How does the PSTN work?

The PSTN works by reserving a dedicated circuit for every call, then using a hierarchy of switches to connect that circuit across town or across the world. When you place a call, the network sets aside a path with enough capacity for your whole conversation and keeps it open until you hang up, which is the defining behavior of circuit switching.[3] That path runs from your telephone, over the local loop to your local exchange, then across shared trunks and higher-level switches to the exchange serving the person you called, and back down to their line.[1][2]

Two things happen in parallel to make that connection. The voice itself travels on a talk path, but the instructions that build and release the call travel separately on a control network called Signaling System No. 7.[7] Keeping the signaling on its own channels, out of band from the conversation, lets a switch pass the dialed number to the next switch, pick a route, ring the far end, and free the circuit at the end without stealing capacity from the call.[6]

A single network made of many

No single company runs the PSTN. It is the combined result of national, regional, and local carriers interconnecting their networks under shared technical standards, most of them written by the International Telecommunication Union, so a call placed on one carrier can reach a phone on another.[1] Those standards govern how numbers are formed, how voice becomes digital, and how switches talk to each other, which is what lets equipment from different vendors in different countries behave as one system.[9][8] The rest of this answer follows a single call through that system, from the moment you lift the handset to the moment the circuit is released.

A call begins the instant you lift the handset, which closes the circuit on your local loop and lets current flow from the exchange.[4] The exchange senses that off-hook condition, marks your line as active, and returns a dial tone to signal that it is ready to receive digits.[4] This first exchange, called a central office in North American usage and historically a Class 5 switch, is the piece of the network that terminates your individual line.[4] One practical detail of the traditional copper loop is that it draws its power from the exchange rather than from the home, which is why an old wired phone kept working during a local power cut. The dial tone does more than reassure you that the line is alive, because it also tells you the switch has a register ready to store the number you are about to send, so hearing it confirms the local exchange has capacity to handle your call at that moment.[4]

The digits you dial reach the exchange as address signaling, which is the process of telling the switch where you want the call to go.[5] On modern lines each key press generates a dual-tone multi-frequency tone, two audio frequencies played together, that the exchange decodes to identify the digit.[5] DTMF, trademarked as Touch-Tone and standardized in ITU-T Recommendation Q.23, replaced the older rotary method that signaled digits by briefly breaking the loop current a set number of times.[5] The dialed number itself follows the international numbering plan defined in ITU-T Recommendation E.164, which gives every line a unique, routable address built from a country code and national digits.[9] The exchange collects the full string of digits before it acts, because it needs the complete destination number to decide whether the call stays local or has to travel to another switch.[4]

Once it has the number, the local exchange decides how to reach the destination, and the answer depends on where that number lives.[1] If the called line sits on the same exchange, the switch connects the two loops directly.[4] If it does not, the call travels over shared trunks to higher tiers of the network, historically tandem switches for regional traffic and long-distance switches for calls between regions.[1] In the North American design these tiers are described as Class 5 switches at the local level and Class 4, or tandem, switches that interconnect central offices for longer-distance calls, so a call to someone on a different local switch may pass through one or more Class 4 switches before it reaches the far end.[11][1] The network is arranged as a hierarchy precisely so it does not need a direct wire between every pair of exchanges, which keeps local traffic local and reserves the costly long-haul circuits for calls that actually cross regions.[1] A switch chooses the route using the dialed digits and a routing plan, and the call climbs the hierarchy only as far as the distance requires before it comes back down to the destination exchange.[1]

The connection is built by a sequence of signaling messages that travel on their own network, separate from the voice path.[7] Signaling System No. 7, introduced in the ITU-T Q.700 series, is that control network, and its call-control part, the ISDN User Part or ISUP, carries the specific messages that set up and release a call.[7][6] The originating switch sends an Initial Address Message, which names the called number and seizes a circuit on the trunk to the next switch.[6] When the far exchange reaches the called line and it starts ringing, it returns an Address Complete Message, and once the person answers it sends an Answer Message, which is normally the point where billing starts.[6] Because this exchange of messages happens out of band, on a data network rather than on the talk path, the call sets up in well under a second and the voice channel carries only the conversation.[7] The same signaling backbone also supports services callers now expect, such as caller ID and number portability, because a switch can query shared databases before it completes a call.[6]

With the circuit connected, your voice travels as a digital signal that the network holds open for the whole call.[3] Near the edge the exchange converts the sound into numbers using pulse code modulation, the method defined in ITU-T Recommendation G.711, sampling the voice 8,000 times a second and coding each sample as eight bits, which produces one voice channel of 64 kbit/s known as a DS0.[8][1] That 64 kbit/s channel is the basic unit the core multiplexes onto shared circuits, so a T1 line combines 24 of these channels into 1.544 Mbit/s, while an E1 carries 32 timeslots at 2.048 Mbit/s, of which 30 carry voice.[10] Time-division multiplexing interleaves many calls onto the same trunk by giving each one a repeating slot, and higher up the network aggregates thousands of calls onto fiber using synchronous systems such as SONET and SDH.[10][1] The switching stays circuit based through all of this, so even a fully digital call holds a fixed slice of capacity from end to end rather than sharing it the way internet traffic does.[3]

The call ends when either party hangs up, which the exchange reads as an on-hook condition and treats as a request to tear the connection down.[6] The switch nearest the person who hung up sends a Release message across the SS7 network, and the switch at the other end answers with a Release Complete, after which the circuit is idle and available for the next call.[6] Freeing the circuit matters because circuit switching reserves that capacity for one conversation at a time, so the network can only reuse the path once the release handshake finishes.[3] Any billing record for the call is closed out at this point, using the answer and release times the signaling network recorded.[6] The whole teardown, like the setup, happens on the control network rather than on the talk path, so neither party hears it and the released trunk is ready for another call almost immediately.[7]

The PSTN's method is the opposite of how the internet moves data, and the contrast explains both its steadiness and its cost.[3] Circuit switching reserves a full channel for each call and keeps it reserved even during silences, which guarantees consistent capacity and steady voice quality from the moment the call connects.[3] Packet switching, the method behind Voice over IP, breaks speech into packets and sends them across links shared with other traffic, so it uses capacity only when there is something to send.[2] The reserved-circuit approach is part of why a traditional phone call has such predictable quality, and it is also why the network uses capacity less efficiently than a packet system, since a held circuit cannot be used by anyone else even when no one is speaking.[3] The two worlds connect through gateways, so a call can start on VoIP and finish on the PSTN, which is how the industry is gradually moving traffic off circuit switching and onto IP.[2]

The same design that makes the PSTN dependable also sets its limits, and naming them explains where it fits and where it is being replaced. Reserving a dedicated path gives very consistent call setup and voice quality, because the capacity is committed the instant the call connects rather than shared with other traffic.[3] The traditional copper loop also carried its own power from the exchange, so it usually kept working in a local blackout, a property that IP replacements have to recreate with backup batteries.[4] Holding a full circuit open for every call, including the quiet moments, uses network capacity less efficiently than packet switching, and the older switches were built mainly for voice, so adding features often meant changing the hardware itself.[2] For an organization still running lines on the PSTN, the practical point is that carriers are retiring this circuit-switched equipment in favor of IP, so the timing of a move is increasingly set by the carrier's schedule rather than left entirely to the customer.[2]