Monday, August 02, 2004

Telirati Newsletter #48

This newsletter, written 4 years ago, describes a lesson still valid today: New technologies have to be an improvement over what they replace - a lesson VoIP has only now applied with the superior sound quality available with Skype.
Telirati Newsletter #48: What Roger Ebert Can Teach Us About Telephony

What can a movie critic, a fat man in a tweed jacket, teach us about telephony? A lot, as you will see. Telephony, because it is so widely used, is about what people want, much the same as making and showing movies. In a recent column, Roger Ebert, a noted movie critic, takes his gaze away from the screen and looks into the projection booth. There he finds digital cinema technology trying to make an unwise leap and landing under the wheels of the juggernaut of refined existing systems:

I have seen the future of the cinema, and it is not digital. No matter what you've read, the movie theater of the future will not use digital video projectors, and it will not beam the signal down from satellites. It will use film, and the film will be right there in the theater with you.

Then there is the hype:

But how good is digital projection? I saw it demonstrated in May at the Cannes Film Festival, and have read reports of those who've attended the custom "Phantom Menace" installations. A system offered by Hughes is not very persuasive, the witnesses say, but the Texas Instruments system is better; reviews range from "85 percent as good as a real movie" to "about as good." The special effects in "Phantom Menace" looked especially sharp, viewers said, and a reason: They were computer-generated in the first place, so they arrived at the screen without stepping down a generation to film. And because they depicted imaginary places, it was impossible to judge them on the basis of how we know the real world looks.

For all the hype, "about as good" is about as good as reviews get. Remind you of anything? Like how H.323 and most other forms of IP telephony is "about as good" as normal switched-circuit telephony? "About as good" should ring alarm bells to management and investors. Nobody wants to get a new technology that is about as good as the old one. In one case, we have a digital cinema technology that delivers less resolution than film, in the case of telephony, a technology that could deliver better phone calls, but generally doesn't, because of poor choices in product formulation. In the case of cinema, what is going to keep digital projection at bay?

"Dijection" offers a wonderful new prospect, if it's for real. But it's not the only possible future. Far from the boardrooms of Texas Instruments, which has unlimited financial resources and wants to grab the world movie distribution market, there is an alternative film-based projection system that is much cheaper than digital, uses existing technology and (hold onto your hats) is not "about as good" as existing film, but, its inventors claim, 500 percent better. That is not a misprint. This system is called MaxiVision48. I have seen it demonstrated. It produces a picture so breathtakingly clear it is like 3-D in reverse: like looking through an open window into the real world. Motion is shown without the jumpiness and blurring of existing film projection, details are sharper, and our eyes are bathed in visual persuasion.

Is Ebert describing something like IMAX? No. IMAX has its own problems. The system he describes is a relatively simple and quite inexpensive optimization of 35mm film projection:

Without getting into labyrinthine technical explanations, here is how MaxiVision48 works: It can project film at 48 frames per second, twice the existing 24-fps rate. That provides a picture of startling clarity. At 48 frames, it uses 50 percent more film than at present. But MV48 also has an "economy mode"... MV48 uses a new system to pull the film past the projector bulb without any jitter or bounce... MV48 completely eliminates the jiggle that all current films experience as they dance past the projector bulb. Watching it, I was startled to see how rock solid the picture was, and how that added to clarity... The result: "We figure it's 500 percent better than existing film or the Texas Instruments video projection system, take your choice," Goodhill (the inventor) told me.

Unfortunately for the digital cinema equipment providers, this refinement is startlingly inexpensive:

It is also a lot cheaper, because it retrofits existing projectors, uses the original lamp housings and doesn't involve installing high-tech computer equipment. MaxiVision's business plan calls for leasing the projectors at $280 a month, but if you wanted to buy one, it would cost you about $10,000. Estimates for the Texas Instruments digital projector range from $110,000 to $150,000 per screen.

In this, we see reflected other aspects of how the new telephony is actually unfolding, in contrast with some predictions. Instead of the telephone infrastructure (and incumbent operators) being replaced, they are being refined. Fully IP-based infrastructures may be on the way, but solutions that focus on true network operator needs are likely to do quite well until IP calls are not just "nearly as good" or even "just as good," but will likely be dominant until anyone offering an alternative technology delivers along with it phone calls that are really better.

Well, more expensive and not better seldom beats cheaper and better, but this is not the most interesting part of the comparison. The really interesting part is the fact that the refinement trumps the new technology by offering a higher quality expeience to the viewer. The system's inventors seem to grasp the needs of the people making the pictures, which is how they arrived at a system that delivers something superior to every point in the value chain: a superior tool to the cinematographer, an economical system to the cimema operator, and a visibly better result to the customer buying the ticket:

One advantage of a film print is that the director and cinematographer can "time" the print to be sure the colors and visual elements are right. In a digital theater, the projectionist would be free to adjust the color, tint and contrast according to his whims. Since many projectionists do not even know how to properly frame a picture or set the correct lamp brightness, this is a frightening prospect... We saw a scene that had been shot for Goodhill by another cameraman who likes the system, Steven Poster, vice president of the American Society of Cinematographers. Poster deliberately assembled a scene filled with technical pitfalls for traditional film and video systems:

We see actor Peter Billingsley walking toward the camera, wearing a patterned shirt. He is passed by another guy, wearing a T-shirt with something written on it. The camera tilts down as Billingsley picks up a hose to water a lawn. The camera continues to move past a white picket fence. In the background, a truck drives out of a parking lot. Not great art, but great headaches for cinematographers, who know that picket fences will seem to "flutter" if panned too quickly, that water droplets will blur, and that the sign on the side of a moving truck cannot be read. All true in the old systems. With MV48, we could read the writing on the shirt, see every picket in the fence, see the drops of water as if in real life and read the side of the truck. Case closed.

This is a devastating case. Simpler, better, cheaper, more compatible. By now you may be thinking "In film, I can see the advantages, literally. But what do you do once you have delivered a full-rate voice telephone call?" And this question is at the crux of why alternative telephone technologies have had such little impact on sales of telephone equipment. Has nobody asked if you (or, more importantly, the customer) would want a better phone call? The very term "full rate" inspires a lack of inspiration. Why go to all the trouble of digital phones, network interfaces, complicated protocol stacks, powerful servers, abundant bandwidth, and the imposition of QoS on a network that mostly spells c-h-a-o-s when all that is delivered at the end is a full-rate phone call, perhaps not even, and almost always with marginal to poor latency?

Can we do better? Why yes. And this answer is not even an outre suggestion, it is right under our noses, at least if we are familiar with ISDN or ATM. "High quality voice" is an ISDN service. It bears reminding that ISDN means Integrated Services Digital Network, integrating the definition of services into the definition of the network. We only need support a long-defined service in order to provide to users voice quality that is more like broadcast-quality audio than a phone call. ATM is even more flexible, allowing us to trade off network bandwidth for signal processing horsepower in situations where that is appropriate. H.323 and IP telephony on 100Mb networks would easily afford the bandwidth. Yet we are stuck with systems that deliver a product that is "almost as good" as the decades-old systems they want to replace. Unlikely.

But can such attention to a quality experience by the user find powerful allies? In film, it can:

The big film companies such as Kodak and Fuji should like the system, since it will help them sell more film. The directors who love celluloid, like Steven Spielberg and Martin Scorsese, should know about MV48. And there are other applications. Retail outlets use "video walls" to create atmosphere. Rainforest Cafes could put you in the jungle. NikeTown could put you on the court with Michael Jordan. No more million-dollar walls of video screens, but a $10,000 projector and a wall-sized picture.

What about telephony? Both IP telephony and voice-over-ATM systems have the prospect of delivering a truly better product, in audio quality, in call information, in lightning-quick connection, in integration with PC and Palm-based information, in the richness and friendliness of voice interface. Why use bells and tones and buzzes when you can explain things to the user and offer alternatives? Still, makers of new-style systems seem to think their work is done when it is "almost as good as" phone calls delivered over the very first, and now decades-old electronic exchanges.

Are the developers of new telephony systems stupid? No, they are merely tired. By the time they have gotten their products to work, with their investors breathing down their necks, they are happy to have something that will satisfy a specification and a business plan, rather than make a customer say "Wow!" But if you really want success, you won't stop until you have something that is really remarkable, not just respectable.

Moreover, the best way to come to an understanding of how to assemble the right set of capabilities into a system that delights not only the end-customer, but every hand along the value chain to that end-customer is to develop as integrated a view of your pursuit as Roger Ebert and the inventors of the MV48 projector have of cinema. Without the ability to translate artistic need, viewing delight, and cameraman's craft, all the way down into the engineering detail of how film is pulled through a projector, they would have produced something only prosaic, incremental, and only a half-step improvement in one place in the value chain. Similarly, without the ability to integrate knowledge of users, network operators, and all levels of system design and implementation, just having the ambition to produce a great telephony product won't get you to actually have one in the end.

Copyright 2000 Zigurd Mednieks. May be reproduced and redistributed with attribution and this notice intact.