The advanced networking course is a joint course between Harvard and Carnegie Mellon. Lectures are transmitted over the NSF vBNS. The transmitted information includes two-way video, two-audio, and control for the lecture notes used by the lecturer. For video and audio distribution, we use the vic and vat tools. These tools were chosen because of their known reliability, and lower latency, allowing fairly smooth interactive discussions between the two sites. The video quality is quite good: CIF resolution (288 by 352) at fairly high frame rates (about 30 frames per second). The video encoding is based on h.261 and both encoding and decoding are done in software. As a result, the frame rate is limited by the speed of the sending and receiving computer systems. Lecture notes are displayed locally at each site, i.e. they are not transmitted over the network, and control of the notes at the remote site is done through ActiveWare's Cosession.

Any particular lecture is typically given at one site (the lecture site) and displayed at the other site (the remote site), although this constraint is primarily driven by our desire to have coherent lectures and not by technical considerations. Each site has two displays: a large projection screen that is typically used to display lecture notes, and a large monitor that displays the remote audience (at the lecture site) or the instructor (at the remote site). This means that people at the lecture site (including the lecturer) can see the lecture notes and the remote audience. People at the remote site normally see the lecture notes on the screen and the instructor on the monitor, although the lecture site can decide to send video of the audience if there are questions. To capture the video, each site has two cameras: one for the audience and another for the instructor (at the lecture site). Each site also has several microphones. The instructor has a wireless microphone and separate microphones are used by the audience to ask questions, allowing fairly interactive question and answer sessions. The instructor uses a laptop to display the lecture notes locally and to control the remote display of the lecture notes.

The above picture shows the equipment that is used at each site. The equipment can be grouped in four groups, supporting different functions. The two cameras, video selection equipment and one computer are responsible for capturing and transmitting the video. The monitor and a second computer are responsible for displaying the video stream received from the other site; it is also possible to display the video on the projection screen, if there is no need to display lecture notes. The instructor's laptop handles the display of the lecture notes on the projection screen; at the remote site, this function is performed by the presentation computer. The final group of equipment deals with audio. Each site uses an audio mixer to mix the input from the different local microphones. This signal is then transmitted to the other site by the sending computer. The audio stream is received by the receiving computer and sent to the speaker system. Echoes are a problem in this set up: the sound of one site can be picked up by microphones at the other site and be piped back to the first site, creating a loop. We avoid echoes by manually switching the audience microphones off when they are unused. The microphone used by the instructor is very directional and is not a problem. The picture below shows how the lecture and audience sites use (a subset of) the computers to distribute video, audio, and lecture notes control.

The above set up can be improved in a number of ways. A first area of improvement is to automatically deal with audio echo cancellation and to improve the quality of the audio in general. Second, the video quality can be improved. For example, we could use video capture cards that do motion JPEG compression, allowing better frame resolution. Third, the lecture site could transmit two video streams, one of the audience and one of the lecturer. The remote site could then display both video streams (for example when no lecture notes are shown, e..g. during questions) or the remote site could select what to display. Finally, the current set up uses point-point communication. Using IP multicast would allow other sites to follow the lecture, although getting active participation from multiple sites would of course be a lot more complex. The first problem, audio, is by far the biggest concern in the current system, although significant improvement would probably require a room especially designed for distance education.