Engineers have talked for years about videoconferencing on the desktop, but the possibility is only now becoming reality. Until now, high cost and relatively poor video quality prevented widespread acceptance of videoconferencing. But using the Texas Instruments TMS320C80 Multimedia Video Processor (MVP) to implement the H.320 and H.324 videoconferencing standards offers an avenue to high-quality audio and video at a price that makes desktop videoconferencing affordable.
H.320 is the only broadly and internationally recognized standard for video conferencing today. The H.320 standard and the soon to be ratified H.324 video phone standard create vast opportunities for equipment makers, software developers, and for the vendors they depend on. Where videoconferencing customers formerly were limited to large corporations that could afford room-sized systems to serve internal needs, desktop videoconferencing can reach every work station and into the home consumer market.
The H.320 standard allows for the interoperability of videoconferencing equipment from different manufacturers, a necessity for widespread, including international, use. The H.324 standard from the same international standards body, the International Telecommunications Union, or ITU, formerly known as CCITT, is expected to have similar acceptance for video phones.
It is for this reason that videoconferencing demonstrations are now becoming the hot spots at technical conferences and trade shows. A wide variety of companies-including Texas Instruments-recently have introduced processor chips for PC videoconferencing.
At the CeBIT conference in Hanover, Germany, March 8-15, TI demonstrated H.320 software designed to work with the company's TMS320C80 to provide the highest quality H.320 videoconferencing implementation yet seen for desktop PCs. By combining the high-performance 'C80 with a comprehensive package of industry-standard H.320 software, TI has created a complete digital signal processing solution for videoconferencing.
Though several other companies also demonstrated videoconferencing during CeBIT, many conference attendees agreed that the TI solution was clearly superior. What's more, unlike most competing solutions, the TI implementation is completely software-based, allowing for further enhancements without any hardware changes.
According to Dr. Viktor Vogt, president and CEO of IAT, one of Europe's leading innovators in videoconferencing systems, "Because of its superior computing power, the TMS320C80 gives IAT the opportunity to offer our customers the best available video and audio transmission quality over ISDN lines."
"Based on the 'C80, we have developed a multimedia codec offering H.320/MPEG-1/JPEG compression and decompression functionality," added Urs Stamm, director of marketing at IAT. "By utilizing the 'C80's superior computing power and flexibility, the codec meets the requirements of our own internal multimedia workstations for video communication as well as the requirements of our OEM partners, one of which is developing a point of sales/point of information system to be available later in 1995."
Although most demonstration systems comply with the H.320 standard, some provide video quality so poor as to be essentially useless. Tiny display areas, jerky motion, blurriness, and many blocky looking artifacts characterize many of the H.320 systems being offered today. Many demos try to hide the poor video quality by showing only very small images.
The question is: Why do various videoconferencing systems exhibit such remarkable quality differences when all are compatible with the H.320 standard? Doesn't the standard dictate common characteristics that should make the performance of all compliant systems virtually the same?
Answering these questions involves a look at what the standard does and what it does not do. It is important to realize that H.320 is intended mainly to create a common language that allows videoconferencing systems from different manufacturers to talk to each other. It does not define the quality of the communication.
The standard supports a variety of compression techniques that may be used to compress different parts of a single video frame. The sophistication of the H.320 implementation determines which compression technique is applied to which parts of a frame and how efficiently each technique utilizes the available bandwidth.
In many H.320 systems, some available techniques are ignored because the core processor does not have the horsepower to handle them. Other systems use certain techniques only minimally or provide poor implementations. TI's H.320 software is able to take greater advantage of available compression techniques because it is designed around the 'C80 processor with its unique combination of processing power and flexibility.
When you see demonstrations of videoconferencing, it is important to keep in mind that not all H.320 systems are alike. Here are three major questions to ask about any system you see:
Compliance with the standard may seem like a given, but that is not always the case. Before H.320, all videoconferencing systems relied on proprietary technology. Conference site A could communicate with site B only if the same make and model equipment was installed in both places.
Today there are a number of proprietary methods being used that may give better quality at the expense of interoperability. The TMS320C80 is unique among video conferencing processor chips in that it provides the processing power to do H.320 well, and at the same time can support a wide range of proprietary enhancements both inside and outside the standard to improve quality and achieve the best of both worlds.
Only when the H.320 standard is employed in demonstrations can a viewer compare apples to apples. Only then will quality differences that arise from differences in the implementation of the standard become apparent.
The fundamental objective of video compression is to provide the best video quality within the available data rate. In the case of videoconferencing where higher data rates require more expensive transmission, video images have to be highly compressed. Typical video conferencing compression rates are in the range of 100-to-1 to 300-to-1 or roughly about 10 times as much as the MPEG-1 playback standard. Invariably, with this very high compression, there will be some loss in video quality, particularly in video sequences with significant changes from frame to frame. A better quality videoconferencing implementation will minimize the impact on video quality within the bounds of the data bits available.
The H.320 standard supports a wide range of transmission data rates, and as should be expected, varying the data rate can result in extreme differences in video quality. Generally, the more data delivered to a single video frame, the more accurately the frame displays. The two most common data rates used in videoconferencing today are 128K bits per second, known as base rate ISDN, and 384K bits per second, or triple base rate ISDN. It should be noted that out of these data rates must come both audio and video. Within the H.320 standard the most commonly used audio compression standard is G.728, which requires 16K bits per second. This must be "taken off the top" from these data rates. Generally, 128K bit data rates is the target for desktop H.320 videoconferencing, while room systems may employ either 128K or 384K data rates.
A new standard that is closely related to H.320 is H.324, now in the ratification process. H.324 is aimed at supporting the lower data rates available on ordinary analog phone lines of 28.8K bits per second. The H.324 standard will improve on H.320 to give better quality at lower data rates.
Many video conferencing demos are run in "loopback mode," where the video is compressed and then decompressed on a local display without actually traveling over a telephone line. These loopback demos can be useful in evaluating video quality, but it is critical to know what data rate is being used. It is possible to run a loopback demo utilizing data rates that would not be economical for use with today's phone lines. If that is the case, the demo will produce video superior to real-world performance.
The third question-how good does the picture look-can determine whether the product is useful. Video quality is highly subjective. We all see pictures in slightly different ways, and we may value different qualities, such as clarity or smooth motion.
Still, there are picture quality characteristics that almost any viewer will consider as deficiencies such as blurry images, appearance of blocky artifacts, jerky motion, and the failure of the audio to synchronize with the video. To some degree, these problems are evident even on expensive, room-sized videoconferencing systems operating at commonly affordable transmission rates. To take off in the marketplace, desktop videoconferencing must minimize these problems while also greatly reducing system cost.
By its nature, videoconferencing technology loses some detail in the data compression and decompression process. But the more processing power available and the better the implementation of the H.320 standard, the higher the overall visual quality will be. With so many variables, video quality can differ a great deal from one H.320-compliant system to another. A system designed around a powerful processor such as the 'C80 can take advantage of H.320 capabilities to produce sharp images and reduce blocky artifacts and jerky motion. Other systems may implement only enough of the standard to gain compatibility without adding the features necessary for good quality video.
Think of the H.320 standard as a toolkit of compression methods. It is up to each implementation to decide which tools to apply where within each image. There can also be significant differences in how well each implementation applies each tool. It is possible to be H.320 compliant and only implement a subset of the tools or do a poor job of applying them, but invariably video quality will suffer. To provide the best video quality, an H.320 implementation must constantly make coding and bit allocation decisions to get the best video quality within the available data rate.
Some of the key variables that may effect video quality include:
H.320 does not specify how motion vectors must be obtained, and implementations vary widely in how the vectors are computed. An implementation with very little motion estimation may comply with the standard, but it will provide poor video quality. On the other hand, there is a point of diminishing returns in motion estimation. TI's H.320 software achieves very good motion estimation by using intelligent algorithms that take best advantage of the special imaging hardware inside the 'C80, saving processing power and/or cost for other functions.
While systems can claim video at up to 30 frames per second, the best overall video quality over affordable transmission lines is generally achieved at lower frame rates. Though frame rate can be limited by the ability of the compression processor, one should be aware that claims of high frame rates are not a good way to tell which system will produce the best overall video quality. When and by how much to vary the frame rate is an implementation decision.
One of the most common reasons to pre-process the image is to eliminate the noise that most video cameras generate. The noise is generally single pixel errors that occur somewhat randomly. Compressing and sending the noise wastes data bandwidth. Simple low pass filtering the video can reduce noise but blur the rest of the image. More sophisticated processing can be used to greatly reduce the noise while preserving the resolution of the original image.
All of these factors, and others, contribute to the video quality achieved by a given H.320 implementation. Some of their effects are interdependent and choosing the best trade-offs in a given video sequence can dramatically improve the appearance of the pictures you see on your screen. Because the 'C80 is both powerful and fully programmable, TI has been able to implement a highly intelligent, case-sensitive H.320 solution that takes desktop videoconferencing to a new level. Of course, continuing research and development will improve many of these compression tools over the next year or two. Adaptive quantization, for example, currently is a subject of major investigation and refinement. As the industry learns more about techniques, improvements in the standard, such as the new H.324, will, no doubt, evolve.
For OEMs, the virtual certainty of change and improvement involves both a problem and an opportunity. It can be an opportunity because it opens new avenues for product differentiation and competitive advantage. It can be a problem because it creates the possibility that a product will be, in effect, obsolete by the time it reaches the market. Keeping up with technological change often requires major redesign and hardware development. TI's fully software programmable DSP-based videoconferencing solutions can help OEMs enhance the opportunities while minimizing the problems brought by technological advances.
The programmability of the 'C80 and related products that TI plans to introduce for videoconferencing permits upgrades in software, rather than hardware. Not only can enhanced versions of the current standard be changed in software, new standards such as H.324 or even radically different algorithms, such as wavelets, can be supported long after the first product has shipped. For OEMs, this feature can mean the difference between success and failure in a rapidly changing market. But for the millions of people who simply want to communicate face-to-face across the miles, it is the quality of the picture and sound that matter most.
The next time you see videoconferencing technology demonstrated at a trade show, put yourself in the consumer's place and ask, "Which desktop system looks best?" Chances are, you'll prefer the advanced H.320 implementation developed by Texas Instruments.
# # #