ELECTRIC LAMPS AND VALVES, MANUFAC TURE OF. The history of the incandescent lamp has centred mainly round improvements in the filament. In the first practical lamps (produced independently by T. A. Edison and J. W. Swan between 1878 and 188o), carbon filaments were used, and carbon had no real rival until the appearance first of the Osmium lamp (1898) and then of the Tantalum lamp (1903) . By 1904 Tungsten had begun to establish itself as the most suitable filament material, and although there was one fundamental improvement in its manufacture (the change from the squirted filament to the drawn wire which began in 1906), its supremacy has not so far been challenged.
Side by side with this improvement in the filament, there has gone improvement in the manufacture of the whole lamp. The earlier lamps were hand made ; gradually automatic methods have been developed, and the cost has fallen progressively while the performance of the lamp has improved.
By 1927 the total world consumption of electric lamps had reached a figure of 95o millions per annum, and manufacture on this scale is inevitably conducted on lines of specialized mass pro duction. Now it is not within the scope of this article to discuss the methods of mass-production—they are mainly matters of organization,—but it is relevant to explain that the success of mass-production depends largely on a knowledge of the scientific principles underlying its processes and products. In what follows, therefore, these principles are explained in detail in their relation to the actual methods and materials used in the manufacture of lamps.
The first essential, then, is to find the filament material which can best stand these conditions, and tungsten, with its exception ally high melting point (3,38o° C) is the outstanding material for the purpose.
The limiting temperature in practice is far below the melting point of the filament ; long before that temperature is reached other factors come into play which tend to shorten the life of the filament ; these are evaporation of the metal (which both thins the filament and obscures the glass bulb), and loss of mechanical strength. Of these, evaporation is the more important, and for a life of i,000 hours, it sets a practical limit of about 2,100° C to the temperature at which a tungsten filament may be maintained in a vacuum. (Carbon has an even higher melting point than tungsten, but it fails at a much lower temperature through disintegration.) The use of a vacuum has this advantage, that the only loss of energy from the filament is by radiation. On the other hand, if the filament is surrounded by an inert gas (i.e., one which will not attack it chemically), its temperature can be raised consider ably higher than is possible in a vacuum before the loss of tung sten by evaporation becomes serious. It is true that the vapour pressure of the tungsten has gone up with the temperature, but in spite of this, evaporation is suppressed by the surrounding gas, the molecules of which prevent the tungsten molecules from leaving the filament permanently as freely as they otherwise would. But this is not the whole story. The use of a gaseous atmosphere instead of a vacuum allows the temperature of the filament to be raised, but at the same time in addition to the heat wasted by radiation, a certain amount is now carried away by conduction and convection.
These terms require some explanation. Heat carried by conduc tion is passed on from molecule to molecule of the gas, just as happens in a solid. Convection on the other hand is the transfer of heat by moving streams of gas. The gas near the filament becomes heated by conduction ; its density decreases and it rises, cool gas taking its place. The hot gas goes on rising till it reaches the walls of the bulb, where it is cooled, and sinks again till it reaches the filament; thus a continuous cycle is set up, which is termed convection.
Now in general the amount of heat carried away from a hot body by a gaseous atmosphere is proportional to its surface area; but in the case of a thin filament the statement requires quali fication. Langmuir showed that any hot body in a gas is sur rounded by a stagnant layer of gas ; heat passes through this layer by conduction and is then carried away from its surface mainly by convection. It is therefore the surface area of this gas layer, and not of the solid body itself which determines the amount of heat lost.
Applying this knowledge to the case of a filament, it is found that the thickness of the stagnant gas layer, which is independent of the diameter of the filament, is very large compared with the diameter for ordinary sizes of lamp, so that for thin filaments the advantage of the extra temperature attainable in a gaseous atmosphere is more than offset by the greater convection losses. The use of thick filaments is in most cases not practicable, as they are suitable only for very large currents, but it was found that a thin filament coiled into a close spiral behaves, so far as con vection losses are concerned, almost exactly like a solid filament of the same diameter as the spiral, while the spiral form allows the wire to be long enough to have the necessary electrical resist ance. It thus becomes possible to use a gas to cut down the evaporation of the filament and at the same time to obtain the increased efficiency due to the higher temperature.
This then is the principle on which the so-called "gasfilled" lamp depends. In order to make fullest use of it, the right gas must be chosen. For all but the largest sizes of lamp, argon (mixed with a little nitrogen) is used. Argon is chosen because it is chemically inactive, and being a heavy gas it carries away less heat from the filament than would a lighter gas. The nitrogen is added to lessen the tendency for discharges to take place in the gas at the higher voltages. In the very largest lamps argon shows no advantage in efficiency over nitrogen, but as the latter re-acts slightly with tungsten, the argon-nitrogen mixture is frequently preferred even for large lamps in spite of its increased cost. The Rating of Lamps.—Lamps are classified according to the voltage on which they are intended to run, and are rated in watts (the watt is the electrical unit of power and equals , 4 s6 horse power). The ratings of lamps for use on supply mains range from about 15 watts to i,soo watts. Smaller lamps (e.g., motor car lamps, flash lamps and so on), and larger lamps (e.g., light house lamps and cinema studio lamps) are also made.
The light given by electric lamps is measured in lumens (a lumen is equal approximately to eight-hundredths of the intensity of the light emitted by a uniform source of one candle power. See article on "Photometry"), and the "efficiency" of a lamp is measured in lumens per watt. This efficiency is, as previously explained, dependent on the filament temperature, and it can always be increased (up to the melting point of tungsten) at the sacrifice of life.
It is therefore necessary to decide on a suitable life, and arrange for the filament to run at the temperature (and therefore effi ciency) which will give that life. It has been found in practice that the most economical life is somewhere between 75o and i,000 hours.
The size of the lamp introduces another factor. Lamps of large wattage can, owing to their thicker filaments and for other reasons, run at higher efficiencies than smaller lamps, and in practice the efficiency of lamps for a thousand-hour life ranges from 8 lumens Thus if a lamp is correctly designed to have a life of i,000 hours at the declared voltage of the electricity supply, its actual life will be boo or 2,000 hours according as the supply voltage is 4% high or 4% low.
Another difficulty is the existence of electricity supplies of many different voltages in different districts; its effect is, of course, to multiply unnecessarily the number of types of lamp which need to be manufactured. There is, however, a strong movement to wards standardisation, which, fairly complete in U.S.A., is mak ing great progress in Great Britain, but less in the rest of Europe.
Most bulbs are made of a "crown" or lime-soda glass, largely because this type of glass has a shorter working range than a lead glass, that is to say it remains soft over a shorter range of tempera ture. This property suits the automatic type of machines used for blowing bulbs and for sealing the bulb on to the internal glass work.
At the same time, there must be constancy of composition from day to day. A skilled glass blower blowing bulbs or assembling lamps by hand can vary conditions to suit the glass as he gets it ; automatic machines cannot do this without resetting; and if this is required too frequently the machines cease to be automatic.
A further factor necessitating constancy of composition is that the lead-in wires which carry the current to the filament have to be sealed through the glass, and therefore the coefficient of ex pansion of the glass must be kept closely within the limits which suit the wires used.
The lime-soda glass used for the bulbs is melted in tank fur naces having a capacity of about 8o tons. The blowing of the bulbs is entirely automatic ; the "gathering" of the necessary amount of glass for one bulb is done by suction applied through a long arm which dips momentarily into the tank and then trans fers the glass to a mould in which it is blown by compressed air into the correct shape. A later development is a bulb-making machine essentially different in principle from the one described. Instead of being a rotating turret indexing machine with ram operated arms moving back and forth from the furnace to the spindles, it is a tractor-operated continuously moving mechanism, which is fed by a continuous flow of glass from the furnace. The glass flows by gravity from the tank and passes through rollers, forming a continuous ribbon of glass.
Moving in synchronism with the glass ribbon and the blow-head conveyor is a conveyor containing the moulds for shaping the seg ments of the glass ribbon into bulbs. The completed bulbs are automatically conveyed, through the various succeeding processes to the inspecting and packing section.
This truly marvelous mechanism can produce as many as 44o bulbs per minute ; and since the machine runs continuously day and night when production from the tank is begun, the daily capac ity is far beyond the half million mark.
The glass tubing and rod used in the internal structure of the lamp is also made by machine; in this case the molten glass is fed continuously on to a rotating mandrel (hollow in the case of tube, solid for rod) from which it is as continuously drawn off in the state of finished tube.
The sequence of operations in the assembly of a lamp ready for pumping is shown in figure i , which is mainly self explanatory. Two points only need further mention. The lead-in wires are made of a nickel-iron alloy, copper plated and then covered with potassium borate.
The composition of the alloy and the thickness of the copper coating are so adjusted that the composite wire has approximately the same coefficient of expansion as the glass. The borate coating is added to allow the glass when melted to "wet" the wire and make a gas-tight seal.
The support wires are made of molybdenum, a metal of very high melting point (2,60o° C) and also tungsten. The molybdenum wire is made by a process similar to that described later for the tungsten filament. Its mechanical properties sometimes make it more suitable than tungsten.
This process is carried out on automatic machines consisting roughly of a turntable rotating step by step, the hollow stems of the lamps being inserted in suitable holders. As the machine rotates, these are connected successively with suitable vacuum pumps. It is not, however, practicable on a mass production scale to pump each lamp by mechanical means to the required vacuum, which is of the order of 0•000i mm. mercury. On the other hand, if the residual pressure is much greater (i.e., anywhere between .005 and I mm.) destructive discharges are liable to take place in the bulb.
The necessary high vacuum is therefore produced by physico chemical means after the lamp has been sealed off from the pump. The method is to apply to the filament, before or after mounting, a thin coat of red phosphorus mixed with a binder and an addi tional material such as cryolite or another fluoride. (The action of the cryolite is to keep as transparent as possible the film formed on the inside surface of the bulb.) The whole mixture is known as a "getter." When the lamp has been sealed off from the pump the operation termed the "clean-up" is carried out, during which the getter ful fils its function of completing the evacuation to the necessary high degree.
After basing, the lamp is tested, and it remains only to wrap the finished lamp.
In the higher wattage lamps the use of cement in fastening the base to the bulb has never given complete satisfaction. When subjected to repeated wide variations in temperature throughout the life of the lamp, and sometimes to other weakening influences such as vibration and moisture, even the high quality basing cement now used cannot always be depended upon to maintain its strength.
A new mechanical base provides much greater strength and much improved general all-around performance. Those factors which weakened the cement type base have no effect on the strength of the mechanical base throughout life.
This new base consists of an inner brass shell and a brass screw base. Four lugs on the inner brass shell snap snugly into four recesses in the neck of the bulb, one of the lead wires is welded to the inner shell, and the brass screw base is screwed onto the inner shell and drawn up flush with the neck of the glass bulb, holding the lugs firmly in the recesses in the bulb. The second lead wire is welded to the end of the base, and the base and inner shell are pierced, and thus locked together.
Although most lamps under 500 watts have the regular threaded base, a new base, termed bi-post, is used on practically all of the high wattage lamps that require the accurate positioning of the light source with relation to reflectors or lenses.
The bi-post base represents a development in high wattage lamp construction that differs radically from the fundamental princi ples of lamp design as followed ever since the first lamp was built. The traditional lamp manufacturing process has been fol lowed in the past because such construction had satisfactorily fulfilled every need. Even when the cinema and aviation made de mands for high-powered lamps in sizes far above standard prac tice, the development followed naturally along old established lines. However, when it is understood that the largest incandescent lamps made contain three pounds of heavy tungsten metal, or enough to make forty thousand 50-watt lamps, the matter of fila ment weight introduced an entirely new factor in lamp design. Ob viously this development did not happen overnight. In reality it was the result of months of intensive research, which in the end resulted in a lamp of fewer parts and a change in the entire opera tion of lamp fabrication.
This new construction insured a lamp of more rugged character through the omission of many of the conventional parts, which were formerly centres of weakness. For instance, the conventional base, which was secured to the bulb by cement or clamping, is eliminated and replaced by a glass cup, shaped like a pie plate, with two metal posts sealed to it. Connection to the socket is made through the lower part of the post, which consists of a cylindrical prong with a shoulder for accurate seating.
Then, too, the filament and lead wires were originally supported by a glass stem structure. This necessitated the use of special glass and special lead material, having a coefficient of expan sion the same as the special glass. To get the heavy current into the bulb, it was necessary, therefore, to have a leading-in wire composed of a section of copper, of tungsten, of nickel, and in some types a section of molybdenum. The new construction eliminated the stem seal and multi-material lead and substituted a supporting structure made of one piece of channel nickel. Such construction gives maximum strength to the long leads carrying the heavy filament and provides a greater exposed surface to dissipate the heat and prevent over-heating. Lastly, the metal prongs which hold the lamp in the socket have the nickel channel leads welded directly to them and they, themselves, carry the weight of the entire metal structure. Formerly the glass assumed the burden.
The development is far-reaching in its significance. For, in addition to increasing the ruggedness of the lamp, it also in sures a lamp that is inherently prefocus in character; the rela tive position of the base and light source being determined entirely by metal-working accuracy.
Very small amounts of water vapour can seriously shorten the life of a gasfilled lamp ; this is due to a cyclic action in which the water vapour acts as a catalyst. The mechanism of the action is worth describing in some detail, because water vapour attack was, and to some extent still is, one of the most difficult problems to overcome in the manufacture of gasfilled lamps. Briefly, the water vapour attacks the hot filament forming hydrogen and tung sten oxide ; the latter volatilises and leaves the spot where it was formed. Its rate of travel through the stagnant layer of gas round the filament is however very slow, and by the time it has travelled a short distance away it has cooled enough for the reverse action to take place; i.e., metallic tungsten is again formed, together with water vapour, the latter being free to start a new cycle. The tungsten usually finds its way to an adjoining part of the fila ment, where it builds up a thick patch. The action is cumulative since the hot spot first attacked tends to become thinner and therefore hotter, so stimulating the attack. The only preventive is very careful pumping—great care in drying and purification of the wash-out and filling gases and extreme cleanliness of metal parts. The clean-up method is useless because of the presence of a gaseous atmosphere, nevertheless phosphorus is frequently used as it can combine chemically with traces of free oxygen.
The Manufacture of the Tungsten Filament.—Although metallic tungsten can be liberated fairly easily from its ores, the high melting point of the metal has necessitated the development of quite special methods for its conversion into wires and other usable forms. The earliest tungsten filaments were made from tungsten powder by mixing it with a binder into a paste which was squirted through dies ; the binder was afterwards burnt off leaving a coherent but fragile filament. The squirted filament was super seded by drawn tungsten wire.
The first step is to produce a pure tungstic oxide, which is re duced in hydrogen leaving tungsten in the form of a fine powder. The powder is pressed hydraulically into rectangular "slugs" which may be anything from 8 in. to 12 in. long and from a in. to 2 in. square; these on removal from the mould are just strong enough to be handled. The slugs are hardened by heating in a hydrogen atmosphere to about 1,200° C, and then "sintered" by the passage of a current great enough to bring them to nearly 3,000° C (400° below the melting point). During the sintering process the grains of metal powder grow together, and the sintered slug or ingot has a coarse crystalline structure.
The size of the grain can be controlled by many factors such as the fineness of the original oxide, and of the tungsten powder, also the temperature and time of sintering, and can vary between a microscopically fine structure and one so coarse that the whole ingot may consist of a few large crystals. At this stage, however, the individual crystals are not the homogeneous close-packed molecular lattices usually associated with crystal structure; under the microscope voids are visible, and the density of the ingot is only 17-18 as compared with 19.32 for a perfect tungsten crystal.
The ingot is next "swaged"—an operation consisting in ham mering the heated ingot by means of a pair of rapid-acting me chanical hammers. The swaging operation is repeated many times, the hammer clearance being gradually reduced as the ingot be comes a thin rod; the swaging temperature is also reduced pro gressively from 1,50o° C heat to a dull red (about 900° C). The final product of swaging is a rod about 0.75 mm. (or o in.) in diameter; the metal has been consolidated and at the same time the crystal axes have been elongated in the direction of the axis of the rod giving a fibrous structure.
The thin rod is now strong enough to be drawn through dies. The drawing process is similar to that normally employed for other metals ; the main difference is that the wire is heated to a temperature which gradually falls from about i,000° C to 60o° C as the diameter decreases; the dies are made of diamond or similarly hard material and are kept heated. As the drawing progresses the metal becomes increasingly ductile, in which re spect it appears to differ from other metals, which if drawn with out annealing become harder. The structure of the finished wire is in effect a bundle of fine elongated crystalline threads.
The spirals used in gasfilled lamps and in some vacuum lamps are made by coiling the wire. Taking the 6o-watt lamp as an ex ample, the tungsten is first drawn into a straight wire 19/10,00o of an inch in diameter, so fine that it is almost invisible to the naked eye. The tungsten wire is then wound 335 turns to the inch, around a mandrel, leaving the coils i/i,000 of an inch apart. The coils must be kept as close together as possible to reduce heat loss but they must not touch each other, for this would result in a short circuit, causing the lamp to fail instantly. A re cent development is the recoiling of the coil. This is done on an other mandrel, 7o turns per inch, with a spacing of 7/1,00o of an inch between the secondary coils. Before the first coiling, the tungsten wire is approximately loin. long. The first coiling com presses it to 3.4 inches. The second compresses it still further to a length of s of an inch with a coil diameter of 310/10,000 of an inch. Following the second coiling the mandrels are dissolved by means of chemicals. When installed in the lamp the coiled coil filament is mounted as a cross bar between two lead-in wires with one support in the centre of the filament instead of being looped around three supports as was the practice with the older type of coiled filament.
Curiously enough, this end can be achieved in two entirely dif ferent ways. One is to add thoria as before, but with an alkali salt such as sodium or potassium chloride in addition. The other is to add to the tungsten a small amount of silica and an alkali salt (such as sodium chloride). In this case the additives do not remain in the finished wire ; apparently they boil out during the sintering process carrying away other impurities, and leaving the tungsten purer than it can be obtained without their use. As a result of these developments filaments with non-sagging character istics became available, thus greatly improving the efficiency and lumen maintenance throughout life.
No comprehensive theory has yet been established to explain this phenomenon of rapid crystal growth in coiled filaments.
In the Rating Test for uniformity of manufacture, the lamps are burnt at the rated volts marked on them, and their consumption in watts and efficiency in lumens per watt are observed. These must fall within definite limits laid down in standard specifications.
Life Tests are naturally carried out on only a small proportion of the lamps made. The lamps so tested are burned both at rated voltage and at high forced voltages; the latter to shorten the time of testing and thus decrease the cost of testing. From such normal and forced life tests the quality of the product can be constantly checked. The lamps are run to burn-out, the lumens being measured at regular intervals. (C. C. PA. ; D. M. WA.)