Technology guide · VoIP
VoIP Technologies
& Protocols
How signalling, media, codecs, routing, quality and security work together to deliver voice over IP.
VoIP is a system, not a single protocol. Reliable calling depends on coordinated signalling, media negotiation, packet transport, routing, security and operations.
The VoIP architecture
A VoIP endpoint converts audio into digital samples, encodes them with a codec, divides the encoded stream into packets and sends those packets across an IP network. Signalling locates the other party and negotiates the session; media usually travels on a separate path.
Signalling
SIP establishes, modifies and ends sessions. H.323 remains present in some legacy environments.
Description
SDP advertises media addresses, ports, codecs and direction attributes through offer/answer.
Media
RTP transports real-time audio; RTCP reports reception and quality information; SRTP protects media.
Transport
UDP, TCP, TLS and WebSocket can carry signalling. Media normally uses UDP because timeliness matters more than retransmission.
Interconnection
SBCs enforce policy, normalise signalling, control topology, secure borders and may anchor media.
Operations
DNS, routing policy, number translation, quality monitoring, fraud controls and resilient failover complete the service.
Core protocols
SIPSession control, registration, routing and feature signalling.
SDPMedia description: IP addresses, ports, codecs and session attributes.
RTPSequence-numbered and timestamped real-time media packets.
RTCPReception reports, loss, jitter, timing and participant information.
SRTPEncryption, integrity and replay protection for RTP/RTCP.
DNSNAPTR, SRV and address records support service and server discovery.
TLSConfidentiality and integrity for signalling between trusted hops.
ICE / STUN / TURNTechniques used to discover and establish media paths across NAT and firewalls.
Codecs and media
Codecs trade bandwidth, computational cost, resilience and perceived quality. G.711 provides narrowband PCM at 64 kbit/s before packet overhead; G.729 uses lower bitrate; Opus supports a wide range of bitrates and audio bandwidths. Packetisation interval, IP/UDP/RTP overhead and encryption must be included when calculating capacity.
Media planning
- Define allowed codecs and transcoding policy.
- Calculate real wire bandwidth, not codec bitrate alone.
- Keep packet loss, delay and jitter within the service design.
- Verify DTMF transport, commonly RFC 4733 telephone events.
- Test early media, hold, transfer and re-INVITE behaviour.
Quality and performance
Latency affects conversational responsiveness; jitter is variation in packet arrival time; loss creates gaps or concealment artefacts. Jitter buffers smooth variation but add delay. Operational teams commonly correlate RTP/RTCP data with ASR, ACD, PDD, SIP responses and route information. Subjective MOS is the mean of listener opinion scores. Operationally calculated or model-predicted MOS is an estimate of perceived quality and should be identified as such; neither replaces packet-level evidence.
Carrier interconnection
Production interconnection requires agreed IPs, transports, codecs, number formats, identity treatment, CLI policy, capacity, routing, failover and test cases. SBCs often provide demarcation. SIP ladder traces, CDRs and media captures should be correlated using accurate timestamps and stable call identifiers.
Security and resilience
Border control
Restrict peers, validate messages, rate-limit traffic and isolate trust zones.
Signalling protection
Use current TLS where appropriate and protect credentials and certificates.
Media protection
Use SRTP with suitable key management where confidentiality is required.
Fraud control
Monitor destinations, velocity, duration, identity, failed attempts and unusual traffic patterns.
Availability
Design redundant routes, SBCs, DNS, carriers, power and monitoring paths.
Observability
Retain privacy-aware signalling, quality and security telemetry for investigation.
Standards references: SIP (RFC 3261), SDP (RFC 8866), RTP/RTCP (RFC 3550) and SRTP (RFC 3711).
