Convergence /

Telchemy hones in on jitter

By Steve Taylor and Larry Hettick
Network World Convergence Newsletter, 09/25/02

NetWorld+Interop in Atlanta was a quiet trade show. This was a blessing in disguise for those who actually showed up, as there was good traffic on the greatly contracted show floor, and some very interesting companies were exhibiting. One of the more interesting voice-over-IP companies was Telchemy.

Although Telchemy doesn't have a product per se for the enterprise, its software is embedded into other products such as IP phones, voice gateways, RMON probes and network analyzers for the enterprise. The software provides VoIP Quality of Experience (QoE) monitoring for an ongoing, nondisruptive voice quality measurement.

Bob Massad, Telchemy's vice president of marketing, says the difference between QoE and quality of service is that QoS deals with priority, queue and bandwidth configurations, while QoE deals with how effectively the QoS controls keep call quality high.

Key to Telchemy's measurement is looking at what's happening with jitter buffers in the end systems. This is so important that Telchemy implements in its product a jitter buffer emulator intended for network probes and analyzers. The jitter buffer is the " smoothing " buffer that deals with the variable delay from packet to packet in the network, providing for a smooth and constant rate play-out of the voice stream.

For voice, jitter is key. Here's why.

The first level of delay in a packet network is the fixed delay. So long as the fixed delay is relatively short, VoIP encounters no problem. But when you're listening to a voice stream at the receiving end, the flow of packets must appear to be continuous. Consequently, " jitter buffers " are introduced with additional fixed delay, usually about 40 milliseconds.

The sizing of the jitter buffer is nontrivial. It really should relate to the way the network behaves. For example, if there are a large number of discards, one would want to increase the jitter buffer size in small increments. Large buffers guarantee smooth playback, but they may introduce too much delay, such that there is a discontinuity in the speech that can cause problems for conversations. Small buffers minimize delay but tend to result in higher packet loss or discards, which are detrimental to listener quality.

Next time we'll dig deeper into Telchemy's voice quality analysis software.