AGOCG logo
Graphics Multimedia VR Visualization Contents
Training Reports Workshops Briefings Index
This report is also available as an Acrobat file.
Back Next Contents

Experiences, good practice

It was found from practice that it was better to use rooms which were solely dedicated to videoconferencing rather than multi use lecture/seminar rooms. Since the equipment in dedicated rooms is permanently set up sound levels and camera positions will always be correct which obviates the need for testing at the beginning of each session.

Most sites involved in LIVE-NET 2 only have one CODEC and Ascend so when using ISDN they can only receive digital video from one site. This is acceptable for two site sessions but for sessions involving three or more sites some form of switching must be provided. It is possible to automatically switch video signals using a Multi Point Control Unit (MCU). When using an MCU the sites involved call the MCU which detects the site with the loudest sound level and transmits its video signal to the other sites. The current 'transmitting' site receives the signal from the previous transmitting site. Each site is fed with a sound mix from the all other sites.

MCUs can also be manually controlled in which case a chairman decides which site is to be seen by the others. This is similar to using the chairman control program in LIVE-NET.

Due to the cost of over £40k LIVE-NET 2 does not have an MCU. If an MCU was needed for a session it would be necessary to book time on one provided by an external service provider, such as BT. For a 384kbps session the cost of booking an MCU is over five times greater than the line charge, £225 per hour as opposed to £43.20 per hour [see Appendix A]. Because of the expense MCU's have only been used for a few LIVE-NET 2 sessions eg. the Photobiology conference sessions held in September 1994.

It is possible to run three site lecture type sessions over ISDN without using an MCU by using the two Ascends at UCL and combining the video signals from the two non lecturing sites into one and sending it to the lecturers site. The combination is performed with a quad split’ which takes four video inputs and produces a single video output with each of the inputs occupying a separate quadrant of output. This type of set up would perhaps not be acceptable for a meeting or seminar but it does seem to work quite well for a straightforward lecture.

As with other video networks the majority of the technical problems encountered with LIVE-NET are caused by sound balancing and the prevention of audio feedback.

LIVE-NET uses a Shure automatic microphone system (AMS) which is fed into a frequency shifter. The AMS will switch when the volume detected by the microphone reaches a preset level. This helps prevent ambient noise from reaching remote sites since the system will only switch when someone speaks

Since LIVE-NET uses microphones sitting in front of, but pointing away from, loudspeakers of the video monitors audio feedback or howlround can occur. The use of a frequency shifter in the audio path increases the sound level before howlround starts. Although the frequency shifters introduce artifacts in the audio signal (a sort of cyclic ringing’) just before feedback starts their use is considered acceptable in view of the extra gain in headroom they provide.

Whilst it is fairly straightforward to adjust the sound levels for two site sessions adjustment for sessions involving three or more sites prove more difficult. This was due to the difficulty in balancing sound levels, the sound input level to the system at the sites would normally be different which meant the audio level of the video monitors at each site had to be individually adjusted.

It was originally intended to use the AMS to provide automatic switching or control of the camera so as to achieve eye contact between the speaker and other participants. This would have been achieved by using the control signals from the AMS to switch the video signal from multiple cameras or by using a pan/tilt head attached to a single camera to physically move it. In practice did not prove to be necessary to use such a system. For teaching sessions a fixed camera pointing towards the lecturer could be used whilst students could be viewed as a group with a fixed wide angle camera.

In the case of meetings some form of control was needed. This was provided by a combination of hardware and software. Multiple cameras at a site could be controlled with pan tilt heads and selected with a video switch box.

To enable multi site meetings to be run with the limited number of inter site video channels available quad splitters were used to display four pictures on a single monitor. Two monitors could be used to display the pictures from up to eight sites. A program running on the LIVE-NET computer would then be used to control which picture would appear full screen on the centre monitor at each of the sites. This would normally be the person who was talking at the time.

This relied on having a nominated chairman at one of the sites selecting the picture to appear full screen. In practise this was acceptable during the formal sections of a meeting such as when a person was making a report since the system only needed to be switched infrequently. During discussion periods the constant switching required proved to be distracting and often the full screen picture was left on the chairman.

Some form of automatic switching of images such as provided by a Multipoint Control Unit (MCU) when using ISDN would have been preferable.

The quality of the analogue and SuperJANET links are fixed whereas with ISDN it is possible to select the desired quality of the received picture. However since the cost of an ISDN call is directly proportional to the data rate there is a conflict between maximising picture quality and minimising cost. The cost of a one hour call between UCL and another London college is £4.80 at 128kbps and £14.40 at 384kbps. A 2Mbps ISDN call (ISDN 30, equivalent to SuperJANET) would cost £72 per hour. One of the LIVE-NET lecture series started off using 128kbps but at the lecturers request switched over to 384kbps after the second lecture due to the poor picture and sound quality incurred whilst trying to lecture. All the other lecture series currently run over LIVE-NET with ISDN are at 384kbps, the maximum data rate possible with our equipment. The Ascends can run at up to 512kbps but the six ISDN 2 lines available at Gower Street are split between the two Ascends limiting the maximum baud rate to 384kbps.

LIVE-NET has been asked to run several ad-hoc one off ISDN sessions by external organisations. Whilst we are willing to help when possible we do insist on testing the link a day or two beforehand. We have had trouble linking our equipment to remote sites despite nominal compatibility. For example we have had trouble getting our CODEC and Ascend to call Japan whilst there have been no problems with the VC7000 videophone.


Back Next Contents

Graphics     Multimedia      Virtual Environments      Visualisation      Contents