If the far end hears your voice cut out during an established VoLTE call, start with the uplink media. A successful SIP call setup confirms the session was established. To explain the interruption, you need to see what happened to the RTP packets at the time the audio dropped.
Compare the same stream at two points
Capture the affected uplink stream at the device’s IP egress and at the ingress of a network-side media capture point. Match the RTP SSRC and sequence numbers on the same media leg. If a media relay rewrites them, use its stream mapping before comparing captures. Use packet capture times to compare arrivals. RTP timestamps represent sampling time and are not wall-clock timestamps.
Here is a synthetic example using a shared time reference and 20 ms packet spacing.
RTP sequence
Device egress
Network-side ingress
3100
0 ms
45 ms
3101
20 ms
67 ms
3102
40 ms
Not observed
3103
60 ms
107 ms
The device capture shows packet 3102 leaving its IP stack. The network-side capture shows the adjacent packets but does not show 3102 within the observation window. After checking for reordering and capture drops, that narrows the investigation to the segment between the two capture points. It does not identify the air interface as the cause.
Choose the next trace based on what you see
If the packet is missing between the two points, add an intermediate user-plane capture if available. Check whether the same symptom aligns with uplink grant gaps or HARQ retransmissions. Use the radio log to test that explanation, rather than inferring it from one counter spike.
If the packet reaches the network-side point, check when it reaches the receiving endpoint. A packet can be present in a trace and still arrive too late for playout.
If the packet is absent at device egress, first verify capture coverage and whether silence suppression was active. Confirm whether the device generated the packet before investigating the radio path.
When you have access to only one network-side capture, what is the first additional trace you request to distinguish radio delay from a transport or media-path issue?
Disclosure: I teach the paid English course VoLTE/IMS & 5G/LTE Throughput. The curriculum includes a lesson on codecs, RTP, and voice-quality analysis.
This is a typical BAU scenario and lets say B stop hearing A but A hearing B very well.
Rather than jumping straight in to RTP lets look at the ways this could happen from core side. Since both Uplink and downlink carries through same Bearer in Core and A side hear B and if both are stationary then first thing to check is your handset accidentally went mute or not. That’s the first thing. If that is cleared out we needs to be sure No Bearer modification visible in control plane. Check sip Update or similar modification happening in control plane if there is then need to check whether they are successful. If that is also fine then you should check Rx interface to see there is any Rx messages related to access Bearer of either side. If that is also fine. then we jump in to RTP. First thing you need to look at in the RTP capture is there any media supervision got enabled due to lack of media stream media supervision is a feature in user plane to identify mute calls and disconnect them. If its enabled then there should definitely be one which and will indicate which stream and where and why that stream got dropped. If that is not there then only we should jump in to RTP. even then you should have a good Idea of media anchoring points for a call in your network. Otherwise it is like searching for a needle in a hay stack. Usually media is anchoring for a call. by PGW/UPF,IBGF/MGW,EBGF/MGW. and sometimes MG for all these three nodes might be only one so that’s why I said you need to have a clear Idea of it. Why I put media RTP packet inspection to the last is normally you don’t capture and store RTP packets by an operator purely because there isa a capacity issue as well as an ethical issue since RTP stream can be reconstructed to listen the conversation easily. Therefore operators sometimes capture RTP header only so they can have some visibility in to deep level of investigation and very rarely an issue go that deep.
The two capture method is the right frame, and the part worth adding is how to prove the fix afterwards. A call that sounds fine is not evidence. Re-run the same two capture points on a new call after the change and confirm the sequence numbers are continuous where the gap used to be.
Before and after also need the same conditions. The gap may appear only on handover, on one codec, or under load. If the second capture is a quiet stationary call, you have shown that the fault is absent in a case that never failed, which proves nothing.
The other habit that saves time is to record the failure window by wall clock at both ends before you start. RTP timestamps describe sampling order, not arrival time, so the two traces are joined by the capture clock. If those clocks are not aligned, fixing that is the first task.
Then the ticket writes itself. It needs the point where packets stop arriving, the sequence numbers either side of that point, and the control plane events in the same window. Say also which side of the boundary the loss starts on, because that decides who owns the next step.