A nice piece of kit was the Alexanderson Alternator. It was one of those solutions that is clever but simple.
First, some background:
Various electrical telegraphy methods were tested through the early 1800s but it wasn’t until 1838 that a commercial telegraph system was in place.
In 1861 Philipp Reis demonstrated a system for transmitting sound over electrical wires but it was not suitable for carrying human speech, because of the limited frequency range.
In 1876, Alexander Bell had invented a means of conveying human speech over wires, and by 1877 had commercialised the process.
In the 1880s, Heinrich Hertz performed experiments proving the existence of radio waves.
In 1896, Marconi developed a practical system of radio telegraphy. This system used spark gap transmission: this produces broad spectrum radio waves bursts, which is suitable for Morse code and similar systems but doesn’t lend itself easily to audio.
Reginald Fessenden is credited with the first use of radio to transmit voice in 1900. He used modulation of a high frequency spark gap (20000 Hz), reasoning that high frequency sparks are a reasonable approximation of continuous wave. Over the next few years he continued to work on this idea but ultimately the sound, though fairly intelligible, was very unpleasant and also quite susceptible to noise. Also the broad frequency band of the output means that the entire spectrum is polluted, and there is a concommitant background noise level at the broad frequency receiver end as well. Nonetheless this early work on amplitude modulation led to greater things.
Fessenden did much of the work in developing truly continuous wave radio output with a view to using an amplitude modulated continuous wave to transmit speech. Intelligible human speech requires frequencies up to at least 4000 Hz, and the bandwidth required for a modulating signal is twice the top frequency of that signal: ie at least 8000 Hz in this case. The carrier signal needs to be much higher frequency than the modulating signal: many tens of thousands of Hz.
Alternating current was common by this time and it was already known that alternators produce sinusoidal electrical signals. These typically ran at frequencies from a few dozen Hz up to the low hundreds. Could you safely run an alternator at tens of thousands of revolutions per second?
Fessenden asked General Electric to work a high frequency alternator for radio transmission. In 1904 Ernst Alexanderson, a GE employee, invented what is now called the Alexanderson Alternator. It included a rotor whose edge is divided into hundreds of magnets, separated by non-magnetic material. This meant that by spinning the rotor at a few hundred cycles per second, sinusoidal radio waves could be generated with frequencies up to 100000 Hz. By 1906, the Alexanderson Alternator was sufficiently well-developed for Fessenden to make a Christmas Eve broadcast from his Massechussets radio station, in which he read from the Bible and played violin, which was picked up by maritime receivers. This 75 kHz broadcast was the first use of AM radio for entertainment purposes.
A competing technology was the Poulsen Arc, named after its inventor, Valdemar Poulsen. The Poulsen Arc used a spark transmitter to turn DC into AC, and it too could produce radio waves suitable for AM at frequencies of tens of thousands of Hz. It was much more portable than the Alexanderson Alternator, but it had disadvantages. The frequency and amplitude of the output tended to be unstable, and the wave was not very sinusoidal: it was rich in harmonics, which complicated matters for receiver design, and also polluted the spectrum, whereas the AA produced output in a nice defined band.
By the 1920s, thermionic valve technology had become the dominant means of producing, modulating and amplifying radio signals, and remained so until the semiconductor revolution. The Poulsen Arc and spark gap transmission generally were banned internationally in the early 1930s, due to their tendency to flood the electromagnetic spectrum with noise. The Alexanderson Alternators remained in use by the US military for long distance, high power communication in the VLF band: pings, not voice.