Calculate latency budgets
Got a VoIP tip of your own? Why not send it in? We'll post it on our Web site, and we'll enter you in our tips contest for some neat prizes.
Everybody knows latency is bad for VoIP. Latency causes pauses in conversation that are annoying at best. At worst, they can even cause the listener to misunderstand you because so much of the meaning in our speech is carried non-verbally, by such things as inflection and tone and pauses in the conversation.
But many VoIP engineers don't know how much latency is too much. The simple answer is 150 ms. According to the ITU-T's recommendation G.114, 150 ms is the maximum allowable one-way delay because 250 ms is the amount of one-way delay that typical users can experience before the conversation becomes uncomfortable. This is usually where the callers begin to speak at the same time, but can't recover gracefully, because by the time they realize the other party is also talking, they're too far into their own statement.
The 100-ms difference in these two values is simply a safety net. Most of the delay is fixed; a result of the digital signal processors' (DSP) work that is specified by the CODEC you're using, and also a result of the distance the signal must travel. However, you can also experience delay as a result of traffic congestion, particularly if you have deep queues. Thus, that 100-ms difference is useful because you can't accurately predict congestion during any given VoIP call and this margin of error will assure your calls aren't affected during peak times.
Although you will find some vendors are more conservative than others, and their suggested maximum delays range from 100 ms to 200 ms, the ITU-T's recommendation of 150 is the generally accepted target value and should be used to calculate your delay budgets.
Thomas Alexander Lancaster IV is a consultant and author with over ten years experience in the networking industry, focused on Internet infrastructure.