VoIP MOS Estimator

Estimate VoIP call quality from round-trip time, jitter and packet loss. Get the E-model R-factor and MOS score for G.711, G.729 and Opus.

Advertisement

Mean Opinion Score (MOS) rates voice quality from 1 to 4.5, and the ITU-T E-model turns network measurements into one. This estimator takes round-trip time, jitter and packet loss, and a codec (G.711, G.729 or Opus), and returns the R-factor and MOS, with a plain verdict: good at 4.0 and above, fair from 3.6, and poor below that.

It uses a simplified version of ITU-T G.107. One-way delay is half the round-trip time plus the codec delay plus a jitter buffer, which is assumed to be the larger of 20 ms or twice the jitter. Delay and codec-dependent loss impairments are subtracted from a base R-factor of 93.2. For example, G.711 with an 80 ms round trip, 10 ms jitter and 0.5% loss scores about 4.3, while G.729 at 300 ms, 30 ms jitter and 2% loss falls to about 3.3.

Use it to check a WAN or internet path before rolling out voice, to see how much loss or delay a codec tolerates, and to set targets for SD-WAN or QoS policies. The model assumes random loss and ignores echo and burst loss, so treat the number as a planning estimate and confirm with live call measurements. To size the bandwidth for the same calls, use the VoIP bandwidth calculator.

How MOS and the R-factor relate

Mean Opinion Score (MOS) rates call quality from 1 to 4.5. The ITU-T E-model works it out from an R-factor between 0 and 100, which starts near 93.2 for a perfect connection and drops as delay and loss add impairment. This estimator reports both numbers.

It labels results good at a MOS of 4.0 or higher, fair from 3.6 to 4.0, and poor below 3.6. Most users notice problems below about 3.6.

What the model includes

The estimate uses a simplified version of ITU-T G.107.

  • Delay: one-way delay is half the round-trip time, plus the codec delay (0 ms for G.711, 15 ms for G.729, 26.5 ms for Opus), plus a jitter buffer assumed to be the larger of 20 ms or twice the jitter.
  • Codec and loss: each codec has a base impairment and a tolerance for loss. G.711 is uncompressed and tolerant, G.729 is compressed and more sensitive, and Opus sits in between.

For example, G.711 with an 80 ms round trip, 10 ms jitter and 0.5% loss scores a MOS of about 4.3. G.729 with a 300 ms round trip, 30 ms jitter and 2% loss falls to about 3.3.

What the model leaves out

The model assumes random packet loss. It does not include echo, bursty loss, transcoding between codecs, headset quality or how well a given phone system conceals lost packets. Treat the score as a planning estimate and confirm with live call measurements from your phone system.

Frequently Asked Questions

What is a good MOS score for VoIP?+

A MOS of 4.0 or higher is good. Between 3.6 and 4.0 is fair and acceptable for many users. Below 3.6, most users notice problems. The maximum in the E-model is about 4.5.

What latency is acceptable for voice calls?+

ITU-T G.114 recommends keeping one-way delay below about 150 ms for most applications. Delays between 150 and 400 ms are acceptable if users are aware of them, and above 400 ms is generally unacceptable. This calculator takes round-trip time and halves it, then adds codec and jitter-buffer delay.

How much packet loss can a voice call tolerate?+

It depends on the codec. Uncompressed G.711 with loss concealment usually stays good under about 1% random loss, while compressed codecs such as G.729 are more sensitive. Run your own numbers through the estimator to see how the score moves.

Why does jitter matter for voice?+

Packets arrive at uneven intervals, and the receiver holds them in a jitter buffer to play them smoothly. More jitter needs a bigger buffer, which adds delay, and packets that arrive too late are dropped and count as loss.

How does codec choice affect call quality?+

Each codec adds its own delay and has its own sensitivity to loss. G.711 is the simplest and most tolerant but uses the most bandwidth. G.729 uses less bandwidth but loses more quality to packet loss. Opus adapts well and is common in modern systems.

Is this the same as measuring a real call?+

No. This is a model based on network measurements. Real calls also depend on the phone system, headsets, echo and how loss happens in bursts. Use the estimate to plan and compare, then test live calls.

Related tools

This tool is provided for informational and educational purposes only. All processing happens in your browser — no data is sent to or stored on our servers. While we strive for accuracy, we make no warranties about the completeness or reliability of results.