The bearer model
Voice over LTE is carried on a dedicated bearer with its own quality class, separate from data. Understanding which class carries what is the prerequisite for every measurement that follows — a KPI computed across all bearers will not show a voice problem at all.
| QCI | Carries | Type | Priority | Packet delay budget | Loss rate |
|---|---|---|---|---|---|
| QCI 1 | Voice media (RTP) | Guaranteed bit rate | 2 | 100 ms | 10⁻² |
| QCI 2 | Conversational video | Guaranteed bit rate | 4 | 150 ms | 10⁻³ |
| QCI 5 | IMS signalling (SIP) | Non-guaranteed | 1 | 100 ms | 10⁻⁶ |
| QCI 8 / 9 | Default data bearer | Non-guaranteed | 8 / 9 | 300 ms | 10⁻⁶ |
Filter by QCI, always
A generic drop call rate mixes voice and data bearers and will hide a voice problem entirely. Every VoLTE KPI in this article is computed on QCI 1 only. Signalling faults, by contrast, appear on QCI 5 — a call that never reaches the media stage fails there, not on QCI 1.
The KPIs that matter
VoLTE DCR = QCI 1 abnormal releases / (QCI 1 normal + abnormal releases) × 100%
Filtered to QCI 1. Handover to circuit-switched at the coverage edge is a major contributor, so a rising figure often points at mobility rather than at the radio conditions where the call was established.
| KPI | What it measures | Indicative target | Source |
|---|---|---|---|
| Accessibility | QCI 1 bearer setup success | ≥ 99.5% | E-RAB establishment counters, QCI 1 |
| Retainability | QCI 1 abnormal release rate | ≤ 0.5% | E-RAB release counters, QCI 1 |
| SRVCC success | Active call handover to circuit switched | ≥ 98% | SRVCC execution counters |
| Post-dial delay | Invite to ringing | < 3–4 s | SIP trace, IMS records |
| Voice quality | Perceived audio quality (MOS) | ≥ 3.5 | RTCP, probe, drive test |
| Packet loss | QCI 1 RTP loss | < 1% | RTCP, PDCP discard counters |
| Jitter | RTP inter-arrival variation | < 20 ms | RTCP, jitter buffer statistics |
Reading radio conditions for voice
The same level-against-quality read applies, but the actions differ — voice is a low-rate, delay-sensitive service, so the levers are about robustness rather than throughput.
| SINR good | SINR poor | |
|---|---|---|
| RSRP good | Healthy voiceConditions are not the constraint. If quality is still poor, look at transport, codec rate and jitter buffer. | InterferenceCheck overshoot, cell overlap and PCI planning before touching any voice-specific parameter. |
| RSRP poor | Cell edgeRobustness levers apply — transmission time interval bundling, header compression, and correctly tuned handover to circuit switched. | Coverage holeHand the call out to the legacy layer cleanly, or add coverage. No parameter will rescue this. |
Behaviour at the coverage edge
Where LTE coverage ends but legacy coverage continues, an active call must be handed to the circuit-switched domain. This is the single most failure-prone moment in a VoLTE call, and it fails in two opposite ways.
- Triggered too late — the radio link fails before the handover completes, and the call drops. The KPI records a drop, not a handover failure, which is why the two must be read together.
- Triggered too early — calls leave a perfectly serviceable LTE layer, degrading voice quality and loading the legacy network unnecessarily.
Both the radio trigger threshold and the readiness of the target network have to be right. A correctly tuned trigger into a congested or unprepared legacy layer still fails, and the failure will appear to be a radio problem.
Troubleshooting playbook
Voice faults present as a small number of recognisable symptoms. Start from the symptom — it narrows the search far faster than starting from the KPI.
| Symptom | Likely causes | Where to look |
|---|---|---|
| Call setup fails | IMS registration failure; QCI 5 signalling problem; QCI 1 rejected by admission control; proxy unreachable | SIP trace, registration status, QCI 1 setup counters, resource block load and admission thresholds |
| Call drops mid-conversation | Radio link failure at the edge; late or failed handover to circuit switched; handover failure | SRVCC attempt and success counters, trigger thresholds, radio link failure causes, neighbour relations, legacy network readiness |
| One-way or muted audio | Media path broken in one direction; uplink grant starvation; address translation or transport fault; bearer mismatch | RTP and RTCP in both directions, uplink scheduling, media path through the core, traffic marking end to end |
| Poor voice quality | Packet loss, jitter, interference, low codec rate, over-aggressive discontinuous reception | RTCP loss and jitter, SINR, codec rate adaptation, DRX timers, transport quality of service |
| Robotic or choppy audio | High jitter, bursty packet arrival, frame erasure | Jitter buffer tuning, scheduler latency, header compression context loss, PDCP discard |
| Long post-dial delay | Signalling latency, SIP retransmission, slow QCI 1 establishment | Invite to ringing timing, signalling transport, dedicated bearer setup time |
Interpreting mean opinion score
| MOS | Quality | Perception |
|---|---|---|
| 4.3–5.0 | Excellent | Wideband or HD voice; impairment not perceptible |
| 4.0–4.3 | Good | Toll quality; users satisfied |
| 3.5–4.0 | Fair | Acceptable — the usual target floor for VoLTE |
| < 3.5 | Poor | Impairment noticeable; complaints expected |
MOS is an outcome, not a cause. It confirms that users are affected and quantifies how badly, but the fix always comes from the packet-level measurements underneath it — loss, jitter, delay and the radio conditions producing them.