Electronics Guide

Articles in This Category

Overview of Traditional Communication Systems

Traditional communication systems span the technologies developed from the mid-nineteenth century through the late twentieth century for transmitting messages and voice across distances. The electric telegraph, commercialized in the 1840s, established the first electrical communication network, encoding text as patterns of current using Morse code. The telephone, patented by Alexander Graham Bell in 1876, added real-time voice, and within a few decades automatic switching, long-distance amplification, and multiplexing turned isolated local exchanges into continent-spanning networks.

Each generation built upon the previous one. Manual switchboards gave way to electromechanical step-by-step and crossbar exchanges, which in turn yielded to stored-program-controlled electronic switches and fully digital networks. Understanding this lineage provides essential context for modern telecommunications, because many concepts—dedicated end-to-end paths, standardized signaling, and graceful degradation under fault—were first solved in the telephone network and later adapted to data and wireless systems.

Three engineering constraints shaped nearly every decision in this field. Copper was expensive, so the network concentrated traffic aggressively and reused facilities wherever possible. Bandwidth was scarce, so voice was band-limited and, later, compressed. Service was regarded as essential, so equipment was designed for continuous operation over decades, powered independently of the commercial grid, and maintained in place rather than replaced. Those constraints explain why the telephone network looks conservative next to the data networks that succeeded it, and why so much of it remains in service.

The Analog Local Loop

The local loop, or subscriber line, is the pair of copper wires connecting a customer's premises to the nearest central office. The two conductors are traditionally called tip and ring, names inherited from the operator switchboard plug. In a conventional plain old telephone service (POTS) line, the central office supplies a nominal −48 volt direct-current battery across the pair. This voltage powers the telephone, which is why basic wired handsets continue to operate during a local power outage.

Supervision, Ringing, and Dialing

Signaling on the loop is elegantly simple. Lifting the handset closes a loop that draws direct current—typically 20 to 50 milliamperes, depending on loop length and the line circuit's current limiting—which the exchange detects as an off-hook condition; replacing the handset opens the loop. To alert the called party, the exchange superimposes a ringing signal of roughly 90 volts RMS at about 20 hertz on the line, applied in a cadence of two seconds on and four seconds off in North America. Because ringers load the line, each is rated with a ringer equivalence number, and a standard residential line supports a total of about 5.0 REN.

Dialing originally used loop-disconnect pulses, in which the rotary dial briefly interrupted the loop once per digit count at about ten pulses per second. Pulse dialing was largely replaced by dual-tone multi-frequency (DTMF) signaling, in which each digit is sent as a pair of audio tones drawn from a low group of 697, 770, 852, and 941 hertz and a high group of 1209, 1336, 1477, and 1633 hertz. The frequencies were chosen so that no tone is a harmonic or simple sum of the others, which makes speech far less likely to imitate a digit. DTMF is faster than pulse dialing and, because it passes through the voice path, remains usable end to end for menu navigation and remote control long after call setup.

Transmission Limits and Loop Engineering

The usable voice band is limited to approximately 300 to 3,400 hertz, a deliberate compromise that preserves intelligibility while conserving bandwidth. Restricting the band costs some naturalness—the low end truncates vocal fundamentals, and the high end blunts the distinction between fricatives such as "s" and "f"—but it permits dense multiplexing and tolerates the loss characteristics of long copper pairs.

Loop length is bounded chiefly by direct-current resistance, which must stay low enough for reliable supervision and adequate transmit current. Design practice caps loop resistance at roughly 1,300 ohms, which corresponds to a few kilometers of ordinary 24 or 26 AWG pair. Longer loops were historically treated with loading coils—inductors of about 88 millihenries inserted at regular intervals of some 1.8 kilometers—which flatten the frequency response across the voice band at the cost of attenuating everything above it. That trade-off returned as a problem decades later, because loading coils must be removed before a pair can carry digital subscriber line service.

Two-Wire to Four-Wire Conversion and Echo

A single pair of wires carries both directions of conversation simultaneously, so the telephone uses a hybrid—originally a transformer, later an electronic circuit—to separate transmitted from received speech. The same circuit deliberately leaks a small amount of the talker's own voice into the earpiece as sidetone, without which a handset sounds dead and speakers raise their voices.

Where the network converts this two-wire loop to the separate four-wire transmit and receive paths of long-distance trunks, imperfect hybrid balance reflects a portion of the signal back as an echo. Echo becomes objectionable as round-trip delay grows, which is why carrier-grade systems employ echo cancellers—standardized in ITU-T G.168—on long terrestrial, satellite, and packet circuits. The related delay budget is equally well established: ITU-T G.114 recommends keeping one-way mouth-to-ear delay below about 150 milliseconds for ordinary conversation, a figure that packet voice designers inherited directly from telephony practice.

Circuit Switching and Network Architecture

Traditional telephony relies on circuit switching: for the duration of a call, the network establishes a dedicated end-to-end path of fixed bandwidth between the two parties. The path is reserved at call setup, held for the entire conversation, and released at teardown. This guarantees constant low latency and predictable quality, at the cost of holding resources even during silence—a trade-off later inverted by packet switching.

The Switching Hierarchy

Switching offices are organized in a hierarchy. Local, or class-5, switches connect to subscribers and terminate the local loop. Tandem and toll switches interconnect local offices and route long-distance traffic; the classic Bell System arrangement ranked offices from the class-5 end office up through toll, primary, sectional, and regional centers, each level aggregating traffic from the level below and providing alternate routes when direct trunk groups filled. Interoffice trunks carry many simultaneous calls between switches, historically over multiplexed copper, microwave radio, and, increasingly, optical fiber. This layered structure allowed the network to scale to hundreds of millions of subscribers while keeping most traffic local.

Traffic Engineering

A telephone network is deliberately undersized. Subscriber lines are concentrated onto a smaller number of switch paths, and trunk groups are sized for busy-hour demand rather than for the theoretical maximum. Engineers size these groups with traffic theory, measuring offered load in erlangs—one erlang is one circuit occupied continuously for the observation period—and applying the Erlang B formula to relate the number of circuits to the probability that a call finds every circuit busy. A common target is a grade of service of one percent blocking in the busy hour. The economics are strongly nonlinear: large trunk groups carry proportionally more traffic per circuit than small ones, which is one reason the network aggregated traffic into hierarchies rather than meshing every office directly.

Switch Fabrics

Inside the switch, the connection is made through a fabric of crosspoints. Early fabrics blocked, meaning that some idle inlet and outlet pairs could not be joined because no free path existed between them. Charles Clos published the classic analysis of multistage networks in 1953, showing how a three-stage design can be made strictly nonblocking with far fewer crosspoints than a single large matrix would require. Digital exchanges implement the same idea in time as well as space, alternating time-slot interchange stages—which move a sample from one time slot to another in memory—with space stages that move it from one bus to another, in the time-space-time arrangement that characterizes digital switching.

From Electromechanical to Digital Switching

The first generation of automatic switching used the Strowger step-by-step system, with the earliest public automatic exchange opening in La Porte, Indiana, in 1892. Electromechanical selectors physically stepped through contacts in direct response to dial pulses, building a connection one digit at a time. The design was ingenious but rigid: the dial drove the switch train directly, so the numbering plan was effectively wired into the hardware, and every call consumed a chain of noisy mechanical selectors.

Crossbar switching, with the first Bell System number 1 crossbar installed in 1938, improved reliability and speed by separating the control logic from the matrix of crosspoints that formed the actual voice path. Common control equipment—markers, senders, and registers—collected the dialed digits, chose a path through the matrix, and then released itself for the next call. Because control was no longer tied to the talking path, exchanges could translate numbers, choose alternate routes, and record billing data in ways step-by-step offices never could.

A decisive shift came with stored-program control. The Western Electric 1ESS, placed into service in Succasunna, New Jersey, in 1965, used a computer with programs held in memory to manage call processing, while the talking path still ran through sealed reed-contact crosspoints. Because behavior was defined in software rather than wiring, operators could introduce features such as call waiting, call forwarding, and speed dialing without rebuilding hardware. Subsequent switches carried calls as digital bit streams internally: the 4ESS entered service as a digital toll switch in 1976, and the 5ESS, introduced in 1982, became a widely deployed digital local exchange with a distributed architecture that placed switching modules close to subscribers. Comparable systems appeared worldwide, among them the Nortel DMS family, the Ericsson AXE, and the Alcatel E10.

The engineering discipline these machines established is as significant as the hardware. Duplicated processors ran in step so that a fault could be isolated without dropping calls, software was patched on live systems, and maintenance routines continuously audited the fabric for faults. Those practices set the expectation of continuous availability that later networks have struggled to match.

Voice Digitization and Transmission

Carrying analog voice over digital trunks requires sampling and quantization. The standard codec, ITU-T G.711, samples the voice band at 8,000 samples per second and encodes each sample with 8 bits using logarithmic companding, producing a 64 kilobit-per-second channel known as DS0. Two companding laws are used: μ-law in North America and Japan, and A-law elsewhere. The 8 kilohertz sampling rate satisfies the Nyquist criterion for the 3.4 kilohertz voice band with margin for filtering.

Companding is what makes 8 bits sufficient. A linear 8-bit coder would sound coarse on quiet passages, because its quantization steps are uniform while speech spends most of its time at low amplitude. Compressing the sample logarithmically before quantization makes the step size proportional to signal level, so the signal-to-quantization-noise ratio stays roughly constant across a wide dynamic range and approaches what a linear coder of about 12 bits would deliver. The cost is that companded samples must be expanded before any arithmetic is performed on them, which is why conference bridges and echo cancellers convert to linear internally.

The Digital Carrier Hierarchy

Individual DS0 channels are combined by time-division multiplexing into standard carrier hierarchies. The North American T1 system aggregates 24 voice channels into a 1.544 megabit-per-second stream—24 channels of 64 kilobits per second plus 8 kilobits per second of framing—while the European E1 system fills 32 time slots at 2.048 megabits per second, of which 30 carry voice, one carries framing, and one carries signaling. Higher orders followed the same pattern: 28 DS1 signals combine into a 44.736 megabit-per-second DS3, and the SONET and SDH optical hierarchies begin at 155.52 megabits per second for OC-3 and STM-1. These digital carriers formed the backbone over which the digitized telephone network ran for decades and still underlie much carrier infrastructure today.

The North American and European systems also differ in how they carry per-call supervision. T1 systems traditionally used robbed-bit signaling, borrowing the least significant bit of each channel in designated frames to convey on-hook and off-hook state; the practice is inaudible in speech but reduces a data channel to 56 kilobits per second, which is why leased digital services were long quoted at that rate. E1 instead reserves time slot 16 for channel-associated signaling, leaving all 30 voice channels clear at the full 64 kilobits per second.

Network Synchronization

Time-division multiplexing only works if every node agrees on the clock. A receiver that runs slightly fast or slow relative to the transmitter eventually reads a frame twice or skips one, producing an audible click and, on data circuits, a burst of errors. Carriers therefore distribute timing through a stratum hierarchy, in which a stratum 1 primary reference source holds an accuracy on the order of one part in 10¹¹ and lower strata lock to it, holding over gracefully if the reference is lost. Synchronization remains a live concern in packet networks, where mobile backhaul and circuit emulation must recover an equivalent timing reference without a continuous bit stream to lock to.

Signaling Systems

Signaling is the exchange of control information that sets up, manages, and tears down calls. Early networks used in-band signaling, sending supervisory and addressing tones—single-frequency and multi-frequency signals—over the same voice path used for conversation. This was simple but inefficient, tied up voice circuits during setup, and proved vulnerable to fraud: because the North American network used a 2,600 hertz tone to mark a trunk idle, a subscriber who reproduced that tone could seize a trunk and then dial it directly with multi-frequency signals, the technique behind the "blue box" fraud of the 1960s and 1970s.

Common-Channel Signaling and SS7

Modern networks adopted common-channel signaling, separating control onto a dedicated data network. Signaling System No. 7 (SS7) was first approved by the CCITT, the ITU-T's predecessor, in its 1980 Yellow Book recommendations and substantially expanded in the 1988 Blue Book, becoming the global standard for out-of-band signaling. By carrying call control on packet links independent of the voice trunks, SS7 enabled faster call setup, more efficient trunk use, protection against tone-based fraud, and advanced services such as caller identification, toll-free number translation, local number portability, and short message service in mobile networks.

An SS7 network is built from three node types. Service switching points are the exchanges that originate and terminate signaling for calls; signal transfer points route signaling messages, always deployed in mated pairs so that no single failure isolates an office; and service control points hold the databases that answer queries. The protocol stack layers accordingly: message transfer part levels 1 through 3 provide the physical links, error-checked link operation, and message routing; the signaling connection control part adds addressing flexibility; ISDN user part carries the call-control messages that set up and release trunks; and transaction capabilities application part carries the database transactions used for number translation and mobile roaming.

The separation of signaling from media produced an unintended consequence that still matters. SS7 was designed for a closed community of trusted carriers and includes little authentication, so as interconnection widened, researchers and attackers demonstrated subscriber location tracking, call interception, and one-time-password theft by injecting messages into the signaling network. Operators now screen signaling traffic at network borders, and this experience directly informed the stronger security posture of newer interconnection protocols.

Access Signaling and the Move to IP

On the subscriber side, ISDN introduced message-based signaling to the access line. Q.921 provides a reliable data link on the D channel and Q.931 carries the call-control messages, giving customer equipment the same descriptive control the network used internally—the origin of features such as calling-name delivery and the immediate call setup that ISDN users noticed at once. Private branch exchanges networked with each other used the related QSIG protocol to carry features between sites.

As carriers moved control onto IP, the SIGTRAN family transported SS7 messages over IP using the Stream Control Transmission Protocol, preserving the application layers while replacing the physical links. Signaling gateways interwork between the two worlds, which is how a call originating on a copper loop can terminate on a SIP endpoint with the caller identification intact.

Numbering, Routing, and Regulation

A global telephone network requires a single, unambiguous addressing scheme. ITU-T Recommendation E.164 defines it: an international number consists of a country code followed by a national number, with a maximum of 15 digits. Country code 1 covers the North American Numbering Plan, in which a ten-digit number divides into a three-digit area code, a three-digit exchange code, and a four-digit line number. The structure was originally geographic and hierarchical by design, because early switches routed on leading digits and had little capacity to look anything up.

Two developments broke that assumption. Toll-free service required the dialed digits to be translated into a routable number by a database query rather than interpreted positionally, which is why 800-series numbers were among the first users of SS7 database transactions. Local number portability went further, severing the link between a number and a switch entirely: an office must now query a portability database to learn where a ported number actually terminates. The number became an identifier rather than an address, a shift that VoIP later depended on completely.

Regulation shaped the network as firmly as engineering did. Universal service obligations required carriers to serve high-cost rural areas, interconnection rules governed how competitors exchanged traffic, and tariffs determined which services existed at all. That regulatory layer explains much of the network's shape—and, as the following sections describe, it also governs the pace at which the copper network may now be retired.

Digital Access and Voiceband Data

Long before packet voice, the copper loop was pressed into service for data. Voiceband modems encoded data into the 300 to 3,400 hertz channel, evolving from a few hundred bits per second to the 33.6 kilobits per second of ITU-T V.34. The V.90 and V.92 recommendations reached 56 kilobits per second downstream by exploiting an insight about the network rather than the loop: where the path from the internet service provider is digital end to end, the downstream direction suffers only one analog conversion, so the modem can decode the codec's quantization levels directly instead of demodulating a waveform. Group 3 facsimile used the same channel under the ITU-T T.30 protocol, and its persistence in medicine, law, and government is a large part of why analog lines have proved so hard to retire; the T.38 recommendation exists specifically to relay fax through IP networks without the errors that speech codecs introduce.

The Integrated Services Digital Network digitized the loop itself. A basic rate interface carries two 64 kilobit-per-second bearer channels and one 16 kilobit-per-second signaling channel over an ordinary pair, using 2B1Q line coding in North America. A primary rate interface fills a T1 with 23 bearer channels and one signaling channel, or an E1 with 30 bearer channels and one signaling channel, and became the standard way to connect a private branch exchange to the public network. ISDN saw uneven adoption—strong in Germany and Japan, weak in the United States, where it arrived expensive and late—but its signaling model and its role in enterprise trunking outlived its subscriber base.

Digital subscriber line technology then abandoned the voice band's constraints altogether, placing data carriers in the spectrum above 4 kilohertz so that a splitter can separate them from an ordinary telephone call on the same pair. Asymmetric DSL exploited the fact that consumers download more than they upload; VDSL2 pushed spectrum use to 17 or 35 megahertz and reached on the order of a hundred megabits per second over short loops, typically from a fiber-fed cabinet rather than the central office. Carriers had already begun shortening loops for other reasons: digital loop carrier systems placed remote terminals in neighborhoods and backhauled digitized voice over a few fiber or T1 facilities, which reduced copper plant but also placed equipment between the subscriber and the exchange that DSL and analog services had to accommodate.

Reliability, Powering, and Emergency Service

The telephone network set a standard of availability that is still quoted as an aspiration elsewhere: five nines, or 99.999 percent, which permits roughly five minutes of outage per year. That figure was achieved by design rather than by luck. Central offices run from a −48 volt battery plant float-charged by rectifiers and backed by engine generators, so a switch continues to operate through a commercial power failure. Processors are duplicated and run in synchronism, circuit packs are replaceable while the system is live, and diagnostic routines audit the fabric continuously. Sizing every element for the busy hour, then adding margin, produced systems that failed gradually and visibly rather than suddenly.

Because the loop is powered from the exchange, a basic analog telephone works during a neighborhood blackout. That property underpins emergency calling. In North America, enhanced 911 service ties the calling number to a location record: the switch passes automatic number identification to a selective router, which uses it to choose the correct public safety answering point and to retrieve the subscriber's civic address from an automatic location identification database. The scheme depends on the fixed association between a copper pair and an address—precisely the assumption that mobile and IP telephony break.

Replacements must therefore reproduce more than the conversation. Next generation 911 architectures replace the analog trunks with IP emergency services networks that route calls on location information supplied by the originating device or provider, and they can accept text, images, and telematics data that the analog system could not carry. Powering is the harder problem: an IP telephone, a modem, and a residential gateway all draw local power, so continuity of service during an outage now depends on customer-premises batteries rather than on the central office plant. Alarm panels, elevator telephones, medical alert devices, and unattended monitoring circuits raise the same question, and migrating them safely has consistently proved the slowest part of retiring copper.

The Transition to Packet Networks

Beginning in the late twentieth century, voice traffic migrated from dedicated circuits to packet-switched networks. Voice over IP digitizes speech, compresses it with codecs, and transports it as IP packets, using session protocols such as SIP for call control. Packet transport uses bandwidth far more efficiently than circuit switching, because no resources are held during silence and the same network carries voice, data, and video together. In exchange, packet voice must actively manage the delay, jitter, and loss that a dedicated circuit eliminated by construction, which is why quality-of-service marking, jitter buffers, and packet-loss concealment are core parts of every VoIP design.

This convergence did not discard traditional telephony so much as absorb it. Media gateways translate between the PSTN and IP domains under the control of protocols such as H.248, signaling gateways bridge SS7 to IP-based control, and session border controllers police the boundary between carriers. Enterprises replaced primary rate interfaces with SIP trunks, usually keeping their dial plans and numbers unchanged. Even the codec of choice, G.711, is the same 64 kilobit-per-second encoding the digital telephone network standardized decades earlier.

Retirement of the legacy network is now well advanced and is driven as much by regulation as by technology. Deutsche Telekom moved essentially its entire German customer base to IP by the end of the 2010s. In the United Kingdom, Openreach is withdrawing the public switched telephone network on 31 January 2027, a date deferred from December 2025 largely to protect telecare and other vulnerable users, with legacy line rental repriced to encourage migration. In the United States, the Federal Communications Commission adopted an order in March 2026 that streamlines copper retirement and service discontinuance, replacing earlier replacement-service tests with a single technology-transition rule and granting carriers blanket authority to grandfather legacy copper voice and lower-speed data services. Fixed-telephone subscriptions worldwide have meanwhile declined by roughly three percent a year, according to International Telecommunication Union figures.

Even so, interoperability with the installed base—copper loops, analog handsets, fax machines, alarm panels, and emergency-calling requirements—keeps the conventions of traditional telephony relevant in the design of contemporary systems. Engineers commissioning a modern voice service still specify echo control, still reason about one-way delay in milliseconds, and still test against a numbering plan and a signaling model that were settled before the internet carried a single call.

Conclusion

Telephony and traditional communications form the foundation of modern telecommunications, establishing principles and infrastructure that continue to influence network design. The dedicated path of circuit switching, the discipline of structured signaling, the digitization of voice into standard bit rates, and the relentless pursuit of carrier-grade reliability all originated here. Their assumptions are visible throughout current practice, from the delay budgets applied to packet voice to the availability targets set for data centers.

For engineers, technicians, and students, understanding traditional telephony supplies essential historical context and reveals fundamental principles that endure across successive generations of communication technology. It also explains the work that remains: as copper is retired network by network, the obligation to deliver a call that always completes, always locates the caller, and always survives a power failure passes intact to the systems that replace it.

Related Topics