Clearboard is focused on providing a high-quality means of collaboration, where the video is the subject of collaboration. This means that more bandwidth will be required in this application than in more typical video conferencing applications.
The following chart illustrates the maximum bandwidth available on a variety of different networking technologies. Note that the Y-axis is a logorithmic scale. Remember that the bandwidth between any two machines is only as large as the smallest link between them.
Clearboard allows you to modify a number of different parameters when sending the video. More parameters are planned for future implementations.
Video Size - NTSC sized video (640X480) is the largest video that can be sent. Full-screen capture mode will be implemented by zooming the screen down to this size, and then zooming it up on the receiver machine. Half resolution (320X240) and quarter resolution are also available.
Video Quality - Clearboard uses JPEG compression to compress each video frame before sending it over the network. JPEG has a concept of image quality called the Quality Factor (Q-factor) to control the quality of the compressed image, and hence the compression ratio. The Q-factor ca range from 0 to 100. The amount of compression that results from a given Q-factor setting varies, because it is dependant on the type of content that you are compressing (more later).
Frame Rate - Clearboard currently has no control over frame-rate, but bandwidth calculations have been made for lower-than-30 frames per second. An application based on Clearboard should provide a frame-rate control, as lower frame-rates can greatly lessen the bandwidth requirements. Certain subjects, especially when captured from the screen don't need a full 30 frames per second for high-quality collaboration to take place. For example, many CAD applications render new frames at a rate much slower than 30fps. In this case there is no need to send screen updates so frequently.
The following charts illustrate the different bandwidth requirements needed by Clearboard, based on the parameters described above. The first type of video measured was standard, O2 camera video, with a typical video-conferencing subject (a talking head). This was measured at a number of resolutions.
 |
||
 |
 |
|
A similar measurement was made using video captured from the desktop. The subject was the model from the Huge Engine Model demo spinning in both wireframe and solid mode. The bandwidth required for this subject was higher than standard talking head-style video. This is probably a combination of factors:
 |
 |
The following images provide a comparison of the image quality for the different Q-factors used in measurements above. In each image, a Q-factor 90 image is compared with a lower Q-factor image at the same resolution (The Q-90 always appears on the left of the image).
 |
|
|
|
||||||||||||||||||||||||
 |
|
|
|
||||||||||||||||||||||||
 |
|
|
|
||||||||||||||||||||||||
Ed Allard, Silicon Graphics
Last Update, Dec 10, 1997
