A spark transmitter made its signal by discharging a capacitor across an air gap hundreds of times a second, and every discharge spread energy across a wide slice of the spectrum. Early wireless stations jammed each other because the transmitters were built that way and the receivers could not sort one signal from another. This piece covers the physics of the damped wave, how stations crowded each other off the air, how tuning, the 1912 and 1927 rules and the vacuum tube moved radio to continuous wave, why spark stayed aboard ships for years afterward, and what the episode says about spectrum occupancy today.

One gap, one burst: how a spark transmitter made its signal

A spark transmitter had very few parts. A power source, either an induction coil running from batteries or a transformer fed by an alternator, charged a capacitor (in the earliest sets, a bank of Leyden jars) until the voltage across a pair of electrodes was high enough to break down the air between them. At that instant the spark became a conductor, and the capacitor discharged through a coil of wire. The capacitor and the coil formed a resonant circuit, so the current rang back and forth at a frequency set by their values, and that ringing current drove the antenna. Each oscillation was weaker than the one before, and within a short time the energy was gone, the spark went out, and the capacitor began charging again for the next discharge. A working set repeated this cycle hundreds of times a second for as long as the operator held the key down.

The gap did several jobs at once in that design. It served as the switch that released the stored energy, it set off the oscillation, and its firing rate determined the tone the receiving operator heard. The operator’s key simply turned the power supply on and off, so every dot and dash was a string of separate decaying bursts. Nothing in the circuit shaped those bursts or limited where their energy landed in frequency, because nobody had yet built a component that could do that at useful power.

Marconi’s earliest practical sets in the late 1890s put the spark gap directly in the antenna circuit, with the antenna and its ground acting as the only capacitance. That arrangement dumped its energy into the air within a very few cycles. The receivers of the day matched that behavior. The coherer, a glass tube of metal filings that clumped together and conducted when a burst of radio energy arrived, responded to any sufficiently strong impulse and had little ability to tell one frequency from another. The system worked as a whole precisely because it did not depend on frequency. A sharp, strong burst at the transmitter produced a response at the receiver, and in 1898 that was the design goal. The difficulty that followed came from the fact that every other station worked the same way.

Damping and decrement: why a spark signal was tens of kilohertz wide

Engineers of the period measured how fast a spark signal died out with a figure called the logarithmic decrement. It is the natural logarithm of the ratio between one oscillation’s amplitude and the next one’s. A small decrement means the wave train rings for many cycles before fading, while a large decrement means it is gone within a few. The decrement is therefore a direct measure of damping, and early radio regulation was eventually written around it.

The connection to bandwidth is basic physics. A signal that lasts only a short time cannot sit at one frequency, because its energy spreads across a band whose width varies inversely with the duration of the signal. For a damped wave, a reasonable approximation puts the half-power bandwidth at roughly the decrement times the operating frequency, divided by pi. I worked that arithmetic for a set at 500 kHz with a decrement of 0.2. The result is a half-power width of about 32 kHz, and a damped wave’s spectrum falls off slowly beyond the half-power points, so meaningful energy reaches well past that figure on both sides. A plain-aerial set with heavier damping spread considerably farther. For comparison, a narrowband analog FM channel in land mobile service today is 12.5 kHz wide, and a hand-keyed Morse signal from a clean continuous-wave transmitter needs on the order of a hundred hertz.

The damping came from several sources. Resistance in the gap and the wiring wasted energy as heat, which engineers worked to reduce. Radiation itself also removes energy from the oscillating circuit, which means that an antenna that radiates well also damps quickly. In a simple spark set, the goal of putting a strong signal into the air and the goal of keeping that signal narrow worked against each other, so a designer could improve one only by giving up some of the other. The coupled circuits described below were the first serious attempt to get around that tradeoff, and they succeeded only in part.

Many stations, one band: how early wireless jammed itself

For the first decade of the twentieth century, American wireless had no license requirement and no assigned wavelengths. Rival commercial companies, the Navy and a fast-growing number of amateurs all built stations, and many of them ran at whatever wavelength their antenna and capacitor happened to produce. Coastal areas such as New York harbor had several stations within a few miles of one another. A strong spark signal nearby occupied most of what a broadly tuned receiver could hear, so one station sending meant everyone else in the area waited. Operators who wanted to work distant ships could find a local station making that impossible for as long as its key was down.

Commercial rivalry made the situation worse. The America’s Cup yacht races off New York in 1901 are commonly cited as an early example, with competing wireless companies trying to report the races to shore and interfering with one another so badly that much of the traffic was lost. Accounts disagree on which company jammed which, and on how much of the interference was deliberate rather than the unavoidable result of several spark sets operating close together. I treat the episode as an illustration of what broadband transmitters in close quarters did to each other and leave the assignment of blame to the histories that have worked through the newspaper record.

In the absence of rules, operators relied on practice and on their own ears. A trained operator could pick one spark station out of several by its tone, since sets with different supply frequencies and gap arrangements produced notes that sounded distinctly different through headphones. Stations in a busy area also took turns by informal agreement, waiting for a break before calling. That system depended on courtesy, and it failed whenever one operator did not cooperate, whenever a distress call needed the air, or whenever a high-powered station keyed up near smaller ones. Navy complaints about amateur and commercial interference became part of the case for federal licensing well before 1912, and the 1910 Wireless Ship Act had already required certain passenger ships to carry radio without saying anything about how stations should share the air.

Coupled circuits and the quenched gap: tuning that narrowed spark without curing it

The first engineering response was tuning, which the period called syntony. Oliver Lodge patented a syntonic arrangement in Britain in 1897, Ferdinand Braun developed coupled transmitter circuits around 1898 and 1899, and Marconi’s British patent 7777 of 1900 covered tuned circuits at both the transmitter and the receiver. The core idea in the transmitter was to move the spark gap out of the antenna and into a separate closed circuit with a large capacitor. That closed circuit stored more energy and lost less of it to radiation, so it rang with a lower decrement. It was coupled to the antenna through a transformer, and the antenna was tuned to the same frequency. Tuned receivers could then favor one wavelength and reject others to some degree.

Coupled circuits brought their own complication. When the closed circuit and the antenna were coupled tightly, energy moved back and forth between them during each spark, and the antenna radiated on two frequencies, one above and one below the intended wavelength. Loose coupling produced a single, narrower emission but transferred less power into the antenna. Operators and inspectors used wavemeters to check where a set was actually radiating, and adjusting the coupling became a routine part of keeping a station in order.

The quenched spark gap, generally credited to Max Wien around 1906 and widely built by the German firm Telefunken, addressed the coupling problem directly. A quenched gap consisted of a stack of short gaps between cooled metal plates, and it extinguished the spark in the closed circuit almost as soon as the energy had passed into the antenna. The antenna then rang by itself, at its own frequency and with its own lower damping, without energy flowing back into the primary. Rotary gaps, and supplies running at several hundred cycles per second, raised the spark rate so that the received signal sounded like a clear musical note rather than a rasp. Many ship installations of the 1910s used rotary or quenched gaps fed from motor-alternators for exactly that reason.

All of these improvements reduced the decrement, but none of them removed damping, because the antenna’s own losses and its radiation always set a floor. A well-adjusted quenched-gap set was far easier to share the air with than a 1900 plain-aerial set, though it still sent a series of decaying wave trains, and those trains still spread across many kilohertz.

1912 and 1927: what the statutes and conventions did about damped waves

The Radio Act of 1912, passed in August of that year in the months after the Titanic, required American stations to be licensed by the Department of Commerce and tied ship traffic to set wavelengths, with 600 meters (500 kHz) as the standard ship wavelength and 300 meters as an alternative. Amateurs were confined to 200 meters and below, which was then considered of little commercial value. The act’s regulations also limited damping directly. As I read the statute, the logarithmic decrement per complete oscillation of the emitted wave trains could not exceed two-tenths, with an exception for distress traffic, and anyone who needs that figure for a citation should check the statute text rather than rely on my summary. The 1912 London international convention addressed ship wavelengths and operating practice at the international level during the same year.

The decrement limit shows that regulators had grasped the physics, since it controlled how quickly a wave died away and therefore, indirectly, how wide the signal was. It did not require continuous wave, and it did not ban spark. My arithmetic in the bandwidth section shows that a set meeting the two-tenths limit at 500 kHz still occupied tens of kilohertz. The 1912 framework made spark more orderly and assigned it wavelengths to share, and the companies kept building spark sets for years under it. The existing article on this site about the 1912, 1927 and 1934 laws covers the licensing side of the act in more depth than I do here.

By 1927 the main interference problem in the United States was no longer spark. It was broadcasting. Federal court decisions, the 1923 Intercity Radio case and the 1926 Zenith case among them, had undercut the Commerce Department’s authority to assign frequencies and power, and broadcast stations, most of them using continuous-wave carriers, moved and multiplied until reception in many cities became unreliable. The Radio Act of 1927 created the Federal Radio Commission and gave it authority over the apparatus stations used, including the “purity and sharpness” of their emissions, and that phrase carried over into the Communications Act of 1934. The International Radiotelegraph Convention signed in Washington in November 1927 dealt with damped waves explicitly. It treated continuous and damped emissions as separate types and set restrictions and phase-out dates for damped-wave transmitters, and ship stations received the longest allowance. I have not reproduced those dates here because I am not certain of each of them, and the convention text in the ITU’s historical archive is the place to check.

Arc, alternator and tube: the ways to make a continuous wave

A continuous-wave transmitter produces a steady carrier at one frequency and sends Morse by starting and stopping it, or by shifting it. The first practical source at high power was the arc converter developed by the Danish engineer Valdemar Poulsen around 1903. It used an electric arc in a hydrogen-rich atmosphere inside a strong magnetic field to sustain oscillation in a tuned circuit. The Federal Telegraph Company built arcs in the United States, and the Navy bought them for long-haul stations during the 1910s. An arc could not easily be started and stopped at keying speed, so many arc stations ran continuously and keyed by shifting frequency, which meant the station emitted a signal on two frequencies the whole time it was on the air. Arcs also produced strong harmonics, so an arc station sat narrower in frequency than a spark station while still producing interference problems of its own.

The second source was the high-frequency alternator, a rotating machine built to generate radio-frequency current directly. Reginald Fessenden pushed for such a machine, and Ernst Alexanderson at General Electric developed the large alternators used for transoceanic long-wave traffic in the late 1910s. The station at Grimeton in Sweden still has a working Alexanderson alternator, is listed as a UNESCO World Heritage site, and transmits on 17.2 kHz on scheduled occasions. Fessenden is also credited with voice transmissions from Brant Rock, Massachusetts, in 1906. The well-known story of a Christmas Eve 1906 broadcast of music and speech to ships, however, rests largely on later recollection, and historians dispute its details, so I treat it as a commonly told account rather than an established event.

The vacuum tube eventually displaced both. John Ambrose Fleming’s diode of 1904 and Lee de Forest’s three-element Audion of 1906 were first used as detectors. Around 1912 to 1914 several inventors, Edwin Armstrong among them, found that feeding part of a tube’s output back to its input made it oscillate, and the priority fight over that discovery ran through the courts into the 1930s. A tube oscillator could produce a clean continuous wave at moderate power from equipment far smaller and cheaper than an arc or an alternator. Quartz crystal control, developed during the 1920s, then held that frequency steady enough for close channel assignments to work.

The receiving side had to change as well. A continuous-wave signal turned on and off carries no audible tone, so a simple crystal detector produces only faint thumps at the start and end of each character. Reception needed a heterodyne, a local oscillator in the receiver mixed with the incoming carrier to produce an audible beat note. Fessenden had patented the heterodyne principle years earlier, but a tube oscillator in the receiver made it practical. Moving to continuous wave therefore required replacing the receivers along with the transmitters, which goes a long way toward explaining why the change took as long as it did.

Spark at sea into the 1930s: why the old sets stayed aboard

Ship stations held onto spark well after shore stations and amateurs had moved on. A marine spark installation was a substantial and durable piece of equipment that had been paid for, licensed and inspected, and the operators trained on it knew how to keep it running at sea. There was little reason to tear out a working set while the rules still allowed it, and the international dates gave ships more time than any other service. Spark also had a practical advantage for distress work at 500 kHz, since any receiver could hear it, including an old crystal set with no beat oscillator, and that made it attractive as a reliable emergency transmitter.

Spark therefore stayed in maritime service through the 1920s, and on some vessels and in some emergency installations into the 1930s, while new installations increasingly used tubes. Battery-powered spark sets driven by induction coils remained in use as emergency transmitters because they needed nothing more than a charged battery. I do not give a date for the last spark transmission at sea, because I have not seen one documented well enough to stand behind. Anyone researching a particular vessel should go to the ship’s registry entries, the period’s government lists of radio stations, and maritime radio archives.

Amateurs changed over faster, largely because tubes became cheaper and easier to get in the early and middle 1920s and because the shorter wavelengths they had been pushed onto suited tube transmitters well. The ARRL’s own history of its December 1921 transatlantic tests credits continuous-wave tube stations with much of the success in getting American amateur signals heard in Britain, and anyone wanting the station-by-station details should look there. As an Amateur Extra licensee I have read a good deal of that history, but I have not found a single date for the end of amateur spark in the United States that I would cite, and I would describe it as having largely disappeared by the late 1920s.

The full transition across all services took roughly two decades. The pace was set by equipment life, the cost of replacing receivers as well as transmitters, the time operators needed to retrain, and international dates negotiated to accommodate fleets already at sea. Anyone who lived through the 2013 narrowbanding deadline in land mobile radio will recognize that pattern, where the rule arrives on one schedule and the equipment and training change on another.

From decrement to emission masks: spectrum occupancy then and now

The 1912 decrement limit is the ancestor of the spectrum rules a radio shop works under now. The ITU Radio Regulations define occupied bandwidth as the band outside which the emission carries 0.5 percent of its mean power on each side, so that 99 percent of the power falls inside it, and necessary bandwidth as the minimum required for the class of emission at the needed quality. FCC licenses state an emission designator describing bandwidth and modulation type, and the equipment rules add emission masks and frequency tolerances that limit how much energy can land outside the channel and how far the carrier can drift. Each of these rules regulates the same thing the 1912 inspectors measured with a decremeter, which is how much spectrum a transmitter actually uses compared with how much it was assigned.

Occupancy covers more than bandwidth, because a channel is also used up in time and in geographic space. Spark wasted bandwidth on a scale no current rule would permit. The broadcasting crisis of the middle 1920s showed that narrow continuous-wave signals collide just as badly when nobody assigns frequencies, power and locations, and that experience is why the 1927 act built a commission around assignment rather than around transmitter technology alone. The 2013 narrowbanding mandate and the move to two-slot TDMA in P25 Phase 2 apply the same reasoning to land mobile radio, and the existing articles on narrowbanding and on P25 phases cover those in detail.

Spark physics still affects public safety receivers. Arcing hardware on power lines produces broadband impulse noise by the same mechanism as a spark gap, and failing ignition systems, switching power supplies and some LED drivers produce wideband noise as well. FCC Part 15 governs these unintentional radiators, but the practical result at a receive site is a raised noise floor, which reduces talk-in range from portables long before anyone notices a distinct interfering signal. Utilities have crews and equipment to locate arcing hardware, and the useful role for a public safety agency is to document the problem and report it rather than to go looking for the source itself.

The move to continuous wave also delivered something early operators valued, which was more distance per watt. Concentrating a transmitter’s power into a narrow band, and receiving it with a selective receiver, lets the signal stand out against far less noise than a wide receiver has to accept. In practice that is a receive-side lesson for a modern system as much as a transmit-side one, since keeping the noise floor at a receive site low is often cheaper than adding transmitter power or another site.

What to do at your agency

• The radio system administrator should pull each of the agency’s licenses from the FCC Universal Licensing System this month, compare the emission designators listed on each with the modes the radios on that frequency actually transmit, and list any mismatch in a memo to the frequency coordinator or license holder of record.
• The radio shop supervisor should take a dated noise-floor reading at each receive site with a service monitor or spectrum analyzer during the next scheduled site visit and record the figure in the site maintenance log, so that a later rise in noise has a baseline to compare against.
• The communications center manager should add one paragraph to the existing interference SOP stating who records the time, channel and affected sites when broadband noise appears, which utility contact receives the report about suspected power-line arcing, and how to reach the FCC’s public safety interference reporting channel listed on the FCC website.
• The radio shop should check deviation and frequency error on a sample of at least ten portables and mobiles with a service monitor this month and keep a list of the serial numbers checked and the results, so that the fleet’s actual occupied bandwidth is documented rather than assumed.
• The training officer who maintains history or orientation material should review any slide that says the Titanic ended spark or that the vacuum tube replaced it overnight, and should correct it to describe the gradual changeover with its sources named.
• The COML or radio administrator should add an agenda item to the next regional radio users or interoperability meeting asking each member agency to report any unexplained broadband noise at shared sites, and should collect the reports in one list kept with the meeting minutes.

Takeaways

• A spark transmitter produced a series of decaying wave trains, and because each train lasted only a short time, its energy spread across a band many kilohertz wide.
• Early receivers such as the coherer responded to strong impulses with little selectivity, so stations in the same area jammed each other whenever more than one was sending.
• Coupled circuits, rotary gaps and the quenched gap reduced the damping and made spark easier to share, but every spark set still sent damped wave trains with a floor set by antenna losses and radiation.
• The Radio Act of 1912 licensed stations, assigned ship and amateur wavelengths and limited the decrement of the emitted wave, but it did not ban spark or require continuous wave.
• By 1927 the main American interference problem had moved to broadcasting, and the 1927 act and the Washington convention of the same year responded with assignment authority and phase-out dates for damped waves.
• Continuous wave arrived through the arc, the high-frequency alternator and finally the vacuum tube, and it required heterodyne receivers as well as new transmitters.
• Spark stayed in maritime service through the 1920s and on some vessels and emergency sets into the 1930s, so the changeover took about two decades rather than happening at one moment.
• Modern emission masks, occupied-bandwidth definitions and narrowbanding descend from the 1912 decrement limit, and arcing hardware still produces spark-like broadband noise that raises the noise floor at public safety receive sites.