In the previous post, Tamar Barzuza helped me explain more about the G.1070 standard, and what we can learn from it on the proper network conditions necessary for a high quality experience in visual communications.
I’ve been asked numerous times regarding “requirements” from the network to support visual communications: What are the minimum network conditions to allow for a high quality experience? Can you suggest thresholds for packet loss and delay?
Automatic tools for quality assessment are the only feasible way for evaluating the end-user experience in real-life systems at all times. They aim to provide an objective quality score, with high correlation as possible to what viewers would note as their personal experience quality score.
Measuring the affect of network conditions, such as packet loss, latency and delay, on video quality is challenging. On one hand, it is very dependent on the way the video is compressed. On the other hand, there are no agreed thresholds in any of the relevant standard bodies for video quality.
In general, the G.1070 standard, as well as most automatic tools for quality assessment assume that the higher the bit rate, the frame rate and the resolution, the more sensitive the viewer is to packet loss. Any amount of packet loss reduces the video quality, and the reduction magnitude is determined by parameters such as the bitrate, the frame rate and the resolution.
Therefore, determining an amount of packet loss that is reasonable is dependent on network, display and compression parameters, as well as on the application standards. And, of course, no packet loss is always the best case.
But in the recent year or so most visual communications providers have introduced advanced techniques to cope with packet loss, most use a layered coding structure and packet protection. RADVISION’s SVC solution is a good example. Using these techniques high-end visual communications systems can now withstand a great deal of packet loss, but for a good user experience I would say “stable” network conditions, with up to %1-%2 packet loss, should be the threshold.
Delay also contributes to the quality of experience, negatively that is. And I’m talking about all types of delays here: video delay (the delay in the video itself), audio delay (delay in the audio itself) and the Lip sync, or skew, (the difference between the audio and video).
And while latency, the difference between the real-time event and that that appears on screen, may be irritating over thresholds of 300ms, and seriously annoying at 500ms, viewers are usually much more sensitive to lip sync and are around 50ms.