Voice over IP (VoIP) is the technology behind most business phone systems today, yet it is often described in vague terms. This guide breaks down what actually happens when you speak into a VoIP handset or softphone, which protocols do the work, and why the traditional telephone network has not disappeared.
What VoIP actually is
Traditional telephony — the public switched telephone network, or PSTN — was built around circuit switching. When you placed a call, the network reserved a dedicated path between the two endpoints for the duration of the conversation. That path was yours whether you were speaking or silent.
VoIP replaces that model with packet switching. Your voice is digitised, compressed, split into small packets, and sent across an IP network alongside every other kind of traffic. There is no reserved circuit; packets find their way independently and are reassembled at the other end. The result is far more efficient use of bandwidth and a phone system that lives on the same infrastructure as your email, file transfers and video calls.
The journey of a single call
- Capture and digitisation. A microphone captures analogue sound waves; an analogue-to-digital converter samples them into a digital stream.
- Encoding. A codec compresses that stream. Common examples include G.711 (uncompressed-quality, higher bandwidth), G.729 (heavily compressed, lower bandwidth) and Opus (adaptive, widely used in modern web clients).
- Packetisation. The encoded audio is chopped into small chunks and wrapped in RTP (Real-time Transport Protocol) packets, which carry sequence numbers and timestamps so the receiver can reorder them.
- Transport. Packets travel over UDP rather than TCP. UDP does not retransmit lost packets — for real-time voice, a late packet is useless anyway, so speed beats reliability.
- Reassembly and playback. A jitter buffer at the receiving end holds packets briefly, reorders them, conceals small losses, and feeds a steady stream to the speaker.
Signalling: SIP and friends
RTP carries the audio, but something has to set the call up first. That is the job of signalling protocols. SIP (Session Initiation Protocol) is the dominant one: it handles registration, ringing, answering, holding, transferring and hanging up. SIP negotiates which codecs both sides support using SDP (Session Description Protocol), then steps back and lets RTP carry the media.
Other pieces you may encounter:
- SIP trunking — replacing physical ISDN/PRI lines with an IP connection to a provider that terminates calls to the PSTN.
- WebRTC — the browser-native stack for voice and video, used by most in-browser calling widgets.
- SRTP and TLS — encryption for the media and signalling paths respectively.
- STUN/TURN/ICE — NAT traversal helpers, because most endpoints sit behind routers that hide their real addresses.
What makes VoIP sound bad
Because VoIP shares a network with everything else, its quality depends on network conditions rather than a dedicated circuit. Four metrics matter:
- Latency — one-way delay. Beyond roughly a couple of hundred milliseconds, conversation starts to feel like a walkie-talkie exchange.
- Jitter — variation in packet arrival times. Buffers absorb some of it; heavy jitter causes choppy audio.
- Packet loss — missing audio. Modern codecs conceal small losses well; sustained loss produces robotic or clipped speech.
- Bandwidth contention — a large upload can starve voice traffic unless QoS rules prioritise it.
Practical fixes are usually mundane: wired connections instead of congested Wi-Fi, QoS marking on the router, adequate upstream bandwidth, and keeping the media path as short as possible.
Where traditional telephony still fits
Legacy telephony has not vanished, and there are reasons for that. Copper lines draw power from the exchange, so a basic analogue phone works during a power cut; a VoIP handset does not unless you have backup power and internet. Emergency calling and precise location reporting are also more complicated over IP, and regulators handle it differently per country.
Mobile networks add another layer: VoLTE and VoNR are, technically, VoIP running inside the operator's own IMS core with guaranteed quality of service — the packet-switched model applied with carrier-grade control. Meanwhile, the last mile to many landline subscribers is still analogue, and gateways bridge the two worlds. In practice, most calls today cross both packet and circuit domains at some point.
Choosing between them
For most businesses the decision is not "VoIP or PSTN" but how much of the estate to migrate and how fast. VoIP wins on cost per minute for long distance, on flexibility (numbers are not tied to a location), and on integration — call data, CRM hooks and recording are far easier when everything is software. Traditional lines are worth keeping for alarm panels, lift phones, payment terminals and any site where internet reliability is genuinely uncertain.
Where automated voice messaging comes in
Not every voice use case is a two-way conversation. Appointment reminders, delivery notifications and payment alerts are one-way messages, and they are typically delivered as automated outbound calls rather than staffed conversations. UIPAPP's IVR voice calling feature works in that space alongside SMS: you write the message text in the panel, it is converted to natural speech, or you upload your own MP3, then upload a number list and review who answered and how long they listened. The feature is disabled by default on new accounts — support enables it and defines per-country call pricing. The underlying delivery still relies on the same voice infrastructure described above, whichever side of the IP/PSTN boundary the recipient sits on.
Understanding the packet path makes troubleshooting far less mysterious: when a call sounds wrong, the answer is almost always somewhere between the codec, the network and the buffer.