnacle Systems, allows NBC to have a centralized storage area that contains all the clips translated by Xenoclip. They configure all the Fx Dekos and Thunder XL units to have the “2-ME” setup as a locally attached drive. This provides the ability to drag and drop clips from this server to a local drive (giving it instant access to playback).

All venues are connected to the graphics section of the IBC using either ISDN or T1 lines. This provides the venues this drag and drop ability as well. After Xenoclip translates the clips, they are sent to the “2-ME,” where anyone, whether local in the IBC or distant at a venue, can acquire either still or moving images for playback.

A significant portion of many remote-production efforts today is graphics. One major network worried as long ago as the mid 70s that graphics would distract the viewer. Today the viewer often gleans as much from visual information presented with the action as the dialog presented by the announcers. The graphics area will continue to expand as many players in the graphics realm are leading the foray into interactive television (ITV). The information gathered and displayed as television signals can now also be “re-purposed” for shipment to “Web servers.” As ITV-aware set-top boxes become available to DTV viewers, the ever-growing graphics “engine” deployed in these boxes will generate the data that provide the viewer’s interactive experience.

Jim Boston and George Hoover are in the process of writing a book about the remote-truck industry. They have established a Web site to promote this project. If you have anecdotes or other information that you think should be in the book, please visit

George Hoover is senior vice president and general manager of NEP, and Jim Boston is a West Coast consultant.