Marconi´s 200kW transatlantic transmitter.

Enigma surrounds Marconi´s massive 200kW wireless station, built nearly a century ago. George Pickworth pieces together the technology behind the world's most powerful spark transmitter.

From Electronics World + Wireless World  January 1994.
Reprinted with permission of the editor of  Electronics World - a Cumulus Business Media publication

arconi's 200kW timed-spark continuous-wave transmitter was the ultimate spark-type transmitter. It was installed at Marconi's Caernarvon transatlantic 'super' station in Wales and came into service in 1916 to handle North Atlantic traffic. This was after the original synchronous-spark, wave-train transmitter was taken over by the military in 1914 for long range strategic signaling. The timed-spark transmitter worked US stations at New Brunswick, Tuckerton, Marion and the Central Radio Station at Long Island. In 1919 it transmitted the first signals directly to Australia. Wavelength was given as 14km, which is approximately 21.5kHz. To take advantage of the Earth/ionosphere wave guide effect, all transoceanic 'super' stations operated frequencies less than about 50kHz. However the lowest useable frequency, typically 20kHz, was set by physical constraints imposed by antenna structures: even the largest practical structures were very inefficient at 20kHz.
Because all transoceanic stations were confined to a 30kHz bandwidth, a high level of selectivity became vital to reduce mutual interference as the numbers of stations progressively increased. The only way of attaining this was with continuous wave systems. With these, oscillations progressively built up in the receiver tuner by virtue of resonance: this was known as syntony - a term invented by Lodge. Remarkably, Marconi's 1906 Clifden transatlantic super-station in Ireland, which originally radiated continuous waves, was a quenched-arc type. It had a plain triple disc discharger that was inherently self cooling, and the draught created by the rotating discs dispersed ionized gases. Fig. 1.

Fig.1. Marconi's 1906 Clifden transatlantic super-station was powered by a DC generator charging 6000 lead acid accumulators. It radiated wave trains.

On the other hand. Poulsen, with his quenched-arc system incorporated 'rod' type electrodes, comparable with an arc lamp. These required elaborate water cooling and a strong magnetic field to drive ionized gases from the arc-gap. The Clifden transmitter was powered by a DC generator which charged 6000 lead acid accumulators. However, the Clifden discharger was modified by attaching transverse electrodes to the main disc, similar to the 1916 synchronous discharger. These electrodes, described in the November issue, radiated wave trains; the official explanation was so that signals could be received by Marconi's magnetic detector which responded only to wave trains. There is some evidence that the magnetic detector will demodulate AM signals but I have not been able to confirm it.

Options.  
Originally, the Caernarvon transmitter was a 200kW synchronous spark type radiating wave trains. It only allowed a very limited degree of syntony since there were too few waves in each train. These declined too quickly for resonance to be effective. In 1916 however, it was replaced with a continuous wave transmitter to increase receiver selectivity. As early as 1906, Fessenden and Golsdschmidt adopted rf alternator-type continuous wave transmitters for their north Atlantic service. Poulsen adopted the quenched-arc continuous wave system for his Hawaii/San Francisco link. Marconi's approach on the other hand was to indirectly produce continuous waves. He used spark systems to generate wave trains in rapid succession so that in effect they overlapped in phase. This led to the development of the Caernarvon timed-spark discharger. Although the waves were continuous, they undulated in amplitude. Provided they remained in phase, this in itself did not significantly effect syntony. Indeed the undulations modulated the transmission with a tone. This article is about generating continuous waves by causing wave trains to overlap.

Fig.2. Tesla's 1899 radio telephony transmitter. There is little information on how it worked but it probably relied on the battery's internal resistance to limit the capacitor charge rate and a critical hub rotation rate.

Fig. 3.  Morrietti's 'hydrothermic' discharge transmitter. Unlike most of its contemporaries, it had no moving parts.

Overlapping wave trains.  
My experiments have shown the number of significant waves in a train to be around 25. When creating continuous waves by progressively reducing the period between trains they eventually overlap. Repetition rate must therefore be at least l/25th of the transmission frequency, but undulation is unacceptable at this rate. The Caernarvon transmitter generated wave trains that overlapped every 13.5 waves which produced slightly undulating continuous waves. A repetition rate equal to the 25th sub-multiple presented no particular problem at the 20 to 50kHz transoceanic frequencies. But timing the wave trains - hence the term 'timed'- so that they overlapped in phase required extraordinary technical expertise. However, at the 500kHz and 1 MHz maritime frequencies used with mechanical dischargers this repetition rate was out of the question.

Tesla systems.  
Radio telephony, as pioneered by Tesla was the original motivation for continuous waves. In 1899 he employed a discharger, or 'break' as Tesla called it. This consisted of a hub with 16 - or sometimes more - radial electrodes rotating at very high speed between a pair of fixed electrodes. The device was energized with DC and a discharge occurred each time a pair of rotating electrodes aligned with a fixed pair. The discharge was quenched as the gaps widened. Fig. 2. I am unsure how the device actually worked; indeed Tesla himself does not make this clear. A possible explanation is that the internal resistance of the battery limited the capacitor charge rate. During discharge, current is drawn from the capacitor faster than it is replaced, so potential falls. Then, as the hub continues to rotate, the discharge is quenched. In this way, the capacitor is charged and discharged synchronously with alignment of pairs of electrodes. At a critical rotation speed, this corresponds to the resonant frequency of the circuit. The 'flywheel' effect of the tuned circuit converted charging and discharging into continuous sine waves. Operation can therefore be compared to the quenched-arc system. The effect of the capacitor-type microphone was to de-tune the circuit so that power output corresponded to sound pressure waves. In later versions, Tesla used a pair of toothed wheels rotating at very high, but at slightly differing speeds, in opposite directions. This was reported to give up to 10.000 discharges a second. For an even greater discharge rate, Tesla added a jet of mercury intercepted by projections on a disc rotating at extremely high speed, but this was a low power device intended as an oscillator for use with his regenerative receivers.

Fessenden's experiments.  
In 1900, during early radio telephone experiments prior to adopting radio frequency alternators, Fessenden built a system whose main elements were a battery, a vibrating-reed type interrupter and a transformer. The interrupter was tuned to 10kHz. in series with the aperiodic primary winding of the hf transformer. In turn, the transformer's secondary winding was tuned to the interrupter frequency. This was the first transmitter to use a vibrating reed to set transmitter frequency. Fig. 4. My replication of Fessenden's experiments showed that the method worked well when the vibrator was tuned to a low sub-multiple of the resonant frequency. Here, the oscillation trains overlapped. Operation seems to have been by each DC pulse shocking the tuned circuit into oscillation. Before thermionic valve type oscillators, there were many ingenious spark systems. Despite these, rf alternators were the only devices capable of producing continuous waves pure enough for practical radio telephony.
Using rf alternators, Fessenden radiated his voice from his Brant Rock station in the USA. As early as 1906, these broadcasts were reported to be heard by operators at his station in Scotland. However, Marconi still rejected alternators, probably because they were still in their infancy and ran at very high speed. Frequency raisers which allowed alternators to run at lower speed had yet to be developed; so had the inductor type alternator, which eliminated windings on the rotor and thereby the major problem with rotor windings flying out of their slots.

Fig. 4.  Fessenden's continuous-wave oscillator was the first to use a vibrating reed interrupter tuned to set transmitter frequency.

Fig. 5.  Early cw spark transmitter with a tuned-reed interrupter had limited power and contact bounce problems when operating above a few kilohertz.

Early spark cw.  
The Marconi company had considerable expertise in the manufacture of spark apparatus. Having progressed so far with arc/spark systems, particularly for maritime use, it is understandable that Marconi should have pursued this path.
Early attempts to produce continuous waves with spark systems were based on tuning the interrupter of an induction-coil type spark-transmitter to a sub-multiple of the oscillator frequency, Fig. 5. However, while tuning the vibrator to a sub-multiple presented no special problem, contact bounce made precise timing extremely difficult and almost impossible to maintain at frequencies above a few kilohertz. Moreover, the vibrating reed was essentially a low power device. In my 7kHz reproduction, fine tuning to bring the wave trains into phase was by adjusting the oscillator frequency, Figs 6(a,b). I found using shock excitation simpler however.

Morrietti's system.
Morretti's 'hydrothermic' discharger, used for experimental radio telephony between Rome and Tripoli around 1910, consisted of a pair of copper discs set horizontally, one above the other. The lower disc had a tiny hole drilled through the. Through this hole, acidulated water was steadily pumped so as to form a jet that impinged on the upper disc, Fig. 3.
Current immediately vaporized the jet. This interrupted the circuit which was then reestablished - to be interrupted again to create current pulses. Operation was therefore automatic and required no moving parts. Each pulse shock excited the secondary circuit at a submultiple of its resonant frequency. In this way, oscillations persisting in the secondary circuit were continually reinforced to produce slightly undulating continuous waves. Morrietti's device drew power from a 500V DC generator via a variable resistor. This component was presumably intended to synchronize the discharges and bring them in phase with oscillations in the antenna circuit.

Marconi's timed discharger.
Marconi's 1913 experimental consisted of a bank of four rotary dischargers. Each comprised four radial electrodes rotating between a pair of fixed electrodes and driven by a common shaft extending from the drive motor. Individual primary circuits, one per discharger, were inductively coupled to a common secondary circuit. The device was powered by a DC generator, Fig. 7. Discharge commenced while the gaps were still narrowing. Further narrowing as the electrodes rotated reduced resistance across the gaps. Amplitude of the oscillations declined and energy was transferred to the secondary circuit. Then the gaps widened and the draught created by the rotating electrodes dispersed ionized gases, thus quenching the discharge and returning the gaps to a high resistance state. Bearing in mind that the device was energized by DC, I am not sure which mode it ran in. The primary circuit could have operated in quenched-arc mode and transferred energy to the secondary by induction. Alternatively, current pulses through the primary circuit could have shock excited the secondary circuit into oscillation. It was most probably a combination of both these modes. Whichever, discharges occurred consecutively, giving a total of 16 discharges per revolution. Assuming operation in quenched-arc mode, the quenching effect caused by the rotating electrodes would limit the number of oscillations in each train to 15 or fewer, insufficient to overlap. As a result, wave trains in the primary circuit were discrete. Provided they were in phase with oscillations in the secondary circuit, they would be reinforced to create continuous oscillations with undulations corresponding to their reinforcement points, Fig. 8. Reinforcement of oscillations in the secondary circuit would be exactly the same with shock excitation of the secondary circuit. Consider for example a frequency of 10kHz reinforced every 13th oscillation. This makes the reinforcement frequency 10⁴/13 which is 769.2Hz. As each revolution produces 16 reinforcements, it needs to run at 769.2/16, or 48rev/s (2888rev/min). Precise speed control was vital to keep the trains in phase. Reinforcement could be made to occur at other points by changing drive speed. By the same token, resonant frequency of the tuned circuits could also be changed, but then drive speed would have to be adjusted to keep wave trains in phase with oscillations persisting in the secondary circuit. However, maximum drive speed was unlikely to have exceeded 50rev/s (3000rev/min) which limited operating frequency to the order of 15kHz. Moreover, the device had an inherent drawback. In order to handle considerable power, the electrodes had large surface area. Because potentials were high, spark-gaps were fairly wide. Changes in atmospheric pressure, humidity and presence of ionized gases significantly altered the dielectric strength of air. Therefore, the point at which discharge occurred, varied and this upset timing.

 

 

Fig.6.  Reproduction of Fessenden's experiments produced the overlapping wave trains of (a). Wave trains overlapping in phase are detailed in (b).

Fig. 7. Marconi's experimental timed discharger comprised four rotary dischargers each with four radial electrodes rotating between a pair of fixed electrodes. Each revolution of the common shaft caused 16 discharges.

Caernarvon 200kW timed discharger.
Unfortunately very little information on the operation of the 200kW Caernarvon timed-spark transmitter has been published. Some information regarding the timer even seems to be misleading. This is understandable as there was great commercial rivalry between exponents of alternator and quenched arc systems. Nonetheless, by gleaning information from various sources, and by making a few assumptions, I believe that the following notes truly explain the operation of this remarkable transmitter, Fig. 9.In essence, operation was similar to the experimental timed discharger, but the Caernarvon timed-spark discharger employed two pairs of discharge assemblies. Each of these consisted of a tuned primary circuit, a power discharger and a timing discharger. The timer had a common drive shaft arranged so that discharges occurred in sequence, but in alternate assemblies, as the shaft rotated. Both primary circuits were inductively coupled to the common secondary circuit. The power dischargers were plain triple-disc types, i.e. with one main disc and two side discs. Discharge was from side disc to main disc to side disc. Independent electric motors rotated the main and side discs. This assisted cooling and created a draught that dispersed ionized gases. Drive speed was not critical and it seems as if they rotated at a relatively low speed, in the order of a few hundred rev/min.The remarkable feature of these systems was that the distance between the main disc and the side discs was so wide that a discharge could not ordinarily occur at the energizing potential of 5kV DC. However, set close to the side discs were a pair of ionizing electrodes.

Fig. 8.  In Marconi's discharger, oscillations in each wave train from the primary were insufficient to overlap but persistence in the secondary circuit caused continuous, albeit undulating, oscillation.

Timing discharger.
Both timing dischargers were connected to a common drive shaft and drive motor so that both timing dischargers rotated synchronously. Each discharger had a number of radial electrodes, probably 16, but because timing current was low, the timing electrodes had small face areas. Also, because the potential was relatively low, gap-width was narrow. This allowed close mechanical tolerances which minimized timing errors caused by changes in the dielectric strength of air. When the electrodes of the timing discharger were aligned, gap-width was less than a millimeter. At each alignment the timer capacitor, charged to 5kV, discharged across the gap and through the primary of an induction-type coil. This induced a very high potential in its secondary winding, which in turn caused a spark across the ionizing electrodes, which created a conductive path for the main discharge. As a result, the timer can be likened to an automotive capacitor discharge ignition system. Because the power discharger electrodes were in the form of discs, discharge could occur at any time so timing could be precisely controlled. The 5kV DC 300kW generator delivered 50A charging current to the primary circuit capacitors. This could be generated with dynamos without undue arcing across the commutator segments. Because of the ionizing spark, the power discharge occurred at this relatively low potential of 5kV; moreover, the wide spark gap caused the discharge to be quenched immediately the ionizing spark ceased. Keying was by interrupting the charging current to the timer capacitor as this circuit took only 300mA. Incidentally, although installed in a brick-silence cabin, the noise caused considerable stress among operators.

Operation.
My original experiments suggest that the time duration of the timing-spark, and consequently the main discharge was probably in the order of 250us. On this assumption, at 21.42kHz, or 47us, each train in the primary circuit probably consisted of about five oscillations. Subsequent research suggested that the primary circuit was effectively untuned and discharge was virtually a DC pulse which shock excited the secondary circuit into oscillation. But as I have already explained, the operating mode does not affect timing.

The reason for the half cycle was probably because the two primary inductors were arranged 180° out of phase so that trains started with alternate negative and positive half cycles. Rotational speed of the timer discharger can be determined in exactly the same way as for the experimental timed discharger. Because the antenna system was in effect a large LC tuned circuit and the linear elements were only a fraction of a wavelength long, it was a poor radiator. This allowed oscillations to persist in the antenna system and progressively build up in amplitude as a result of periodic reinforcement. These aspects made the system workable. But, because of the significant time taken for oscillations to build up, signaling speed was limited to about 100 words a minute. Antenna current was 280A and efficiency from generator to antenna was given as 66%. Although requiring careful adjustment and maintenance, the timed-spark transmitter was reliable. It proved so successful that the manufacture of a duplicate machine was put in hand immediately it came into service, but this was delayed because of difficulty in obtaining materials during the war. It does however seem to have been prone to radiating harmonics. In the meantime, a Poulsen quenched arc was installed as a reserve, but this was incapable of delivering antenna current of more than 170A and could not be worked for long periods without giving trouble.

	Fig. 9. Principles of Marconi's 200 kW Caernarvon discharger. Distance between the main disc and side disc would have been too great for discharge were it not for ionizing electrodes.

Alternators.
In 1921, it was decided that in order to increase keying speed and take advantage of improvements in receiver selectivity, it was necessary to replace the timed-spark transmitter with one that produced continuous waves of constant amplitude. Thermionic valve transmitters could meet this criterion with low power installations, but valves that could handle high power had yet to be developed. The only option was the radio frequency alternator. In September 1921, a pair of 200kW Alexanderson rf alternators were installed at Caernarvon and remained in service until about 1923.In hindsight, Marconi might just as well have adopted alternators in 1906, rather than build the Clifden quenched-arc transmitter. Unfortunately, history never gives the alternative, so we will never know what the outcome would have been if he had. However one thing is certain - neither the synchronous-spark nor the timed-spark transmitters would have been made, and technical historians like myself would not be trying to figure out how these remarkable machines actually worked.

Further reading
Fahie, J. )., A history of wireless telegraphy, Dodd-Mead & Co, New York 1901.
Stanley, Rupert, Text book on wireless telegraphy, Longmans-Green & Co., Vols I & II, 1914/19.
Baker, W. J., A history of the Marconi Company, Methuen, 1970.
Laughter, V. H., The operator's wireless, telegraphy and telephone handbook, Frederick J. Drake, Chicago, 1909.
Tesla, Nikola, Colorado Springs notes, Beograd, Hungary, 1899.
Vyvyan, R. N., Wireless over thirty years, 1933.
Constable, A., Early wireless, 1980.

Index

Next page