the curve depends upon the sum of the decrements of the sending and receiving circuits. By the term "decrement" of a circuit is meant the logarithm of the ratio of the amplitudes of two successive oscillations in the train.

To obtain very sharp tuning we have therefore to employ either undamped oscillations or very feebly damped oscillations in the transmitter, and also a receiving circuit in which there is as little dissipation of energy by resistance and other causes as possible. It is then possible to cause a change of even less than one-half of 1 per cent, or 5 parts in 1,000 in the wave length of the received waves to cease to actuate the receiver. This means that we can distinguish between two waves 1,000 and 1,005 or 1,010 feet in length, respectively, and that our receiver may be tuned to respond to one and not to the other. The persistent or undamped oscillations created by the arc transmitters have, therefore, an advantage in this respect over spark transmitters, in that the damping or decrement of the transmitter is less; but it should be borne in mind that the damping of the receiver circuit has also a large influence on the form of the resonance curve, and that good isolation can not be obtained unless the receiving circuit also has a small decrement. Under favorable conditions we can employ a sending key, which does not interrupt the production of the electric waves at the sending station, but simply alters the wave length slightly by about one-fourth per cent. If, then, the corresponding receiving station has a feebly damped receiver, this change will be sufficient to cut up the continuous record or telephone sound at that station into Morse dots and dashes, and so transmit signals. But another station not so tuned will either receive nothing at all or else a continuous unbroken line or sound not having any meaning. There are other methods by which signals not intended for a particular receiver can be rejected by it. Fessenden has described for this purpose an interference detector, in which the impulses it is not desired to receive are made to divide between two paths, the oscillations in which are then caused to neutralize each other's effect on the oscillation detector. On the other hand, the waves of the wave length it is desired to receive do not so neutralize themselves, but produce a signal by their operation on the detector. We must pass on to notice in the next place some improvements in oscillation detectors, and means of testing them. As already explained, the ether waves sent out by the transmitting antenna fall on the receiving antenna and create in it or some other circuit connected to it very feeble oscillations. These oscillations being very feeble alternating currents of high frequency, can not directly affect either an ordinary telegraphic instrument or a telephone, but we have to interpose a device of some kind called an oscillation detector, which is affected by oscillations in such a manner that it undergoes some

W

M

0.3 volt

change which in turn enables it to create, increase, or diminish a local current produced by a local battery and so affect a telephone or telegraphic relay. One kind of change the oscillations can produce in certain devices is a change in their electric resistance, which in turn is caused to increase or diminish a current through a telephone or telegraphic relay generated by a local battery. To this type belong the well-known coherers of Branly, Lodge, and Marconi, which require tapping or rotating to bring them back continually to a condition of sensitiveness. Coherers, however, have been devised which require no tapping. Thus it has been found by Mr. L. H. Walter that if a short length of very fine tantalum wire is dipped into mercury there is a very imperfect contact between the mercury and tantalum for low electromotive forces. This may perhaps arise from the fact that tantalum, like iron, is not wetted by mercury. If, however, feeble electric oscillations act between the mercury and tantalum, the contact is improved whilst they last. If, then, the terminals of a circuit containing a telephone in series. with a shunted voltaic cell are connected to the mercury and tantalum, respectively, and if damped or intermittent trains of electric waves fall on an antenna and excite oscillations which are allowed to act on the mercury tantalum junction, then at each train the resistance of the contact falls, the local cell sends current through the telephone and produces a short sound, FIG. and if the trains come frequently enough this sound is repeated and will be heard as a continuous noise in the telephone (see fig. 15). This sound can be cut up dot and dash signals by a key in the sending instrument. If the transmitter is sending persistent oscillations, then some form of interrupter has to be inserted in the receiving circuit to enable us to receive a continuous sound in the telephone which can be resolved into Morse dot and dash signals by the key in the transmitter. The operator usually wears on his head a double telephone, and listens to these long and short sounds in the telephone and writes down each letter or word as he hears it. The reception of signals in modern radiotelegraphy is most usually effected by ear, by means of some type of oscillation detector capable of actuating a telephone. It is important then to notice that, to obtain the highest sensitiveness when using the telephonic method of reception, the spark frequency or number of oscillation trains or the number of interruptions of the persistent train per second must take place at such a rate that it agrees with the natural time period of the diaphragm of the telephone

E

FIG. 15.-Walter's tantalum de

tector.

into

used. An ordinary telephone receiver is most sensitive, according to the researches of Lord Rayleigh and M. Wien, for some frequency lying between 500 and 1,000. Thus Lord Rayleigh (see Phil. Mag., vol. 38, 1894, p. 285) measured the alternating current in microamperes required to produce the least audible sound in a telephone receiver of 70 ohms resistance at various frequencies, and found values as follows:

TABLE II.

Frequency

Least audible current in microamperes.

22828

192 256 307 320 384 512 640 768 2.5 0.83 0.49 0.32 0.15 0.07 0.04 0.1

M. Wien found for a Siemens telephone somewhat different results viz:

Frequency

Least audible current in microamperes..

64 128 256 512 720 1,927 1,500 12 1.5 0.13 0.027 0.008 0.013 0.024

Both, however, agree in showing a maximum sensitiveness for currents of a frequency between 600 and 700. This is due to the fact that the frequency of the actuating current then agrees with the natural frequency of the ordinary telephone diaphragm. Hence, alternators for large power radiotelegraphic stations are now designed to give currents with a frequency of about 300 or 600 alternations per second, so that, when producing discharges of a condenser, the number of sparks per second may be at least 600, and fulfill the conditions for giving maximum sound in the telephone of the receiver per microampere. Another class of oscillation detector recently discovered comprises the crystal detectors which depend on the possession by certain crystals of the curious property of acting as an electrical valve, or having greater conductivity in one direction than the other, and also on not obeying Ohm's law as conductors. It was discovered by General Dunwoody of the United States Army, in 1906, that a mass of carborundum, which is a crystalline carbide of silicon formed in electric furnaces, can act as a detector of electric oscillations if inserted in the circuit of an antenna, the crystal mass being held strongly pressed between two spring clips, which are also connected by a shunted voltaic cell in series with a telephone. When feeble oscillations are set up in the antenna, a sound is heard in the telephone. This property of carborundum has been carefully investigated by Prof. G. W. Pierce, of Harvard, and he showed that a single crystal of carborundum has remarkable unilateral conductivity for certain voltages when held with a certain contact pressure between metallic clips. Thus for a crystal held with a pressure of 1 kilogram, and subjected to an electromotive force of 30 volts, the conductivity in one direction through the crystal was 4,000 greater

Current in Micro-amperes

1200

1000

800

600

400

200

25

20 15 10 5 0 5 10 15 20 25 Applied Voltage.

than in the opposite direction (see fig. 16). The result of these experiments was also to show that the current voltage curve or characteristic curve of a carborundum crystal is not linear-that is to say, the crystal as a conductor does not comply with Ohm's law, for the resistance of the crystal decreases as the current is increased. Hence the conductivity of the crystal is a function of the voltage acting on it (see fig. 16A). Accordingly, if we pass a current from a local cell through a crystal under a voltage say of 2 volts, a telephone being inserted in series with the cell, and if we apply an oscillatory voltage also to the crystal, which varies say between +0.5 and 0.5 volt, then the crystal is alternately FIG. 16.-Characteristic curves of carborundum subjected to a voltage of 2.5 and 1.5 volts, but the corresponding currents would be say 8.4 and 1.8 microamperes, as shown by an experiment with one particular crystal employed by Professor Pierce. The mean current would then be 5.1 microamperes, whereas the steady voltage of 2 volts would only pass a current of 4 microamperes. Hence, apart from the unilateral conductivity, and merely in virtue of the fact that the characteristic curve is not a straight line, we find that such a crystal or

Current.

crystal.

even a confused mass of crystals can act as a radiotelegraphic detector. There are, therefore, two ways in which a crystalline mass of carborundum can be used as a radiotelegraphic detector. It consists of a conglomeration of crystals arranged in a disorderly manner, or not so symmetrically as to neutralize one another's unilateral conductivity. Hence the mass of crystals, like the single crystal, possesses unilateral conductivity, and also a conductivity which is a function of the voltage applied to it. We may then use it without a local cel', and avail ourselves of its valve property to rectify the trains of oscillations in the antenna and convert them into short unidirectional trains which can affect a galvanometer or telephone; or secondly, we may place the crystal between the ends of a circuit containing a telephone and a shunted voltaic cell, and then on passing oscillations

i2

Applied Voltage.

FIG. 16A.

through the crystal we hear sounds in the telephone due to the fact that the conductivity is a function of the voltage, and is therefore increased more by the addition than it is diminished by the substraction of the electromotive force of the oscillations to or from the steady voltage of the local cell. The telephone, therefore, detects this change in the average value of the current by a sound emitted by it. Professor Pierce has discovered that several other crystals possess similar properties to carborundum, for example, hessite, which is a native crystalline telluride of silver or gold; an anatase, which is an oxide of titanium; and molybdenite, which is a sulphide of molybdenum. As regards the origin of this curious unilateral conductivity, it seems clear that it is not thermoelectric, but at present no entirely satisfactory theory of the action has been suggested.

A number of forms of oscillation detector have recently been invented which depend on the curious fact that a slight contact between certain classes of conductors possesses a unilateral conductivity, and can therefore rectify oscillations. One such detector now much used in Germany consists of a plumbago or graphite point, pressed lightly against a surface of galena. It has been found by Otto von Bronk that a galena-tellurium contact is even more effective. To the same class belongs the silicon-steel detector of Pickard. If such a contact is inserted across the terminals of a condenser placed in the receiving circuit, and if it is also in series with a telephone, the trains of oscillations are rectified or converted into more or less prolonged gushes of electricity in one direction through the telephone. These coming at a frequency of several hundred per second, corresponding to the spark frequency, create a sound in the telephone, which can be cut up by the sending key into Morse signals. According to the researches of Professor Pierce and Mr. Austin it seems clear in many cases that this rectifying action is not thermoelectric, since the rectified current is in the opposite direction to the current obtained by heating the junction.

I may, then, bring to your notice some recent work on another form of radiotelegraphic detector, which I first described to the Royal Society about five years ago under the name of oscillation valve. It consists of an electric glow lamp, in the bulb of which is placed a cylinder of metal which surrounds the filament but does touch it. This cylinder is connected to a wire sealed through the glass. Instead of a cylinder, one or more metal plates are sometimes used. The filament may be carbon or a metallic filament, and I found some year or more ago that tungsten in various forms has special advantages. The bulb is exhausted to a high vacuum, but of course this means it includes highly rarefied gas of some kind. When the filament is rendered incandescent it emits electrons, and these electrons or negative