The Limits of Audio/Visual Perception

While there are limits to what networks can deliver economically, there are also limits to what humans can perceive. Human perception is not infinitely sensitive, so some coarseness of data can be accepted, which tends to make life easier for the network developer.

For example, the limit of human auditory sensitivity is in the range of 20 Hz to 20 kHz per second. So there is no point in a network providing audio service above that pitch.

Television in the United States is shown at 30 frames per second. Generally, there is no need to show at higher rates because viewers can't tell the difference anyway. One exception where humans can tell a difference is for flashing lights. Flashing lights at 30 frames per second is noticeable. Human eyes are much more sensitive to brightness than to color.

Other important picture rates are listed here:

  • Film, 24 frames per second (fps)

  • European television, 25 fps

  • Computer monitors, 72 fps

Human motor response for interactive games is less well calibrated. Some people are quicker than others. If human motor response is less than 1/75th of a second (about 15 milliseconds), then computer monitors will fail to deliver interactive experience for interactive games. If less than 1/30th of a second, then TVs will fail.

Latency for interactive games, therefore, must be less than 33 ms on an NTSC monitor, and less than 15 ms on a proscan monitor. The latency budget for the network is even more stringent because processor delay of the game player must be added.

The threshold of human voice response used by the phone industry is 100 ms. If voice takes longer than 100 ms round-trip, then the conversants can't differentiate a speaker's pause from network delays. This is also a useful number to use as an upper boundary for any latency that requires hand-eye coordination. If you subtract the time it takes the monitor to display an image, then network and computer processing delays are bounded by 67 ms (100 ms–33 ms) for TV sets and 85 ms (100 ms–15 ms) for PC monitors.

At Issue: Bandwidth Guarantees Versus Prioritized Packets

A corollary to the viewpoint that signaling may be the bottleneck in RBB networks is the view that it is preferable to prioritize packets rather than provide bandwidth guarantees in the network to provide reliable transport for video.

In software terms, this raises the question of whether it is worthwhile at all to support the Resource Reservation Protocol (RSVP) and ATM Switched Virtual Circuits or whether to abandon bandwidth guarantees and simply route individual packets based on a priority setting in the packet.

This is an important question because bandwidth guarantees require signaling, an end-to-end exchange of information among network elements as to the availability of resources (bandwidth, buffers) at the time the call request is made by the consumer. If resources are not available, the call is blocked, just as the phone system operates.

The end-to-end resource identification takes time and, as indicated previously, can be the bottleneck of network processing. So far, the industry is supporting bandwidth guarantees. At low levels of consumer acceptance, as during the rollout of new RBB services, end-to-end signaling may not be an issue. But if transport schemes such as packet or Sonet develop, then the question will likely recur.


..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
18.191.62.122