Jitter

Telecoms glossary • J

Jitter

Call-quality fault • measured in milliseconds of variation

Packets that all arrive, but not evenly. The fault you hear as choppy audio on a line that tests clean.

What is jitter?in VoIP and business telephony

1. Jitter is the variation in how evenly voice packets arrive. A call sends a packet every 20 milliseconds; if the gaps at the far end stretch and shrink, the receiving device has to hold packets in a buffer and release them at a steady rate. Jitter is that variation, measured in milliseconds.

2. Jitter is not audible in itself. What you hear is the buffer failing. A packet that turns up after its slot has passed is discarded, so severe jitter reaches your ear as packet loss even though nothing was lost in transit.

<30 msCisco target, average one-way
20 msGap between packets on a call
RTCPWhere jitter is reported
DiscardedWhat a late packet becomes

The jitter buffer, and why jitter becomes loss

Every VoIP endpoint holds incoming audio for a few milliseconds before playing it, so that uneven arrivals can be smoothed back into an even stream. Cisco’s guidance targets average one-way jitter under 30 ms, the same document that sets 150 ms for delay and 1% for loss.

What the buffer does, and what it cannot do

Sent one every 20 ms Arrived spacing varies Played out steady rate arrives late discarded

The buffer absorbs small variations at the cost of a little delay. Anything outside its window is thrown away and heard as a gap.

Most endpoints run an adaptive buffer that grows and shrinks with conditions, and can be pinned to a fixed size instead. Either way the trade is the same: a deeper buffer tolerates more jitter and adds more delay, so jitter quietly spends the latency budget you were relying on for a natural-sounding conversation.

What makes packets arrive unevenly

Jitter is a timing problem, so its causes are the things that make one packet wait longer than the packet in front of it.

Queueing behind bulk traffic
A voice packet that lands behind a large file transfer waits for it to clear the interface. The wait changes packet by packet, which is exactly what jitter measures. This is what QoS exists to prevent.
Wi-Fi retries
Wireless devices contend for airtime and re-send frames that collide. A retry is invisible to the phone but arrives late, so a Wi-Fi handset produces jitter even when nothing is congested.
Two paths of different length
A link that balances traffic packet by packet across two circuits delivers a stream that alternates between a short path and a long one. Balancing per call rather than per packet fixes it.
Power saving
Access points and handsets that sleep between transmissions batch packets and release them together. The audio arrives in clumps rather than a stream.

Note what is missing from that list. A saturated line, a failing switch port or a bad patch lead cause loss more than variance, and belong to packet loss instead.

Read the number instead of guessing

Jitter is one of the few call-quality faults that reports itself. RTP carries an interarrival jitter estimate, and the receiving end publishes it in RTCP receiver reports, which is where every quality dashboard gets its figure from.

Three places to look before you open a ticket. Most desk phones show a call statistics screen with live jitter during a call. Your provider’s portal usually keeps a per-call quality record. A continuous test from the office over a working day exposes the pattern that a single reading cannot.

Averages flatter a line. A call that sat at 8 ms all day and spiked to 90 ms twice reports a comfortable average and still produced two unusable calls, so ask for the peak alongside the mean.

What to put to a provider

Ask where their jitter figure is measured. A number taken at the edge of the provider’s network says nothing about the last stretch through your router, your switch and your office cabling, which is where most of it is created.

Then ask for jitter, loss and latency together. A provider who can produce all three is measuring the call; a provider who answers with an uptime percentage is measuring the platform, which is a different question.

About this entry

Definitions are written for South African business buyers and checked against primary sources – provider documentation, standards bodies and ICASA – not vendor marketing. Reviews are independent and sponsors are always disclosed. Read our editorial policy and scoring methodology.

Spotted something out of date, or think we have this wrong? Tell the editors and we will re-check it.

Now hiring Hire your AI receptionist