|
Below a certain limit of wavelength it is physically impossible to use valves of ordinary type, and even at wavelengths considerably above this limit there is a distinct advantage in using special patterns. How these valves differ from those designed to operate at normal wavelengths is explained in this article.
From 150 to 15,000 metres nobody thinks of varying the choice of valves to suit the wavelength; While all-wave sets make one lot of valves cover everything down to 15 metres. But try taking the next decimal step to 1.5 metres!
It is just possible to make some ordinary types of triode valve 'go down' to 1.5 or 2 metres as oscillators, by taking special care about it. It may even be possible, but not very comfortable, to use such valves in simple receivers down to 3, or perhaps 2 metres. The tetrode or pentode valve of normal design can only with difficulty be persuaded to give the slightest trace of amplification below 10 or 12 metres. More elaborate valves, such as frequency-changers, are not beyond criticism at longer wavelengths still. Even where a standard valve performs quite well at, say, 5 metres, a specially selected type of valve is likely to do considerably better.
Why?
There are a number of ways in which valves of ordinary design increasingly withdraw from useful business as the wavelength becomes shorter. The chief of these concerns the associated tuning circuit. This (to run over familiar ground once more) is made up of inductance L and capacity C, which between them decide the wavelength A to which the circuit responds, according to the famous old formula λ = 1885 √LC where L is in microhenrys (μH) and C in microfarads (μF). In ordinary (that is to say, medium- and long-wave) practice, L and C are considered to reside chiefly in the tuning coil and capacitor respectively. It is true that one often has to take account of 'stray' capacity and perhaps inductance due to the wiring and to the electrodes of the valve, but these are generally subsidiary in effect.
To fit circuits for shorter waves, L and C must be appropriately reduced and so in shortwave sets we have coils with fewer turns of wire, and capacitors with fewer or smaller plates. But obviously if these quantities are progressively reduced, whereas the valve is retained unaltered, the proportion of L and C residing within the valve must grow from a small and almost negligible one to a condition of equality, then of predominance, until finally the valve's inductance and capacity alone are greater than are needed for the wavelength.
Valve Capacity
Take some actual figures. To tune to 400 metres one may conveniently use an inductance of 150 μH and a capacity of 0.0003 μF, or 300 pF. Thinking in μμF, or μF will become clumsy in a minute or two, the L/C ratio is 150/300, or 0.5. Of the 300 pF the valve contributes perhaps 10 at most; only about 3% of the whole. Now drop to 40 metres. To preserve the same L/C ratio for if it is greatly reduced the impedance of the circuit becomes too low, and self-oscillation (if required) impossible - the tuning circuit must comprise 15 μH and 30 pF, so that now the valve capacity is 33% of the whole. Taking another drop to 4 metres, on the same assumption the valve capacity is more than three times as great as the whole tuning capacity. Which, like certain steps in Euclid, is absurd.
To get out of this impasse we must sacrifice L/C ratio and obtain the short wavelength by disproportionate reduction in L. The tuning capacity can hardly be got below 20 pF allowing for wiring, self-capacity of coil, and a mere trace for tuning control. The inductance must then be only 0.225 μH. By this time the inductance of the valve is quite an appreciable part, and, unless the length of leads is reduced almost to nil, there very little change out of our allowance of 0.225 μH with which to provide the tuning coil itself. The absolute limit is reached when all the L and C exist in the valve and its irreducible minimum of external connections; and this, of course, varies with the type of valve, but for ordinary receiving valves is in the region of 1 metre. In practice it is impossible actually to reach this limit, for a valve provides quite a lot of capacity but very little inductance; in other words, the L/C ratio is so low that it is impossible to set up any useful HF voltage across it.
This restriction could be at least partly offset if it were possible to make a considerable reduction in the HF resistance, or losses. Unfortunately, it is difficult to prevent the loss resistance from being actually greater than usual, because, instead of the capacity being concentrated in a tuning capacitor where the dielectric (insulating material between the plates) can be confined chiefly to air, which causes negligible loss, a considerable proportion is made up of miscellaneous items such as the capacity be tween valve pins, with comparatively high-loss Bakelite dielectric. Hence the practice of valve de-capping indulged in by short-wave enthusiasts.
Somewhere around the wavelengths where these difficulties appear in acute form we run into another trouble. The normal action of a valve depends on a time-table which allows nothing, or next to nothing, for the journey made by the electrons across the small internal gap between filament and anode. Now, while the electrons shoot across this small fraction of an inch at a pace which makes a high-velocity bullet seem like a fatigued snail in comparison, it must be remembered that at a wavelength of 3 metres a complete up-and-down cycle of oscillation is all over in 1/100,000,000th of a second. This certainly does not leave any time for dawdling on the way. Actually, it is found that the electrons arrive at the anode late by an appreciable fraction of this period, and at still shorter wave-lengths may not turn up until the right moment for assisting the oscillation to keep going has passed. So the oscillations fail to go. Even at a wave as long as 22 metres it has been shown that this effect is equivalent to shunting the tuned circuit with a resistance, lowering its efficiency.
For the very shortest waves (usually called micro-waves, as an escape from some such dreadful term as ultra-ultra-short waves) below one metre, special types of valves - magnetrons. are used which work entirely differently from the common sort, and, in fact, actually depend for their existence on the time lag that is the cause of the others downfall. 'One manbs meat . . ., etc.', again. This leads to rather amusing results; for example, the tuning is done by varying the HT Volts instead of a capacitor. But it is not a type of valve which has any great appeal for the amateur as yet, so we pass on.
The straightforward way of arriving at a valve more suitable for short waves is to look at a standard model through the wrong end of a telescope, and produce a reduced scale model. Suppose every linear dimension is reduced to a half. Then the distances between the electrodes, and hence the time taken by electrons in crossing them, are halved. That is one good thing. But when the plates of a capacitor are spaced by half the distance, the capacity is doubled, which seems to be not so good. In this case, however, the area of the plates is reduced to a quarter; so, taking both these factors into account the inter-electrode capacities are halved. Good again.
The inductances are reduced in still better proportion. And the smaller clearances make up for the reduced cathode area in maintaining the valve characteristics at a high level. The only drawback (apart from the nimble fingeredness required in manufacture) is that the power - handling capacity is lowered. But except for transmitting nobody need worry about that. Another point is that with the generally miniature components - that are appropriate for ultra-short waves, tiny valves lend themselves to the construction of very neat portable equipment.
The Hivac Midget valves have proved to be well suited for work at ultra-short wave lengths.
For these reasons the Hivac Midget valves have obvious advantages. They are available in three varieties of triode of high, medium, and low impedance (XD, XL, and XP respectively), a screen-grid tetrode (XGS), and an output pentode (XY).
Figures showing how the inter-electrode capacities compare with those of full-size types are seen in the table.
By going the whole hog in valve diminution some remarkable results have been achieved, such as 'straight' reception at 0.4 metres, and a several-fold HF amplification per stage at 1 metre!
The Acorn miniature valve, as first produced for operation on ultra-short wavelengths.
The valves responsible for this performance are described, for reasons obvious on inspection, as Acorns; and besides being small compared with normal types they are laid out rather differently in order not to throw away in the leads and mountings any advantages obtained in the valves proper. Two varieties are obtainable, a triode and a HF pentode, both indirectly heated. Being of American origin, the heaters are for 6.3 Volts (a 6-Volt accumulator will do) and 0.15 Amp. The characteristics, judged merely for ordinary valves, are remarkably good; but especially so in view of the conservative characteristics of most American types. In fact, as the table shows, they excel the full-size valves.
One of the most decisive points of superiority is not listed above. The equivalent shunt resistance due to the electronic time lag, according to measurements that have been published, is fourteen times as great in the case of the 954 Acorn, and the loss therefore correspondingly smaller.
The tiny coil on the right is appropriate for use in conjunction with an Acorn valve for tuning to a wavelength of 0.8 metres. The comparison with a normal broadcast coil is interesting.
It is rather amusing to consider the size of coil suitable for tuning to 0.8 metre, as shown in the accompanying illustration, where it is compared with an ordinary broadcast receiver type.
Acorn valves are now obtainable in this country, but only for bona-fide amateur or experimental use.
It must not be forgotten that the original object of the superhet was to dodge the difficulty of short waves by 'converting' them to longer ones. But having made full use of this expedient there is always the irreducible minimum of the input portion of the frequency-changer, which must function at the original wavelength, and also the oscillator, whose wavelength must be very near.
It is just that nearness, which must never degenerate into actual identity, that is part of the difficulty. When two circuits are tuned to very nearly the same wavelength the slightest coupling between them produces a tendency for the circuits to pull one another, shifting the wavelengths. This is a most undesirable tendency in a short-wave superhet, because a very small change of this sort produces a very large proportional change in the intermediate frequency. As both of these tuned circuits input and oscillator are connected. to the same frequency-changer valve, it is difficult to avoid a trace of capacity between the two. The coupling via such capacity increases in proportion as the wavelength is reduced. An important point to consider in a short-wave frequency-changer valve, then, is smallness of capacity between oscillator and input control grid.
Next, for reasons given earlier, there is increasing difficulty in obtaining self-oscillation as the wavelength is shortened. So the oscillator section of the valve should have a maximum of mutual conductance and a minimum of HF loss.
Thirdly, the wavelength of the oscillator at any setting of the tuner should be very stable. It should not, for example, change as the bias on the control grid varies during the process of AVC, or fading would have the effect of mis-tuning the required programme. Nor should it be subject to drift as the valve slowly warms up, or a station tuned in at the start will gradually be lost.
Although it is not impossible to use the heptode type of frequency-changer down to fairly short waves - 13 metres or thereabouts - the limitations just set forth are not negligible. The oscillator section being formed by two of the grids in the electron stream, one is not so free to design it for maximum 'oscillability'. And for the same reason it is difficult to reduce the capacity to the remainder of the electrodes as much as one would like.
The triode-pentode is undoubtedly the best for short waves, according to the above criteria. The two sections, being as much separate as they possibly can be within one bulb, are freer from interactions than any other type, and the triode characteristics can be made to measure. It is unfortunate, therefore, that it is just when these merits are of outstanding advantage at the shortest waves that difficulty is experienced in arranging the mixing of oscillator and signal waves. One proposed solution is that described by W T Cooking in The Wireless World (March 8, 1935), in which two heptodes are used in push-pull. To avoid the necessity for two valves the triode-hexode, exemplified by the Osram X41, has been introduced, which is designed to combine the merits of the separate oscillator section in the triode-pentode with those of the electronic mixing principle of the heptode.
 
The X41 valve (triode-hexode) is especially suited for operation as a short-wave frequency-changer.
The capacity between sections is not quite as small as in the triode-pentode, but the interaction is much less than in the heptode, and it can, therefore be used effectually at a shorter wavelength. Unfortunately, a battery valve of this type requires a higher HT current than the heptodes for equal conversion gain, Since the HT consumption is of such importance the triode-hexode is not so likely to be favoured for battery receivers.
A Graham-Farish short-wave SG valve. The base is of low-loss material, and inter-electrode capacity is reduced by leading out the grid connection to the top cap.
|
