Previously we addressed measuring the end-user's opinion of VoIP call quality. It's also very important to differentiate between call quality types. Often overlooked is that there are at least 2 prevalent types of call quality: 1) listener call quality (LQ) and 2) conversational call quality (CQ). They are very different and, when problems occur in one type or the other, they may have different causes and fixes involved.
LQ relates to call clarity and the presence or absence of disturbing effects such as noise or drop-outs. CQ relates to the degree of support for or interference with a dialog or conversation. A typical conversational difficulty is double-talk, where a response collides with new or repeated speech. There can be excellent call clarity but because the speech from each side has collided it becomes unintelligible.
To measure VoIP LQ, focus on network packet loss and jitter buffer discards. In "VoIP-unprepared" metering instrumentation very often only the network packet loss number is available, which is a significant problem when trying to measure call quality because network packet loss is rare. VoIP-capable instrumentation provides access to and includes jitter buffer configuration and operational information so that call quality can be accurately assessed. If LQ is poor, is it likely due to discards, due to excessive jitter, which in turn would likely come from congestion on the access link or from problems with the jitter buffer such as its
To measure CQ, add a delay model to the mix. Delay affects conversational smoothness or cadence. An example of delay effects is double talk. It generally occurs because a speaker is waiting for a response to something said, the response is delayed, and the speaker assumes that the other end didn't hear what was said and repeats it. Where CQ scores are low but LQ scores are high, it is most likely a delay-related problem, which may point to specific problems such as route flapping.
Thus, VoIP network managers need network instrumentation that provides both LQ and CQ. Having only misses a big part of the picture. Understanding and using both scores is needed in order to determine if there are problems, what the problems are and what are the most probable causes.
Bob Massad is a VP at Telchemy, Inc., a provider of voice quality management solutions.
This was first published in February 2003