The Gallery of Electro-Mechanical Amplifiers.

A celebration of an almost forgotten technology.
The title of this page has been changed to "Electro-Mechanical amplifiers", to distinguish them from mostly mechanical ones, such as the Frenophone, which can be found in the Mechanical Amplifiers gallery.

Gallery opened May 2000

Page updated: 30 Jan 2022

New arc amplifier added

The year is 1900. Cost and attenuation are seriously limiting the growth of long-distance telephony, particularly in the USA.This needs a little explanation. You can always reduce the attenuation per mile of a telephone line by using thicker copper to reduce the resistance. Copper is however expensive, and there are limits to what is practical in this direction.

A Digression on Line Losses in Telephony.

To us, the obvious solution is to apply some amplification to boost signal strength; but it is long before the invention of the transistor, and even the valve (the vacuum tube to US readers) is still years away.


Mere lack of resource has never blocked human ingenuity, and there was a handy fact to exploit. Carbon microphones, as used universally in telephones until the mid-1960's, are not mere transducers that turn sound power into electrical power, but actually give a power gain of about a hundred times. The microphone is a variable resistance, made up of fine carbon granules, that controls the flow of current from a DC power supply; it does not merely convert acoustic energy into electricity. This power-amplification technology was the essential basis of the first practical telephone systems.

From this, it is but a short step to the concept of coupling a telephone receiver to a carbon microphone to make an amplifier, and several people took it. Within a few years of the original Bell invention, patents on this notion had been taken out by Edison, Houston, Lodge and Hughes, not to mention others less famous. By 1896, 27 patents for a mechanical-electrical amplifier- though it was then called a "repeater"- had been taken out.

Carbon microphones are not high-fidelity devices. Anyone who has used an old-fashioned telephone will recall that they would intermittently go low-gain or noisy, due to the granules packing, and a sharp rap of the handset against the wall was required to restore normal service. It was clear that the mechanical amplifier was far from perfect, but it was the only amplification option that looked practical.

Left: The basic operation of a mechanical amplifier is very simple.

The electromagnet attracts the armature in response to the incoming signals. This varies the pressure on the carbon granules in the transmitting part, which varies their electrical resistance, and hence the current flowing in the output circuit.

Left: The first Shreeve mechanical amplifier: approx 1904.

In 1903 Herbert E Shreeve of the American Telephone & Telegraph Company (AT&T) was given the job of designing a practical mechanical amplifier. He discarded the microphone and earpiece diaphragms, and replaced them with a simple plunger that was driven by the receiving coil and pressed against the carbon granules of the transmitter.

One of the requirements of a practical amplifier was stability over time, which meant a solution to the problem of granule packing was needed. Shreeve found that this was due to thermal expansion caused by the heat liberated in the carbon chamber. This reduced the resistance, causing an increase in DC current but reduced audio output. This problem was controlled by designing the transmitter so that expansion of its parts did not increase the pressure on the carbon; and this led to the 1904 model below.

The first successful test was in 1904 on a circuit from Amesbury, Massachusetts to Boston. This model saw commercial use on a New York - Chicago circuit between August 1904 and February 1905. Note the cooling fins for the carbon microphone section at the right. Repeaters were originally used on non-loaded open-wire lines, with not more than one repeater in the circuit. (Bell 1944)

Left: The Shreeve amplifier patent: 1905

  • A= support standard
  • B= base
  • C= casing
  • D= casing end closure (insulating)
  • M= receiving magnet
  • m= magnetising coil
  • V= variable resistance medium (carbon)

Source: US Patent 791,655 of 1905

Left: Herbert E Shreeve: circa 1915

Herbert Shreeve (1873 - 1942) was a British telephone engineer who emigrated to the United States around 1895. He worked for AT&T as an engineer, and served in the US Army Signal Corps during the first World War.

He worked for AT&T for most of his career. He was at the San Francisco end when the first transcontinental (New York to San Francisco) telephone call was made in January 1915. (using valve repeaters) He died in 1942, in New Jersey.

Source: Bell Telephone magazine 1941

Left: The Shreeve 1A amplifier.

The 1A worked reasonably well, and saw some use. However it was not possible to run more than two in series without the audio degrading too much, and three were required for transcontinental USA operation.

The degradation was usually referred to as "distortion" but this seems to have meant a poor frequency response rather than non-linearity. (Though there was plenty of that as well)

The Shreeve 1A included a simple form of feedback control to control the mechanical pressure on the microphone element, and so the current through it. The current heated a zinc strip which withdrew the rear electrode to act against increases in this current; the thermal inertia prevented the signal being cancelled out as well as the long-term current trend. This was twenty years before the formal invention of negative feedback, though the principle had been in use for many years in the form of Watt governors.

The cooling fins appear to be at the bottom. Horizontal fins are not efficient; they should be vertical.

Left: A cartridge type amplifier, Model 3A: 1914

The existing 1A repeaters were not good enough to be operated in series, and in 1911 a major effort was launched to find a better design. It was found that the 1a had a significant resonant peak in the response "about midway in the telephone talking range" which I assume means around 1 kHz. Non-linear distortion was also a serious problem with low levels being amplified less than high levels. There were also variations in gain from moment to moment. It had pretty much every fault known to audio amplification.

The Model 3A was developed in 1912 and standardised in 1914. The mass of the moving parts was reduced so the resonant frequency was at 2.2 kHz, at the top of the telephone frequency range and less of a problem; it was reduced in amplitude by a shunt LC filter. Non-linear distortion remained a problem. The parts likely to need replacement were packaged in an easily replaceable cartridge. It became the standard type of amplifier until valves were introduced, but performance was still limited by poor frequency response and a rapid falloff of gain at low levels; three repeaters in series was the practical limit.

Here fixed annular cooling fins surround the base of the cartridge. The two round things on top appear to be bolts holding the can down. In use the repeater was installed so the cylinder was horizontal.

Note the five connection terminals to the left.

Source: Bell System Technical Journal 1944

Left: The internal construction of the Model 3A

The basic principle was to minimise the mass of the vibrating electrode, pushing the fundamental resonance to as a high a frequency as possible. The function of the "compensating winding" is presumably to regulate the current through the carbon element; since there appear to be only 5 terminals, it appears the magnetising and compensating windings share a common terminal.

Left: Another Model 3A repeater: 1919

This version has a slightly different appearance. It is not known why the cylinder is tilted.

Left: A bank of Model 3A repeaters: October 1919

This shows the repeater bank for the AT&T Boston-to-Washington DC long distance trunkline at the Providence, Rhode Island switching center in 1919. The vertical panels are identical (apart from the one nearest the camera) and each contains the repeaters and associated equipment for one wire in one direction on the trunkline. The repeaters are the horizontal cylinders in a double row along the top of the panel. Each bidirectional line required a repeater for each direction. Below the repeaters are rheostats for adjusting the DC current on the line, and the power supply for the repeaters.

Even the 3A electromechanical repeaters were not very satisfactory; even under good conditions they gave less gain that a valve equivalent. Replacement with valve amplifiers began in 1920.


I do appreciate that these are not strictly electromechanical amplifiers. But the history arc amplifiers is closely linked to them, so they're staying here.

There were other attempts to achieve amplification before the valve. Peter Cooper Hewitt, who introduced mercury-vapour discharge lamps in 1901, suggested an amplifier in which beams of mercury ions were deflected by electromagnetic coils. He was granted US patent 749,792 in 1904.The amplifier worked, but suffered badly from variable gain, noise, and distortion. Hewitt went on to introduce the very successful mercury-arc rectifier, which was widely used for heavy-current rectification until the 1970s.

Left: The Hewitt mercury-arc amplifier

A mercury-arc discharge runs between the negative electrode 10 and the positive electrode 11. This was modulated by a deflecting coil 5, with the input typically provided by a carbon microphone 4. The purpose of the curly wire is uncertain; it is not mentioned in the patent text. It may have been for starting the discharge initially.

Source: US patent 749,792 of 1904

In 1912 Dr H D Arnold produced an improved amplifier on this principle; it had a good frequency response, but had such severe starting and maintenance difficulties it was unsuitable for unattended operation. It was used experimentally on the transcontinental circuit, but never commercially.

Left: The Arnold mercury-arc amplifier.

Here is my take on how it works:

Normally an arc is running from anode A to mercury cathode C, powered from Batt 2 and current limited by resistor R. The function of inductor L is obscure, but it might be something to do with promoting arc stability.

The arc is deflected sideways when currents flow through the deflection coils (green) mounted on the D-shaped iron core (in pink). These currents come from the carbon microphone on the right, energised by Batt 1 and coupled through transformer T1.
The arc thus moves between two subsiduary electrodes, in a way that is not too clear from this drawing, and the audio output is taken from the balanced to unbalanced transformer T2 to the earphone at lower left.

K appears to be the starting key. When it is pressed, starting anode A' is connected to Batt 2 and presumably an initial arc is struck through the shorter path A'-C.

Left: The Arnold mercury-arc amplifier

This model looks somewhat different to the patent drawing; there is only one reservoir of mercury at the bottom, so presumably the starting arrangements were different. This is a laboratory version.

Another possible amplifier investigated was the cathode-ray amplifier of Robert von Lieben; it proved to be no more promising. It controlled the flow of cathode rays rather than a mercury arc.

Left: The von Lieben amplifier

The 1905 von Lieben amplifier was a cathode-ray tube, its electron beam being steered towards or away from a target electrode by magnetic coils. Patent 179,807 was granted for a "cathode ray relay" in 1906.

The wire filament k is the electron source, heated by battery b1. The electron beam is focused (theoretically at least) so that it is narrow when it passes through small apertures in the cylinders f and f1. When the beam is undeflected it passes through the upper aperture and impacts the inner cylinder f, which is at a positive voltage relative to the cathode, and a current flows through the load a1. When the beam is deflected part or all of it hits the outside cylinder f1, also at a positive voltage, and is diverted away from load a1.

This does not sound like it was a very linear business even if all was working properly, but in addition there were problems making the electron beam the right shape.

Later a control grid that acted on the electron stream was found to work better than the deflection coils, the result resembling a conventional valve. Based on this, the two German companies AEG and Siemens founded Telefunken AG to build electron tubes.

Source: Patent 179,807. It is not currently clear if it was an Austrian or a German patent.

Dieckman & Clag (of Strasbourg) filed for a patent on a similar device in 1906; this used electrostatic rather than magnetic deflection of the beam. No further information has been found so far. Dieckman & Clag are unknown to Google.

Meanwhile, the vacuum tube as we know it was being created. Fleming produced a thermionic diode in 1904, and in 1907 a US Patent was granted to Lee de Forest for adding a control grid, to make the first triode. By October 1912 de Forest was demonstrating a valve audio amplifier to Bell officials. By October 1913 improved valves had been built and tested on commercial circuits between New York and Baltimore, and at the end of 1914 valve amplifiers were in use on the transcontinental circuit. It was clear that valves were the way ahead, and the mechanical amplifiers built for this service were placed on standby.

So, was that the end for this technology? Indeed not. Now read on...


S G Brown Ltd, of London, were well-known makers of headphones. They manufactured several models of electromechanical amplifier, working as above with the input coils moving an armature connected to a carbon microphone. The microphone worked in push-pull; in other words one carbon capsule was compressed while the other was released, and the anti-phase outputs were combined in a transformer. What they called the "Microphone Bar Amplifier" was first manufactured during WW1 (1914-18) and was used by the British Army; after the war a large number were sold off as government surplus. I think 'Bar' is probably an acronym for 'Brown Audio Relay' but this is unconfirmed at present.

By the early 1920s, S G Brown were advertising their 'Crystavox' loudspeaker, which was an electromechanical amplifier combined with a horn loudspeaker in a single unit.

Left: A Brown electromechanical amplifier: 1914

This picture comes from an article in Wireless Age, which claims the amplifier (it calls it a relay) operates by moving a partial contact, set up by very delicate adjustment. No carbon microphone is involved. The article goes on to say that the instrument can be turned upside down without affecting its working, and "carried on board ship and worked in all weathers".

The moving part was some sort of reed, and it appears that it was intended to be resonant at say 1 kHz for the reception of Morse code. "Damping the reed with a piece of rubber" presumably killed this resonance and made the reproduction of speech and music possible.

The device has a hinged lid; the thing sticking up has a pivot at the left but is currently unidentified. It will have to be swung down if the lid is to be closed.

The article claimed it was used in the Electrophone system, "by adding a loudspeaking device with trumpet the sounds may be heard at some distance in the room."

Source: Wireless Age Feb 1914, pp420-421

Left: Diagram of Brown electromechanical amplifier: 1914

  • N Permanent magnet
  • H,K Soft iron pole extensions with coil windings
  • P Steel reed
  • M,O Top and bottom metal contacts
  • W Fine adjustment screw

K is described as 'the regulating winding' but it is not currently clear what that means.

Source: Wireless Age Feb 1914, pp420-421

Left: A Brown electromechanical amplifier: 1914

Another picture of the amplifier shown above. Here the lever has been swung over to its other position.

Just below the lever can be seen the green receiving coil.

Left: A Brown Type V electromechanical amplifier: about 1924

Four models were available; this is a Type V (for valve sets) with a 2000 ohms input impedance, and driving a 2000 ohms output load.

Note the useful circuit on the lid. The carbon microphone is a push-pull type, hopefully reducing second-harmonic distortion, feeding a step-up output tranformer.

Left: A Brown Type V electromechanical amplifier: about 1924

Close-up of the amplifier internals. The two receiving coils can be seen on the left.

The carbon microphone element is behind the brass disc at centre-right.

Left: A contemporary advertisement for Brown amplifiers, showing the Four models were available. The one shown above is a R53/3 Type V.

Interestingly the amplifier is claimed to be "without distortion", which is quite untrue; distortion was high. The prices shown were a great deal of money for the time.

Rather strangely, the instrument is said to be suitable for all classes of communication, despite having a 1 kHz resonant frequency clearly chosen with CW (morse) in mind.

These two pictures are reproduced by kind permission of Lorne Clark, whose very fine website can be seen at

Left: A 1927 advertisement from the New Wilson Company.

By 1927, Brown appears to have lost interest in this technology and passed the rights over to the New Wilson Company. The amplifiers also appear to have been marketed by the Empire Electric Company of London NW1. Interestingly, New Wilson claim to be the patentees of the idea; this may simply mean they bought the patent rights from Brown.

The only report I have unearthed states that the gain was reasonable but the audio quality poor, as would be expected. Once again a very dubious claim of "no distortion" is made.

Image from The Radio Times for 23rd Sept 1927

You can see a double Brown amplifier here. It is described as a 'crystal amplifier' but that means it was intended to amplify the output of a crystal radio set; crystals were not involved in the amplification. The power gain was said to be about 1000 times, implying a voltage gain of some 30 times, which sounds plausible. Each of the two amplifiers therefore had a voltage gain of about 5.5 times. There is an adjustment lever for each amplifier.


Left: A policeman demonstrating the power microphone PA system in 1937.

A related technology was the Tannoy power microphone, which could drive a loudspeaker directly, without the need for an intervening amplifier of any kind. This consisted of eight push-pull carbon microphones mounted in a heat-dissipating aluminum casting; they drove a centre-tapped transformer which in turn drove an efficient horn-loaded loudspeaker. It was therefore possible to make a battery-powered public-address system without the fragility and complexity of a valve amplifier. In this early model the horn loudspeaker is mounted on a tripod with the accumulators slung underneath to give stability.

This apparatus was produced in anticipation of war. In 1938 about 800 were made for the Civil Defence Corps, the Auxiliary Fire Service (my mother was in that) and the Police for crowd control during bombing raids. There appears to be no information on whether they were ever used.

Even the final production model was not very efficient, delivering about 8 Watts to the speaker while drawing about 50 Watts from the accumulators. This was not really a problem because horn speakers are efficient at turning electricity into acoustic output, and the volume produced was quite enough. The apparatus would only have been used intermittently; there was an on/off pushbutton built into the microphone handle.

The same technology was used for intercoms in harsh and noisy environments, such as tanks in WW2.

Info from The Tannoy Story by Julian Alderton, pub Gaskell 2004


As I said above, a carbon microphone in itself is an amplifier. Until miniature valves became available in the late 1930's, a valve hearing-aid was not very practical. Early valves were rather large (imagine a hearing-aid constructed with octal valves) and battery drain was high for the filaments. There were a few large hearing aids with valves, but the majority of the electrical hearing aids sold between 1900 and 1938 were electromagnetic/carbon.

Left: circuit of the Western Electric No 38A Audiphone hearing-aid.

The rheostat is presumably a gain control, and appears to alter both the microphone energising current and the amplifier energising current. This no doubt saves battery power when the full gain is not required.

The circuit shows three cells in the battery. If this may be taken literally the operating voltage was about 4.5V

Left: internals of the Western Electric No 66B hearing-aid. This appears to have the same circuit as above. The rheostat is on the left, and the microphone, earphone and battery are all external to this little box.

Mr Hugh Hetherington (who kindly supplied this image) tells me that the hearing-aid carbon microphone employed a carbon ball technique rather than the carbon dust used in telephone transmitters; these apparently gave more amplification than the carbon dust transmitters, but were never used in telephones.

This topic was originally an external link to a very informative webpage. Sadly, this link seems to have died, so I give a much condensed account below. I have been unable to locate the original site builders- if anyone feels I have purloined information, please let me know.

In early 1912 several businessmen from New York were traveling in Austria-Hungary and while in Budapest, they were surprised to learn that they could listen to concerts or lectures without leaving their rooms. Budapest had an early cable distribution system that drove headphones.

These gentleman saw business potential in this, and formed the New Jersey Telephone Herald Company. It was decided to install the system in Newark, New Jersey, with the idea that if it was successful there, it would be introduced in New York. Wires were leased from the telephone company, installation was started early in the Spring of 1912 ,and regular programs were being broadcast by July. The programs were produced in a suite of rooms very similiar to a present day studio for a radio broadcast station, with acoustic treatment on the walls and Erickson microphones.

What is not currently clear is how the signal from the microphones was amplified so it could drive hundreds of headphones across a city.

For the first two or three months the subscription department was swamped with orders for installations, and within the first three months there were about 5,000 subscribers. The service cost $1.50 a month. However, as with everything else, people soon tired of their new toy, mainly because loudspeaker reception was not available, although the signals that were received were very clear and of excellent head-phone volume.

The management of the company realized where the difficulty lay and Mr. Rainbault and his chief engineer, Mr. J. L. Spence, worked on the perfection of a "mechanical amplifier". No details are currently known but this probably worked on the same electromagnetic/carbon scheme as the telephone repeaters above. However, the results obtained were far from satisfactory, so in December of the same year it was decided not to fight any longer against such odds. The New Jersey Telephone Herald Co. was closed and the headsets removed from the homes of Newark. There had been an outlay of over $200,000.

Thus electromechanical amplifiers once more proved unequal to the task of amplification. Radio broadcasting in the USA did not begin until 1916.

I would be very glad to hear from anyone who has any more information on these remarkable byways of technology.

Acknowledgement: This article draws heavily on the book "Engineering and Science in the Bell System". All speculation is mine.

Back to Home PageBack to The Museum EntranceTop of this page