A New Amplifier for Centimetric Wavelengths.
The work forming the subject of this article has been carried out for the Admiralty and is published by permission. It was initiated at the Physics Department, Birmingham University in I942 and carried on, after a move in I944, at the Clarendon Laboratory, Oxford University. All the investigations described were completed before the end of I944, unless otherwise stated.
A W7/2D travelling wave tube seen without the guide magnets and collector heat sink.
Two main urges lie behind the trend to ever higher frequencies: the need for a wider range of frequencies in the transmission of intelligence of ever-increasing complexity, and the need for narrow beams of radiation. For instance, in order to transmit intelligence such as television signals, which at present comprise practically all frequencies up to four or five MHz, we have to use carrier frequencies many times higher than the highest signal frequency. When the definition of the television picture is increased, as it surely will be, carrier frequencies will have to go up, too, in order to accommodate the wider band of frequencies which will have to be transmitted. The other main reason is that very high frequencies, or wavelengths of the order of centimetres, are essential for radar or radar-like applications if the structures which are to be used as aerials for transmitting and receiving are to be of reasonable size. Such structures have to be of a size of many wavelengths if a beam of radiation sufficiently narrow is to be produced. The valve is the heart of modern wireless, and the key to all these developments.
The need for centimetre waves became apparent some time before the war, and considerable effort was directed towards developing valves which could be used as amplifiers at these wavelengths. Oscillators, it was thought, would follow more or less automatically from amplifiers, since to convert an amplifier into an oscillator all that would be needed would be some feed-back from the output to the input in the correct phase to sustain oscillations.
Fig. 1 - Operation of the Klystron amplifier.
The Klystron (see Fig. 1) was Klystron as an oscillator and as an amplifier. As an oscillator not enough power was forthcoming. The conversion of the DC power of the beam (i.e., the product of beam current and beam voltage) into available RF power was considerably less efficient than simple theory predicted, and further, the DC power itself is very limited, due to the necessity for a long and narrow electron beam.
An early Klystron local oscillator - NR89.
Space-charge forces set a definite and rather low limit to the beam current, which can be passed through a number of small apertures such as occur in the Klystron. Thus the power output from Klystrons is frequently reckoned in Watts rather than in kilowatts.
Fig. 2 - The Magnetron oscillator.
This was not good enough for radar and the invention of the multi-resonator Magnetron (see Fig. 2) superseded the Klystron in an incredibly short time as a high-power oscillator.
Apart from the utilization of a novel principle, the reasons for the enormous RFpowers which can be obtained from the magnetron are: the circulating electron current in the valve is very largely due to its cylindrical geometry and the efficiency as a converter of DC into RF power is very high. Thus the power output of magnetrons is usually reckoned in hundreds of kilowatts, and it is certainly true that, without the magnetron, radar as we know it now could never have arisen. How the Battle of the Atlantic, and the Battle of the Ruhr would have gone without radar, is not difficult to estimate, and how the war would have gone without winning these two battles is left to the reader's imagination. The writer apologises for this digression, but believes that it has some bearing on the story.
An early cavity magnetron - CV76.
It is due to the enormous RF powers obtainable from the magnetron as transmitter, that comparatively little attention was paid to the receiving end of radar. Putting it in a somewhat exaggerated way: it just was not important enough to put any great effort into improving the sensitivity of radar receivers, since there was enough power - and some to spare - available from the radar transmitters to do most of the jobs radar was required to do. As an RF amplifier, the Klystron is very insensitive, and when it was shown that the ordinary crystal-and-catbs-whisker detector, used as a mixer or converter in a superheterodyne receiver, was better in respect of sensitivity than the Klystron, or any other valve, interest in RF amplification dwindled to next to nothing.
One of the main reasons for the lack of sensitivity of the Klystron as an amplifier was the inevitable inefficient energy exchange between the electron beam and the electric field in the rhumbatrons. Before examining this question, it will be advantageous to define what is meant under sensitivity. The sensitivity of a practical receiver is defined in terms of how many times the noise power at the input exceeds the noise power at the input of an ideal receiver (i.e., one with the hypothetical minimum noise power). This factor is called the noise factor; radar receivers-using crystal mixer input have noise factors somewhere between 10 and 100, while if a Klystron is used as RF amplifier at the input, the noise factor is somewhere between 1,000 and 10,000.
Little was therefore to be gained by improving radar receivers in respect of noise factor, since, at the most, an improvement of a factor 3 or 4 could be expected over a noise factor of, say, I0. (It seems unlikely that the ideal will ever be reached at the centimetric wavelengths.) On the transmitter side, with magnetrons, an increase in output power of 3 or 4, however, was relatively easily obtainable, and hence most of the research effort went in this direction.
Therefore it is not surprising that it was relatively late when it was realized that one of the basic principles of the magnetron, namely, that of interaction between a travelling field and an electron stream travelling at about the same velocity, could also be applied to the amplification of weak signals at the receiving end, making possible an amplifying valve of sensitivity comparable to that of the best crystal mixer receivers.
A detailed examination of the Klystron as RF amplifier showed that the inefficient energy exchange between the electron beam and the electric fields in the rhumbatrons is due to the long time taken by the electrons in crossing these fields. Once this time, often called the transit time, approaches the time of a period of the oscillation, it is clear that the electron will gain during one-half of the time as much energy as it will lose during the other half, and the net result of the energy exchange will be nil. This is the same difficulty that lies behind the decrease in efficiency of the conventional valve at the higher frequencies.
It was therefore a very inviting thought to use the signal in the form of a travelling electric field (instead of a stationary one) and utilize the energy exchange between the travelling field and electrons which travel at about the same velocity.
Here we can find an analogy between valve development and wireless on the one hand, and engine development and flying on the other; it can, perhaps, be said that the step from interaction between stationary fields and electrons to interaction between travelling waves and electrons is reminiscent of the step from the reciprocating internal combustion engine to the gas turbine. And just as in the case of engines, the idea of travelling waves had been suggested many times in many different forms, but little had been done about it. However, with the advent of the centrimetric wave technique the time was ripe and the logical development followed.
The first point to be noted about waves travelling with about the velocity of electron beams is that 'reasonable' electron beams travel much slower than ordinary electro-magnetic waves. Under 'reasonable' we understand electron beams of voltages between a few hundred and a few thousand volts. The actual relation between velocity ν of an electron and voltage V needed to give it that velocity, for non-relativistic velocities, is:-
ν =5.95 &mult; 107 √V cm/sec
where V is the beam voltage.
Thus 2,500 Volts will give electrons a velocity of about 3.109 cm/ sec; just about one-tenth of the velocity of light in free space.
Means therefore had to be developed for slowing down waves by roughly a factor 10, and the simplest and the most readily available structure was found to be a helix of conducting wire. Within wide limits, a wave will follow the wire, clinging to it, as it were, and therefore the progress of the wave in the axial direction is considerably slowed down as compared with the velocity along the wire, which is near enough equal to the velocity of light.
Fig. 3 - Lines of force of a wave travelling along a helix (not to scale).
A picture of the actual lines of electric force, which constitutes the travelling wave, is given in Fig. 3. It has to be realized that these lines of force move in the following way: they rotate about the axis of the helix and they progress from left to right. However, if we imagine ourselves also moving from left to right, with the same velocity in the axial direction as the wave, we will continuously experience the same force. For instance, if we choose to travel with the point A, we will always experience an accelerating force (if we are negatively charged), whereas at the point B we would always experience a retarding force. At other points we would experience appropriate other forces as indicated by the direction and density of the lines of force.
Now let us consider what will happen to a thin electron beam travelling along the axis of the helix. Some portions of the beam will continuously be accelerated, others will be continuously retarded and the regions in between these two will experience neither one nor the other force. The inevitable result of the action of these forces is a displacement modulation of the beam, or, as it can be regarded, a density modulation. There are regions of the beam to the left of which the electrons are accelerated and to the right of which they are retarded. In these regions the density of charge will increase above the average. Correspondingly, there will be regions in which the converse conditions apply and the density of the charge will decrease below average. Thus the beam can be considered to have become amplitude modulated - the actual amplitude of the AC component of beam current growing at first approximately with the square of distance, reckoned from the beginning of the helix.
Experiments which were carried out with a helix of 18-SWG copper wire of 9 mm outside diameter, and an electron beam of a few micro-amperes and 2,400 Volts, showed that this view of the action of wave on beam was substantially correct, and that the resulting modulation of the beam could be made considerably more effective than the modulation to be expected from any practicable rhumbatron.
Now since it has been shown how AC energy can be imparted to an electron beam by means of a wave travelling along a helix, the opposite process, namely the extraction of AC energy from a modulated beam by means of a helix can be expected to occur with equal efficiency.
A rough picture of this process is as follows: Imagine a charge brought suddenly into the space within the helix. This charge will connect to the helix by means of lines of force, and these lines of force will spread out in time, one lot travelling along the helix in one direction, say, from left to right, the other lot travelling in the opposite direction. In other words, two waves are excited. Now, let the charge move along the axis of the helix, from left to right, with the axial propagation velocity of the waves. Then fresh lots of waves are continuously being excited; the ones travelling from right to left annihilate each other. The ones travelling from left to right, however, reinforce each other, building up a wave front of continuously increasing amplitude.
Now an amplitude-modulated beam can be regarded as a continuous uniform beam upon which there is superimposed a system of alternate positive and negative charges distributed in a sinusoidal manner. Each of these positive or negative charges will excite a wave as described before; the complete 'induced' wave can be synthesized from the contributions due to the individual charges.
The pictures given so far, of the action of the wave on the beam, and the action of the beam on the wave, are only first-order approximations. They are nearly true when either the beam is very weak, or the helix very short; in reality these two actions are always present simultaneously, and the real picture is one of interaction between beam and wave. Suppose we start with a wave and an unmodulated beam. Almost at once there will be an amplitude modulation in the beam, and this will, also almost at once, induce a wave in the helix. This induced wave will again cause a beam amplitude modulation, and so forth, ad infinitum.
The theory which describes the complete process gives as result a wave which increases exponentially with distance, after some initial deviation from the exponential law.
In any real helix, in the absence of an electron beam, a wave will be attenuated; that is, its amplitude will decrease exponentially. In the presence of an electron beam travelling at about the same velocity as the wave, the wave will increase exponentially. Hence one is justified in saying that the presence of the beam causes negative attenuation to be introduced into the helix.
Thus, by making either the helix long enough, or the beam current large enough, one can obtain a wave amplitude at the end of the helix which is substantially larger than that at the beginning. Hence the system helix - electron beam is an amplifier. Some of the DC energy of the beam is converted into AC energy in the wave by means of the interaction between wave and beam. A detailed examination shows that more electrons are being retarded than speeded up and thus the law of conservation of energy is satisfied.
Fig. 4 - Layout of the travelling wave amplifying tube.
The travelling wave tube, (see Fig. 4), as the complete device is called, consists in practice of nothing but a long and straight helix of wire supported in an evacuated glass envelope, containing also an electron gun for producing an electron beam and a collector for collecting as much as possible of the beam. Outside the tube proper are devices for matching the input and output leads to the helix, which may take many forms and are only indicated symbolically. It is, of course, important to procure proper matching (that is, reflectionless transitions) from input and output to the helix, as otherwise oscillations are easily excited when the helix is effectively an integral number of half-wavelengths long. Even if oscillations are not excited reflections at the ends will cause selectivity of the tube - that is, enhanced amplification at particular frequencies, destroying the broad-band amplification characteristic of the travelling wave tube, which may be a desirable property on occasion.
With Klystrons, amplification and bandwidth are conflicting requirements, each being roughly inversely proportional to the other. This is chiefly due to the fact that rhumbatrons are resonant structures, usually having rather high 'Q' values, in order to get high field strengths across the gap; but this unfortunately means that they will only amplify a relatively narrow band of frequencies.
The travelling wave tube, on the other hand, is in principle, a completely untuned device and the range of frequencies over which it amplifies is mainly determined by the 'broadness' of the matching arrangements at input and output. Therein lies the importance of the travelling wave tube for the communications of the future.
The initial work undertaken here in England was mainly concerned with the travelling wave tube as a sensitive amplifier and the first tube - a demountable model continuously evacuated - was first tested in December, I943, at the Nuffield Laboratory, Physics Department, Birmingham University.
The helix, 60 cm long, of 18-SWG copper wire, 5 turns/cm, was wound on a 0.25 in mandrel, and a power amplification of 6 was given with a noise factor of 20. This happened at a beam voltage of 1,830 Volts and with a beam current of 110 micro-amperes. The signal wavelength was 9.1 cm.
A short magnetic coil was used to focus the beam so that it impinged on the collector with the loss of only a few micro-amperes on the helix. Thick soft-iron shielding was necessary to keep away stray magnetic fields which are very troublesome with such a long and narrow beam.
A later tube, also continuously evacuated, gave a power amplification of I4 with a noise factor of 12, thus coming very close to the performance of a good crystal when used as a mixer.
The physical reason for the good noise factor of the travelling wave tube lies in the superior efficiency of energy transfer between electron beam and wave. Thus sufficient amplification is obtained with but a fraction of the beam current that would be required for a Klystron giving comparable gain. Less beam current means less shot-noise, and therefore less noise is added to the signal in the process of amplification than in a Klystron.
Work similar to that reported on above has since been undertaken by Bell Telephone Laboratories, New York, and results obtained there with travelling wave tubes have been quoted recently by J R Pierce and L M Field on the occasion of an Electron Tube Conference convened by the Institute of Radio Engineers of America.
Using a helix of iron wire of 30 cm length, with an initial attenuation of 33 dB, and an electron beam of 1,6oo Volts and 10 milli-amperes, a power amplification of 200 was obtained over a bandwidth of 800 MHz to points 3db down on the gain/frequency curve. No particular attention was paid to noise factor. However, an RF output of 200 milliwatts was obtained. A long solenoid was used to get the beam through to the collector. The mid-frequency was 4,000 MHz.
At this early stage of the development of the travelling wave tube, it is difficult to foresee all the possible applications and the consequences which will undoubtedly follow in their train. However, one particular field seems to have been waiting for just such a tube as the travelling wave tube, and that is television and multichannel communications transmission via centimetre-wave links or waveguides. Here the travelling wave tube might eventually play a part not unlike that of the magnetron in centi-metric wave radar.
The writer is indebted to many for very helpful discussions and interest; at Birmingham chiefly from Professor P B Moon, Dr R R Nimmo, Professor J Sayers and Dr G Voglis, and at Oxford Dr J H E Griffiths, Dr A H Cooke and Dr B Bleaney. Very able assistance in the actual work, experimental as well as theoretical, was given by Mr E E Vickers, Mr J Hatton and Mr H Ashcroft at various times.