1/5/2005 -- In large, new installations there frequently are experts in audiovisual equipment who install the equipment needed when computer graphics have to be run a distance or interconnected with more traditional video sources. In many circumstances, though, there may not be dedicated expertise available and so a networking generalist is called upon.

In this article I’ll lay out a number of the technologies available for extending raw computer video – the connection between a CPU and monitor. This is relatively straightforward and achievable without too much difficulty. Please note that designing and implementing a complete audiovisual network with numerous sources, a variety of controls, and a plethora of outputs is beyond the scope of this article and is generally best left to those who have learned of the many gotchas through real-world experience.

The Basics
The connection between a computer’s graphics controller and monitor was designed back in the days when VGA (640 horizontal by 480 vertical pixels) was considered state of the art. While some high-end graphics systems use high-bandwidth connectors such as BNC or 13W3, the PC standard HD15 connector has become near-ubiquitous. The video connectors and cable between them have bandwidth limitations that are a function of distance; in effect, the cable acts as a low-pass filter. The problem is that modern, high-resolution graphics controllers and displays can have very high frequency components in the video signal, and the low-pass filtering of the cabling can have an impact on the image clarity.

Table 1, below, shows the pixel clock rates for a number of commonly used display resolutions. For a correctly reproduced image, the graphics controller’s output amplifiers and the signal at the end of the cable assembly must slew about 0.7 volts from full black to full white in a fraction of the pixel time. The critical factors in this are the cable capacitance per foot and the quality of the terminations between cable and connectors. Computer video “VGA” cables are rarely detailed for these parameters; “high performance” or “premium” are what pass for specifications. Maybe the best rule of thumb is that the thicker the cable, the better the performance. At a minimum, any decent VGA cable should use mini-coax inside (rather than twisted pair) to carry the red, green and blue video signals; steer clear of really cheap or thin cables and you probably will be OK.

Table 1. Video Data Rates vs. Resolution

So, how far can you go between CPU and monitor with a really good VGA cable? Well, the answer (unsatisfying as it may be) is “it depends.” The thing it depends on most is the resolution and hence data rate of the video. However, as the high frequency components in the video are attenuated the most, it also depends on the type of data that is being displayed and how sensitive the viewers are to video fidelity.

The quickest way to see if bandwidth limitations are impacting the video is to display a loose horizontal and vertical crosshatch of one pixel wide lines at the desired operating resolution and refresh rate. The vertical lines require the system to turn on and off very rapidly to light the single pixel, while the horizontal lines are on for a relatively long period of time. Hence, the vertical lines represent the highest frequency components of the video. If horizontal and vertical lines appear equal in brightness, the system bandwidth should be adequate to support that resolution. If the vertical lines are less bright than the horizontal lines, that indicates that the graphics controller, monitor or most likely cabling bandwidth, cannot handle the highest-frequency components in the video. Depending upon the resolution/refresh rate and cable quality, direct connections with copper VGA cables become limitations at distances from 3 to perhaps 10 to 20 meters.

Once you have determined that you need to do something above and beyond a simple cable to connect a CPU and monitor, there are a variety of technical alternatives available. Let’s take a look at some of these and their advantages and disadvantages so that you can make an informed decision if ever put in this predicament.

Cat 5 Everywhere
The simplest and least expensive RGB video extenders are those that utilize ordinary Category 5 cabling. They are comprised of a small transmitter module connected to the source and a receiver at the display, some simple passive transformer baluns and some using active electronics, with up to several hundred feet of Cat 5 cable for connection. As Cat 5 is everywhere, this can provide a quick and easy solution if the performance is adequate for the pixel rate involved.

These devices typically work by converting the 0-0.7 volt RGB video input to an analog differential signal for transmission, then recovering and restoring the video at the receiver end. Some models also have a digital back channel for keyboard and mouse, though few support other peripheral connections such as serial ports. For relatively short distances and moderate pixel rates this works quite well and installation is plug-and-play, with no adjustments or settings required (try to choose one that supports the monitor ID lines, used by the system’s graphics controller to select appropriate resolutions for that particular display).

The limitations of this technology come into play as the pixel rate and distance increase. The Cat 5 copper cabling has a certain amount of capacitance per foot, and that acts as a low-pass filter that reduces the signal to noise ratio of the differential analog video signal; electrical interference from motors and fluorescent lights also can be coupled into the line. Eventually the signal at the receiver end deteriorates to the point that random noise (“snow” effect) or bandwidth limitations become visible. Earth ground differences between transmitter and receiver can result in slow-moving hum bars on the display.

Another more subtle problem comes from the characteristics of the cable construction itself. At high pixel rates the small differences in length between pairs used for the red, green and blue video signals may be enough to shift the relative phase of the color primaries. Particularly on a CRT device where there is no signal retiming (as opposed to a flat panel), the display is relying on all three colors to be in very close synchronism, otherwise they will not combine correctly. It is the same as a convergence problem on a CRT display or CRT-based projector: white dots or vertical lines will have colored fringes at either edge. There are devices marketed specifically to fix this problem by allowing relative phase adjustment of the individual primaries and specialized low-skew cable is available, but that increases cost and system complexity. Better cabling (Cat 5e, 6, or 7) has more closely matched pairs, but terminations similar those required for gigabit Ethernet are needed. However, if your extension application is not particularly demanding, the cost of a couple of hundred dollars for Cat 5 extender sets make them a good way to go.

Fiber Enrichment
As with digital networks, once you get beyond a hundred meters or so with your video extension project it becomes time to step up to fiber optic cabling. As we have discussed in previous articles (See "Fiber Optics 101" and "Fiber Optics 102"), fiber provides much greater bandwidth than copper and is also effectively immune to electromagnetic interference and eavesdropping.

The least expensive fiber optic video extension products digitize the incoming video, serialize it, and then transmit it like any other data, using cost-effective gigabit-Ethernet optoelectronic components. At the receiving side the data is reassembled into video in a frame buffer, and read out for display with a standard graphics controller. This architecture has some advantages and some drawbacks. Traditionally, one fiber strand has been used for each of the red, green and blue primary colors, possibly with another fiber for sync. Recent advances in coarse wavelength division technology have allowed all of these signals to be combined on one fiber. A reduction in fiber count has minimum effect in a new installation where the labor cost is constant and incremental fibers cost little. However, in an existing installation this can be a big advantage if existing fibers can be used and no new fiber pull is required. The downside to using CWDM for this particular application is that the optoelectronic transmitters used are gigabit Ethernet VCSELs sorted to give about 25-nm spacing in the 850-nm band used with multimode fiber. These devices are not standardized and are available only from a few small suppliers.

Another advantage to digital transmission of the video is that conversion between different formats and interfaces can take place. For example, the graphics controller in an SGI system may generate sync-on-green while the display requires separate horizontal and vertical sync. Or perhaps the remote display may require a different refresh rate from a local monitor. Since the receiver has its own frame buffer and graphics controller, its output video characteristics can be entirely different from the source.

And of course with digital transmission there is no concern over signal-to-noise ratio or analog signal degradation in the communication channel. Within the limits of some bit-error rate, the video either gets there or it doesn’t. The power budget of digitizing video extension systems using gigabit-Ethernet optics is generally more than adequate for any reasonable fiber installation. However, there is another limitation and it is a big one.

As you can see from Table 1, once the video is digitized the data rates rapidly become quite high. If all three primaries are electrically multiplexed together onto one fiber (rather than using CWDM) this triples the rate. So, assuming a system is designed to utilize low-cost gigabit-Ethernet components operating at 1250Mbps, it could support SXGA (1280×1024) at 75Hz refresh rate over three fibers or only SVGA (800×600) on one fiber. The problem comes from the modal bandwidth distance limitations inherent in multimode fiber. The IEEE specified maximum distance for gigabit Ethernet using 850-nm optics over the commonly installed FDDI-grade multimode fiber (1000Base-SX over 62.5/125 µm 160MHz-km fiber) is just 220 m. Many of the digitizing extenders on the market claim to be able to reach significantly longer distances. As the official specifications are pretty conservative this may indeed be the case in many installations; however, many, many years of those most intimately familiar with the technology went into the derivation of the IEEE parameters and so you exceed them at your own risk. This situation is not as bad with 50/125 µm fiber (GBE 550 m to 2,000 m) so this could be a wise alternative if you want to run across a campus. And some suppliers offer a single-mode option, which effectively removes any fiber bandwidth or distance problem at an increased cost.

A couple of other provisos regarding digitizing extenders: The analog-to-digital converters (ADCs) used require knowledge of the detailed timing parameters of the incoming video in order to do their job correctly. In general, if you are using one of the standard VESA video formats, they are able to recognize and lock onto the signal, although there may be a delay when the format changes (such as when Windows boots). However, if you have a nonstandard video rate you may need to manually train the system to recognize it; hopefully, you have detailed timing specs, as the ADCs and their associated control logic can be quite particular.

Another little gotcha in the digitizing is the sampling of the video signal should be done at the precise center of each pixel for maximum color accuracy and automatic techniques for doing so may need to be supplemented by some human adjustment, particularly at higher data rates. Undesirable artifacts can also result when the color being sampled is right on the edge of a digital threshold; the ADC output can jitter back and forth between the two values if filtering is not applied. None of these are overwhelming hurdles to overcome (after all, it is the exact same process whereby a flat-panel display takes analog RGB input and turns it into a digital display) and the components are getting better every year, but it is something to be aware of if you are operating at or above SXGA resolution. Because of the availability of the ADC integrated circuits, 1280x1024 resolution at 75Hz has been the maximum operating limit for digitizing extenders for years, but newer devices now on the market should soon result in extenders working up to 1600×1200 at 60Hz, nicely matching the latest high-end flat panel displays.

Oldies but Goodies
Interestingly enough, the very highest performance fiber optic video extenders for RGB video are analog, not digital systems. By avoiding conversion into and from the digital domain, digitizing artifacts and issues with sampling along with the high serial bitstream rates that limit distances on multimode fiber can be avoided. In the simplest sense, high-performance analog video extenders retain the native signal waveform, converting electrical voltage levels to light power and back again. If done well, the output video levels are close to identical to the inputs, with the extension subsystem acting just like a very long VGA cable.

Of course, as with any real-world implementation, the devil is in the details, and proper design is critical. First, the entire system, and particularly the optoelectronics, must be highly linear; nonlinearities will result in loss of color resolution at the full-on and full-off ends of the brightness scale. Differential nonlinearities between the three color channels can also result in shades of pure gray taking on incorrect color tints. Second, as unlike digital systems the signals are not retimed at the receiver the same rules with matching fiber lengths and hence delays that were discussed above with analog copper extenders apply. Third, since the longer the fiber, the more the light is attenuated, some type of automatic gain control must be employed to ensure that the output voltage levels match the inputs.

Assuming that the product and fiber connection systems are well designed, analog video extenders can perform very well indeed, transparently replicating video at the highest resolutions supported by current displays (such as 2048×1536 or 2048×2048) at distances ranging from hundreds to thousands of meters. The main limiting factor comes from the bandwidth of the overall system including electronics, optoelectronics and fiber. As the signal levels are attenuated either via high frequency rolloff or fiber modal dispersion, the signal-to-noise ratio decreases, and upon re-amplification (white-level restoration) electrical or optical noise becomes visible as “snow.” This type of random high-frequency noise is less visible on CRT displays than on flat panels due to the digitizing that takes place in the latter, but in any event it should not be a problem if manufacturers’ application guidelines are followed. Even in extreme cases where distances are pushed to the maximum, this sort of low-level noise is likely to be less objectionable than more coherent artifacts (just as “snow” in broadcast TV reception is less annoying than multipath “ghosting”). In addition to performance, all-analog fiber extenders share with their less expensive analog copper brethren the ability to instantly adapt to different resolution/refresh rate timing; this is important in installations where video switching between different sources is employed.

A Little Primer on TV-type Video
While this article has dealt with extension of raw RGB computer video from graphics controller to display, many displays now support standard or high-definition TV formats as well. Another whole category of extension products from more traditional audiovisual equipment suppliers deals with what I’ll loosely call TV-type video. Sources for this come in many forms. Broadcast or analog-cable standard-definition TV (SDTV) has a signal modulated by separate luma (gray-scale) and chroma (color) signals which eventually are turned into RGB signals for driving CRTs or flat panel TVs. In the United States, the NTSC standard for doing this has been around since the 1950s, with another, incompatible standard called PAL used in many countries (a third standard, SECAM, is similar to PAL.) These standards take advantage of the nature of visual perceptivity to minimize the overall system bandwidth required by reducing color definition relative to gray-scale resolution. NTSC uses a 6MHz broadcast channel with usable bandwidth of 4.2MHz to carry roughly VGA resolution, with each frame comprised of two interlaced “fields” each displayed at 60/1.001=59.94Hz. Within U.S. studios and broadcast facilities, SDTV these days is almost always digitized to a standardized 270 Mbps serial stream; copper and fiber extenders for this are readily available and inexpensive. High-definition TV (HDTV) in the U.S. comes in two formats: 720-line progressive scan and 1080-line two-field interlaced. Raw HDTV in broadcast facilities is digitized to a serial bitstream of 1.485Gbps; this too can be transported on either copper coax or fiber media. For terrestrial broadcast, cable or satellite distribution, HDTV is digitally compressed to fit within part or all of (typically) a 19.2Mbps channel; the set-top box decompresses the video and outputs composite (luma plus chroma), component (separate luma and chroma) or computer-type RGB video to the display. Studio- and consumer-grade extenders for all of these formats are sold, along with a plethora of converters for intermingling standards.

To a Digital World
When you stop and think about it, it seems like a lot of effort goes into converting what is almost always a digital source (computer graphics in a frame buffer, or digital SDTV or HDTV) to an analog output then back again to digital to drive modern flat-panel displays (which are inherently digital). In order to avoid all of the complexity and problems that come with this A/D and D/A conversion, many video sources such as graphics controllers and set-top boxes now come with digital outputs in the form of DVI (digital video interface) or the newer HDMI (high-definition multimedia interface, basically DVI plus embedded audio.) These are standards that use four differential copper pairs to carry encoded RGB at pixel rates up to 1.65Gbps, adequate for 1600×1200 resolution at 60Hz refresh.

For higher resolutions, three additional pairs can be activated. The DVI specification calls for a maximum distance over copper of 5 meters. Based on personal experience, I can assure you that DVI at resolutions of 1280×1024 and above is very sensitive to the quality of the cables used; interesting flickering pixel artifacts are seen when using inexpensive DVI cables beyond about a meter or two. Fiber extenders can carry DVI over distances of several hundred meters but beware – the old GIGO rule applies and the extenders are no better than the copper cables used to connect to the source and display. If you are using high-resolution DVI over any distance, with or without fiber extension, plan on discarding the inexpensive cables that come with displays and buying the best cables that you can find or afford; brand name DVI cables such as Monster are excellent but can run a hundred bucks a pop!

Go Forth and Extend
To summarize, connecting computer video sources to remotely located display devices is not difficult with the extension products that are on the market, but tradeoffs in cost and performance must be made. For a relatively short distance at moderate resolutions, better-quality standard copper VGA cables can suffice. For distances of up to a few hundred meters, the most cost-effective solution would be passive or active transmitter/receiver pairs utilizing Cat 5 cable. For longer distances, digitizing fiber extenders work well and reliably up to their resolution limits. For the longest extensions at the highest resolutions, all-analog fiber extenders are still required. For point-to-point RGB or DVI connections under all of these conditions, no particular expertise is required and manufacturers of extension products can provide any needed application assistance. However, once you get into switching and throwing audio and TV-type video into the mix, it may be time to call on an A/V integrator with experience in the area.

More articles by Randy Bird:

• SONET Architecture 101: A Tutorial for Datacommers
  
• Emerging Optical Technologies: Resilient Packet Ring and Passive Optical Networks
  
• Cranking It Up Again: 10 Gigabit Ethernet Arrives
  
• Understanding WANs: A Technical Primer