Mobile network QoE measurement: VoLTE MOS, video quality, and 5G experience

Quality of Experience (QoE) is the ultimate measure of mobile network performance — not throughput, not signal strength, but what the subscriber actually perceives. This guide covers the complete methodology for measuring QoE on LTE and 5G networks, with a focus on voice and video quality scoring.

QoS vs QoE: why the distinction matters

QoS (Quality of Service) describes network-layer parameters:

• Throughput (Mbps DL/UL)
• Latency (RTT ms)
• Jitter (ms)
• Packet loss (%)

QoE (Quality of Experience) translates these into perceived quality:

• Voice call MOS (1–5 scale)
• Video streaming MOS (1–5 scale)
• Application responsiveness
• Session success/failure rates

The QoS → QoE mapping is non-linear. A network delivering 15 Mbps can produce excellent QoE for voice (requires < 0.1 Mbps) and poor QoE for 4K video streaming (requires > 15 Mbps). Engineers who optimize QoS metrics without measuring QoE may miss the actual subscriber experience.

VoLTE voice quality: MOS measurement methodology

The MOS scale

Mean Opinion Score (MOS, ITU-T P.800) is the international standard for voice quality:

• MOS 5: Excellent (transparent, like being in the same room)
• MOS 4–4.5: Good, slight attenuation perceptible
• MOS 3.5–4: Fair, acceptable for professional use
• MOS 3–3.5: Degraded but intelligible
• MOS < 3: Poor, unacceptable for commercial use

Automated MOS algorithms

PESQ (ITU-T P.862): First standardized objective MOS algorithm. Legacy, limited to narrowband (NB) codecs. No longer recommended for VoLTE.

POLQA (ITU-T P.863): PESQ successor, supports NB/WB/SWB. Accepted for regulatory reporting. Requires commercial license.

ViSQOL (Virtual Speech Quality Objective Listener, Google Research): Open-source (Apache 2.0), supports speech (16 kHz) and audio (48 kHz) modes. Excellent performance on AMR-WB and EVS codecs. Preferred for field measurement due to open-source licensing and native Android integration via JNI (libvisqol_jni.so).

Field VoLTE MOS measurement protocol

Setup:

1. Measurement smartphone (Qualcomm-based, rooted, with measurement suite)
2. Reference endpoint: IVR server or reference smartphone
3. Reference audio: ITU-T P.501 standardized speech files

Sequence:

1. Establish VoLTE call (verify QCI 1 bearer setup in Layer 3)
2. Confirm codec negotiated (check SIP SDP: AMR-NB, AMR-WB, or EVS)
3. Play reference audio from endpoint B → record received audio on measurement device
4. Calculate ViSQOL MOS score
5. Export CSV report (MOS score, timestamps, aligned WAV files)

Correlate with RF KPIs:

• RSRP / SINR during the call
• RTP jitter and packet loss
• Layer 3 events (Bearer Modification, SRVCC trigger, SIP re-INVITE)

Minimum session length: 30 seconds of active speech (excluding silence periods) Recommended repetitions: 5 calls per test zone

Codec MOS ceilings

Codec	Band	Theoretical MOS Max	Usage
AMR-NB 12.2 kbps	300–3400 Hz	4.1	Basic VoLTE
AMR-WB 12.65 kbps	50–7000 Hz	4.5	HD Voice
EVS 13.2 kbps	50–14400 Hz	4.5+	Super HD

A VoLTE call using AMR-NB when both devices support AMR-WB indicates a codec negotiation failure or sub-optimal network configuration. Verify via SIP INVITE/200 OK SDP analysis.

Video QoE: ITU-T P.1204.3 measurement

The video MOS challenge

Video QoE is more complex than voice — it depends on:

• Bitrate and resolution of the adaptive stream (ABR/DASH)
• Frame rate
• Compression artifacts (blocking, blurring)
• Stalling events (buffering = most destructive to QoE)

ITU-T P.1204.3 standard

P.1204.3 (Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport) provides a model for predicting video MOS from:

• QP (Quantization Parameter) extracted from the encoded video
• Bitrate, resolution, frame rate
• Network parameters (throughput, latency)

A Random Forest model (20 decision trees) trained on perceptual evaluations predicts MOS scores with high correlation to human panel results.

Field implementation: extract QP values via Android MediaCodec API during video playback, compute P.1204.3 model, output MOS score.

Video MOS benchmarks

MOS Score	Quality	Subscriber experience
> 4.5	Excellent	Transparent, no artifacts
4.0–4.5	Good	Minor quality variations, not disturbing
3.5–4.0	Fair	Visible but acceptable compression artifacts
3.0–3.5	Poor	Noticeable degradation, blocks visible
< 3.0	Bad	Severe artifacts, stalling likely

Key video QoE thresholds

• Throughput < 2 Mbps: HD video stalling likely (buffer underrun)
• Throughput < 8 Mbps: 1080p video degraded
• Latency > 200 ms: startup buffering time increases
• Jitter > 50 ms: adaptive bitrate algorithm instability

5G NR specific QoE considerations

VoNR (Voice over New Radio)

In 5G SA deployments, voice traffic runs on VoNR (Voice over New Radio) using QoS Flow 5QI 1 instead of LTE QCI 1. The measurement methodology is identical to VoLTE, but the diagnostic Layer 3 messages differ:

• 5G NAS SM: PDU Session Establishment (instead of LTE EPS Bearer setup)
• NR RRC: no SRVCC to 2G/3G (SRVCC to VoLTE instead)

5G NR latency and gaming QoE

5G NR targets < 10 ms latency for ultra-reliable low-latency communications (URLLC). For gaming and real-time applications:

• RTT > 30 ms: perceptible lag in real-time games
• RTT > 100 ms: unacceptable for competitive gaming
• Jitter > 20 ms: causes desynchronization in live streaming and video calls

5G fixed wireless access (FWA) QoE

5G FWA deployments (replacing home broadband) require higher QoE benchmarks than mobile:

• DL throughput > 100 Mbps median (household average with concurrent users)
• Latency < 20 ms
• Video MOS > 4.3 for 4K streaming

Common QoE problems and Layer 3 diagnostics

Problem: MOS < 3.5 despite good RSRP/SINR

Check Layer 3: look for Bearer Modification or SRVCC trigger during the call. If QCI 1 bearer is modified or released unexpectedly → core network QoS issue, not radio.

Also check: RTP jitter in the IMS/VoLTE signaling. High jitter (> 20 ms) despite good radio conditions points to core transport issues.

Problem: Video stalling despite adequate throughput

Check Layer 3: look for PDU Session Modification (5G) or Bearer Modification (LTE) that reduce the guaranteed bit rate. Also verify PRB utilization — if the cell is loaded > 80%, scheduler doesn’t guarantee throughput even with good SINR.

Problem: VoLTE call drops at specific locations

Check Layer 3: locate RRC Connection Re-establishment Request with cause handoverFailure or radioLinkFailure. Map to GPS coordinates to identify the exact coverage hole or handover configuration issue.

QoE measurement in Gulf and MENA markets

Operators in the Gulf region (e&, STC, Ooredoo, Zain) apply high QoE standards, with regulatory QoS/QoE frameworks from TRA (UAE), CITC (Saudi Arabia), CRA (Qatar), and CTRA (Bahrain).

Key QoE targets in Gulf regulatory frameworks:

• VoLTE MOS median ≥ 4.0
• CSSR ≥ 99.5%
• CDR ≤ 0.5%
• DL throughput median ≥ 20 Mbps (LTE), ≥ 80 Mbps (5G NR)

Regular drive test campaigns with QoE reporting are typically required for license compliance in these markets.

Conclusion

QoE measurement is the bridge between RF optimization and subscriber satisfaction. A network with excellent KPIs that still generates complaints has a QoE gap — and only direct QoE measurement (MOS voice, MOS video, latency) reveals where that gap exists.

Modern measurement tools that combine RF KPIs + Layer 3 decoding + automated MOS calculation enable field teams to diagnose QoE issues in hours rather than weeks, with objective evidence rather than subjective subscriber reports.

Further Reading

• VoLTE MOS Voice Quality
• Mobile Video Quality ITU-T P.1204
• Voice MOS: ViSQOL vs POLQA
• VoLTE Testing App — HiCellTek
• Android Network Diagnostic — HiCellTek