Measuring VoLTE voice quality: MOS, POLQA, and field diagnostics

VoLTE voice quality is a critical concern for mobile operators. Unlike data throughput, voice is a real-time service where any degradation is immediately perceived by the user. Measuring this quality objectively in the field requires specific methods and a solid understanding of available metrics.

MOS: the reference scale

MOS definition

MOS (Mean Opinion Score) is a score from 1 to 5 representing the perceived quality of a voice call:

MOS	Quality	User experience
4.3 to 5.0	Excellent	Clear, natural voice, no perceptible degradation
3.8 to 4.3	Good	Satisfactory quality, slight imperfections possible
3.1 to 3.8	Acceptable	Perceptible degradation but communication possible
2.5 to 3.1	Poor	Significant degradation, listening effort required
1.0 to 2.5	Bad	Difficult or impossible communication

For operators, the critical threshold is generally set at 3.5. Below this value, user complaints increase significantly. For a detailed look at MOS score interpretation, see our guide on VoLTE MOS score meaning.

Objective vs subjective MOS

MOS was historically a subjective score assigned by a panel of human evaluators according to ITU-T P.800. This method is costly and slow. In field practice, MOS prediction algorithms analyze the audio signal and estimate the score:

• POLQA (ITU-T P.863): the current standard for objective voice quality evaluation. Version 3.0 supports HD voice (AMR-WB, EVS) and super-wideband
• ViSQOL: open-source algorithm developed by Google, based on spectral similarity. Less accurate than POLQA but free and adaptable
• PESQ (ITU-T P.862): predecessor to POLQA, still used but obsolete for HD voice

For a detailed ViSQOL vs POLQA comparison, see our article mobile voice MOS: ViSQOL vs POLQA.

VoLTE field measurement method

Test architecture

A field VoLTE MOS measurement requires:

1. A calling terminal (UE A) connected to the target operator’s network
2. A receiving terminal (UE B) or a reference audio server
3. A reference audio signal: a standardized voice sequence (typically ITU-T P.501 sentences)
4. A capture tool that records the received audio and computes MOS

The principle: UE A transmits the reference signal, UE B records what it receives. The POLQA or ViSQOL algorithm compares the received signal to the reference and assigns a MOS score.

Measurement protocol

Before the test:

• Verify that VoLTE is active on both terminals (no CSFB fallback)
• Ensure the codec in use is the target one (AMR-WB for HD voice, EVS for super-wideband)
• Disable terminal noise cancellation if possible, as it interferes with measurement

During the test:

• Make calls of at least 60 seconds (codec stabilization takes 5 to 10 seconds)
• Repeat at least 5 calls per measurement zone for statistical significance
• Record GPS position, RSRP, RSRQ, and SINR at the time of each call
• Capture SIP messages (INVITE, 200 OK, BYE) to correlate MOS with protocol events

After the test:

• Calculate MOS for each call
• Correlate MOS with simultaneously captured radio KPIs
• Identify calls with MOS < 3.5 and analyze the cause

MOS and radio KPI correlation

VoLTE voice quality depends directly on radio conditions. Here are the most frequent correlations:

SINR and MOS

SINR is the KPI most strongly correlated with VoLTE MOS. Below 0 dB, the AMR-WB codec switches to robust mode (6.6 kbps instead of 23.85 kbps) and MOS drops below 3.0.

SINR (dB)	Typical AMR-WB MOS	Comment
> 15	4.0 to 4.5	Optimal quality
10 to 15	3.8 to 4.2	Good, codec in HD mode
5 to 10	3.5 to 4.0	Acceptable
0 to 5	3.0 to 3.5	Perceptible degradation, codec adaptation
< 0	2.0 to 3.0	Poor quality, drops likely

RSRP and MOS

RSRP has an indirect impact on MOS through SINR. Low RSRP does not necessarily cause low MOS if SINR remains adequate (absence of interference). However, RSRP below -115 dBm generally triggers a handover or RLF, causing audio interruption.

Jitter and MOS

Jitter (transmission delay variation) is a critical factor for voice. The codec’s dejitter buffer absorbs variations up to a certain threshold (typically 50 to 80 ms). Beyond that, packets arrive too late and are discarded, causing “gaps” in the voice. Jitter above 30 ms degrades MOS by 0.3 to 0.5 points.

Packet loss and MOS

RTP packet loss is the most destructive factor for MOS. Each percent of loss reduces MOS by approximately 0.2 points. Above 3% loss, MOS consistently falls below 3.5. Common causes of packet loss in VoLTE:

• Interference (low SINR) causing PDCP errors
• eNodeB scheduler congestion
• QCI 1 bearer issues (voice bearer not prioritized)
• Poorly executed handover (interruption during transfer)

For a deeper dive into QoS/QoE measurement on mobile networks, see our network QoE measurement guide.

Diagnosing VoLTE degradations

Case 1: consistently low MOS across a zone

Symptom: MOS < 3.0 on all calls in a geographic area.

Diagnosis: check SINR. If SINR < 0 dB, the problem is likely interference (overshooting, PCI collision, external interference). If SINR is adequate, check the QCI 1 bearer configuration and eNodeB scheduling policy.

Case 2: fluctuating MOS during a call

Symptom: MOS oscillates between 4.0 and 2.5 during a single call.

Diagnosis: correlate MOS dips with L3 events. A handover causes a 50 to 200 ms interruption that locally degrades MOS. Multiple successive handovers (ping-pong effect) severely degrade perceived quality.

Case 3: good MOS but frequent call drops

Symptom: MOS is good when the call is active, but calls are frequently dropped.

Diagnosis: VoLTE call drops are often caused by RLF (Radio Link Failure) or SIP issues. Analyze Layer 3 messages to identify whether the cause is radio (RLF, T310 expiry) or protocol (SIP timeout, IMS error). See our VoLTE troubleshooting guide for detailed procedures.

Automating MOS measurements

Manual MOS measurement is time-consuming. In 2026, field measurement tools enable process automation: loop call initiation, audio recording, automatic MOS calculation, and radio KPI correlation.

For automated MOS measurement methods in field conditions, see our article on automated VoLTE MOS measurement and visit our QoS/QoE measurement page for an overview of performance indicators.