Cisco TelePresence Launches New Era of Video Conferencing

October 22, 2007

13 Min Read
Design News logo in a gray background | Design News

Scroll down for a photo gallery of Cisco's TelePresence system!

While testing the TelePresence video conferencing system in 2006, Cisco engineers experienced an unexpected, telling moment. A child, seeing his mother on the system’s life-size video display, crawled beneath a conference room table, hoping to find her when he came up on the other side. But she wasn’t there. “I can’t get to you, Mom,” he called out as he looked between the chairs for her.

Video: Telepresence in Action

TelePresence’s engineering team demonstrates the teleconferencing system to DN Editor-in-Chief John Dodge.

Although it didn’t particularly surprise Cisco engineers at the time, that tale has evolved into a handy bit of corporate folklore, to be passed along to first-time TelePresence users. Prepare yourself. The system you’re about to use is that good.

Indeed, the system’s high-definition video images are so un-television-like that a few curious corporate executives have strolled behind the TelePresence display wall, the better to make
sure the people in those images aren’t really hiding back there.

“Most people are agape when they walk in,” says Jim Kittridge, senior vice president for Wachovia Corp., the giant financial services company that has purchased three TelePresence systems and plans to buy two more by year’s end. “They literally gasp; they can’t believe what they’re seeing.”

That, of course, is a far cry from the expectations most people have of video conferencing today. Since 1964, when AT&T debuted the Picturephone at the New York World’s Fair, few people
have actually purchased such systems. Even fewer have been impressed by the technology. For them, the term “video conferencing” conjures up thoughts of jerky, fractured images and
out-of-sync audio, even today.

“There’s a general market bias against video conferencing,” says Nora Freedman, a senior analyst for IDC, a market research firm. “Most people don’t know how to use it. And most of those, even if they did have it available, wouldn’t use video conferencing because of their own personal hang-ups.”

A New Era?

If engineers at Cisco Systems, Inc. have anything to say about it, however, that era may finally be coming to an end. TelePresence is more reminiscent of the huge screens that hung on the walls of the starship Enterprise in “Star Trek” or in the home of George Jetson of “The Jetsons” than it is of AT&T’s Picturephone.

To be sure, Cisco isn’t alone in its creation of Jetson-like video conferencing systems. Hewlett Packard, Lifesize, Polycom, Teliris and others have jumped in with big, strong new products, thus reinvigorating the video conferencing space.

For Cisco, however, the move to video conferencing was a classic case of a technology powerhouse entering a new field and re-thinking the products from the ground up. Although the company didn’t make cameras or video displays, two of the key links in the technical chain that comprises a video conferencing system, it had gathered a mountain of expertise in Voice over Internet Protocol (VoIP), which would provide the knowledge to break down camera images, send them over a network and receive them on the other end.

“We’re in the VoIP business and this is just a video version of our phone,” says Dave Mackie, principal engineer on the TelePresence project. “It’s a natural extension for us. I tell people that ‘TelePresence is just a really big phone.’”

Cisco’s effort began with a question from its chief development officer, Charles Giancarlo. “Charlie asked me, ‘Is something changing in two-way collaboration?’” recalls Phil Graham, senior director of engineering for Cisco. For Graham, the question turned into a technology hunt, followed by the writing of a technical brief. At the time, Graham’s brief noted, HDTV was gaining momentum and new codecs were coming on line for both audio and video.

Not long after, Graham’s brief spawned a project and a project team. Thus, engineers began the process of defining a video conferencing system that would actually be appealing to corporate clients. Their definition turned into a list of approximately 20 “Rules of TelePresence,” which included edicts involving the size and appearance of images, as well as its sound and ease of use.

“Our business unit was formed by people who hated video conferencing,” Mackie says. “It wasn’t that they hated the idea of video collaboration; it was just that they hated the implementations of video conferencing that had existed to date.”

While the team began to form, Graham and director of marketing Randy Harrell attended trade shows to learn about cameras and HDTV screens. They visited potential customers.

“I remember thinking, ‘I’m terrified of whether I can deliver the quality,’” Graham recalls. “’I’m going to have to tear the image apart, put it on an IP network and have it appear on the other side. Can I do that?’”

Targeting Latency

In essence, Cisco engineers wanted to create an experience, not a product. For users, that experience involved seeing, hearing, picking up subtle details and feeling comfortable. For Cisco engineers, it also involved designing rooms, picking out colors, selecting furniture and lighting and, of course, sending images over the Internet. Doing all that meant they had to gain expertise in areas outside the company-prescribed boundaries. And in cases where vendors couldn’t supply the proper technologies, it meant they had to create their own specifications.

“Cisco had never built a camera before,” Graham says. “Cisco had never intended to be in the display business. We weren’t in the lighting business or furniture business. But we wanted to provide a great user experience, so we needed to build things to our own specs.”

While they worked on all those facets, they also faced the intimidating prospect of sending 1080p images (1,920 x 1,080 progressive scanning pixels) at 30 frames per second with low latency.

“Nine out of 10 people we spoke to at the time thought low latency HD couldn’t be done,” says Mackie. “They thought we were nuts.”

Cisco engineers, however, didn’t want to back down from the goal of low latency. They deliberately set the bar high in that area, targeting 250 msec as the goal to ensure good user experience. Without that 250-msec target, they reasoned, video conferencing sessions would be characterized by walkie-talkie-like communications, in which participants on both ends would be unable to speak simultaneously.

“With long latencies, you can’t interrupt,” Graham says. “You can’t interact. You can’t do all the things people do in a real meeting.”

By the time such targets had been set in January 2005, Cisco had begun to assemble a strong team of experienced engineers. Mackie, for example, came over from Apple Computer, where he’d worked on the software side. He was joined by Michael Dhuey (two-time Design News Engineer of the Year candidate and hardware designer of the iPod), who had also came from Apple. Graham, who had migrated from an internal research group called Cisco Technology Center, had worked with Mackie at Precept Software and at Network Computing Devices, where they’d gained expertise in streaming video. Meanwhile, the project’s director of engineering, Rich Wales, was plucked from Sun Microsystems.

That experienced core, eventually joined by about 20 other hardware and software engineers, canvassed the technological landscape and learned that low latency 1080p HD was on the horizon. One key: H.264, an emerging standard for video compression (also known as MPEG-4 Part 10), which enabled Cisco engineers to achieve higher video quality at a lower bit rate. That technology was augmented by the rapid emergence of 1080p large screen displays, enabling engineers to meet their goal of life-size images and new CMOS-based 1080p sensor technology. The new CMOS technology, an alternative to charge-coupled device (CCD) camera sensors, gave them the ability to more effectively collect and present 1080p images.

Moreover, the new sensors allowed Cisco engineers to present the 1080p data at 30 frames per second, a much higher frame rate than that of typical video conferencing systems, especially those on the desktop. “The biggest enabler in going to 30 frames per second is the transition from CCD to CMOS,” says Director of Engineering Wales. “That’s the range at which CCD becomes problematic, especially from a noise perspective.”

Still, the engineering team faced a monumental computing challenge in expecting the system to process all that imagery. In essence, signals from the 1080p sensor had to be transferred to video encoders, compressed, packetized and transmitted over an IP network. On the other end, they had to be received, handed to a video decoder, decompressed and sent from processing hardware to the display, where it would be presented at 1080p.

To accomplish that, the engineering team built so-called “codec boxes” containing digital signal processor (DSP) arrays with on-board software algorithms. To handle the extraordinary amounts of data traveling back and forth, they endowed each codec box with 32 ADIBlackfin DSPs, each of which handles chunks of the large processing tasks, such as encoding. In all, a typical room-based TelePresence system uses three screens and more than a hundred DSPs.

Graham’s “back-of-the-envelope” calculation puts the system’s overall processing capabilities at about 0.5 tera-instructions per second. “It’s sort of like having 15 (Intel) Pentiums in there,” he says.

Finishing Touches

Endowing TelePresence with such exceptional computing power enabled the team to minimize latency, but engineers knew that was only half the battle. To finish the product, they needed to design a room that would give participants the sense they were sharing a common area with other users, no matter how far away.

They did that by using common colors and cutting meeting room tables in half so they appeared to continue into the screen and re-emerge on the other side. To deal with lighting and color issues, Cisco even called on famed movie cinematographer Janusz Kaminski, a Steven Spielberg cohort. Kaminski helped them change the color palette from blue to a warmer brown and gave lighting tips in an effort to keep participants more comfortable.

“From the earliest days, our motto was, ‘It’s all about the experience,’” Mackie says. “That motto came up again and again whenever we discussed technical trade-offs.”

By April of 2006 — just 15 months after launching the project — Cisco engineers found themselves demonstrating the system for no less than the President of the United States. President Bush’s visit was followed three months later by a similar in-person examination from British Prime Minister Tony Blair. When those demonstrations met expectations, engineers knew their October 2006 launch would be successful.

Since that time, Cisco has implemented about 110 of the TelePresence systems — many within its own company — and expects to install about 60-70 more in the coming year. Inside the company, Graham says the system has enabled him to reduce his number of visits to engineering colleagues around the country. One colleague in Pennsylvania, he says, has cut his trips to San Jose from twice-monthly to twice-yearly.

Customers at Wachovia say they’ve experienced similar results. There, executives have used TelePresence as a means of reducing business trip frequency between the company’s offices in Richmond, St. Louis and Charlotte, NC. Company executives say the product has allowed them to cut air travel bills and hotel costs, while eliminating productivity losses typically associated with airport delays. Moreover, the environmentally conscious Wachovia says TelePresence has helped reduce its energy footprint.

“The most difficult thing has been getting executives into the room because of what they had previously believed about video conferencing,” says Kittridge of Wachovia. “But once we get them in the room, it totally changes their perspective.”

Industry analysts say broader success could depend, in part, on the creation of standard communication protocols that would enable Cisco’s TelePresence to talk to similar systems from such companies as Hewlett Packard and Polycom. “Really wide-scale adoption will only happen when it becomes easier to talk to people in other companies,” says Zeus Kerravala, senior vice president of enterprise research for the Yankee Group.

Cisco executives, however, are unflinchingly optimistic about the technology’s growth. Despite the product’s huge price tags ($299,000 for a TelePresence 3000 and $80,000 for a TelePresence 1000), engineers foresee it reaching homes within three years. Cisco CEO John Chambers already has one and the company expects TelePresence to find more at-home customers among big-company executives with a need to make calls around the world at all hours of the day.

“We’re trying to change the way people work, live, play and learn,” says Graham. “And we’re starting to see it all happening.”

Timeline for TelePresence Design

April 2005 — Cisco assembles team, examines supplier technology.September 2005 — Operating system, first circuit boards, operational.December 2005 — First video IP call.April 2006 — President Bush tests TelePresence.October 2006 — Twelve-hour product rollout links systems in London, New York, San Jose, Hong Kong.October 2007 — 110 TelePresence systems in place.

Rules of TelePresence:Product edicts for video conferencing

When Cisco engineers launched their video conferencing project, they created “Rules of TelePresence” to define the user experience. A few of their rules: 1. People appear lifesize. Images on the TelePresence screen appear in the exact proportions that they would in real life. Phil Graham, senior director of engineering, insisted images of people not be broken into “Hollywood Squares” or appear as “eight-pixel heads.”“With an eight-pixel head, I can’t tell if you’re smiling,” Graham says. “We wanted life-size people sitting across the table from you.”2. Good lighting. Knowing they’d be employing high definition 1080p images, Cisco engineers wanted lighting that would bring out human detail without making subjects uncomfortably hot. Users, they say, should see “the gleam or the tear in the eye,” as well as the sweat on the forehead. In essence, they wanted their subjects to look like TV newscasters, only without the preparation.“We have normal people using this system,” Graham says. “They’re not wearing makeup.”3. Deliver audio that appears to emanate form the person who’s speaking. TelePresence screens are augmented by their own speakers. As a result, when a person who appears on the left screen speaks, the voice comes from that screen.“If a voice comes from the middle of the room, yet their image is on the left, that’s confusing,” Graham says. “Your brain works and works on it and it exhausts you in the end.”4. Make it simple. Although TelePresence signals traverse the Internet, users don’t have to type in IP addresses to make the calls. Instead, they dial a phone handset, just as they would at home. No training is required.“We want people to experience the meeting, not the technology,” Graham says.

TELEPRESENCE GALLERY

Find a supplier on oemsuppliersearch.com

Products/Services Companies

Sign up for Design News newsletters

You May Also Like