ITHACA, N.Y. -- New technology being developed at Cornell University could bring multimedia communications to your desktop computer a lot sooner -- and at a much lower cost -- than anyone expected.

Cornell is testing an idea called "Cells in Frames," which allows computer data to be transmitted over existing computer networks via a system called Asynchronous Transfer Mode (ATM), without expensive new hardware at each workstation. ATM is a way of sending data faster and more smoothly, designed to carry audio and video as well as text. Cornell will use the new system at first to replace its existing campus telephone system with computer telephony.

While the average user probably hasn't heard the news, the people who run computer networks, from the LAN (Local Area Network) in a small office to the worldwide Internet, are gearing up for a major changeover to ATM. The conventional wisdom has been that this will require the installation of new "ATM cards" in every desktop computer, replacing the existing Ethernet cards that connect the computer to the network. But if Cells in Frames works as expected, that may not be necessary.

"We are saying to the datacom industry, 'We don't like the direction you're going, this is what we want to buy, who will make it?'" said Dick Cogger, assistant director for advanced technology and planning for Cornell Information Technologies, the university's computer service. Cogger was one of the first to propose the concept of Cells in Frames, and helped develop the version of it that Cornell will test.

The current generation of computer networks transmit data in "packets." A packet is just a short stream of data bits with special combinations at the beginning and end that tell computers along the way, called "routers," what the packet is and where it's going. A short e-mail message may be broken into several packets, each traveling on its own, mixed in with lots of other packets being sent by lots of other computers. All the packets in the message may not follow the same route, or arrive at the same time, but the computer at the other end knows how to put them back in order and reconstruct the original message. The sender keeps a copy of the data until receipt is acknowledged, and, if the acknowledgment doesn't come, sends it again.

That's fine for e-mail, but not so good for voice or video. If parts of a signal are out of order or missing the result may be unintelligible, and if the receiving computer has to wait for resends, the delay would interfere with normal conversational behavior.

ATM was created by the computer industry to overcome these problems and pave the way for a new generation of "multimedia" computing. It puts data into very small strings of binary digits, called "cells," that can be switched and moved much faster, and in organized groups. If you connect to someone else's computer for a phone conversation using ATM, the computers will set up a route through all the switches out in cyberspace for your messages to follow, and keep that route open throughout the conversation so the data flows smoothly. This is called a "virtual circuit," because once it's open it works just like the hardwired circuit of the old-fashioned telephone system.

Audio and video would be digitized for transmission over this system, just as music is for compact disks. Cogger believes Cells in Frames will allow computers to use ATM with their existing network Ethernet cards, at a saving of several hundred dollars per station. To oversimplify, Cells In Frames will pack a number of ATM cells inside each of the larger chunks of data, called frames, that current Ethernet cards and wiring use. Some new software will be added to the desktop computer to enable it to decode the data. Each LAN will still need a piece of hardware that does the conversion, but that turns out to be vastly cheaper than buying a new ATM card for every computer. In Cornell's case, it also replaces paying millions of dollars for new PBX switching equipment as the campus phone system grows.

Once the system is in place it will offer much more than computer telephony, Cogger said. It will, for instance, make it possible for certain desktop workstations to connect at the high speeds needed to work with supercomputers, without having to make expensive updates to every workstation on a LAN.

The focus right now is on the telephone system just because "That's where the money is," Cogger admits. "Those of us in the computer network business have had this vision of a single high-speed network that provides everything -- voice, data and video -- but spending on multimedia networks is zero," he said.

Cornell has received a $700,000 grant from the National Science Foundation to develop Cells in Frames and test it on about 200 workstations. Most of the money will go to a company called Connectware to design a new LSI (Large Scale Integration) computer chip that will be the heart of the hardware that packs ATM cells into frames. Connectware is headed by Dr. Lawrence G. Roberts, who is credited with creating the packet switching protocol that originally made the Internet possible.

Cogger expects field trials will last through this year, with deployment of a real system starting in 1997. "How long deployment will take is not clear," he said. "Certainly less than three years. Our ambition is in the year 2000 to have shrunk to 12,000 phone lines on the [conventional phone system] and be on the way down." Cornell now has about 16,000 phone lines on its system, and would grow to around 20,000 by the end of the century.

For the technically-minded, detailed information on Cells in Frames is available at on the World Wide Web.