Successful demonstration of Real-time Multichannel Audio over the Internet

First real-time Multichannel Audio Internet demo takes place at the 107TH AES Convention in NY City, NY.

On Sunday, September 26, members of the Audio Engineering Society's Technical Committee provided the first ever real-time demonstration of high-fidelity, multi-channel audio distribution over the internet. The demonstration involved streaming a five-channel audio signal from a live performance by the McGill University Swing Band, playing at McGill's Redpath Hall, to an audience located at New York University's Cantor Film Center, in New York City.

In conjunction with the audio, a live video feed of the Swing Band was streamed simultaneously and projected onto the screen at the Cantor Film Center, where a dancer from NYU performed along with the music.

The audio transmission comprised five channels of 48kHz, 16bit, AC-3 compressed (640Kbps) audio, encapsulated in an AES/EBU stream at 1.5 Mbps. The video transmission was done using Cisco's IP/TV system, employing MPEG-1 compression at an approximate bit rate of 1.1 - 1.5 Mbps, and synchronized manually with the audio stream at the receiving end.

In total, four demonstrations took place, the first employing a 23 second buffer to allow for robustness to network congestion, and the final three using a far less conservative three-second buffer. During the final performance, the network administrators turned on the news feed coming into the New York University computer network, subjecting the demonstration to intense competition for bandwidth.

Although the data travelled from McGill University to NYU over the high-speed Canarie-2 (Canada) and Internet-2 (USA) backbones, no guarantees on available bandwidth could be provided and no special tuning of the links was done. The only concession made by the network administrators was that the incoming Usenet news feeds at both universities were disabled during the first part of the demonstration. During trials leading up to the actual demonstration, some of the links were highly congested with other data traffic, leading to losses of audio data lasting several seconds.

Despite occasional packet loss, the error recovery mechanism built into the transmission software was able to sustain an uninterrupted stream of music throughout the duration of the demo. At three occasions during the final performance, the video froze briefly but quickly recovered.

A detailed analysis of the performance will now be made, comparing digitally recorded audio tracks from the site of the actual performance at McGill with the received audio data at Cantor Hall. Plans are also underway to scale the demonstration for six channels of uncompressed audio at 96kHz, 24bit for a total of 13.8 Mbps. Stay tuned for further details in the weeks ahead...