Disclosure: I am the author of the paid English Udemy course linked below. Here is a practical example of separating call setup from media delivery.
An INVITE 200 OK followed by the corresponding ACK confirms a successful SIP setup exchange for that call leg. It does not prove that either endpoint received audible speech. Distinguish this transaction from a REGISTER 200 OK, and correlate the Call-ID, CSeq and time interval before comparing captures.
Worked example: which direction is missing?
For a simplified direct-media call with negotiated sendrecv audio:
- Endpoint A specifies 192.0.2.10:40000 as its receive address/port in SDP.
- Endpoint B specifies 198.51.100.20:50000 as its receive address/port.
- A-to-B media should be checked against B’s receive endpoint; B-to-A media against A’s. These SDP receive endpoints do not by themselves specify the source port of outgoing RTP.
Suppose A’s capture contains outgoing RTP but no incoming RTP. That observation identifies a missing direction at A’s capture point; it does not yet identify the component at fault.
| Additional observation | Next check and justified conclusion |
|---|---|
| No B-to-A RTP is seen at either B’s egress or A’s ingress | Check B’s negotiated media direction, media generation and capture coverage before assigning a network-loss cause. |
| B-to-A RTP is seen leaving B but not arriving at A | Compare synchronized captures across the intervening path, including any media anchor. This narrows the investigation; it does not identify a particular failing hop. |
| RTP is present in both directions at A, on the expected media flow | Check payload/codec compatibility, decoding, mute and the audio output path. Packet arrival alone does not establish audible speech. |
Use the current negotiated SDP, including any UPDATE or re-INVITE changes. In an actual IMS deployment, compare the appropriate media leg: SBC anchoring, SRTP, negotiated direction and bearer/QoS can change what is visible and decodable at each capture point.
The useful troubleshooting result is a bounded conclusion with a next measurement, rather than “SIP succeeded, so the network is fine” or “one capture is missing RTP, so the network dropped it.”
Free illustrated walkthrough and call-flow diagram
For structured SIP/SDP/RTP and throughput analysis, see the English course curriculum and preview on Udemy (paid).
Technical references: RFC 3261; RFC 3264, section 5.1; RFC 8866; RFC 3550.