International Congress at Munich
A somewhat unusual general feature of the recent Munich Congress was the relatively large number of survey papers given. These should lend an added interest to the forthcoming publication of all the congress papers (in their original languages) as volume 22 of Nachrichtemtechnische Fachberichte. Since as many as 145 papers were given, three simultaneous sessions had to be run. Fortunately, by choosing what he hopes were the most novel and important papers, your reporter generally managed to avoid having to be in several places at once. For reasons of space, this report must be even more selective.
One of the problems of magnetron design is ensuring complete stability in the desired mode of oscillation. This is because the RF waves associated with the many unwanted modes (and indeed also with the wanted one) have velocities which remain constant round the anode structure. Electrons with a constant and suitable velocity can thus interact favourably with all modes. A general way of avoiding such unwanted interaction which was described in a paper by D A Wilbur et al is to make the velocities of the unwanted modes vary round the anode so that these modes cannot interact with electrons of constant velocity. (At the same time the velocity of the desired mode must of course be kept constant round the anode.) These conditions can be satisfied by making the anode up out of two or more sets of resonators whose frequency/phase shift characteristics are the same only for the desired mode. Two examples of anode structures which can be made to satisfy this criterion are a combination of forward- and backward-wave sections or alternatively, a number of different-sized cavities arranged in a random sequence. Tapering the cavity dimensions towards the end of a travelling-wave tube to suppress backward-wave oscillations in a somewhat similar way was discussed in a paper by M Chodorow et al.
A paper by W. E. Willshaw mentioned a way of facilitating third harmonic operation of magnetrons. This is done by making the anode up out of two sets of cavities - one set resonating at the fundamental frequency and the other at the harmonic. Twice as many harmonic as fundamental cavities are provided, so that electron bunches produced at the fundamental frequency have the correct angular velocity for interacting with the harmonic.
A paper by I Favalier described an unusual method of electrically varying the frequency of a klystron. This method uses an external tuning cavity in which a gas discharge is set up. To microwaves such a discharge looks rather like a dielectric whose permitivity depends on the electron density and thus also on the discharge current. By varying the discharge current one can thus also vary the tuning-cavity resonant frequency and consequently also the klystron frequency. The frequency of an X-band (∼ 9 GHz) klystron could thus be varied by more than 500 MHz by increasing the discharge current from zero to 250 mA.
A paper by O Doehler and G Mourier discussed the theory of valves in which an electron beam interacts with a slow-wave structure which is periodic in two dimensions. Such a structure is a more general version of a number of ordinary structures in parallel (in which case the distance between adjacent structures is the second periodicity) so that it should be capable of dissipating high powers. The gain was calculated for the case of the transversotron in which the RF energy is propagated in a direction at right angles to that of the beam. This device is reciprocal, i.e., can amplify forward- or backward- going signals equally well. Thus in order to prevent feedback and consequent oscillations, a non-reciprocal device such as an isolator must be included.
A paper by W Wendrich described a way of using such valves as mixer oscillators. (For a description of the backward-wave valve see, for example, the article by C H Dix in Wireless World, November 1959 (p. 478)). In this modification, the output (at the electron gun end of the slow-wave structure) is terminated in a matched load, and the input is fed to the other end of the slow-wave structure. Both the input and oscillator thus modulate the electron beam and, since the modulation is non-linear, combinations of the oscillator and input frequencies are also produced at the same time. The microwave sum frequency is absorbed in the output load, and the low difference frequency may be picked up at the input.
A paper by A W Trivelpiece showed that a longitudinally magnetized ferrite rod can propagate backward waves with velocities of the order of one hundredth of that of light. Interaction with such waves and a beam passing through an axial hole in the ferrite was observed. Such a simple slow-wave structure should be very easy to make.
Crossed-field (M-Type) Valves
Some of the work on the best-known example of this type of valve - the magnetron - has already been described in this report.
A paper by E Dench et al discussed the bitermitron. This is similar to a crossed-field backward- wave oscillator (for a fuller description of which see, for example, the previously mentioned article by C H Dix in our November 1959 issue (p. 480)) except that, instead of one end of the slow-wave structure being internally terminated in a matched load, both ends are taken out to external circuits. This difference accounts for the first part of the name, the iron being added, as the authors put it, because this is essential for satisfactory operation. The bitermitron can be operated as a conventional voltage-tuned backward-wave amplifier or oscillator. More interesting is the possibility of operating it as a locked oscillator at high power levels. An advantage of this arrangement is that in the event of failure of the driver oscillator, the bitermitron will continue to oscillate.
A paper by I E Orr described a way of controlling the beam current by means of a closely packed parallel-wire grid in front of the cathode emission surface. Normally, any electrodes near the cathode are used for shaping the beam, and the current is controlled by varying the voltage of an accelerator electrode near the entrance to the slow-wave structure. However, the use of a grid control enabled a seven times greater charge in the beam current to be obtained for a given control-voltage change. Little change in the RF characteristics was produced by the addition of the grid. An interesting method was used to observe the effect of different electrode positions. A model scaled-up in size (and electrode voltages) was placed in a large evacuated chamber. This was big enough for a man in a space suit to go in and adjust the electrodes to give a suitable beam shape, the beam position being photographed from the light produced through excitation by the beam of the residual gas in the chamber.
A paper by G E Weibel and R H Bartram described the theory of a device for converting microwaves to sub-millimetre waves at high powers. Electrons are first injected and trapped inside a chamber in a magnetic field. By making the frequency of a microwave input resonate with that due to cycloidal motion in the magnetic field (cyclotron resonance) orbital motion of the electron cloud is next induced. It is then proposed to multiply the orbital frequency and its associated rotational energy by a factor of between 100 and 1,000 by means of a very large (≥ 100,000 Gauss) pulsed magnetic field. Power at sub-millimetre wavelengths should then be radiated directly from the rapidly swirling electron cloud. This cloud actually swirls on its own axis as well as round its orbit - hence the name of the device.
In spite of, or perhaps because of, the recent discovery of new types of valve - such as the parametric amplifier - with very low noise characteristics, many new ideas for reducing noise in conventional valves are being investigated. Moreover, earlier pessimistic theoretical calculations of the minimum possible noise figure in such valves (which was not much less than what had already been achieved in practice) are no longer regarded as valid. A survey paper by M R Currie included a discussion of noise reduction in conventional valves. In such valves it can be shown that nearly all the noise is produced at the beginning of the device, near the cathode and potential minimum. It is on modifying the electron velocities in the multi-velocity region just beyond the potential minimum that many of the newly proposed methods of noise reduction depend. These electron velocities can be modified by depressing the potential minimum by the space-charge effect of high-density injected electron beams or, more usually, by modifying the beam profile either by means of special electrodes near the cathode or by shaping the cathode emissive surface itself. The noise can also be reduced by confining the emission to the edge of the cathode. In general, the use of two- rather than single- dimensional concepts is an advantage in this field. Space-charge effects can also reduce the noise, and since the extent of such reduction varies with the frequency, it may be advantageous to operate at certain frequencies to obtain low noise.
Modulation in an electron beam is usually propagated along it in two waves. One - the fast wave - has a velocity greater than that of the beam itself and the other - the slow wave - a velocity less than (but usually similar to) the beam velocity. Until recently microwave valves have relied on interaction between the slow electron beam wave and a circuit, and it has usually been thought that it is impossible to completely remove the noise from this slow electron beam wave. A paper by P A Sturrock, however, not only suggested flaws in the reasoning by which this impossibility is deduced, but also proposed a possible general method of removing the noise from a slow wave. In this method the fast and slow waves are parametrically coupled together by a pump signal in such a way that any noise on the two waves is interchanged periodically along their lengths. The noise can thus be removed from the slow wave if the noise on the fast wave has previously been removed. This can be done by a somewhat similar interchange process which is already known and which will be briefly described later in this report in connection with transverse-field valves.
A method of noise reduction which should be applicable to both slow and fast electron beam waves was described in a paper by R Adler and G Wade. This depends on the fact that for an electron beam spiraling in a longitudinal magnetic field, the noise temperature for transverse modulation (at right-angles to the direction of the beam motion and magnetic field) can be shown to be proportional to the ratio of the signal to the spiraling (cyclotron) frequencies. By using a large magnetic field to give a high cyclotron frequency, the noise can thus be reduced. Moreover, the field need not be kept large throughout the valve since it can be gradually reduced beyond the initial noise-reduction region without increasing the noise. Experimental results showed a reduction in the fast wave transverse noise temperature from 1,100 K to 180 K for a cyclotron frequency nine times that of the signal. The main problems in applying this method arise in designing a transverse-field input coupler which will operate at a fraction of the cyclotron frequency (rather than, as is usual, at the cyclotron frequency) and, of course, in providing a magnetic field high enough to give a cyclotron frequency several times that of the signal.
Periodic Electron Beams
Until recently microwave valves have relied for their operation on synchronism between RF waves and the longitudinal velocity of an electron beam. Thus, since the velocity which can be attained by electrons is necessarily limited, special slow-wave structures which are capable of propagating RF waves with velocities as slow as those of the electrons have to be provided. As the wavelength is decreased, such structures become correspondingly smaller. They are then more difficult to make and cannot dissipate so much power. In addition, the RF fields also fall off more rapidly away from them and so interaction with electron beams is more difficult to achieve. Slow-wave structures are generally either periodic, like the helix, or resonant, like the cavities in a magnetron. Aperiodic, non-resonant, smooth-wall circuits (such as waveguides), which would be easy to construct, can normally only propagate fast RF waves, i.e., waves with phase velocities equal to or greater than that of light and thus greater than that of any possible electron beam. Electrons can, however, interact with a fast wave by traveling not in a straight line but rather in a periodic path which is made to correspond to the spatial or time periodicity of the RF wave in such a way that any given electron always sees the same value of the RF field.
A paper by C K Birdsall and L Haas discussed beams which are made to follow a zigzag path along a waveguide by means of repelling electrodes outside the guide. Up to 21 crossings have been successfully induced, about 40% of the beam being transmitted through six crossings. In fact, however, only a small number of crossings is necessary for interaction, for similar reasons as in the case of a reflex klystron where the beam passes only twice through the cavity. The beam can interact with either forward or backward waves, possible output frequencies being those at which the RF wave slips a whole number of wavelengths per zigzag ahead of or behind the beam.
Birdsall and Haas also discussed beams which are made to follow a helical path between two concentric cylinders by balancing the outward centrifugal force against an equal inward force produced by an electrostatic field applied between the two cylinders. Here for interaction, the frequency of helical rotation must be equal to the required output frequency, neglecting any motion of the electrons in a direction along the cylinders axis.
Papers by A Reddish and by A H Beck and R F Mayo described beams which are made to follow a cycloidal path by means of a magnetic field. (It is interesting to remember that such motion was in some cases responsible for the operation of early magnetrons.) In such motion the cycloidal radius depends on the electron velocity. Thus the unfavourable electrons which gain energy from the RF field are automatically separated from the favourable electrons which lose energy to the RF field. For interaction, the cycloidal frequency must equal the required output frequency, neglecting any electron motion along the field.
A disadvantage of this system is that to produce short wavelengths high magnetic fields are required (field ≈ 10,000 Gauss divided by wavelength in centimetres). This could be avoided if the cycloidal motion could be at a sub-harmonic of the required output frequency. The paper by Beck and Mayo showed that in this case interaction is still possible, provided that the RF field has an azimuthal spatial periodicity corresponding to the subharmonic used. Thus, for example, the TEm1, waveguide mode can interact to produce output at m times the cycloidal frequency.
These are valves in which the RF field, and consequently also the electron modulation motion, are at right angles to the electron beam motion. Since a transverse-field parametric amplifier with very low noise was developed by R Adler (see the Technical Notebook section of our November 1958 issue (p. 555) or for a full account ProcIRE for June 1958 (p. 1,300) and October 1958 (p. 1,756)) much interest has been shown in such valves, and a number of devices more or less similar to the Adler tube were described at the conference. Two of these were simply microwave versions of the Adler tube (which has a relatively low operating frequency around 600 MHz) - one due to A Ashkin for 4.14 GHz and the other due to T J Bridges for 2.7 GHz.
The orginal Adler tube used input and output couplers of the Cuccia type, which were first des- cribed in RCA Review as 'long' ago as ]une 1949 (p. 270). In such couplers the RF input signal is applied via suitable circuits across two deflection plates on opposite sides of the beam. This produces transverse RF fields across the beam. A longitudinal magnetic field is also applied to produce spiral motion of the beam along the field. When the input frequency across the coupler is made equal to the cyclotron spiraling frequency, the phenomenon of cyclotron resonance produces an increase in the spiral radius. In this device the coupling is between the fast electron-beam and RF waves. It has the great advantage that under certain conditions (such as, for example, for a particular length of coupler) the beam noise is, in theory, totally removed to the coupler (and in practice nearly so) as the RF input signal is transferred to the beam.
One disadvantage of the Cuccia coupler is the fact that the coupler signal input frequency is tied to the magnetic cyclotron frequency. A paper by P A H Hart discussed theoretically a method of obtaining a somewhat wider choice of possible signal frequencies relative to the cyclotron frequency. In this method, slow-wave deflection circuits are used to produce transverse RF signal fields which travel along the electron beam. The phase velocity along the beam of the RF signal is chosen relative to the longitudinal beam velocity so that an observer travelling with this beam velocity would see, because of the Doppler effect, a transverse field not at the signal frequency but at the (in this case lower) cyclotron frequency. Resonance between the apparent signal frequency and the cyclotron frequency, fast-wave coupling and noise removal are then possible as in the Cuccia coupler. A greater bandwidth should, however, be obtainable than with a Cuccia coupler. (Doppler shift concepts similar to that just described were of importance in several of the interaction processes discussed at the conference).
A paper by W R Beam described experimental results on the use of a slow-wave helix to couple to a hollow beam. Here again the input frequency was not tied to the cyclotron frequency. Unfortunately the results obtained did not agree with those to be expected on conventional coupling theories and have not yet been explained.
A paper by R H Pantell discussed the theory of a transverse-field coupler which uses a DC electrostatic rather than a magnetic field. In this coupler, as in the periodic beam device described by Birdsall and Haas which we have already mentioned, spiral motion of electrons is produced between two concentric cylinders by balancing the outward centrifugal force against that produced by an electrostatic field applied between the two cylinders. For interaction the spiraling frequency must equal the signal frequency as in the Cuccia coupler, and again fast-wave coupling and noise removal are possible for a certain critical length of coupler.
Couplers such as we have just been describing can only put spiral modulation motion on to the beam and are not capable of amplifying this motion. One or more quadrupoles were used to produce amplification in almost all the transverse-field valves described at the congress-indeed one author said he would not have dared to describe his device if it had not used a quadrupole. Such quadrupoles each consist simply of four deflection plates spaced round the electron beam. Adjacent plates are given opposite electric potentials so as to produce fields which are tangential to the spiral motion of the beam. Amplification of the spiral modulation motion is produced by arranging for any given electron to continually see an accelerating tangential field as it spirals along the tube. This can be done by means of either a time- or space- varying quadrupole field, i.e. either by applying a pump RF field to a single quadrupole or alternatively, by applying direct voltages to a number of quadrupoles which are rotated and suitably spaced relative to each other. The latter system was used in several of the new devices described at the congress: the former was used in the original Adler tube.
When an RF pump field is used, this provides the energy necessary to produce amplification. In the original Adler tube, to correspond to the two pairs of plates in the quadrupole, the pumping had to be at twice the spiral (signal) frequency. A paper by I E Carroll discussed the possibility of pumping at a rather lower frequency - two thirds of that of the signal. In this case the pumping process can be considered as an initial interaction between the pump and signal to produce a difference signal (at one third of the original frequency), followed by pumping of this difference signal as in the Adler tube at twice its own frequency (i.e., at two-thirds of the original frequency). With this method of pumping the noise should be as low as in the Adler tube. Other effects may, however, prevent its successful practical realisation. Pumping at the signal frequency was discussed in a paper by E I Gordon and A Ashkin.
Amplification by means of a number of spaced relatively rotated DC quadrupoles may be loosely described as being with a zero pump frequency. The energy required to produce amplification is, however, no longer obtained from the pump field, but rather from the DC beam energy, longitudinal beam kinetic energy being converted into spiral rotational energy by the DC quadrupoles. In the example described by J C Bass, adjacent quadrupoles are spaced one quarter of a cyclotron wavelength apart and rotated through ninety degrees relative to each other. Identical results are obtained by not rotating the quadrupoles but instead reversing the signs of the applied direct potentials between adjacent quadrupoles. Apparently DC quadrupoles will not give low noise, but should on the other hand be capable of producing higher power outputs. The absence of an RF amplifying structure would, of course, be an advantage in constructing very high-frequency devices.
Possibilities of getting away from the quadrupole as an amplifier by using a helix and hollow beam were described in the paper by W R Beam already mentioned. In this case also pumping need not be at twice the signal frequency. The possibility of using a number of spaced magnets as a spiral motion amplifier was mentioned in a paper by K Blotekjaer and T Wessel-Berg.