Magnetron - Thesis - LiliMA

3D Computer Modeling of Magnetrons
by Lili Ma
A thesis submitted to the University of London in partial fulllment of the requirements for the degree of Doctor of Philosophy
Department of Electronic Engineering Queen Mary, University of London United Kingdom

December 2004
TO MY FAMILY
Abstract
The research work presented in this thesis is primarily based on computer modelling of magnetrons using a 3D FDTD/PIC code called MAGIC. The usual modelling process has been substantially extended to include realistic details of two e2v technologies S and X band magnetrons. Considerable insight into the magnetron operation has thus been obtained. An investigation into the melting of the output coupling rods in the S band MG5193 magnetron revealed that multipacting taking place between the coupling rods and the central support rod is the most likely cause, rather than an RF mismatch or RF current overheating. Also, a good agreement between the simulated and measured frequency spectrum of the S band MG5193 anode has been achieved by suitably adjusting the dimensions of the bare anode, the strapping system and the tuner. Possible reasons for the high overall eciency of the X band MG5241 magnetron model have been investigated by modelling the end hats emission, the output coupling circuit, the copper surface losses and the rate of rise of anode voltage, but none proved to be relevant. However, the investigations revealed some important features of magnetron operation. Computer modelling revealed for the rst time non-uniform electron emission and back bombardment along the length of the cathode. Also it was shown that in computer modelling it is essential to use a real rather than a uniform magnetic eld prole. Simulation runs with dierent cathode emission proles have indicated marked inuence of the emission distribution on the overall eciency of the magnetron. Computer modelling also indicated that high energy back bombardment electrons are present in the X band magnetron and that an articially high initial velocity of emitted electrons signicantly reduces the overall eciency. It is therefore concluded that the high energy of secondary electrons is the likely reason for a relatively low eciency of i
the X band magnetron. Perhaps one ought to add that all the new and additional information could not have been obtained without a powerful 3D computer code.
ii
Acknowledgments
I would like to express my gratitude to my supervisor, Dr. X. Chen, for his supervision, guidance and support throughout this project. Many thanks are also due to the following people who have kindly oered their help and support: - Mr. M. Esterson specically, who guided me through my four years PhD study. His knowledge of magnetrons and intelligent way of analyzing has been great help. - Professor P. A. Lindsay for his inspiring suggestions, from the view of fundamental magnetron theory. - Dr. L. Ludeking for his support on MAGIC simulation techniques. - Mr. P. Burleigh, Mr. M. Brady, Mr. K. Saleem and Mr. D. Wilson for their suggestions regarding engineering problems and many useful discussions. - Dr. Z. Wang for his help on computer modelling. - All my colleagues at the Electronic Engineering Department, Queen Mary College, for their assistance. Finally I would like to convey my gratitude to my husband and my parents, for their understanding, encouragement and everlasting support.
iii
Table of Contents
Abstract Acknowledgments Table of Contents List of Figures List of Tables List of Abbreviations 1 Introduction 1.1 Introduction to Microwave Tubes . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 1.1.2 1.2 Fundamentals of microwave tubes . . . . . . . . . . . . . . . . . . Dierent microwave tubes and their applications . . . . . . . . . . i iii iv ix xviii xx 1 1 1 2 5 5 6 7 8 9 11
Motivation and Contributions . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 1.2.2 1.2.3 Characteristics of magnetron modelling . . . . . . . . . . . . . . . Challenging problems . . . . . . . . . . . . . . . . . . . . . . . . . Contribution to industrial design . . . . . . . . . . . . . . . . . . .
1.3
Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Operating Principle of Magnetrons
iv
2.1 2.2 2.3
Early Magnetrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Negative Resistance Magnetron . . . . . . . . . . . . . . . . . . . . . . . . 12 Principle of Transit Time Oscillation . . . . . . . . . . . . . . . . . . . . . 13 2.3.1 2.3.2 Kinetic motion of electrons in a planar diode . . . . . . . . . . . . 14 Kinetic motion of electrons in a cylindrical diode . . . . . . . . . . 20
2.4
Resonant Cavity Magnetron Operation . . . . . . . . . . . . . . . . . . . . 22 2.4.1 2.4.2 2.4.3 2.4.4 Equivalent circuit and oscillation modes . . . . . . . . . . . . . . . 23 The Hartree/Hull conditions . . . . . . . . . . . . . . . . . . . . . 26 The Hartree/ Hull conditions for the S/X band magnetrons . . . . 29 Interaction between the RF Field and the electron cloud . . . . . . 31
2.5
Frequency Stability and Mode Separation . . . . . . . . . . . . . . . . . . 33 2.5.1 2.5.2 2.5.3 2.5.4 Frequency jumping . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Strapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Rising sun anode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Other factors aecting frequency stability . . . . . . . . . . . . . . 36
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3 The MAGIC Code and Its Associated Algorithms 3.1 3.2 38
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Finite Dierence Time Domain Technique . . . . . . . . . . . . . . . . . . 39 3.2.1 3.2.2 3.2.3 3.2.4 FDTD algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Numerical stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Uncertainty analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Absorbing boundary condition . . . . . . . . . . . . . . . . . . . . 43
3.3
Particle In Cell (PIC) Technique . . . . . . . . . . . . . . . . . . . . . . . 44 3.3.1 3.3.2 3.3.3 The Computational cycle . . . . . . . . . . . . . . . . . . . . . . . 44 Integration of the equations of motion . . . . . . . . . . . . . . . . 45 Particle and force weighting . . . . . . . . . . . . . . . . . . . . . . 47
3.4
Emission Models in MAGIC Code . . . . . . . . . . . . . . . . . . . . . . 48
3.4.1 3.4.2 3.4.3
The space-charge-limited emission . . . . . . . . . . . . . . . . . . 48 The beam emission . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 The secondary emission . . . . . . . . . . . . . . . . . . . . . . . . 49
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4 Modelling of an S band Magnetron 4.1 52
Modelling of The Output System . . . . . . . . . . . . . . . . . . . . . . . 53 4.1.1 4.1.2 4.1.3 4.1.4 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Cuto frequency and possible modes in the output system . . . . . 54 Modelling of the output system . . . . . . . . . . . . . . . . . . . . 56 Preliminary study on possible multipactor eects inside the output system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2
Cold Test Modelling of the Anode System . . . . . . . . . . . . . . . . . . 66 4.2.1 4.2.2 4.2.3 4.2.4 4.2.5 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 Cold test modelling techniques . . . . . . . . . . . . . . . . . . . . 68 Investigation of the bare anode . . . . . . . . . . . . . . . . . . . . 69 Investigation of the standard anode . . . . . . . . . . . . . . . . . 72
Investigation of the tuner . . . . . . . . . . . . . . . . . . . . . . 73
4.3
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 5 3D Modelling of an X band Magnetron 5.1 5.2 77
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Optimization of the Computer Model . . . . . . . . . . . . . . . . . . . . . 79 5.2.1 5.2.2 5.2.3 The parallel vane model . . . . . . . . . . . . . . . . . . . . . . . . 80 The thick tapering vane model . . . . . . . . . . . . . . . . . . . . 84 Measurement of the external Q value . . . . . . . . . . . . . . . . . 90
5.3 5.4
Real Magnetic Field Distribution . . . . . . . . . . . . . . . . . . . . . . . 92 Hot Test Modelling with Dierent Cathode Emission Models . . . . . . . 94 5.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
vi
5.4.2 5.4.3 5.5
The beam emission model . . . . . . . . . . . . . . . . . . . . . . . 95 The space-charge-limited emission model . . . . . . . . . . . . . . 102
Electric Field Distribution between Cathode and Anode . . . . . . . . . . 107 5.5.1 5.5.2 5.5.3 The calculation paths . . . . . . . . . . . . . . . . . . . . . . . . . 108 Electric eld distribution in the beam emission model . . . . . . . 109 Electric eld distribution in the space-charge-limited emission model112
5.6 5.7 5.8
Varying the Output Coupling Loop . . . . . . . . . . . . . . . . . . . . . . 115 Copper Surface Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 The End Hat Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 5.8.1 5.8.2 Emission models of end hats . . . . . . . . . . . . . . . . . . . . . 121 End hats emission with space-charge-limited model and uniform magnetic eld of 0.405T . . . . . . . . . . . . . . . . . . . . . . . . 126 5.8.3 Reduced spurious current and the real magnetic eld distribution . 129
5.9
Slow Rate of Rise of Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 6 The Eects of Non-Uniform Cathode Emission and Electron Back Bombardment 6.1 6.2 6.3 137
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 Investigation of the space-charge-limited Emission Model 6.3.1 . . . . . . . . . 139
Characteristics of cathode emission and back bombardment distribution along the cathode length . . . . . . . . . . . . . . . . . . . 139
6.3.2 6.4 6.5
Electric and magnetic eld variation along the cathode surface . . 142
Investigation of the Beam Emission Model . . . . . . . . . . . . . . . . . . 147 Inuence of a Non-Uniform Cathode Emission on the Operation of a Magnetron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.6
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
vii
7 The Eect of Cathode Emission Voltage 7.1 7.2 7.3 7.4
153
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Energy Distribution of Electron Back Bombarding . . . . . . . . . . . . . 154 High Initial Cathode Emission Velocity . . . . . . . . . . . . . . . . . . . . 159 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 8 Conclusions and Future work 8.1 8.2 163
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 167
Appendix A Denitions of Qs and Circuit Eciency
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 Appendix B Authors Publications 169
viii
List of Figures
1.1 1.2 1.3 1.4 2.1 The beam-wave interaction is 2D in O type devices. . . . . . . . . . . . . The beam-wave interaction is 3D in M type device. . . . . . . . . . . . . A diagram of a travelling wave tube [3] . . . . . . . . . . . . . . . . . . . . An eight cavity magnetron (type NT98) [5] . . . . . . . . . . . . . . . . . Early magnetrons (a) Cylindrical anode magnetron, (b)Split anode magnetron (c) Multi-segment magnetron (d) Resonant cavity magnetron . . . 12 2.2 2.3 2.4 Electron trajectory in a split anode negative resistance magnetron . . . . 13 Electron trajectory in a planar crossed eld diode . . . . . . . . . . . . . . 14 Electron trajectories in a planar crossed eld diode for dierent magnetic elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.5 Electron trajectories with dierent intial velocities and directions. Electron trajectory 1: the initial velocity is zero; Electron trajectory 2: vx |t=0 = vy |t=0 = E/(2B); Electron trajectory 3: vx |t=0 = 0, vy |t=0 = E/(2B); where E = Vdc /d, Vdc = 5.7kV , d = 0.92mm. . . . . . . . . . . . . . . . . 18 2.6 2.7 2.8 Electron trajectories in the presence of AC eld . . . . . . . . . . . . . . . 19 Geometry of a cylindrical diode [1] . . . . . . . . . . . . . . . . . . . . . . 20 Electron trajectories in a cylindrical diode for dierent magnetic elds. 1. B = 0; 2. B < Bc ; 3. B = Bc ; 4. B > Bc . . . . . . . . . . . . . . . . . . . 22 2.9 The cross section of an eight cavity hole and slot magnetron. The hole and slot structures provide inductance and capacitance for individual resonators. 23 2.10 Equivalent circuit of an eight-cavity magnetron [2] . . . . . . . . . . . . . 24 ix 2 3 3 4
2.11 Field and charge distribution of dierent modes in an eight-cavity magnetron (a)n=1 (b)n=2 (c)n=3 (d)n=4 . . . . . . . . . . . . . . . . . . . . 26 2.12 The Hartree diagram of an eight-cavity magnetron [2] . . . . . . . . . . . 28 2.13 Operating voltage and magnetic eld for dierent modes in S/X band magnetrons (Solid line represents mode, mode number n=6 for S band, n=9 for X band; dashed line represents -1 mode, mode number n=5 for S band, n=8 for X band). (a). e2v type S band MG5193 (b). e2v type X band MG5241 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.14 Electron trajectories when the system is oscillating in the mode. . . . . 31 2.15 Electron spokes in an eight-cavity magnetron when the system is operating in the mode [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.16 The strapping system. (a)Wire strapping system of an S band magnetron. (b)Ring strapping system of an X band magnetron. . . . . . . . . . . . . . 34 2.17 Geometry of a rising sun anode [4] . . . . . . . . . . . . . . . . . . . . . . 36 3.1 3.2 3.3 3.4 Yees cell, for electric and magnetic eld vectors . . . . . . . . . . . . . . . 39 The leap-frog time marching scheme . . . . . . . . . . . . . . . . . . . . 40 A typical cycle in PIC processing [7] . . . . . . . . . . . . . . . . . . . . . 45 Sketch of leap-frog integration method showing time centering of force F while advancing v , and of v while advancing x [7] . . . . . . . . . . . . . 46 3.5 3.6 4.1 4.2 4.3 The scheme for the zero order charge and current weighting [7] . . . . . . 47 Coordinate system illustrating primary particle and secondary electron . . 50 Image of an S band magnetron (e2v type MG5193) . . . . . . . . . . . . . 52 The geometry of the output of the S band MG5193 magnetron . . . . . . 53 Computer model of the rectangular output of the S-band MG5193 magnetron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.4 4.5 4.6 The cuto frequencies of the circular wave-guide (a=35mm) . . . . . . . . 55 The cuto frequencies of the circular wave-guide (a=42.9mm) . . . . . . . 55 Cuto frequencies of the rectangular wave-guide . . . . . . . . . . . . . . 56
4.7
(a)Current distribution along the coupling rod. (b) Current distribution along the radiating dipole. (: glass dome; : ceramic dome) . . . . . . 58
4.8
Transmitted and reected power at the input port with a glass dome (Power reection: 11.5%) . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.9
Transmitted and reected power at the input port with a ceramic dome (Power reection: 4.0%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.10 Transmitted and reected power at the input port without a dome (Power reection: 17.8%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.11 A cross section of the electric vector eld just below the dipole antenna where the melting points are situated. (a) t=1.660ns (b) t=1.820ns . . . . 62 4.12 Voltage oscillation between the central support rod and one of the coupling rods. The applied signal operates at 2.998GHz with peak power of 4MW. 4.13 Snap shots of electron clouds (RF power: 4MW; Phase of RF eld while primary electrons are released: 0.5; Ceramic dome:
r
62
= 9 ; Red colour
represents primary electrons; Purple colour represents secondary electrons.) (a) t= 1.940ns (b) t=2.500ns (c) t=2.900ns (d) t= 3.140ns (e) t=3.300ns (f) t=4.000ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.14 Snap shots of electron clouds (RF power: 4MW; Phase of RF eld while primary electrons are released: 0.5; Glass dome:
r
= 4 ; Red colour rep-
resents primary electrons; Purple colour represents secondary electrons.) (a) t= 1.940ns (b) t=2.500ns (c) t=2.900ns (d) t= 3.140ns (e) t=3.300ns (f) t=4.000ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4.15 3D geometry of the anode system of MG5193. (a). The bare anode with the wire strapping system removed. (b). The standard anode including strapping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.16 Cross section view of the bare anode of MG5193, the slots are parallel. . . 69
xi
4.17 Two dierent computer models of tapering anode slots. (a) Slot geometry 1: the width of the slot is equal to the real value in the middle length of the slot. (b) Slot geometry 2: the width of the slot is equal to the real dimension at their ends near the cathode. . . . . . . . . . . . . . . . . . . 70 4.18 The electric eld distribution in and 1 modes using slot geometry 2 71
4.19 The oscillation spectrum of the bare anode with slot geometry 2. . . . . . 71 4.20 A computed oscillation spectrum of the standard anode including the wire strapping system and the end slots . . . . . . . . . . . . . . . . . . . . . . 72 4.21 The geometry of the tuner . . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.22 Spectrum of oscillations when the tuner is included . . . . . . . . . . . . . 74 5.1 5.2 Image of an X band magnetron (e2v type MG5241) . . . . . . . . . . . . . 78 Cross section view of the parallel sided vane models with (a) Coarse mesh (180 cells in azimuthal direction) (b) Fine mesh (360 cells in azimuthal direction) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 5.3 (a) Spectrum of the broad band signal. (b) Spectrum of the responding signal. Blue curve represents ne mesh model (oscillation peaks at: 9.294GHz, 9.776GHz, 10.325GHz and 10.889GHz); Red curve represents coarse mesh model (oscillation peaks at: 9.360GHz, 9.840GHz, 10.380GHz and 10.920GHz). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 5.4 The geometry of the thick tapering vane model and the electron cloud of a mode (a)Transverse cross section (b)Axial cross section . . . . . . . . 85 5.5 Oscillation spectrum of the thick tapering vane model by broad band frequency response method (oscillation peaks at: 9.472GHz, 9.908GHz, 10.350GHz and 10.926GHz) . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5.6 Magnetron transient evolution. (a) The anode DC voltage. (b) The anode current. (c) The output power. . . . . . . . . . . . . . . . . . . . . . . . . 88
xii
5.7
Magnetron transient evolution. (d) The voltage oscillation between adjacent anode vanes. (e) The FFT of the vane voltage oscillation. (f) The total system stored energy evolution. Before 14ns, the energy is only DC energy, which is 54.12e-6J; after 14ns, the total energy, including the DC energy and RF energy, is 106.06e-6J. . . . . . . . . . . . . . . . . . . . . . 89
5.8
Ephi distribution of mode in (a) The transverse section of the anode. (b) The axial cross section of the anode . . . . . . . . . . . . . . . . . . . 90
5.9
Measure the external Q value from system energy decay rate. (a) The system energy evolution while excited in mode distribution. (b) The instant external Q of 260. . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.10 The radial magnetic eld component distribution used in MG5241P (e2v measurement) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.11 The axial magnetic eld component distribution used in MG5241P (e2v measurement) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.12 radial distribution of the axial component Bz of the real magnetic eld, Bmax =0.418T. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.13 Comparison of the performance of beam emission model with the experimental performance of MG5241P. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Anode voltage vs anode current. (b) Output power modied by circuit eciency of 85% vs anode current. (c) Overall eciency vs anode voltage. . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.14 Comparison of the performance of space-charge-limited emission model with the experimental performance of MG5241P. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Anode voltage vs anode current. (b) Output power modied by circuit eciency of 85% vs anode current. (c) Overall eciency vs anode voltage. . . . . . . . . . . . . . . . 104 5.15 Electric eld is measured along the solid lines indicated in drawing . . . . 109 5.16 Electric eld components are measured along (a) r (rc , ra ), = 300 , z = 0; (b)r (rc , ra ), = 200 , z = 0; . . . . . . . . . . . . . . . . . . . . . . . . 110 xiii
5.17 Electric eld components are measured along (a)r (rc , ra ), = 300 , z = 1.5mm; (b)r (rc , ra ), = 200 , z = 1.5mm . . . . . . . . . . . . . . . . 111 5.18 Electric elds are measured along (a) r (rc , ra ), = 300 , z = 0; (b)r (rc , ra ), = 200 , z = 0; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 5.19 Electric elds are measured along (a)r (rc , ra ), = 300 , z = 1.5mm; (b)r (rc , ra ), = 200 , z = 1.5mm . . . . . . . . . . . . . . . . . . . . . 114 5.20 The geometry of the output coupling circuit . . . . . . . . . . . . . . . . . 115 5.21 Magnetron performance as a function of the output coupling loop. (a) External Q vs output coupling loop length d (b) Anode current vs external Q (c) Overall eciency vs external Q (d) Vane voltage vs external external Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 5.22 The geometry of the lumped lossy elements is shown in dark blue (The conductivity of the two lossy rings is 5.99e5mhos/m) . . . . . . . . . . . . 119 5.23 The geometry of the cathode end hats . . . . . . . . . . . . . . . . . . . . 122 5.24 Cut o voltage corresponding to dierent cathode radii . . . . . . . . . . 123
5.25 Comparison of anode current evolution. (a) Anode current evolution without end hats emission. Anode voltage 5.7kV, magnetic eld uniform at 0.405T. (b) Anode current evolution with end hats emission from cells 4, 5, 6. Anode voltage 5.66kV, magnetic eld uniform at 0.405T, pre-oscillation anode current 1Amps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 5.26 The electron cloud including the end hats emission from the bottom three layers of cells 4, 5, 6 at the beginning of the hot test simulation. Anode voltage 5.66kV, magnetic eld uniform at 0.405T. (a) t=0.525ns (b) t=1.275ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 5.27 Investigation of the end hats emission from the bottom three cells 4, 5, 6. Uniform magnetic eld at 0.405T. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Output power vs anode current (b) Anode voltage vs anode current (c) Overall eciency vs anode current . . 128
xiv
5.28 Investigation of the end hats emission with the beam emission model for both the cathode and the end hats. Real magnetic eld distribution Bmax = 0.418T . Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Output power vs anode current (b) Anode voltage vs anode current (c) Overall eciency vs anode current . . . . . . . . . . . . 131 5.29 The contribution of the cathode emission and end hats emission to the anode current as a function of the anode voltage . . . . . . . . . . . . . . 133 5.30 The simulated transient evolution of the main parameters obtained with the slow RRV of 100kV /s. (a) The eective anode voltage 5.87kV. (b) The anode current 5.26A. (c) The system stored energy, the DC energy is 58.5e-6J; the RF energy is 73.1e-6J. (d) The voltage oscillation between two adjacent vanes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 5.31 Oscillation spectrum of the vane voltage during the start-up of magnetron oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 6.1 6.2 The length of the cathode is divided into fourteen rings . . . . . . . . . . 138 Cathode emission current density distribution along the cathode length when using the space-charge-limited emission model . . . . . . . . . . . . 139 6.3 6.4 6.5 Back bombardment Voltage distribution along the cathode length . . . . . 141 Back bombardment current density distribution along the cathode length 141 Electric eld distribution along the cathode length at dierent azimuthal positions. Vdc = 5.7kV , uniform magnetic eld 0.405T. (a) Opposite the vane, = 0 (b) Opposite the cavity, = 10 . . . . . . . . . . . . . . . . 143 6.6 Electric eld distribution along the cathode length at dierent azimuthal positions. Vdc = 5.62kV , real magnetic eld Bmax 0.418T. (a) Opposite the vane, = 0 (b) Opposite the cavity, = 10 . . . . . . . . . . . . . . 144
xv
6.7
Electrons emitted from dierent cathode sections at the beginning of the emission (a, c, e and g) and when the magnetron is oscillating in a stable mode (b, d, f and h) when the magnetic eld is uniform. Vdc=5.70kV; Uniform B=.405T (a)(b) Electrons from rings 1 and 14 (c)(d) Electrons from rings 2 and 13 (e)(f) Electrons from rings 3, 4, 11 and 12 (g)(h) Electrons from rings 5 to 10. . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.8
Real magnetic eld (Bz and Br) distribution along the cathode length, Bmax =0.418T. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.9
Electrons emitted from dierent cathode sections at the beginning of the cathode emission (a, c and e) and when the magnetron is oscillating in a stable mode (b, d and f) using the real magnetic eld distribution. Vdc=5.62kV; Bmax =0.418T. Very few electrons are emitted from the two rings 1 and 14, therefore no electron cloud diagrams are displayed for those two sections. (a)(b) Electrons from rings 2 and 13 (c)(d) Electrons from rings 3, 4, 11 and 12 (e)(f) Electrons from rings 5 to 10. . . . . . . . 147
6.10 Back bombardment current density distribution with beam emission model149 6.11 Back bombardment voltage distribution with beam emission model . . . . 149 6.12 Cathode emission current density distributions (Total emission current is about 5.89A for all emission proles) . . . . . . . . . . . . . . . . . . . . . 150 6.13 (a) Anode current vs Peak cathode emission current density. (b) Overall eciency vs Peak cathode emission current density. . . . . . . . . . . . . . 151 7.1 7.2 Energy distribution of secondary electrons emitted by silver [1] . . . . . . 154 The back bombardment energy distribution at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a). t1=39.930ns (b). t2=t1+0.14RFperiod (c). t3=t1+0.28RFperiod (d). t4=t1+0.42RFperiod . . . . . . . . . . . . . . . 155 7.3 The back bombardment energy distribution along the cathode at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a) t1=39.930ns (b) t2=t1+0.14RFperiod157
xvi
7.4
The back bombardment energy distribution along the cathode at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a) t3=t1+0.28RFperiod (b) t4=t1+0.42RFperiod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.5
The RF voltage between adjacent anode vanes within one RF cycle. Vdc=5.87kV, Ia=5.14A. Electron back bombardment energy distributions shown in Figures 7.2, 7.3 and 7.4 have been calculated at points indicated by a red star . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.6
The electron back bombarding energy distribution measured over 10RF cycles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.7
Electron clouds in stable mode oscillations with dierent cathode emission initial velocities. (a). The average initial velocity of cathode emission is 4eV. (b) The average initial velocity of cathode emission is 167eV. . . . 161
xvii
List of Tables
2-A Hartree/Hull condition for e2v MG5193 (S Band) and MG5241 (X Band) magnetrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 4-A Comparisons of the power reections with and without domes . . . . . . . 60 4-B Frequency comparison between simulation results and experimental results of the bare anode with dierent computer slot geometries . . . . . . . . . 71 4-C Comparison of the computed and measured spectrum of oscillations in a bare anode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4-D Oscillation spectrum comparison between simulation and experimental results (standard anode) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4-E Comparison of simulation and experimental spectra (MG5193) . . . . . . 73 5-A e2v type MG5241 magnetron: dimensions and typical performance . . . . 80 5-B Hot test simulation results of the parallel sided vane model . . . . . . . . 83 5-C Hot test simulation of the thick tapering vane model . . . . . . . . . . . 86 5-D Investigation of beam emission model with uniform magnetic eld 0.405T 5-E Investigation of beam emission model with real magnetic eld, Bmax = 0.418T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 5-F Investigation of beam emission model with real magnetic eld, Bmax = 0.405T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5-G Investigation of beam emission model with varied cathode emission density102 96
xviii
5-H Investigation of space-charge-limited emission model with uniform magnetic eld B=0.405T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 5-I Investigation of space-charge-limited emission model with real magnetic eld, Bmax = 0.418T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 5-J Investigation of space-charge-limited emission model with real magnetic eld, Bmax = 0.405T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5-K Investigation of the dierent output coupling with space-charge-limited emission model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 5-L Measurement of the output mismatch (e2v) . . . . . . . . . . . . . . . . . 118 5-M Calculation of Q0 and circuit eciency . . . . . . . . . . . . . . . . . . . . 120 5-N Investigation of the copper surface losses . . . . . . . . . . . . . . . . . . . 120 5-O Investigation of possible emission from dierent areas of the end hats . . . 122 5-P Preliminary investigation of the end hats emission . . . . . . . . . . . . . 124 5-Q Investigation of the end hats emission with space-charge-limited emission model and uniform magnetic eld at 0.405T . . . . . . . . . . . . . . . . . 127 5-R Investigation of the end hats emission with beam emission model and real magnetic eld distribution Bmax = 0.418T . . . . . . . . . . . . . . . . . . 130 5-S Simulation runs with slow RRV of 100kV /s and space-charge-limited emission model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 6-A Simulation results with the space-charge-limited emission model and real magnetic eld distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 6-B Simulation results with space-charge-limited emission model and uniform magnetic eld distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 6-C Simulations of beam emission with uniform and non-uniform magnetic elds148 6-D Simulation runs with dierent cathode emission current density distributions150 7-A The average electron back bombardment energy at dierent anode currents156 7-B Magnetron performance as a function of the initial velocity of the emitted electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
xix
List of Abbreviations
ABCS CFA CW FDTD FFT HPM PIC RF RRV TWT
Absorbing Boundary Conditions Cross-Field Amplier Continuous Wave Finite Dierence Time Domain Fast Fourier Transform High Power Microwave Particle-In-Cell Radio Frequency Rate of Rise of Voltage Travelling Wave Tube
xx
Chapter 1
Introduction
1.1 Introduction to Microwave Tubes
The magnetron was the rst truly practical microwave device. It was developed in England in the 1930s and it provided the impetus for the development of microwave radar during World War II. Since then, many microwave devices have been developed for the generation and amplication of microwave radiation. Although in recent years solid-state devices have been occupying many areas that were once reserved for microwave tubes, microwave electron tubes have certain unique properties that are unlikely to be surpassed by solid-state devices, especially in the domain of high power (10KW 10MW) and high frequency (1GHz and higher) applications [1, 2]. In this chapter general operation mechanisms and various applications of microwave electron tubes will be described.
1.1.1
Fundamentals of microwave tubes
There are various types of microwave tubes, but they all share some features and fundamental operating mechanisms. A typical electron tube consists of two or more electrodes enclosed in a glass or metal-ceramic envelope that is evacuated. A cathode is used to
Chapter 1. Introduction
emit electrons either by thermionic emission or electric eld emission, or some other methods, such as secondary emission. Fundamental to all electron tubes are the dynamics of charged particles under dierent electric- and magnetic-eld congurations. The principle of operation is to aect the kinetic or potential energy of electrons so as to generate electromagnetic radiation; for example when an electron is decelerated, it loses its kinetic energy to the RF eld. Thus, the DC energy stored in an external power supply can be converted to microwave energy in the load circuit. Another feature is that there must be a way for the RF energy to be coupled out of the tube. This is always accomplished with the help of transparent windows, capacitive or inductive coupling probes or loops. There are several important parameters characterising microwave tubes such as power output, frequency, bandwidth, tuning range, eciency and noise.
1.1.2
Dierent microwave tubes and their applications
Depending on the dierent interaction mechanism between the electron beam and the electro-magnetic eld, microwave tubes can be grouped into two categories: the O type and the M type. The O type is also called a linear-beam type. A simple schematic is shown in Figure 1.1. In this type of tube, the electron beam traverses the length of the tube and is parallel to the magnetic eld, which is used for focusing only.
Figure 1.1: The beam-wave interaction is 2D in O type devices. The M type are also called crossed eld devices. A simple schematic is shown in Figure 1.2. In this type of tube, magnetic eld is perpendicular to both the electron beam and the accelerating electric eld and plays a direct role in the interaction process.
Figure 1.2: The beam-wave interaction is 3D in M type device. Apart from the above two categories, there are microwave tubes that belong to neither, such as the vircator, also called a space charge type device. Most microwave tubes can be designed to operate either as oscillators or ampliers. The traveling wave tube is a typical O type microwave device. The traveling wave tube (TWT), used as an amplier, consists of an electron gun, a static magnetic eld parallel to the axis of the structure which focuses the electron beam, a slow-wave structure and a beam collector. Figure 1.3 is an example of a TWT in which the slow wave structure is a helix.
Figure 1.3: A diagram of a travelling wave tube [3] The RF input is coupled to the slow wave structure at the electron gun end. The slow-wave structure slows down RF wave so that it travels at nearly the same speed as the electron beam, which makes the beam-wave interaction possible. During this process, some of the energy of the electron beam is transferred to the RF eld and amplies it. The amplied RF signal is coupled out at the end of the slow wave structure. The
remaining energy of the electron beam is dissipated in the collector. To prevent the tube self-oscillating an attenuator is added along the slow wave structure. A traveling wave tube is a broadband device because there exist no inherent resonances in the structure. Traveling-wave tubes are generally used to amplify microwave signals over broad bandwidths up to 0.8GHz. The eciency of TWT is relatively low, typically about 20% [4]. For the above reasons an important application of TWTs is in satellite communication. Microwave magnetron is an M-type device and an oscillator. It consists of a cylindrical cathode and a coaxial anode, which contains several individual resonant cavities. The most ecient oscillation mode is the mode, when the adjacent cavities are 180 out of phase. A static magnetic eld is parallel to the axis of the cathode.
Figure 1.4: An eight cavity magnetron (type NT98) [5] Figure 1.4 shows the drawing of the rst cavity magnetron designed by Randall and Boot [5], which is an eight cavity, non-strapped magnetron. What dierentiates cross-
eld tubes from linear beam tubes is that magnetic eld participates in the beam-wave interaction. During the start-up of oscillations the electron cloud rotates around the cathode surface and generates noise. When the RF eld grows and the anode voltage exceeds a certain value, electrons begin to transfer their energy to the RF eld. When the anode voltage and the static magnetic eld are properly selected, the mode oscillations will dominate. The magnetron has the advantage of being able to generate high power (from several kilowatts to megawatts) at high eciency (40 to 70%) [4]. Magnetrons are widely used in radar systems, medical X-ray sources and microwave ovens. Other cross-eld tubes include cross-eld ampliers (CFAs) [4], amplitrons [6], gyrotrons [7], etc. A detailed mechanism of operation and typical characteristics of magnetrons will be discussed in chapter 2.
1.2
1.2.1
Motivation and Contributions

Characteristics of magnetron modelling
Compared with other microwave tubes, magnetrons are more dicult to model. Firstly, magnetrons require 3D rather than 2D modelling. In O type devices, the magnetic eld is parallel to the axis of the tube, see Figure 1.1. If the electron trajectories are symmetrical about the axis, only two dimensions are required for modelling, namely z and r. In M type devices, the DC electric eld and the magnetic eld are perpendicular to each other, see Figure 1.2. The DC electric eld, together with the magnetic eld control the electron trajectory in the y-z plane. Because the electron beam must be constrained in the direction perpendicular to the surface of the paper, i.e. x direction, three dimensions are necessary for adequate modelling of the interaction process, namely x, y, z.
Secondly, for example magnetrons always have complicated anode and cathode end structures which cannot be represented in 2D, such as anode strapping, which is required for adequate mode separation and end hats, used for constraining electron ow at the two ends of the cathode. Lastly, the axially non-uniform magnetic eld distribution invariably present in real tubes also requires 3D modelling. Hence, 2D simulation is not sucient for a proper study of the interaction process in most magnetrons. However, all the early magnetron modelling was in 2D and therefore it could not simulate magnetron operation realistically. With the development of more powerful computer codes, it is now possible to take into account all the eects mentioned above [810].
1.2.2
Challenging problems
The work presented in this thesis is solely based on 3D simulation using a powerful FD-TD (Finite-Dierence, Time-Domain) and PIC (Particle-In-Cell) code called MAGIC [11]. A number of challenging problems of magnetron operation such as multipactor eect, moding and power dissipation inside the magnetron, which could not be modelled in 2D, can now be investigated by 3D computer modelling. In addition to the above problems, in this thesis an eort has been made to investigate various factors that aect the operating eciency of a magnetron. Another novel contribution is the modelling of electron emission energy and back bombardment energy distribution along the length of the cathode. I. Multipactor eect Multipactor is a vacuum resonance discharge, which can be destructive. To our knowledge, the eect has not been investigated in magnetron modelling before. The electromagnetic conditions for multipactor to take place are very critical and thus dicult
to simulate. In chapter 4, preliminary modelling work has been done on the multipactor eect in the output system of an S band magnetron. II. Moding problem Moding is the competition between various resonant frequencies in the magnetron. When moding occurs, it spoils the operation of a magnetron. One of the tasks is an investigation of mode competition during the start-up of oscillation. III. Various factors aecting the magnetron eciency Various factors that may aect the magnetron operating eciency, such as the end hats emission, the output coupling circuit, the RRV (Rate of Rise of Voltage), the surface copper loss, the initial velocity of cathode emission, etc, have been investigated. Some of these eects, such as the end hats and cathode emission, are very dicult to study using theoretical or experimental methods. In practice such investigations become possible only by using 3D computer modelling. IV. Emission and back bombardment distribution An important contribution reported in this thesis is that for the rst time the nonuniform distribution of cathode emission using a space-charge-limited emission model has been demonstrated. The important back bombardment distribution has also been investigated.
1.2.3
Contribution to industrial design
The usual trial and error process in the development of magnetrons is time consuming and expensive. Through computer modelling, this process can be shortened and made more cost ecient. Also, computer modelling can be used to obtain some important information on magnetron operation such as the starting-up characteristics and power dissipation on the anode and cathode due to electron bombardment. Some of these aspects, such
as the axial distribution of electron energy in cathode bombardment are very dicult to investigate experimentally. Thus our computer simulation may contribute directly to the design of new magnetrons. It is expected that from the manufacturers point of view this would substantially reduce the time and cost of development of new devices.
1.3
Organization of the Thesis
Chapter 2 gives an introduction to the development and the operation mechanism of magnetrons. Detailed discussion has been presented on the beam-wave interaction, various operation modes and the required relationship between the anode voltage and the magnetic eld. Finally the theory of anode voltage-magnetic eld relation has been applied to S/X band magnetrons investigated in this project. Chapter 3 describes the FDTD (Finite Dierence Time Domain) and PIC (Particle In Cell) computer simulation techniques. This covers the FDTD algorithm, its numerical stability, uncertainty and absorbing boundary conditions (ABCs). The PIC technique has been discussed with emphasis on several issues such as the basic processing cycle, leap-frog integration that forwards the particles and the weighting of the particles on discrete grids. Chapter 4 presents the investigation of an S band magnetron (e2v MG5193). In particular the RF power transmission in the output circuit has been studied. Also a preliminary study of the multipactor eect has been discussed. Chapter 5 presents an extensive simulation work on an X band magnetron (e2v MG5241). Dierent meshing schemes have been tested and an optimized computer model has been developed. In cold test modelling the resonance spectrum and the corresponding Qs have been obtained. In hot test simulations dierent cathode emission models have been compared and various factors that may aect the magnetron eciency have been investigated.
Chapter 6 investigates the non-uniform emission and back bombardment of the cathode in MG5241 magnetrons, including the eect of uniform and non-uniform magnetic eld distribution. Chapter7 investigates the energy distribution of cathode back bombardment, again in MG5241 magnetrons. It is found that high energy secondary emission electrons are present in large numbers. An additional investigation of initial electron velocities has shown that in all cases high energy secondary emission signicantly aects the eciency of the magnetron. Chapter 8 contains nal conclusions and suggestions for future work.
References
[1] D.M.Pozar, Ed., Microwave engineering. Addison-Wesley, 1990.
[2] R.S.Symons, Modern microwave power devices, IEEE AESS Systems Magazine, pp. 1926, January 2002. [3] http://www.tpub.com/neets/book11/45d.htm, Ed., Electrical Engineering training series. Integrated publising, 2004. Prentice Hall Internationsl, 1990.
[4] S.Y.Liao, Ed., Microwave devices and circuits.
[5] W.E.Willshaw, L.Rushforth, A.G.Stainsby, R.Latham, A.W.Balls, and A.H.King, The high power pulsed magnetron development and design for radar application, The Journal of the Institution of Electrical Engineers, Proceedings at the Radiolocation Covention, pp. 9851005, March-May 1946. [6] L.Ludeking, R.Smith, and R.Wortman, An examination of the performance of backward wave CFAs in simulation and experiment, in Pulsed Power Plasma Science, ser. Digest of Technical Papers, vol. 1, no. 17-22, June 2001, pp. 229231. [7] V.A.Flyagin and G.S.Nusinovich, Gyrotron oscillators, Proceedings of the IEEE, vol. 76, no. 6, pp. 644656, June 1988. [8] X.Chen, M.Esterson, and P.A.Lindsay, 3D simulation of microwave magnetrons,
10
Digest, IEE Colloquium on High Frequency Simulation II, pp. 14, Nov 1997. [9] X.Chen, P.A.Lindsay, and M.Esterson, 3D modelling of high power, explosive emission magnetrons, in Displays and vaccum electronics, ser. ITG-Fachbericht 150/IEEE, Berlin, 1998, pp. 453458. [10] Z.Wang, X.Chen, and M.Esterson, Characterisation of resonance modes of a magnetron cavity using MAGIC- a FD-TD code, Eective Microwave CAD Tools, IEE Coloquium, pp. 14, 1999. [11] L.Ludeking, Ed., Manual of MAGIC Tool Suite. 2002. Mission Research Corporation,
Chapter 2
Operating Principle of Magnetrons

Introduction
In this chapter, the various types of magnetrons will be introduced. The operating mechanism of these magnetrons, especially the modern resonant cavity magnetron will be discussed. Theoretical analysis will be applied to the two magnetrons in S/X bands, which were studied in the later modelling work.
2.1
Early Magnetrons
Magnetron is a kind of cross-eld device, in which crossed electric and magnetic elds are used. The rst magnetron was mentioned by Hull in 1921, who used a cylindrical anode with coaxial lament (see Figure 2.1(a)). In 1929 Okabe developed the spilt anode magnetron, in which the anode was divided into two halves (see Figure 2.1(b)). Later on other workers further developed the split anode magnetron by dividing the anode into more segments (see Figure 2.1(c)). In all these magnetrons, the anode segments 11
Chapter 2. Operating Principle of Magnetrons
12
(a)
(b)
(c)
(d)
Figure 2.1: Early magnetrons (a) Cylindrical anode magnetron, (b)Split anode magnetron (c) Multi-segment magnetron (d) Resonant cavity magnetron are connected to outer oscillation circuits. In 1940, Randall and Boot introduced the rst modern resonant cavity magnetron [1], in which the resonant cavities replaced the resonant circuits connected to the anode segments (see Figure 2.1(d)). The power output of this type of magnetron was dramatically increased compared with the previous types. It is with this type of magnetron that the development of modern radar became possible. All these magnetrons rely on two types of operation mechanisms, the negative resistance oscillation and the electron transit time oscillation.
2.2
Negative Resistance Magnetron
The split-anode negative resistance magnetron is operated with its two halves of the anode at dierent RF potentials. The equipotentials in a negative resistance magnetron
13
Figure 2.2: Electron trajectory in a split anode negative resistance magnetron tend to point towards the anode segment with lower potential. Under the combined inuence of the electric and magnetic elds, the electron leaving the cathode follows a cycloid path, with its rolling center along the equal potentials and nally strikes the lower potential anode segment, see Figure 2.2. In practice, only few electrons reach the higher potential segment of anode; most of the electrons strike the lower potential segment of the anode, resulting in more current ow in the lower potential segment. This explains the eect of the negative resistance oscillation. The limit of this type of magnetron is that the RF oscillation period must be considerably greater than the electron transit time. The reason is that throughout the time of the electron journey from the cathode to the anode, the potentials of the two anode segments must keep polarity unchanged. When the operating frequency is increased, the eciency drops. This type of magnetron is capable of operating at wavelength down to 100cm and gives output in the order of 100watts with an eciency of 25% [1].
2.3
Principle of Transit Time Oscillation
The cylindrical anode magnetron, the multi-segment anode magnetron and the resonant cavity magnetron all depend on electron transit time oscillation, which makes higher
14
frequency oscillation possible. The beam-wave interaction inside a magnetron is very complicated. In this section a simplied planar diode and cylindrical diode will be discussed rst to explain the basic kinetic motion of electron between the cathode and the anode under the inuence of the crossed electric and magnetic elds. On this basis the operation of resonant cavity magnetron will then be discussed in section 2.4.
2.3.1
Kinetic motion of electrons in a planar diode
Figure 2.3 demonstrates a planar diode, which consists of two innite large plane electrodes with potential dierence of Va . The distance between the two electrodes is d0 . Thus the electric eld between the cathode and anode is E = Va /d0 . A uniform magnetic eld is applied perpendicular to the electric eld and parallel to the two electrodes.
Fe
v
Rc
Figure 2.3: Electron trajectory in a planar crossed eld diode
An electron emitted from the cathode moves under the inuence of electric force Fe = eE and magnetic force Fm = evB between the anode and the cathode. In order to solve the trajectory of the electron, the kinetic motion equation is combined with the two force equations as below: dr2 = e(E + vB) dt2
(2.1)
15
F e = eE
(2.2)
Fm = evB
(2.3)
In the Cartesian coordinate system, the electron velocity can be resolved into the horizontal component dx/dt and the vertical component dy/dt. Eqs 2.1 to 2.3 can then be rewritten as below:
eE eB
d2 y dx =m 2 dt dt
(2.4)
eB
dy d2 x =m 2 dt dt
(2.5)
The solution of the above equations, which neglect the space-charge eects, shows that the electron trajectory is a cycloid:
x = A1 sin( c t + ) + A2 t + x0
(2.6)
y = A1 cos( c t + ) + y0
(2.7)
dx = c A1 cos( c t + ) + A2 dt
(2.8)
dy = c A1 sin( c t + ) dt
(2.9)
Where A1 , A2 , x0 , y0 and c = eB/m are all constants, which can be determined by the initial condition. If we let the electron start from the origin, the initial conditions can
16
be expressed as: x|t=0 = 0; y|t=0 = 0; dx/dt|t=0 = vx0 ; dy/dt|t=0 = vy0 . Applying the initial condition to Eqs 2.6 to 2.9, the constants can be determined as shown:
A2 = A1 = 1 c
E B
2 2 vy0 + (A2 vx0 )
x0 = y0 =
vy0 c
(A2 vx0 ) c x0 A1 y0 A1
sin = cos =
For the convenience of calculations, constant is not given in an explicit form. Rearranging the solutions Eqs 2.6 to Eq 2.9, the following equations can be obtained:
x = y0 sin(c t) x0 cos(c t) + A2 t + x0
(2.10)
y = y0 cos(c t) + x0 sin(c t) + y0
(2.11)
dx = c y0 cos(c t) + c x0 sin(c t) + A2 dt
(2.12)
dy = c y0 sin(c t) + c x0 cos(c t) dt
(2.13)
17
Considering a simple condition that the electron starts with zero velocity (vx = 0, vy = 0) from the cathode, it can be seen in Figure 2.3 that the electron travels like a point on a circle rolling along the cathode surface, its angular frequency is c = eB/m. Within one period T = 2/c , the electron will complete a revolution and go back to the cathode. Since the magnetic eld does not contribute to the electron energy change, the maximum velocity occurs at the top of the cycloid orbit as shown in Figure 2.3, where vy = 0 , y = ymax . The radius of the rolling circle is determined by Rc = ymax /2. Since ymax = y(T /2), substituting the initial conditions into Eq 2.11, the radius of the rolling circle can be obtained as Rc = mE eB 2 (2.14)
From Eq 2.14 the maximum distance an electron travels in x direction within one cyclotron period T is xmax = x(T ) = 2Rc = 2mE eB 2 (2.15)
For a given electric eld, the radius of the electron orbit will vary with the magnetic eld. For a cuto condition, when an electron trajectory just grazes the anode, that is ymax = d0 , the magnetic eld is Bc . Solving the Eq 2.14 and substituting for E the value Va /d0 , the cuto magnetic eld is given by B = Bc = 2mE 1 = ed0 d0 2m Va e (2.16)
The trajectories of electrons for dierent magnetic elds are shown in Figure 2.4. Curve 1 stands for electron orbit when the magnetic eld is zero; if magnetic eld is smaller than Bc , the electron will strike the anode, curve 2; otherwise, the electron will miss the anode and continue its cycloid orbit, curve 4; curve 3 represents the cut o condition when B = Bc , the electron just grazing the anode. In Figure 2.4, the elecrton is assumed to have zero initial velocity, vx0 = vy0 = 0.
18
Figure 2.4: Electron trajectories in a planar crossed eld diode for dierent magnetic elds When the intial velocity of the electron is non-zero, the elecron trajectory will be quite dierent. Figure 2.5 demonstrates the electron trajectories with dierent initial velocity components, vx0 and vy0 . The magnetic eld is 0.41T, the DC voltage is 5.7kV, the distance between cathode and anode is 0.92mm, which are the operating parameters and dimension used for the X band magnetron, which has been investiagted in Chapter 5. In this diagram, the electrons leave the cathode surface with zero velocity (curve 1), velocity normal to the cathode surface (curve 3) and velocity at 45 degrees to the normal (curve 2) respectively.
Figure 2.5: Electron trajectories with dierent intial velocities and directions. Electron trajectory 1: the initial velocity is zero; Electron trajectory 2: vx |t=0 = vy |t=0 = E/(2B); Electron trajectory 3: vx |t=0 = 0, vy |t=0 = E/(2B); where E = Vdc /d, Vdc = 5.7kV , d = 0.92mm.
19
Now consider an AC eld applied between the cathode and the anode. The magnetic eld is bigger than the cut o magnetic eld Bc , thus the electron will not strike the anode directly. Electrons emitted at various positions of the RF period of the AC eld will be either accelerated or decelerated. Figure 2.6 demonstrates the trajectories of the two kinds of electrons.
Figure 2.6: Electron trajectories in the presence of AC eld When an electron is accelerated by the AC eld, the rolling circle of its cycloid path will increase; it will be deected by the magnetic eld more sharply and strike back to the cathode quickly as demonstrated by curve 1. When an electron is decelerated by the AC eld, it will experience weakened magnetic force, thus its rolling circle will be reduced in size. This electron will nally move horizontally with drifting velocity E/B as demonstrated by curve 2. In the rst case, the electron absorbs energy not only from the DC eld but also from the AC eld. Since this kind of electrons will strike back the cathode quickly, they will not aect the AC eld much. In the latter case, the electron will give up some of its energy absorbed from the DC eld to the AC eld. Thus the AC eld will experience a net gain of energy. This energy is obtained from the DC eld through the electrons.
20
2.3.2
Kinetic motion of electrons in a cylindrical diode
All the real magnetrons consist of coaxial cylindrical electrodes. Based on the above planar diode, a cylindrical diode will be discussed here. Figure 2.7 is the geometry of a cylindrical diode, with cathode radius rc and anode radius ra . The magnetic eld is parallel to the axis of the diode and perpendicular to the electric eld.
E
rc
ra
Figure 2.7: Geometry of a cylindrical diode [1] Using the same analysis as in section 2.3.1, the kinetic equations of the electron in cylindrical coordinates are given by d d2 r d = m[ 2 r( )2 ] dt dt dt
eE eBr
(2.17)
eB
1 d 2 d dr =m (r ) dt r dt dt
(2.18)
Rearranging Eq 2.18 results in the following equation d(r2 eB d )= rdr dt m
(2.19)
Considering that the electron leaves the cathode with zero initial velocity, d/dt|r=rc = 0, integration of Eq 2.19 yields c r2 d = (1 c ) dt 2 r2 (2.20)
21
where c = eB/m is the electron cyclotron frequency. The angular velocity of the electron is then obtained. 0 = c r2 (1 c ) 2 r2 (2.21)
In the cut-o condition, when the electron can just graze the anode, its radial velocity is zero, dr/dt|r=ra = 0 . Since we assume the electron leaves the cathode with zero initial velocity, according to the energy conservation, when the electron reaches the anode, its kinetic energy can be written as 1 mv 2 = eVa 2 t where the angular velocity is vt = ra d dt (2.23) (2.22)
Substituting Eqs 2.20 and 2.22 in Eq 2.21, the cut o condition between the magnetic eld and DC voltage can be obtained 8mVa /e 2 2 ra (1 rc /ra )
B = Bc = or Va =
(2.24)
e 2 r2 2 ra (1 c )2 Bc 2 8m ra
(2.25)
where Va is the DC voltage between anode and cathode. For a given Va , when B < Bc the electron will strike the anode; otherwise it will miss the anode. If we keep the magnetic eld constant, the electron orbit will strike the anode when Va > (Va )c ; otherwise, it will miss the anode. Figure 2.8 shows the dierent electron orbits in a cylindrical crossed eld diode (smooth anode magnetron). When an RF eld is present, the interaction between the electron and the RF led under the inuence of the DC voltage and the magnetic eld is similar to that of the planar diode. Electrons will be either accelerated or decelerated by the RF eld. When the DC voltage and the magnetic eld are properly selected, the energy will be transferred
22
from the DC eld to the RF eld via the electrons.
E
rc
Figure 2.8: Electron trajectories in a cylindrical diode for dierent magnetic elds. 1. B = 0; 2. B < Bc ; 3. B = Bc ; 4. B > Bc .
2.4
Resonant Cavity Magnetron Operation
The operational mechanism of an eight-cavity, hole and slot magnetron is discussed here. The anode block of this kind of magnetron is composed of eight coupled resonators surrounding the cathode. Figure 2.9 shows the cross section of the magnetron anode. In the axial direction the resonant cavities open into end spaces . The inner surface of the anode block consists of alternate vanes and gaps. RF elds excite inside the magnetron can be assumed to consist of two waves, travelling in opposite directions. Electrons are emitted from the cathode. Through the interaction between the electron cloud and the electro-magnetic eld, DC input power is converted to RF output power. The resonant system is the most important part for the magnetron design. It determines the frequency and conguration of the RF eld and the interaction between the electron cloud and the RF eld, leading to an ecient energy conversion. A good resonant system should have the characteristics of making the magnetron work in a stable manner. This includes stability against small frequency changes and frequency jumps between dierent operation modes. This aspect of beam-wave interaction will be discussed in more detail presently.
23
Figure 2.9: The cross section of an eight cavity hole and slot magnetron. The hole and slot structures provide inductance and capacitance for individual resonators.
2.4.1
Equivalent circuit and oscillation modes
Here the circuit theory is used to explain the range of possible RF oscillations. In a resonant cavity magnetron individual cavities replace resonant circuits. Usually the individual cavities are operating at their lowest oscillation mode. An equivalent oscillating circuit is shown in Figure 2.10. Complicated electro-magnetic coupling exists between the individual cavities. L and C are inductance and capacitance representing a single cavity. C is the capacitance between the individual anode segment and the cathode. The lines of magnetic ux in one individual cavity couple to another cavity in the end spaces; this kind of magnetic coupling is represented by M in the equivalent circuit of Figure 2.10 [2]. The oscillation frequency of an individual resonator is 0 = 1/ LC . Because the eight cavities are symmetrically distributed, the phase dierences between adjacent cavities are the same, say . In each cavity, the oscillating voltage between the anode vanes can be described by,
V1 = Vm sin(0 t)
(2.26)
24
Figure 2.10: Equivalent circuit of an eight-cavity magnetron [2]
V2 = Vm sin(0 t )
(2.27)
V3 = Vm sin(0 t 2) ......
(2.28)
V8 = Vm sin(0 t 7)
(2.29)
V9 = Vm sin(0 t 8) where Vm is the amplitude of the RF voltage.
(2.30)
Since magnetron is a closed circuit, V9 should be equal to V1 when the system is oscillating. This leads to 8 = 2n (n=0, 1, 2...). In general, the number of resonator cavities is N, the phase condition for oscillations to exist is given by
N = 2n
(2.31)
25
thus, = 2n N (2.32)
where n is an integer from zero to N-1. From Eq 2.32, it is obvious that there are several dierent modes that can satisfy the same phase condition for the oscillations to exist. Modes n = N/2 + m and n = N/2 m (m = 1, N/2) are pairs of degenerate modes. These modes are also called doublets. If the magnetron is a perfect symmetrical structure, the two components of a doublet would have the same wavelength. Thus the total possible modes in a N cavity system are from n=1 to n=N/2. When n=N/2, = , its corresponding doublet is the n=0 mode; this means zero charge distribution on the anode segments, and the mode does not exist. Thus, the mode is a non-degenerate mode, or a stationary wave mode. All the other modes are traveling waves or having a traveling wave component superimposed on a standing wave. In reality exact symmetry is not attainable. Also there is asymmetry caused by the output-coupling loop. Thus the two components of a doublet have slightly dierent wavelengths. For an eight-cavity magnetron, the possible modes are n=1, 2, 3, 4, n=4 corresponding to the mode. The charge and electric eld distribution of the four possible modes are illustrated in Figure 2.11. If is the resonant frequency of mode n, the time needed for traveling wave to travel around the anode for one round would be nT, where T = 2/ ; therefore its rotating frequency would be = 2 = nT n (2.33)
So the rotating velocity around the anode is slowed down by a factor n. It is this that makes it possible for the electron to rotate in synchronism with the RF eld, which enables the energy transfer to the RF eld.
26
(a)
(b)
(c)
(d)
Figure 2.11: Field and charge distribution of dierent modes in an eight-cavity magnetron (a)n=1 (b)n=2 (c)n=3 (d)n=4
2.4.2
The Hartree/Hull conditions
For a magnetron to operate properly, special conditions between the applied anode voltage and the static magnetic eld must be satised. In this section, the two equations regarding these conditions will be introduced and applied to the two magnetrons (S/X band), which are the subject of the simulation work in the later chapters. It is discussed in Section 2.3.2 that in the cylindrical diode magnetron, for an electron to touch the anode, the cuto condition is
2 rc era 2 [1 ( )2 ]2 B0 8m ra
V0 =
(2.34)
27
Where V0 ra rc B0 Cuto anode voltage; Radius of anode; Radius of cathode; Cuto static magnetic eld.
In the absence of RF eld, with a given magnetic eld, when the anode voltage is smaller than V0 , the electron will not reach the anode. This voltage V0 is called the cuto voltage, or Hull voltage. When a magnetron is operating below the cuto voltage, the electron cloud will be formed between the cathode and the anode. For the magnetron to start oscillating, it is required that the electrons should rotate synchronously with the RF eld at the anode surface, so that
e = where e /n n The The The The
(2.35)
angular frequency of electron; angular frequency of RF eld; rotating frequency of the corresponding traveling wave; mode number of the traveling wave.
Applying this synchronous condition to the two equaions Eq 2.17 and Eq 2.18, which govern the electron movement in a cylindrical coordinate system, the threshold voltage for RF oscillation to start in a magnetron can be obtained.
Vt =
2 ra r2 r2 m (1 c ) B a ( )2 2 2 ra n 2 e n
(2.36)
The threshold voltage is also known as the Hartree voltage after Hartree who developed the theory. Rigorous solutions of this problem can be found in [2, 3]. This voltage
28
is the condition at which oscillations should start provided at the same time that B is suciently large so that the undistorted space charge does not extend to the anode. Assuming a constant ratio of rc /ra , consider the eect of keeping B and xed, Vt
2 is proportional to ra , so that operating conditions remain unchanged if the operating
voltage is increased proportional to the square of the anode radius. Eq 2.34 combined with Eq 2.36 determine the operation zones of a magnetron, as shown in Figure 2.12, which describes the situation in an eight-cavity magnetron.
Figure 2.12: The Hartree diagram of an eight-cavity magnetron [2] Eq 2.34 is a parabolic curve, above which, magnetron cannot operate. Eq 2.36 is a straight line that is tangent to the parabolic curve. In the Hartree diagram shown in Figure 2.12, mode has the lowest Hartree voltage. Below this voltage, magnetron also cannot operate, because in this region, the DC eld is too weak to move the electrons into synchronism with the slowest rotating RF eld, i.e. the mode. When the anode voltage is slightly above the Hartree voltage, the magnetron starts to oscillate. It has been observed that the region between the Hull voltage curve and the mode Hartree voltage line is the usual magnetron operating zone. However, in practice, the anode voltage of a magnetron is always set below the 1 mode line (n=3 in Figure 2.12) to avoid mode competition.
29
2.4.3
The Hartree/ Hull conditions for the S/X band magnetrons
In this section, the theory discussed above has been applied to the simulation models of an S band hole and slot magnetron and an X band strapped magnetron. Since in practice the main competing mode is the 1 mode, other higher order modes have minor eects on mode oscillations; only and 1 modes have been calculated and plotted in the following Hartree/Hull graphs, see Figure 2.13. The anode and cathode dimensions, the static magnetic elds applied, the oscillation frequencies of and 1 modes, which are described in Chapters 4 and 5 separately and their corresponding Hartree/Hull voltages of the two magnetrons are presented in the table 2-A. Table 2-A: Hartree/Hull condition for e2v magnetrons Magnetrons B0 (T) ra (mm) rc (mm) VHull (kV) VHartree() (kV) VHartree(1) (kV) f (GHz) f1 (GHz) MG5193 (S Band) and MG5241 (X Band) S Band 0.16 17 9.25 80.49 42.89 52.58 2.99 3.25 X Band 0.405 2.82 1.9 8.54 4.84 5.66 9.50 10.42
Magnetrons can only operate between the Hull parabola and the straight Hartree lines corresponding to dierent modes. The cuto voltages and magnetic elds of dierent modes are the tangential points in the diagrams, marked by *. mode has lower cuto voltage and magnetic eld than -1 mode. With the same magnetic eld the mode oscillation requires a lower threshold voltage than -1 mode. In real magnetron operation, the -1 mode always competes with the mode, potentially spoiling the magnetrons operation. Simulations in Chapters 4-5 will demonstrate that when the input voltages is above the Hartree voltage of -1 mode, the system always skips mode oscillations and moves into -1 mode. Mode competition is a complicated and
30
challenging problem, it is related to many factors. A detailed discussion will be carried out in the following section.
Hartree/Hull relationship between anode voltage and magnetic field 140
120
100
Anode voltaeg, KV
80
60
40
20
0.02
0.04
0.06
0.08 0.1 0.12 Magnetic field, Tesla
0.14
0.16
0.18
0.2
(a)
Hartree/Hull relationship between anode voltage and magnetic field 14
12
10
Anode voltaeg, Volt
0.05
0.1
0.15
0.2 0.25 0.3 Magnetic field, Tesla
0.35
0.4
0.45
0.5
(b)
Figure 2.13: Operating voltage and magnetic eld for dierent modes in S/X band magnetrons (Solid line represents mode, mode number n=6 for S band, n=9 for X band; dashed line represents -1 mode, mode number n=5 for S band, n=8 for X band). (a). e2v type S band MG5193 (b). e2v type X band MG5241
31
2.4.4
Interaction between the RF Field and the electron cloud
The mechanism of interaction between electrons and the RF eld in a resonant cavity magnetron is similar to that in a cylindrical diode magnetron. Usually the operating mode in a magnetron is the mode. In this section the interaction mechanism will be explained whwn the system is oscillating in the mode.
12 3 4
Figure 2.14: Electron trajectories when the system is oscillating in the mode.
The RF eld distribution of mode is illustrated in Figure 2.14. There are four typical initial positions of electrons inside the cavity, labeled 1, 2, 3, and 4 in the diagram. The RF eld aects these electrons dierently. Electron 1 is in the accelerating RF eld, so it will be speeded up, and the curvature of its trajectory will be increased. Under the combination eects of static electric eld and magnetic eld, electron 1 will strike back the cathode quickly and thus be removed from the interaction region. Electron 1 wont contribute energy to the RF eld; on the contrary, it will consume a small amount of RF energy. Electron 3 is in the decelerating RF eld. Its angular velocity will be reduced, resulting in the reduction of its curvature. Such electrons will move towards the anode. If the frequency of the RF eld is appropriate, this electron will always be in the decelerating RF eld when it passes through the successive anode segments. This electron will keep giving out energy to the RF eld until it nally strikes the anode. Electrons 2 and 4 exchange little energy with the RF eld. The eect of the RF eld is to tune the velocity of their guiding centers. For electron 2, this velocity is increased, thus its guiding center will catch up that of electron 3; but for electron 4,
32
its guiding center velocity is reduced, thus it will fall back towards electron 3. The nal result is that electron 2 and 4 move into the decelerating region, and bunch together with electron 3 forming so-called spokes. This kind of phenomenon is called the phase focusing action of the RF eld. The four dierent electron trajectories are also shown in Figure 2.14. Electrons giving up energy to the RF eld are called favorable electrons. Electrons absorbing energy from the RF eld are called unfavorable electrons. Once oscillations start inside the cavity, unfavorable electrons tend to be removed from the interaction region rather quickly; the favorable electrons, which are in the majority, tend to bunch and move towards the anode while rotating around the cathode at an angular frequency , Eq 2.33. When the oscillations are stabilized, the electron cloud will be shaped into several spoke-like ridges rotating around the cathode. The number of spokes is determined by the RF mode. For the mode, in an eight-cavity magnetron, the number of spokes is 4. Figure 2.15 shows electron spokes in an eight-cavity magnetron. This electron cloud rotates around the cathode with an angular velocity , thus keeping it in step with the phase velocity of the RF eld and continuously giving up energy to the RF eld. Electrons at the end of the spokes land on the anode with velocities well below the DC value of v = eVa /2m
Figure 2.15: Electron spokes in an eight-cavity magnetron when the system is operating in the mode [2]
33
2.5
2.5.1
Frequency Stability and Mode Separation

Frequency jumping
For dierent modes, the charge density on the anode segment is dierent, which results in the dierent eective capacitance and inductance, thus the dierent resonating frequency. Usually the mode has the lowest resonating frequency and the highest power eciency, largely because capacitance is maximum in that mode. The mode is always selected as the operating mode in a magnetron. Unfortunately the frequency dierence between and other modes is not as great as would be desired, therefore frequency jumping from one mode to another can happen. This is undesirable in practice because eciencies of dierent modes vary considerably. Low eciency means low power output and severe heating of the anode and the cathode, and the life shortening of the latter. Frequency jumping can also happen when the magnetron is operating in a doublet mode. The frequency may jump from that of one component of the doublet to that of the other more easily than it jumps from one mode to another, since the two components of a doublet normally have nearly the same frequencies. Because the two components are coupled to the output to a dierent extent, slight changes in the output load can cause a jump.
2.5.2
Strapping
In order to improve the mode stability, several methods of mode separation have been developed. One of them is the use of strapping. Alternate anode segments are connected together to help make the phase dierence between them to be zero or 2, thus enforcing the anode to work in the mode. There are two main forms of strapping, ring strapping and echelon strapping. Figure 2.16(a) shows echelon strapping of an S band magnetron and ring strapping of an X band magnetron is shown in Figure 2.16(b). These two magnetrons will be investigated in Chapter 4 and Chapter 5 respectively.
34
(a)
(b)
Figure 2.16: The strapping system. (a)Wire strapping system of an S band magnetron. (b)Ring strapping system of an X band magnetron. According to the circuit theory, strapping aects the capacitance and inductance of the equivalent circuit, which determines the resonating frequency of the system. When operating in mode, all the segments connected to one set of straps are at the same potential. There will be no current owing along the straps. But since the two sets of straps are at dierent potentials, there will be eective capacitance in parallel with the resonator capacitance, which increases the wavelength. When operating in non- modes, the two ends of a strap are at dierent potentials. There will be net current owing along the straps. The magnitude of the current depends on the potential dierence between the ends. The straps therefore act as inductive path, connected in parallel with the eective resonator inductance, the wavelength is therefore reduced. The eect of strapping is
35
therefore that the frequencies of the non- modes are increased and that of the mode is decreased. So, by tuning the strapping system, the mode and non- modes can be separated appreciably. Also the Q factors of dierent modes can be changed by adjusting the straps so that the mode can dominate the oscillation and other modes are suppressed. Since it is impossible to make the straps of zero impedance, the possibility for operation in a non- mode still exists since the strap impedance may cause potential dierence between two connected segments. Strap breaks have been introduced to further suppress the non- modes oscillations. It is found that by introducing some asymmetry into the strapping system, such as removing one or two straps in the case of echelon strapping, Figure 2.16(a), or introducing one or more breaks in the ring straps, Figure 2.16(b), the loaded Q of the lightly coupled component of the doublet can be adjusted to a value nearly equal to that of the heavily coupled one. This makes the non- modes oscillations less likely. This phenomenon can be explained in another way. When operating in non- modes, current ows in the straps. By introducing strap breaks, the current ow is stopped, therefore the non- modes oscillations are suppressed.
2.5.3
Rising sun anode
When the operating frequency is further increased, the dimensions of a strapped magnetron are correspondingly reduced. This makes it very dicult to construct straps. Also small spacing results in a large copper loss and reduced magnetron eciency. The rising sun anode has been developed to overcome such diculties. Figure 2.17 shows the geometry of a rising sun anode, which alternately has large and small cavities. A rising sun anode can be considered as having two resonating systems, one consisting of small cavities and the other of large cavities. The two resonating systems have dierent resonating frequencies corresponding to all the possible modes as in the equal size cavity magnetron. Naturally coupling exists between the two resonant systems. The
36
resonant frequency of a rising sun anode is between those of the small and large cavities. By adjusting the ratio of the frequencies of the small and large cavities, the optimum frequency separation between mode and other modes can be obtained. Another advantage of rising sun anode is that it is no longer necessary for the anode length to be small compared with the wavelength in order to avoid instability. This permits the use of a relatively long anode, enabling higher power generation.
Figure 2.17: Geometry of a rising sun anode [4]
2.5.4
Other factors aecting frequency stability
Besides strapping, there are other methods for mode separation. In Chapter 4, a mechanical tuning mechanism is introduced in an S band magnetron to detune the -1 mode and change its Q factor in order to encourage mode oscillations. Ceramic resonators are used in some tubes to absorb -1 oscillations. Other factors aecting the mode jump include cathode to anode ratio, the regulation of the modulator and the output coupling circuit, etc. By varying the cathode to anode ratio, the operating conditions of Hartree/Hull curves will be changed. The rate of rise of the anode voltage (RRV) is also critical to the start-up of oscillation. A rise which is too steep can cause the magnetron to jump into a non- modes oscillation. Besides the strapping, it is observed in Chapter 5 that the Q factors of and -1 modes can also be tuned by modifying the output coupling circuit, thus suppressing -1 mode oscillations. Detailed discussion of these items is included in the following chapters.
37
References
[1] R.S.H.Boulding, Ed., The Resonant Cavity Magnetron. The Whitepriars Press Ltd., 1952. [2] G.B.Collins, Ed., Microwave Magnetrons. Radiation Laboratory Series, 1948. [3] P.A.Lindsay, M.Esterson, and X. Chen, The magnetron threshold volatgerevisited, Phys. Plasmas, vol. 4, no. 2, pp. 463468, February 1997. [4] T.A.Treado, W. Doggett, and G.E.Thomas, Operating modes of relativistic risingsun and A6 magnetrons, Plasma Science, IEEE Transactions, vol. 16, no. 2, pp. 237248, April 1988. Massachusetts Institute of Technology,
Chapter 3
The MAGIC Code and Its Associated Algorithms

3.1 Introduction
In this report, all the simulation work is carried out using a code called MAGIC. The MAGIC code is a Finite-Dierence, Time-Domain (FDTD) and Particle-In-Cell (PIC) code for simulating plasma physics processes, i.e., those processes that involve interactions between space charge and electromagnetic elds. Beginning from a specied initial state, the code simulates a physical process as it evolves in time. The full set of Maxwells time-dependent equations is solved by FD-TD algorithm to obtain electromagnetic elds. Similarly, the complete Lorentz force equation is solved to obtain relativistic particle trajectories and the continuity equation is solved to provide current and charge densities for Maxwells equations. This approach, commonly referred to as the electromagnetic Particle-In-Cell (PIC) technique, provides self-consistent solutions of the interaction process between charged particles and electromagnetic elds [1]. Brief introduction to FD-TD method and PIC algorithm will be given below. The MAGIC algorithm has been developed in three typical curvilinear coordinate 38
Chapter 3. The MAGIC Code and Its Associated Algorithms
39
systems, i.e. Cartesian, Cylindrical and Spherical. The choice of the coordinates usually depends on the geometry of the boundary conditions. In modelling magnetron, a cylindrical coordinate system is the best, since it conforms to the geometry of the cathode and the anode.
3.2
Finite Dierence Time Domain Technique
In 1966 Yee [2] introduced FDTD method to solve electromagnetic problems. This method uses simple central dierence approximations to evaluate the space and time derivatives in Maxwells time dependant curl equations. The region being modeled is represented by two interleaved grids of discrete points as shown in Figure 3.1. In this diagram, each magnetic eld vector component is surrounded by four electric eld components, vice versa.
Figure 3.1: Yees cell, for electric and magnetic eld vectors
3.2.1
FDTD algorithm
The Maxwell equations are given by:
D =
(3.1)
40
B =0
(3.2)
E =
H t
(3.3)
H =
E + Js t
(3.4)
Js = E Where: D = eE: B = H: E: H: : : Js : : Electric ux density Magnetic ux density Electric vector eld Magnetic vector eld Magnetic permeability Electric permittivity Electric conduction current density Electric conductivity
(3.5)
The leap-frog scheme (see Figure 3.2) is used to update the E and H elds alternately in the time domain. This scheme uses the values of the E eld at time t to nd the values of H eld at time t+t and then uses the same method to nd the values of E at t+2t, thus allowing an electromagnetic wave to march forward over the space grid.
E k 1 B k 1 / 2 ( k 1) t ( k 1 / 2 ) t
Ek kt
B k +1 / 2 ( k + 1 / 2) t
Figure 3.2: The leap-frog time marching scheme
41
The rst order partial derivative, central dierence approximation used in FDTD, is given by a Taylor series expansion.
F (x) F (x + (x/2)) F (x (x/2)) = + O((x)2 ) x x
(3.6)
Following Yees notation, a grid point in the region under investigation can be expressed as
(i, j, k) = (ix, jy, kz)
(3.7)
Thus, any function of space and time can be expressed as An (i, j, k) = Am (ix, jy, kz, nt) = Am (x, y, z, t) m
(3.8)
Where A: m: t: x: y: z: i,j,k : n: Vector electric E or magnetic eld H Vector eld component (x, y, z) Temporal increment x spatial increment y spatial increment z spatial increment Cartesian spatial grid coordinates time step
The 2nd and higher order derivatives of the series, which are usually not computed, are treated as a rounded up error of the order O((x)2 ), as indicated in equation 3.6. Applying the above expansion to Maxwells curl equations, the electric and magnetic
42
vector eld components can be obtained in following equations 3.9 and 3.10.
Ez |n+1 = i,j,k
1 1+
i,j,k t 2 i,j,k i,j,k t 2 i,j,k
Ez |n + i,j,k
i,j,k
Hy |i+ 12,j,k Hy |i 12,j,k

2 2
n+ 1
n+ 1
1+
i,j,k t 2 i,j,k
2 2 Hx |i,j+ 1 ,k Hx |i,j 1 ,k 2 2
n+ 1
n+ 1
y (3.9)
n+ 1 2 Hy |i,j,k
1 1+
i,j,k t 2i,j,k i,j,k t 2i,j,k
n 1 2 Hy |i,j,k +
t i,j,k i,j,k t + 2i,j,k
Ez |n 1 ,j,k Ez |n 1 ,j,k i+ i
2 2
Ex |n Ex |n i,j,k+ 1 i,j,k 1
2 2
z (3.10)
The equations for other eld components, Ex , Ey , Hx and Hz can be obtained using the same method.
3.2.2
Numerical stability
The FDTD algorithms for Maxwells curl equations require that the time increment t have a specic bound relative to the lattice increments x, y, z. This bound is necessary to avoid numerical instability, which can cause spurious results increasing without limit as the time marches on. The bound on the time interval is given below 1 c
1 (x)2
1 (y)2
1 (z)2
(3.11)
When cubic cells are used (x = y = z = ), equation 3.11 can be simplied as
t c 3
(3.12)
where c is the velocity of light. The above two equations state that the maximum allowable time step is determined by the smallest cell size, so that electromagnetic wave
43
cannot traverse across a grid in one time step. The above discussion focused on the numerical stability of the basic Yee algorithm in Cartesian coordinates. However, the stability of the entire FD-TD procedure for Maxwells equations depends upon more than the stability of the Yee algorithm. In fact, a generalised stability problem arises due to interactions between the Yee algorithm and augmenting algorithms used to model: boundary conditions; variable and unstructured meshing; lossy, dispersive, nonlinear, or active materials.
3.2.3
Uncertainty analysis
The numerical algorithm for Maxwells equations dened by the FDTD method can cause numerical dispersion of the simulated wave modes in the computational lattice. That is, the phase velocity of the simulated wave in the FDTD grid can dier from the vacuum speed of light, varying with the modal wavelength, the direction of propagation in the grid, and the grid discretization. So a proper grid discretization, usually smaller than one tenth of the wavelength, is used to minimize this error. Staircase within rectilinear FDTD grids related to material boundary can also cause uncertainty. The well-known contour path modelling, or fuzzy technique, deriving eective permittivities by looking at electric ux eld values represents a simple alternative method compared to sophisticated gridding to avoid staircase eects without additional computational cost during the time-stepping algorithm.
3.2.4
Absorbing boundary condition
In order to model open region problems, an absorbing boundary condition (ABC) is often used to truncate the computational domain, since the tangential components of the electric eld along the outer boundary of the computational domain cannot be updated using the Yee algorithm. Most of the popular ABCs can be grouped into those that are
44
derived from dierential equations or those that employ a material absorber. The absorbing boundary condition has been introduced by Taylor et al [3], who use a simple extrapolation method. A partial dierential equation that permits wave propagation only in certain directions is called a one-way equation. When applied at the outer boundary of an FDTD grid, a one-way wave equation numerically absorbs impinging scattered waves. Engquist and Majda derived a theory of the one-way wave equation suitable for ABCs in Cartesian FDTD grids [4]. Based on Engquist-Majdas formulation, Mur developed a second order ABCs [5], which annihilates the normal incident wave quite accurately. However for oblique incident waves, large reections can occur. Taove introduced losses in the region that surrounds the structure being modeled, which absorbs both the outgoing waves and the waves that are reected by the boundaries of the mesh. The use of material ABCs made a signicant advance in 1994 with the work of Berenger [6]. Berengers ABC, termed the perfectly matched layer (PML) absorbing boundary condition, employed a split-eld artice to ensure that the phase velocity tangential to the PML region was the same as in free space for all frequencies and for all angles of incidence.
3.3
3.3.1
Particle In Cell (PIC) Technique

The Computational cycle
The general computational cycle for interaction between electrons and electromagnetic elds is demonstrated in Figure 3.3. This cycle starts with some appropriate initial conditions on the particle positions and velocities. The particle quantities are known at the particle and may take on all values in phase space. The eld quantities will only be obtained on the spatial grids, that is, discrete points. In order to calculate the eld quantities on the grids, the charge and current densities on the grids must be known rst. This process of assigning particle charge and current on the grid is a kind
45
Fi vi xi
( E , B) j F j
( x, v ) i ( , J ) j
( E , B) j ( , J ) j
Figure 3.3: A typical cycle in PIC processing [7] of weighting, which depends on the particle position. Once the charge and current densities are established on the grid, various kinds of numerical methods can be used to obtain the electric and magnetic eld components. With the elds known on the grid, the electromagnetic forces acting on particles can be obtained through interpolating the grid elds and the particle position. Thus the electron can be marched forward. This process includes another weighting of the eld components and the particle [79].
3.3.2
Integration of the equations of motion
The Leap-frog method is used to solve the particle motion equations because of its simplicity and accuracy. A rst-order example is illustrated here. The two rst-order dierential equations to be integrated separately for each particle are v =F dt
(3.13)
dx =v dt
(3.14)
Where
46
F: m: v: x:
the electromagnetic force acting on the particle particle mass particle velocity particle position
These equations are replaced by nite dierence equations vnew vold = Fold t
(3.15)
xnew xold = vnew t
(3.16)
The ow in time and notation are shown in Figure 3.4. The program is time centering and will advance the velocity and position alternately. That is, v and x cannot be known at the same time. That is why this method is called leap-frog method.
t 2
t+
t 2
t + t
Figure 3.4: Sketch of leap-frog integration method showing time centering of force F while advancing v , and of v while advancing x [7] Two issues must be addressed here. First, the initial conditions for particle velocities and positions must be modied to t in the ow. Using the force F at t=0, velocity v(t = 0) is pushed back to v(t = t ). Second, the energies calculated from v(kinetic) 2
and x(potential, or eld) must be adjusted to appear at the same time.
47
3.3.3
Particle and force weighting
As described in Figure 3.3, the computing cycle of PIC performs two weighting calculations. One is the charge and current weighting, and the other is the eld or force weighting. In order to calculate the force on the grid points, it is necessary to calculate the charge and current density on the discrete grids from the particle. This process is performed by interpolating the particle charge between neighbouring grids. A zero order example is shown in Figure 3.5.

X j 1
X
j x 2
Xj
xi
X j +1
j+
x 2
n j (x j )
Xj
xj
Figure 3.5: The scheme for the zero order charge and current weighting [7] The grid charge density n(j) at grid point j is obtained by assigning all the particles N (j) within the j th cell (from j x/2 to j + x/2), that is n(j) =
N (j) x ,
as shown in
Figure 3.5. This weighting is called nearest-grid-point weighting or NGP. When particles move in or out of the j th cell, the grid density will jump up or down, making the observed particles to have a nite size rather than a point size. One eect of this jumping is that it introduces associated density and eld random jumps or noise, both in space and time. This noise may become intolerable in many plasma physics problems. Higher order weighting can avoid this problem. The eld or force weighting can be conducted in the same manner. The zero order of the NGP force comes from the eld at the nearest grid point, that is E(x i ) = E(xj ). Where xi is the position of the particle, xj is the nearest grid point to xi . Discussion for
48
higher order weighting can be found in [7, 9].
3.4
Emission Models in MAGIC Code
The MAGIC code provides several emission models to simulate dierent emission processes, such as the thermionic emission, the photoelectric emission, the space-chargelimited emission, the beam emission, the secondary emission or the high eld emission. The emission models used in our modelling work are the space-charge-limited emission, the beam emission and the secondary emission. They will be described briey below. Detailed explanation of the emission models can be found in the MAGIC manual book [1].
3.4.1
The space-charge-limited emission
In MAGIC, the space-charge-limited emission is given the name explosive emission. In the computer model, the physical details of the explosive emission, which results from the plasma formation on a material surface, have been ignored. Instead, the computer model relies on a phenomenological description. The particle emission itself is based upon Childs law, specically, the normal electric eld vanishing at the emitting surface. In the phenomenological treatment of plasma formation, breakdown can occur only if the normally directed eld at the half-cell, Ec , exceeds some specied breakdown eld threshold, Ec > Ethreshold . If the eld at a particular cell exceeds the threshold, then that cell is said to break down. A single, non-emitting cell between two emitting cells is also allowed to break down, even if the threshold is not exceeded. Once initiated by breakdown, plasma formation in a cell is assumed irreversible. Thus, a cell that has broken down may emit charged particles due to the inuence of the ambient electric eld. The calculation of the charge that would be drawn away from a surface is based upon application of Gausss law to the half-cell, allowing for a small residual eld at the surface, dQ/dA = 0 (Ec Er ) , where dQ/dA is the emitted charge, Er is the residual
49
eld at the surface, is the existing charge density at the surface.
3.4.2
The beam emission
With the beam emission model, the emission from a surface has the emission current density and the initial beam voltage, specied by the user. The initial kinetic energy and momentum for each particle are computed relativistically from the beam voltage. The beam emission model also provides options to vary the beam direction and distribution by either thermalizing the beam or giving it a specic direction (dieren from the normal) to the surface. For example, the THERMALIZATION option allows you to specify a fraction of the beam energy to be distributed randomly in the transverse coordinates. The energy in each coordinate is determined randomly, and the dominant (normal) component is re-normalized to preserve the exact specied beam voltage.
3.4.3
The secondary emission
With this emission model, secondary electrons are emitted from the surface of a conducting spatial object when it is hit by primary particles. Figure 3.6 illustrates the coordinate system. The polar angles of the primary particle and the secondary electron, p , s , are measured with respect to the surface normal. The secondary emission yield depends upon the primary particle energy Ep and angle of incidence p , and upon the material-dependent parameters, the peak yield peak and the primary energy at peak yield Epeak . In MAGIC, the secondary emission coecient, , is assumed to be a product of two functions, one in the primary polar angle and one in primary energy. (Ep , p ) = 0 (Ep ) cos(p ) (3.17)
50
Figure 3.6: Coordinate system illustrating primary particle and secondary electron
0 (Ep ) = peak
Ep exp(2 Epeak
Ep ) Epeak
(3.18)
Here, 0 is the secondary emission coecient at normal incidence for the primary particle. An articial cuto in the polar angle dependence (at 85 degrees) is employed to prevent the coecient from going to innity for grazing angles. On default, the secondary energy distribution is peaked at 7.5 eV with a full-width at half-maximum of 10 eV. The angular distribution is homogeneous, having no preferred direction.
References
[1] L.Ludeking, Ed., Manual of MAGIC Tool Suite. 2002. [2] K.S.Yee, Numerical solution of initial boundary value problems involving maxwells equations in isotropic media, IEEE Trans on Antennas and Propagation, vol. 14, no. 4, pp. 302307, 1966. [3] C.D.Taylor, D.H.Lam, and T.H.Shumpert, Electromagnetic pulse scattering in timevarying inhomogeneous media, IEEE Trans on Antennas and Propagation, vol. 14, Mission Research Corporation,
51
no. 4, pp. 302307, 1966. [4] B.Engquist and A.Majda, Absorbing boundary conditions for the numerical simulaiton of waves, Math. Comp, vol. 31, pp. 629651, 1977. [5] G.Mur, Absorbing boundary conditions for the Finite-Dierence approximation of the Time-Domain electromagnetic-eld equations, IEEE Trans. Electromagnetic Compatibility, vol. 23, pp. 377382, 1981. [6] J.P.Berenger, A perfectly matched layer for the absorption of electromagnetic waves, Journal of Computational physics, vol. 114, pp. 185200, 1994. [7] C.K.Birdshall and A.B.Langdon, Eds., Plasma Physics Via Computer Simulation. McGraw-Hill, New York, 1985. [8] C.K.Birdshall, Particle-in-cell charged-particle simulations, plus monte carlo collisions with neutral atoms, IEEE Transactions on Plasma Science, vol. 19, no. 2, pp. 6585, April 1991. [9] R.W.Hockney and J.W.Eastwood, Eds., Computer Simulation Using Particles. Adam Hilger, Bristol, 1988.
Chapter 4
Modelling of an S band Magnetron

An S band slot and hole magnetron (e2v technologies type MG5193) has been modeled using the 3D version of the MAGIC code. The anode structure of the S band magnetron (MG5193) is illustrated in Figure 4.1. The underlying symmetries are two perpendicular polar systems, the output system and the anode system. For the convenience of respectively computer modelling, these two parts will be investigated and analysed separately in sections 4.1 and 4.2.
Figure 4.1: Image of an S band magnetron (e2v type MG5193)
52
Chapter 4. Modelling of an S band Magnetron
53
4.1
4.1.1
Modelling of The Output System

Introduction
Simulation of the output system of the S band magnetron has been carried out because when the glass dome was replaced by a ceramic one the two metal rods forming the output coupling loop would always melt in operation and thus destroy the magnetron. Here, the dome is a vacuum tight RF window, see Figure 4.1. The computer model of the output system is shown in Figure 4.2, where the melting points are marked in red. The two coaxial coupling rods are connected by a dipole antenna forming a double loop. The output port of the system is a rectangular waveguide, which is connected to the circular waveguide through a kind of water-fall transition. Figure 4.3 shows a 3D picture of transition. The RF power generated in the anode-cathode cavity reaches the output system through two coaxial line ports, which feed the dipole antenna. In the circular part of the output system the RF power travels as a T E11 wave and is then transformed to a T E10 wave as it enters the rectangular waveguide.
Figure 4.2: The geometry of the output of the S band MG5193 magnetron In order to explain the melting phenomenon, various possible reasons have been looked at. Computer modelling has been conducted to check the resonance spectrum, input and output power, current and voltage distribution. Also, an investigation of the
54
Figure 4.3: Computer model of the rectangular output of the S-band MG5193 magnetron possible multipactor eect between the input coupling rods has been carried out.
4.1.2
Cuto frequency and possible modes in the output system
The cuto frequencies of various modes in the circular wave-guide and the rectangular wave-guide of the output system have been calculated in order to check all possible modes that could be supported. Polar coordinates have been selected, with the axis of the output waveguide in the z direction. For the TM (Hz = 0) modes inside the circular waveguide, the boundary conditions require Ez and E to be zero at the outer wall (r=a), resulting in Jn (kc a) = 0. Jn (x) is a Bessel function of the rst kind, n denotes angular variation.
2a kc a = c a = = pnl c For a given n, the cuto frequency of a TM mode is given by pnl 2a
(4.1)
fc =
(4.2)
where l indicates the root number and refers to radial variations. The cuto frequency of the TE(Ez = 0) mode in a circular waveguide can be calculated in the same way. In
55
this case the boundary conditions at r=a require E = 0, so that 2a = pnl kc a = c a = c (4.3)
where Jn (pnl ) = 0. Jn (x) is the rst derivative of Bessel function, again n denotes angular and l denotes radial variation. For given n and l, the cuto frequency of a TE mode is given by fc = pnl 2a (4.4)
For the circular waveguide the radius of the waveguide is a=35mm between the input ports and the dipole antenna. The possible modes and their corresponding cuto frequencies are shown in Figure 4.4.
Figure 4.4: The cuto frequencies of the circular wave-guide (a=35mm) The radius of the middle section (beyond the dipole antenna) of the circular waveguide is a=42.9mm. Its possible modes and corresponding cuto frequencies are shown in Figure 4.5.
Figure 4.5: The cuto frequencies of the circular wave-guide (a=42.9mm) Since the operating frequency is 2.998GHz, the only possible mode in the small radius section of the waveguide is the T E11 mode. But in the middle section, two possible modes may be present, T E11 and T M01 . In the rectangular part of the waveguide, the cuto frequencies are the same for TE
56
and TM modes, except that for a TM wave neither m nor n (Eq 4.5) can be zero. So, there are no T M01 or T M10 modes. fc = 1 2 m 2 n 2 + a b (4.5)
where m and n are integers. The dimensions of the rectangular waveguide are a=34mm, b=72mm. Calculations shows that only the T E10 mode can propagate, its cuto frequency being 2.08Ghz, see Figure 4.6.
Figure 4.6: Cuto frequencies of the rectangular wave-guide
When the system is operating at 2.998GHz, besides the operating mode T E11 , T M01 may also be excited in the circular waveguide. But in the rectangular section, only the T E10 mode is possible. The T E11 mode in a circular waveguide has the same eld distribution as the T E10 mode in a rectangular waveguide. Thus, the rectangular output can transmit RF power in the T E10 mode. The eld distribution of the mode is discussed in next section.
4.1.3
Modelling of the output system
The resonance spectrum and current distribution on the input rods and the antenna have been investigated through computer modelling. The results are presented here in the form of tables and diagrams. All the simulation has been completed using the MAGIC code, which has been introduced in Chapter 3.
57
4.1.3.1
Current distribution on the input rods and the dipole
As introduced at the beginning of this chapter, when the glass dome is replaced by a ceramic one, the two coupling rods always melt during operation. Therefore the objective of the investigation of the output system was to nd the reason for the melting. The current and voltage distribution along the two coupling rods and the dipole antenna have been measured to see if RF overheating causes the melting. The power level of this magnetron is 4MW, pulsed. The input voltage and current on each input port are respectively 20kV and 100Amps (peak values). Figure 4.7(a) shows the current distribution on the coupling rod with both glass dome and the ceramic dome. It drops sharply where the dimension of the rod changes. This is due to the discontinuity of the impedances at that point. The current distribution along the dipole is shown in Figure 4.7(b); it shows that the current decreases from the center of the dipole towards the two ends. The currents in these diagrams are all peak values. Figures 4.7(a) and 4.7(b) present the current distributions on the rod and the radiating dipole with ceramic and glass domes respectively. It indicates that with the ceramic dome, the dipole current is higher than with the glass dome, but the current distribution on the rod is atter. It is obvious that for both glass and ceramic domes the highest current is in the feeding coax lines near the input port, which are thinner in dimension than the rods and the bar. Considering the melting problem addressed at the beginning of this chapter, conclusion can be reached that RF overheating cannot be the reason for the melting of the metal rod, because if it is the current that causes the melting, the two coaxial lines should melt rst.
4.1.3.2
Eect of the dome
The eects of the dome have been investigated. Simulations have been carried out with the glass dome, the ceramic dome and without the dome. It is found that when the dome
58
(a)
(b)
Figure 4.7: (a)Current distribution along the coupling rod. (b) Current distribution along the radiating dipole. (: glass dome; : ceramic dome) is removed, the reection is getting bigger. Also from the simulation it is observed that the ceramic dome has smaller reection than the glass dome. The results are shown in Figures 4.8 to 4.10. In these diagrams, instant power is obtained by integrating Poynting vector at the input port. The positive values represent the transmitted power; the negative values represent the reected power. The rate of reection is obtained by dividing the reected
59
Figure 4.8: Transmitted and reected power at the input port with a glass dome (Power reection: 11.5%)
Figure 4.9: Transmitted and reected power at the input port with a ceramic dome (Power reection: 4.0%) by the transmitted power. The magnitudes in these diagrams have no absolute meaning, therefore they should not be compared with each other. Table 4-A compares the reections with and without the domes. The reason of the reection increase when without the dome is that the dome acts as a kind of impedance match to the connecting waveguide. When the dome is removed, the impedance matching of the waveguide system deteriorates, therefore the reection increases.
60
Figure 4.10: Transmitted and reected power at the input port without a dome (Power reection: 17.8%) Table 4-A: Comparisons of the power reections with and without domes Dome type Ceramic dome Glass dome Without dome r =9 r =4 r =1 Power reection 4.0% 11.5% 17.8%
4.1.4
Preliminary study on possible multipactor eects inside the output system
It is concluded from the previous section that the reason for the melting of the two coupling metal rods when the glass dome is replaced by a ceramic one cannot be due to the RF current, which is only of the order of 40 Amps. It seems that if the RF current in the coupling loop was the reason for the loop melting the small diameter input leads should melt rst. Other possible explanations for the observed melting could be a spurious resonance due to the ceramic window, anomalous eld distributions around the loop with higher order modes, multipactor eects and plasma discharge, possibly aected by any stray magnetic eld in the region of the loop. The present hypothesis is that the localised multipactor causes some gas evolution and a gas discharge is sustained by the RF power generated by the magnetron and this melts the surface of the rods. The computer modelling has therefore been concentrated on investigating the possibility of multipactor
61
in the output system. The Multipactor discharge is a resonant vacuum discharge frequently observed in microwave systems, such as RF windows, accelerator structures, microwave tubes and other devices. The underlying mechanism is an avalanche caused by secondary electron emission. Primary electrons accelerated by RF elds upon impact on a surface, release a large number of secondary electrons, which may in turn be accelerated by the RF elds and impact again, releasing even more electrons, and so on. The discharge can take place on a single surface or between two surfaces [14]. Multipactor is usually undesirable. It can dissipate substantial amounts of energy or detune a microwave signal. The discharge also heats the surface, possibly increasing noise level and causing damage by destroying vacuum. What concerned us most is how the multipactor may develop on the metal rods of the output system of the S band magnetron and the electro-magnetic conditions which accompany it. Figures 4.11(a) to 4.11(b) show vector plots of the electric eld excited inside the output system when a 3GHz RF signal is applied through the input ports. Figure 4.11(b) is half RF period later than Figure 4.11(a). The multipactor possibilities between dierent metal rods and surfaces can be analyzed by calculating the electron transit time in the rst instance. Because the electric eld distribution between the metal rods is not uniform, strict analytical calculation of the electron transit time in this area is not feasible. A simple calculation has been carried out to estimate the electron transit distance within half RF period. The RF voltage oscillation between the central support rod and the coupling rod is shown in Figure 4.12, with peak value of 4.7kV. The distance between the two rods is 10.98mm. Assuming the electric eld between the central support rod and the coupling rod is uniformly distributed. We obtain Emax = 4.7kV /10.98mm = 428kV /m. Within one half of the RF period, the distance the electron
62
(a)
(b)
Figure 4.11: A cross section of the electric vector eld just below the dipole antenna where the melting points are situated. (a) t=1.660ns (b) t=1.820ns
Figure 4.12: Voltage oscillation between the central support rod and one of the coupling rods. The applied signal operates at 2.998GHz with peak power of 4MW. can travel is calculated using Eq. 4.6.
1 s = v0 t + at2 2
(4.6)
where v0 is the initial velocity, t equals to half RF period and a is the acceleration caused by the electric eld.
63
In computer modelling, secondary electrons are released from the coupling rod with initial energy 0.5eV. Therefore the initial velocity can be obtained, v0 = 4.2e5m/s. With operating frequency 2.998GHz, the half RF period is t = 1/(2f ) = 0.167ns. For a crude estimation, considering only the eect of the electric eld, the acceleration speed of the electron is a = eE/m = 7.5e16m/s2 , where e is the electron charge and m its mass. The approximate distance an electron can travel within half RF period is therefore s 1mm. Due to the complexity of the electromagnetic eld distribution, computer modelling seems to be the only feasible approach to identify the areas where multipactor might take place and to discover the required electro-magnetic eld conditions. In the output system, all the surfaces of the metal rods and the waveguides are set to be able to emit secondary electrons when struck by primary electrons. A typical value for the secondary emission coecient of copper has been used for the metal rods and the waveguides, its maximum value is 1.3 with a primary energy of 500 eV. Simulations with dierent levels of RF input power and dierent emission phases of the primary electrons have been conducted. Simulations have shown that the primary electrons released from the melting regions seem to knock o a large number of secondary electrons at the point where the central support rod joins the dipole; also the secondary electrons then back bombard the melted region. Simulations have also shown that the yield of the secondary electrons is more sensitive to the phases of the RF eld than to the power level of the RF eld. The phase of 0.5 seems to be a favorite one, which generates the largest number of secondary electrons which then keep bombarding the coupling rods. This process is shown in Figure 4.13, where red dots represent primary electrons and purple dots the secondaries. The dome is ceramic, with
r
= 9.
In order to compare the eect of dierent domes, simulation with a glass dome has been carried out with the same conditions as those for the ceramic simulation. Figure 4.14 shows the electron bombardment when a glass dome is present. Comparing Figure 4.13(f) with Figure 4.14(f), it seems that the secondary electron cloud is more
64
(a)
(b)
(c)
(d)
(e)
(f)
Figure 4.13: Snap shots of electron clouds (RF power: 4MW; Phase of RF eld while primary electrons are released: 0.5; Ceramic dome: r = 9 ; Red colour represents primary electrons; Purple colour represents secondary electrons.) (a) t= 1.940ns (b) t=2.500ns (c) t=2.900ns (d) t= 3.140ns (e) t=3.300ns (f) t=4.000ns
65
(a)
(b)
(c)
(d)
(e)
(f)
Figure 4.14: Snap shots of electron clouds (RF power: 4MW; Phase of RF eld while primary electrons are released: 0.5; Glass dome: r = 4 ; Red colour represents primary electrons; Purple colour represents secondary electrons.) (a) t= 1.940ns (b) t=2.500ns (c) t=2.900ns (d) t= 3.140ns (e) t=3.300ns (f) t=4.000ns
66
condensed when the ceramic dome is used than in the case of the glass dome. If this form of electron bombardment is allowed to continue, it is quite likely to result in multipactor eect and thus the melting of the metal rods. According to the analysis above, the most likely reason for the melting is multipacting inside the structure. The Multipactor eect is a complicated physical phenomenon. The simulation conducted here is only a preliminary study.
4.2
4.2.1
Cold Test Modelling of the Anode System

Introduction
The chosen magnetron (e2v type MG5193) is a wire echelon strapped, slot and hole magnetron, as shown in gures 4.15(a) and 4.15(b). There are strap breaks on each end of the anode. Four end slots are introduced to compensate the eld distortion caused by the strap breaks, as shown in gure 4.15(b). The theory of the strapping system has been discussed in Chapter 2. In this design, the inductance of the oscillator resides mainly in the circular hole, and the capacitance mainly between the parallel walls of the slot. DC voltage is applied through the coaxial input waveguide. RF power is coupled out by two coaxial wires through the two output ports on the circular wall of the anode, which are then connected to the matched output system, which has been investigated in the previous section. For this S band magnetron, very condensed mesh has been used to simulate the hole and slot structure, and the wire stapping system, which results in a large number of cells (1,468,320 cells). With our current computer resources, the hot test modelling, which involves the interaction between the electron and the RF eld, requires a prohibitively long simulation time. So only the cold test modelling has been carried out for this particular S band tube. In the cold test modelling, the oscillation spectrum, the pattern of specic modes and the Q values of various modes can be obtained. The cold test
67
(a)
(b)
Figure 4.15: 3D geometry of the anode system of MG5193. (a). The bare anode with the wire strapping system removed. (b). The standard anode including strapping. modelling is a very useful way to test the basic operation parameters of a magnetron and thus to validate the computer model.
68
4.2.2
Cold test modelling techniques
The simulation techniques used in the cold test modelling are the eigenmode solver and the cold test response methods. The eigenmode algorithm applies an operator to a given eld pattern which grows eigenmodes in a dened frequency range [5]. In general the accuracy of the eigenmode simulations depends on the scanning windows being dened, the operator iteration times and the mesh size of the computer model. In this chapter, the eigenmode simulations are carried out with default iteration setup. It usually takes no more than one hour (depending on the geometry) for one run. So, eigenmode simulations are very quick for searching and predicting the existence of possible modes. The cold test response technique excites the system with a broad band signal and then lets the signal ring down in a system with output coupling. The mode spectrum can be obtained after a fast Fourier transform (FFT) of the response signal. The resolution of the spectrum is inversely proportional to the simulated time window of the evolving signal. For this S band magnetron, a 40ns time window requires more than 30hs CPU time, giving the spectrum resolution of 20MHz, which is accurate enough since the system operates around 3GHz and the frequency dierence between and 1 modes is more than 100MHz. There is a 3% frequency shift between the simulation results of the two techniques. Basically three reasons account for this discrepancy. One is that the geometries are dierent in these two simulations. Eigenmode simulation requires an all-sealed geometry, while the cold test response simulation requires an open geometry for output coupling. The second reason is that in eigenmode simulation the scanning window, determined by the numerical algorithm of MAGIC, is always broader than the frequency dierence between two neighboring modes. This may introduce inaccuracy. The third reason is that in eigenmode simulation, the iteration times also aect the mode accuracy. It is realised that in order to obtain the right spectrum of MG5193 anode, the geometry of the bare anode model (i.e. without strapping and end slots) should be set up rst
69
of all. With the bare anode model, eorts have been made to optimise the slot geometry. Then, complete computer model with straps has been set up. The computer model has been further improved by including the tuner geometry. The eect of the tuner is to introduce RF asymmetry and change the polarisations of all the degenerate modes, thus splitting and suppressing them. Cold test simulations have been carried out and have shown a close agreement with the experimental results.
4.2.3
Investigation of the bare anode
The function of the strapping system is to bring down and separate the oscillating modes. It is realised that in order to obtain the right spectrum for the MG5193 anode, the spectrum of the bare anode without the strapping and the end slots must be obtained rst. Accordingly eigenmode simulations have been carried out while the straps and output coupling wires have been removed and the four end slots at the end of the cavity holes excluded. Figure 4.15(a) shows the 3D structure of the bare anode.
Figure 4.16: Cross section view of the bare anode of MG5193, the slots are parallel. In the actual tube, the sides of the slots are parallel as shown in Figure 4.16; however in the computer model, due to the limitation of the polar coordinate system, they are slightly tapering, in order to avoid a staircase eect. Two dierent slot geometries have been investigated. Firstly the simulations have been carried out with the width of the slot being equal to the real value in the middle length of the slot 4.17(a). The results are presented in
70
(a)
(b)
Figure 4.17: Two dierent computer models of tapering anode slots. (a) Slot geometry 1: the width of the slot is equal to the real value in the middle length of the slot. (b) Slot geometry 2: the width of the slot is equal to the real dimension at their ends near the cathode. Table 4-B as slot geometry 1. With this geometry, the computed oscillation frequencies of both and 1 modes are lower than the measured results. Since the slots act as capacitors in the oscillation circuit, this approximation to the shape of the slots can introduce capacitance errors in the equivalent circuit, thus shift the resonating frequencies. In the second geometry being investigated the width of the slot is equal to their real dimension near the cathode, see Figure 4.17(b). The computer model of the second slot geometry consists of two tapering segments connected together, showing a small staircase between the two segments. With this modied geometry, the and 1 mode frequencies obtained by simulation moved much closer to the expected values, as shown in Table 4-B, slot geometry 2. Figures 4.18(a) and 4.18(b) show the electric eld distribution of the and 1 modes, using slot geometry 2 and obtained with the help of MAGIC eigenmode solver. Investigations of dierent slot geometries demonstrate that the width of the slots near the cathode is more important in determining the equivalent capacitance in the resonant circuit. Exciting the anode system with a broad band signal, the mode spectrum can be also obtained, see Figure 4.19. Compared with e2v measurements (see Table 4-C), the computed frequencies are close to the measured results; in addition to and 1 modes,
71
Table 4-B: Frequency comparison between simulation results and experimental results of the bare anode with dierent computer slot geometries Modes MAGIC eigenmode solver (GHz) e2v (GHz) 4.13 (Slot geometry 1) 4.192 4.19 (Slot geometry 2) 1 4.11 (Slot geometry 1) 4.141 4.15 (Slot geometry 2)
(a)
(b)
Figure 4.18: The electric eld distribution in and 1 modes using slot geometry 2
Figure 4.19: The oscillation spectrum of the bare anode with slot geometry 2. the presence of other higher order modes can also be revealed.
72
Table 4-C: Comparison of the computed and measured spectrum of oscillations in a bare anode. Modes MAGIC (GHz) e2v (GHz) 1 4.188 () 4.192 () 2 4.153 ( 1) 4.141 ( 1) 3 4.072 4.104 (possible 2 4 3.942 4.015 (possible 3) 5 3.703 3.891 (possible 4) 6 3.855 (possible 5)
4.2.4
Investigation of the standard anode
Putting back the wire strapping system and the four end slots, cold test simulations have been conducted and compared with the measurements. The geometry of the standard anode is shown in Figure 4.15(b).
Figure 4.20: A computed oscillation spectrum of the standard anode including the wire strapping system and the end slots Figure 4.20 shows the computed mode spectrum of a standard anode. The mode oscillation is obtained at 2.998GHz and 1 mode at 3.251GHz. Table 4-D compares the experimental and simulation results. The simulation results agree with the experimental results very well, except only one 1 mode frequency as obtained from computer simulation. This is because the computer model doesnt include the tuner, which is used in the real magnetron to split the degenerate modes. The eects of the tuner will be investigated in the next section.
73
Table 4-D: Oscillation spectrum comparison between simulation and experimental results (standard anode) modes MAGIC (GHz) e2v (GHz) 2.998 3.000 1 3.251 3.150/3.250
4.2.5
Investigation of the tuner
The tuner is a segment of coaxial waveguide penetrating into the anode from the radial direction as shown in Figure 4.21. The adjacent two cavity holes are then connected together. As discussed in Chapter 2, magnetron can be analysed by an equivalent circuit theory. The principle of the tuner is that it introduces variable capacitance into the anode system. The RF distortion produced by the tuner changes the polarisation of all the degenerate modes, thus splitting them. By pushing the tuner in and out, the capacitance of the circuit can be changed, resulting in the change of the resonating frequencies of the circuit. The main eect of the tuner is to split the polarisation of the 1 modes, which are the strongest competing modes with the mode. In experiments, by adjusting the tuner, the external Q factor of one of the 1 modes is usually tuned to around 300 and the other to around 1000. This arrangement can cause the over coupling of the low Q 1 mode to the output and at the same time suppress the oscillation startup of the high Q 1 mode, thus help mode oscillations to dominate; this improves the operation eciency and stability of the magnetron. In computer modelling, the frequency and polarization tuning function is realised by altering the tuner gap as shown in Figure 4.21. Table 4-E: Comparison of simulation and Modes MAGIC (GHz) 1 2.986 () 2 3.132 ( 1, lower) 3 3.266 ( 1, upper) 4 3.318 5 3.350 6 3.653 7 3.785 experimental spectra (MG5193) e2v (GHz) 3.000 () 3.160 ( 1, lower) 3.220 ( 1, upper) 3.347 ( )
74
Figure 4.21: The geometry of the tuner
Figure 4.22: Spectrum of oscillations when the tuner is included The cold test response simulation has been carried out. All the degenerate modes ( 1 and other higher order modes) and their splitting have been observed. Changing the tuner gap, it is found that the dierent dimensions of the tuner gap can not only split and tune the degenerate modes but also change the mode frequency substantially. Table 4-E compares the simulation results and the experimental results. Very good agreement with the experiments has been obtained after including the tuner inside the cavity. Figure 4.22 shows the cold test oscillation spectrum.
75
4.3
Summary
Various possible causes for the melting of the coupling rods inside the output system of the S-band magnetron have been studied by using MAGIC. The preliminary study has indicated that the melting is most likely caused by multipacting between the coupling rod and the central rod supporting the dipole. The anode system of the magnetron has been analysed and simulated using MAGIC 3D code. The computer model has been built and validated. Cold test simulation results agree very well with the experimental results. It has been demonstrated that the tuner plays an important role in the mode competition. The modelling work in this chapter shows that careful approximation has to be made in order to get adequate accuracy and practical computation times. It also shows that MAGIC is a powerful software tool, which can model the ne details of magnetron geometry and operation, such as wire strapping, hole and slot structures and tuner. Therefore MAGIC has been proved to be a reliable tool to carry out further modelling work on an X band magnetron, as discussed in Chapter 5.
References
[1] R.A.Kishek and Y.Y.Lau., Multipactor discharge on metals and dielectrics: Historical review and recent theories, Phys. Plasma, vol. 5, no. 5, p. 2120, 1998. [2] Valfells, A. Verboncoeur, J.P, and Y.Y.Lau, Space-charge eects on multipactor on a dielectric, Plasma Science, IEEE Transactions, vol. 28, no. 3, p. 529, 2000. [3] L.-K. Ang, Y.Y.Lau, R.A.Kishek, and R.M.Gilgenbach, Power deposited on a dielectric by multipactor, Plasma Science, IEEE Transactions, vol. 26, no. 3, pp. 290295, June 1998. [4] R.A.Kishek and Y.Y.Lau, Multipactor discharge on a dielectric, in Physics Review Letter, vol. 80, 1998, p. 193.
76
[5] L.Ludeking, Ed., Manual of MAGIC Tool Suite. 2002.
Mission Research Corporation,
Chapter 5
3D Modelling of an X band Magnetron

5.1 Introduction
As mentioned in chapter 4, with our current computer resources, the hot test modelling of the S band magnetron, which has a very complicated structure, proved to be unrealistic. Therefore a magnetron with a structure that more easily ts with the symmetry of the polar coordinate system has been chosen. An X band magnetron, which has eighteen vanes, sector shaped cavity and ring straps (e2v technologies type MG5241), has therefore been selected for our investigations (see Figure 5.1). Compared with the S band (MG5193) magnetron, the X band magnetron with its vanes and ring straps better tted the polar coordinate system and required fewer cells for modelling. The total number of cells used for modelling the S band magnetron (MG5193) was 1,468,320, while the equivalent number of cells for the X band magnetron (MG5241) is only 586,950, 40% of the previous gure. This reduces the simulation time for the X band magnetron dramatically. A typical hot test simulation of EM run time 40ns, which is the time required for the X band magnetron model to settle down to stable oscillations, can be completed
77
Chapter 5. 3D Modelling of an X band Magnetron
78
in 40 hours on a 2.5GHz CPU computer.
Figure 5.1: Image of an X band magnetron (e2v type MG5241) In this Chapter, a complete set of modelling techniques and their use will be described as applied to the X band magnetron. First, dierent computer models have been evaluated to choose the most suitable one for further hot test modelling. Several dierent methods have been developed to measure the external Q value of the magnetron, to achieve consistent results. Based on the validated model of the X band magnetron, hot test simulations have been carried out with dierent cathode emission schemes and various input parameters. In general, the computed performance agrees well with the experimental performance, but there is an approximate 10% discrepancy between the computed and measured values for the overall eciency. Several possible reasons for this discrepancy have been investigated, including the output coupling circuit, the end hats emission, the copper surface loss and the RRV (rate of rise of anode voltage). At this stage of the computer modelling work, cathode secondary emission has not been included. The cathode secondary emission in magnetrons is a very complicated process. How it aects the magnetron operation is not fully understood. In the MAGIC modelling, numerical instability involving cathode secondary emission is always a problem. There-
79
fore in the thesis, the direct simulating of cathode secondary emission is not carried out. Instead, through investigation of the back bombardment energy distribution and simulating the eects of the secondary emission by primary emission with an equivalent energy distribution, an important eect of the secondary emission has been revealed. This work will be presented separately in Chapter 7.
5.2
Optimization of the Computer Model
The X band magnetron has a long established production history enabling a ready comparison of computed and measured results, so that the model can be meaningfully benchmarked. Table 5-A presents the dimensions of the tube and its typical performance. Ideally the magnetic eld should be uniformly applied along the axial length of the magnetron. But in reality, due to the eects of the pole pieces, which are located outside the two ends of the cathode, the magnetic ux is non-uniform. In this section, the main objective is to optimise the computer model; for simplicity, a uniform rather than real magnetic eld distribution is used in the computer modelling. The eects of the real magnetic eld distribution on magnetron performance will be investigated separately in section 5.3. In Chapter 6, further investigations will be carried out on the important eects of the real magnetic eld distribution on the cathode emission and back bombardment. The vanes in this X-band magnetron are parallel in shape. However, this presents a problem since computer modelling is carried out in terms of polar coordinates, so strict adherence to the physical geometry would require a ne mesh structure leading to long computed runs. One important aspect was to t the model to the symmetry imposed by the code, in order to maximise the computing eciency. Three dierent vane models have been investigated: parallel vanes (in practice step-like parallel due to the nite size of the cells used in computing), thin tapering vanes, the model and the real vanes having the same thickness half way down their length and thick tapering vanes, the model and
80
real vanes having the same thickness at the end facing the cathode. The thin tapering vane model had been given up because the very thin vane tips can alter the electric eld near the vane tips and thus destroy accurate representation of the magnetrons interaction process. Table 5-A: e2v type MG5241 magnetron: dimensions and typical performance Cathode radius (mm) Anode radius (mm) Frequency bore radius (mm) Width of vanes (mm) Number of vanes Pi mode frequency (GHz) Qext of mode Unknown mode (GHz) -1 lower (GHz) Qext of -1 lower -1 upper (GHz) Qext of -1 upper Maximum magnetic eld (T) Anode voltage (kV) Anode current (A) Output power (kW) Overall eciency 1.90 2.82 8.20 0.40 18 9.4100.030 236 10.406 10.137 220 11.063 323 0.418 6.10 5.0 12.2 40.0%
5.2.1
The parallel vane model
The so called parallel vane model is actually step like parallel due to the polar coordinate system and the nite size of the mesh. Both ne mesh (360 cells in azimuthal direction) and coarse mesh (180 cells in azimuthal direction) have been tried on order to examine the eect of the staircase. Figure 5.2.1 shows the cross section view of the parallel vane model in both ne mesh and coarse mesh. A broadband frequency response method has been used to obtain the system oscillation spectrum. With this technique, a broad-band signal, shown in Figure 5.3(a), is applied using a probe located near the output port. Several oscillation frequencies have been excited inside the magnetron. The oscillation spectrums for the two models are
81
(a)
(b)
Figure 5.2: Cross section view of the parallel sided vane models with (a) Coarse mesh (180 cells in azimuthal direction) (b) Fine mesh (360 cells in azimuthal direction) shown in Figure 5.3(b). In both cases four main oscillation peaks could be observed (see Table 5-A): one mode frequency, two 1 frequencies and one unknown mode frequency. The simulation results show that there are frequency peaks of higher order modes around 12GHz and 13GHz, but those frequencies have a minor eect on the magnetron operation since they are far away from the mode. Dierent mesh densities cause a discrepancy in the spectra. In the ne mesh model the mode is excited at 9.294GHz; in the coarse mesh model the mode is excited at 9.360GHz, both are lower than the measured 9.410GHz. This is because in the computer model the space between
82
two adjacent vanes, which are step like parallel, is smaller than that of the real tube, correspondingly the anode capacitance is higher, thus the lower mode frequency. In the ne mesh model, this space is even smaller than that of the coarse mesh model, which causes its even lower mode frequency than that of the coarse mesh model.
10
0
Volt/GHz
10
10
10
15
Frequency (GHz)
(a)
(b)
Figure 5.3: (a) Spectrum of the broad band signal. (b) Spectrum of the responding signal. Blue curve represents ne mesh model (oscillation peaks at: 9.294GHz, 9.776GHz, 10.325GHz and 10.889GHz); Red curve represents coarse mesh model (oscillation peaks at: 9.360GHz, 9.840GHz, 10.380GHz and 10.920GHz). Hot test simulations have also been carried out with the two models respectively and
83
mode oscillations around 9.28GHz have been observed. It is reasonable for the hot test mode frequencies to be dierent from their cold test values, since in hot test the space charge inductance also aects the oscillation frequency. Table 5-B presents the results of the two hot test simulations and their operation conditions. Comparing with the experimental results, the simulated values of the anode current are low and the overall eciencies are quite high, 58.2% and 54.1% for the ne mesh and coarse mesh models respectively, as against the experimental value of 45%. For the sake of simplicity copper surface losses have not been included in the simulation processes. Instead, the overall eciency is calculated by including overall circuit eciency of 85%, the gure being derived from measurements at e2v. Calculation of circuit eciency cold is described in section 5.7 in terms of Q0 , Qext and Qload . The hot test Qext are 145 for the ne mesh and 206 for the coarse mesh, both lower than the experimental 236. Table 5-B: Hot test simulation results of the parallel sided vane model Cathode emission model: space-charge-limited Frequency bore radius: 8.2mm Eective anode voltage: 5.75kV Magnetic eld: 0.405T Circuit eciency: 85% Azimuthal mesh density (degree/cell) 360 mode frequency (GHz) 9.287 Anode current (Amps) 3.3 System RF energy (Joule) 32.2 hot test Qext = 2f0 PW 145 o ut Output power (kW) 13.0 Overall eciency 58.2%
180 9.284 4.1 53.2 206 15.0 54.1%
Generally speaking the ne mesh and coarse mesh models give consistent results. The overall eciencies are 10% higher than the experimental results. Considering the Qext and the overall eciency, the coarse mesh model is more close to the experimental results. This indicates that improving the mesh density of the computer model does not necessarily cause the calculated performance getting closer to the measured performance. The ne mesh structure can reduce the errors caused by staircase, but it may also cause more round-o errors. The azimuthal mesh density of the coarse mesh model (2degrees/cell)
84
is therefore quite sucient. The computed mode frequencies of both models are about 1% lower than the measured mode frequency 9.410GHz. Since in computer modelling, the mode frequency can be tuned easily by adjusting the frequency bore radius (back of the cavity) without aecting the magnetron performance signicantly, therefore in this section, we do not bother to correct the frequency discrepancy, but concentrate on the test of the dierent vane models. A ne frequency tuning will be carried out in later work.
5.2.2
The thick tapering vane model
The geometry of the thick tapering vane model, with its vane tips having the same thickness as the real model, is shown in Figure 5.4(a). Also in Figure 5.4(a) together with Figure 5.4(b), the transverse and axial cross-sections of the model are presented, showing the cathode and the anode together with the relevant parts of the input and output circuit. Because of the thick vane model, the space between two adjacent vanes is actually smaller than the real model, which will change the oscillation frequency. An increase in the diameter of the frequency bore had been made to compensate for this area loss between adjacent vanes thus tune the resonant frequency close to the measured value. When tuning the radius of the frequency bore to 8.68mm, mode oscillations have been observed at 9.472GHz; 1 modes at 9.908GHz and 10.926GHz; unknown mode frequency at 10.350GHz. Figure 5.5 shows the oscillation spectrum obtained from the broad band frequency response. These results are very close to the experimental oscillation frequencies presented in Table 5-A, which are mode at 9.410 0.030GHz, 1 modes at 10.137GHz and 11.063GHz, unknown mode at 10.406GHz. Hot test simulation with this model has been carried out and a very ne mode oscillation has been obtained. In Figure 5.4(a) and 5.4(b), the snapshots of electron cloud of mode oscillation are also shown. The important role of the cathode end hats in conning the electron cloud can now be seen quite clearly. Here the end hats are the two metal edges on the two ends of the cathode, see Figure 5.4(b). The operating
85
(a)
(b)
Figure 5.4: The geometry of the thick tapering vane model and the electron cloud of a mode (a)Transverse cross section (b)Axial cross section conditions of the magnetron and the computed performance are shown in Table 5-C. Comparing with the experimental results, this thick tapering vane model still yields a very high overall eciency of 55.6%. In table 5-C, the Qext of the mode has been calculated in two ways, one is from the cold test energy decay rate and the other is from the hot test simulation using the denition formula of external Q [1]. The two methods have given consistent results, which are close to the experimental value. In computer modelling, several methods have been developed to measure the Qext , which will be discussed in detail in the next section.
86
Figure 5.5: Oscillation spectrum of the thick tapering vane model by broad band frequency response method (oscillation peaks at: 9.472GHz, 9.908GHz, 10.350GHz and 10.926GHz) Table 5-C: Hot test simulation of the thick tapering vane model Cathode emission model Frequency bore radius (mm) Eective anode voltage (kV) Magnetic eld (T) Circuit eciency Azimuthal mesh density (degree/cell) Cold test Qext Oscillation frequency (GHz) Vane voltage (kV) Anode current (A) hot test Qext = 2f0 PW out Output power (kW) overall eciency space-charge-limited 8.68 5.75 0.405 85% 180 260 9.461 5.2 3.8 244 14.3 55.6%
The transient state of the magnetron before it starts oscillating in the mode (thick vane model) is shown in Figures 5.6 and 5.7. Figure 5.6(a) shows the anode voltage evolution. It can be seen that once the oscillations start, the eective anode voltage will drop a bit because of the change in magnetron impedance. Therefore in the diagram, the eective anode voltage is 5.75kV when the system is in stable mode oscillation, which is a little bit lower than the applied voltage of 5.8kV. Figure 5.6(b) shows the anode current evolution, which starts building up after 6ns. Figure 5.7(a) shows the RF voltage between
87
two adjacent anode vanes. After 6ns, oscillations settle down smoothly into the mode. Figure 5.7(b) shows the FFT of the RF voltage between the adjacent vanes, where the spectrum is a very clean mode. The evolution of the output power (Figure 5.6(c)) and the total system energy stored in the magnetron (Figure 5.7(c)) present similar behavior, consistent with the evolution process of the vane voltage oscillations and the anode current. In the above investigations, the space-charge-limited emission model has been applied to the cathode emission for both the parallel vane model and the thick tapering vane model. The magnetic eld being used is a uniformly distributed eld of 0.405T along the axis of the magnetron. At the beginning of our modelling work, the MAGIC code could not allow for the real magnetic eld distribution. After Aug. 2003, MAGIC has been updated to include the real magnetic eld distribution, which makes the computer modelling much closer to real life magnetron operation. In this section, since the objective of preliminary modelling is to optimise the computer model, only one set of results is presented for each vane model. Therefore they do not represent a complete range of performance of the magnetron, but it is hoped enough information has been provided for our initial objective. Concerning the above investigations, the parallel vane model and the thick tapering vane model produced consistent results, but they both exhibit higher overall eciency. This discrepancy in eciency will be investigated in detail in the following sections. The thick tapering vane model has been chosen for further modelling because it best represents the conditions near the vane tip, where the important interaction process takes place. The slight change in the cavity outer radius of this model has little eect on the whole performance. It also avoids the stair-case on the vane structure. Its oscillation spectrum and external Q value of the mode agree very well with experimental results.
88
(a)
(b)
(c)
Figure 5.6: Magnetron transient evolution. (a) The anode DC voltage. (b) The anode current. (c) The output power.
89
(a)
(b)
(c)
Figure 5.7: Magnetron transient evolution. (d) The voltage oscillation between adjacent anode vanes. (e) The FFT of the vane voltage oscillation. (f) The total system stored energy evolution. Before 14ns, the energy is only DC energy, which is 54.12e-6J; after 14ns, the total energy, including the DC energy and RF energy, is 106.06e-6J.
90
5.2.3
Measurement of the external Q value
Two dierent methods have been developed to measure the magnetron Qext in computer modelling. First it is calculated from the systems energy decay. With this method, an eigenmode simulation needs to be completed rst to obtain the mode eld distribution inside the cavity. By presetting this eld distribution inside the cavity and letting the eld ring down, the Qext can be obtained from the systems energy decay rate. Usually when excited in the mode, the system energy will drop quickly below the half value of the original energy within 10ns. A cold test simulation of 10ns in EM time can be nished in 2 or 3 hours with a 2.5GHz CPU computer. Therefore this method is very quick. Figures 5.8(a) and 5.8(b) show the Ephi distribution of the mode in the cross and axial sections of the anode, obtained from the eigenmode solver of MAGIC. Figure 5.9(a) illustrates the system energy evolution while excited in the mode. Here, for the convenience of calculations, the total stored energy is normalized to 1 Joule. Figure 5.9(b) illustrates an instant external Q value of 260.
(a)
(b)
Figure 5.8: Ephi distribution of mode in (a) The transverse section of the anode. (b) The axial cross section of the anode The second method is to use the results obtained from the hot test simulation. The denition of the Qext is 2f0 W/Pout [1]. Where f0 is the mode frequency, W is the system stored RF energy, Pout is the output power. All these parameters can be obtained from the hot test simulation. In computer modelling, the total system stored energy can be calculated. Before oscillations start, the systems stored energy is mainly the DC
91
(a)
(b)
Figure 5.9: Measure the external Q value from system energy decay rate. (a) The system energy evolution while excited in mode distribution. (b) The instant external Q of 260. energy; when the system starts oscillating in the mode, the stored energy is the sum of both DC and RF energy, as shown in Figure 5.7(c). It can be obtained from this gure that before the oscillations start the DC energy is 54.0e 6J; while the system is in stable oscillations, the total energy is 112.6e 6J. Therefore, the systems stored RF energy while in stable oscillations can be obtained by subtracting the DC energy from the total energy, which is WRF = 58.6e 6J. The nal output power is 14.3kW and the oscillation frequency of the mode is 9.461GHz. The Qext can then be calculated; it is 244 and very close to the cold test measurement. Therefore since the two methods of Q measurement give very close results, the thick tapering vane model has been validated again as the optimum model for further hot test investigations of the X band magnetron.
92
5.3
Real Magnetic Field Distribution
At the early stage of our modelling work, all calculations were carried out assuming a uniform magnetic eld. Later in August 2003, MAGIC was updated to allow for a 3D magnetic eld distribution. This enabled the comparison between a uniform and a real magnetic eld. Before the discussion of the simulation results, explanation will be given of the dierence between the uniform magnetic eld and the real magnetic eld as obtained from e2v measurements. Figures 5.10 and 5.11 show the two components Br and Bz of the actual static magnetic eld as they vary in the radial and axial directions. The coordinate origin is located on the axis and in the center of the anode block. The magnetic eld distribution is symmetric in azimuthal direction, that is, no B component is present. Also the magnetic eld distribution is symmetric in the axial or z direction. Thus the magnetic eld distribution is 2D in nature. Figures 5.10 and 5.11 represent the magnetic eld distribution in one half of the longitudinal cross section of the anode. A spike occurs in both Br and Bz near the polepieces, which are just outside the interaction space and near the two end hats, see Figure 5.4(b). The anode structure has been described at the beginning of the cold test modelling. Within the interaction space, which is between the anode vanes and the cathode, Bz varies from 0.405T to 0.418T; the Br component is nearly zero in the centeral region of the cathode, but it has nonnegligible values near the end hats. Figure 5.12 illustrates the Bz variation between the cathode and anode in the centeral cross section; the maximum value of Bz in the z=0 plane is given by Bmax = 0.418T . In the following work, the use of the real magnetic eld distribution will be indicated by the value of Bmax .
93
Figure 5.10: The radial magnetic eld component distribution used in MG5241P (e2v measurement)
Figure 5.11: The axial magnetic eld component distribution used in MG5241P (e2v measurement)
94
Bz between cathode and anode in the centre plane of the magnetron

4200
4150 cathode surface 4100
4050
Bz (Gauss)
4000 anode surface 3950
3900
3850
3800
3750
0.5
1.5
2.5
3.5
4.5
r (mm)
Figure 5.12: radial distribution of the axial component Bz of the real magnetic eld, Bmax =0.418T.
5.4
Hot Test Modelling with Dierent Cathode Emission Models
5.4.1
Introduction
With the computer model being optimised, more hot test modelling will be described in this section in order to investigate the detail of the operation of the X band magnetron. The hot test modelling is based on two cathode emission models, the space-charge-limited emission model and the beam emission model, which have been described in Chapter 3. The eects of the real magnetic eld distribution has been investigated using both emission models. The anode copper surface losses have been simulated in order to be able to study their eect on the start-up of oscillations. The output coupling loop has been investigated in order to obtain a complete set of results, which reveal the eect of the output circuit on the magnetron performance. So far, the cathode secondary emission has not been included, but its eect will be considered in Chapter 7.
95
5.4.2
5.4.2.1
The beam emission model

Uniform magnetic eld 0.405T
The beam emission has rst been modeled with a uniform static magnetic eld of 0.405T. In the real magnetron, the cathode emission current density is about 15A/cm2 . With this value simulations have been carried out with the anode voltages varying from 5.17kV to 5.89kV. Table 5-D presents the simulation results. In Figures 5.13(a) to 5.13(c), the computed V-I curves, the output power against average anode current and the overall eciency against the anode voltage are compared with e2v measured values, for three dierent magnetic eld distributions. Here an explanation must be given of the type of units used in the performance diagrams. In the hot test modelling, the common EM run time is 40ns. The computed anode current and output power are peak values. The X band magnetron is pulsed, with a duty cycle of 1/1000. Thus, the average anode current in mA is equivalent to the peak anode current in amps; the average output power in W is equivalent to the peak output power in kW. With the anode voltage varied from 5.17kV to 5.89kV, anode current increased from 1.26A to 3.98A. Compared with the set of experimental results obtained with Bmax = 0.398T , the computed currents are higher at lower anode voltages 5.17kV and 5.30kV; close to the measured values at anode voltages 5.53kV and 5.71kV; and much lower at a high anode voltage of 5.89kV. The computed point 5.89kV, 3.98A is a rogue result, which deviates from the performance line. This indicates that the beam emission model cannot provide enough current for operation at high anode voltage and high anode current levels. The calculation didnt include the anode copper surface loss. According to the e2v measurements, the circuit eciency is about 85%. Therefore in Table 5-D and the diagram 5.13(b), the output powers are modied by this amount. The modied output powers are still much higher than the experimental results. The computed overall
96
Table 5-D: Investigation of beam emission model with uniform magnetic eld 0.405T Magnetic eld: uniform 0.4050T Qext from cold test: 244 Cathode emission model: beam voltage=5eV; beam current density=15A/cm2 Cathode emission current (A): 6.26 Eective anode voltage (kV) 5.17 5.30 5.53 5.71 mode oscillation start time (ns) 25 Va step up 7 Va step up mode frequency (GHz) 9.454 9.455 9.451 9.448 Back bombardment current (A) 5.05 4.50 3.65 2.51 Anode current (A) 1.26 1.80 2.64 3.70 Total system energy (J) 59.9e-6 69.0e-6 85.5e-6 98.7e-6 DC energy (J) 43.6e-6 45.8e-6 50.5e-6 53.8e-6 System RF energy (J) 16.3e-6 23.2e-6 35.1e-6 44.9e-6 Hot test Qe xt = 2f0 PW 218 220 222 225 o ut Output power (kW) 4.43 6.26 9.40 11.83 Output power modied by circuit eciency 85% (kW) 3.77 5.32 7.99 10.06 Power dissipation on cathode (kW) 0.345 0.429 0.581 0.864 Power dissipation on anode (kW) 1.76 2.74 4.75 8.80 Overall eciency (%) 57.9 55.8 54.7 47.6 Vane voltage (kV) (amplitude) 2.93 3.38 4.25 4.80 Output voltage (kV) (amplitude) 0.65 0.75 0.95 1.07 Output current (A) (amplitude) 13.06 15.06 18.97 21.39
5.89 Va step up 9.452 2.25 3.98 106.0e-6 57.3e-6 48.7e-6 227 12.74 10.83 0.930 10.14 46.2 5.00 1.11 22.29
eciency is shown in Figure 5.13(c), which increases as the anode voltage decreases. The legends of Figure 5.13(b) and Figure 5.13(c) are same as that of Figure 5.13(a). The mode oscillation frequency does not change much with dierent anode voltages. The hot test Qext s have been calculated with the knowledge of the systems stored RF energy and the output power, which values are slightly lower than the cold test measured Qext , but are consistent with dierent anode voltages. The rate of rise of anode voltage is another important parameter in magnetron operation. In reality, the RRV (rate of rising voltage) is 100kV /s, which is not used in the computer simulation because of the unbearably long simulation time it leads to. In
Figure 5.13: Comparison of the performance of beam emission model with the experimental performance of MG5241P. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Anode voltage vs anode current. (b) Output power modied by circuit eciency of 85% vs anode current. (c) Overall eciency vs anode voltage.
! ! " " # $ # % & $

(b)

Ia
"D 1)( C B! @9) (873 25 () 4 1)( ! 20 6 5 A# 74 3 (26 14) 3 20 ' "D C B! @9) (873 25 () 4 1)( ! 5 A# 74 3 (26 14) 3 20 ' "#C B! @9) (873 25 () 4 1)( & 5 A# 74 3 (26 14) 3 20 ' " % &$ "# $ " !
Ia
(a)
(c)

Va (kV)
97
98
modelling the anode voltage is driven to its full magnitude in 2ns. An investigation of the magnetron operation with slow rate of rising voltage of 100kV /s will be carried out in section 5.5. It can be seen from Table 5-D that at certain anode voltages, magnetron can quickly reach mode oscillations. For example when Va = 5.53kV , the magnetron develops mode oscillations within 7ns. At other anode voltages, especially at higher levels, due to a rapid increase of the anode voltage, mode competition becomes very strong and the magnetron may never develop mode oscillations. Thus in computer modelling, in order to obtain information about the operation of the magnetron at a given anode voltage, when it is known that is dicult in reaching a mode oscillation, the simulation will be usually started at a voltage that is known to be suitable for mode oscillations; while the system is stabilised in the mode, the voltage can then be stepped up or down to reach the desired value.
5.4.2.2
Investigation with the real magnetic eld distribution
The real magnetic eld distribution is introduced into the computer modelling at this stage. Table 5-E presents the simulation results with real magnetic eld distribution of Bmax = 0.418T . With this magnetic eld prole, the Bz component at the cathode surface in the center plane of the magnetron is close to 0.405T, making it convenient to compare with the simulation results of the uniform magnetic eld distribution. The cathode emission density is kept unchanged. The anode voltage is varied from 5.56kV to 6.15kV. For the convenience of comparison, all the computed performance of the beam emission model with uniform and real magnetic elds are plotted together in Figure 5.13. At given anode voltages, the computed anode currents with Bmax = 0.418T are lower than the measured values with the real magnetic eld distributions Bmax = 0.425T and Bmax = 0.413T ; the computed output power, even when modied by the circuit eciency, is much higher than the value measured at the same anode current. Here
99
Table 5-E: Investigation of beam emission model with real magnetic eld, Bmax = 0.418T Magnetic eld: real magnetic eld distribution Bmax = 0.418T Qext from cold test: 244 Cathode emission model: beam voltage=5eV; beam current density=15A/cm2 Cathode emission current (A): 6.26 Eective anode voltage (kV) 5.56 5.75 5.93 6.15 mode oscillation start time (ns) 16 7 30 Va step up mode frequency (GHz) 9.456 9.452 9.452 9.453 Back bombardment current (A) 5.07 4.19 2.83 2.32 Anode current (A) 1.20 2.08 3.42 3.90 Total system energy (J) 66.5e-6 80.7e-6 101.8e-6 113.3e-6 DC energy (J) 50.7e-6 54.2e-6 58.3e-6 62.7e-6 System RF energy (J) 15.8e-6 26.5e-6 43.5e-6 50.6e-6 Hot test Qe xt = 2f0 PW 216 221 223 225 o ut Output power (kW) 4.36 7.12 11.58 13.35 Output power modied by circuit eciency 85% 3.71 6.05 9.84 11.35 Power dissipation on cathode (kW) 0.381 0.491 0.685 0.821 Power dissipation on anode (kW) 1.90 4.34 8.18 10.20 Overall eciency 55.6% 50.6% 48.5% 47.3% Vane voltage (kV) (amplitude) 2.94 3.73 4.73 5.05 Output voltage (kV) (amplitude) 0.66 0.83 1.05 1.13 Output current (A) (amplitude) 13.10 16.63 21.08 22.05 again the copper surface losses are not included in the simulation, instead we use the measured circuit eciency of 85% to modify the simulated output power and eciency. Just as in the uniform magnetic eld model, the anode current is much lower than the experimental value at high anode voltage of 6.15kV. Though in general the overall eciency is about 10% higher than the experimental value, it decreases when the anode voltage increases. On this point, it is consistent with the experimental performance. Modelling has also been carried out with the real magnetic eld distribution, but scaled to Bmax = 0.405T in order to test the magnetron performance at a lower magnetic eld level. The anode voltage is varied from 5.54kV to 6.09kV. The simulation results
100
Table 5-F: Investigation of beam emission model with real magnetic eld, Bmax = 0.405T Magnetic eld: real magnetic eld distribution Bmax = 0.405T Qext from cold test: 244 Cathode emission model: beam voltage=5eV; beam current density=15A/cm2 Cathode emission current (A): 6.26 Eective anode voltage (kV) 5.54 5.71 5.90 6.09 mode oscillation start time (ns) 7 Va step up Va step up Va step up mode frequency (GHz) 9.453 9.453 9.452 9.456 Back bombardment current (A) 4.06 2.72 2.30 2.19 Anode current (A) 2.20 3.51 3.91 4.05 Total system energy (J) 77.2e-6 96.0e-6 105.1e-6 111.7e-6 DC energy (J) 50.5e-6 53.6e-6 57.3e-6 61.0e-6 System RF energy (J) 26.7e-6 42.4e-6 47.8e-6 50.7e-6 Hot test Qe xt = 2f0 PW 219 226 226 228 o ut Output power (kW) 7.26 11.15 12.55 13.20 Output power modied by circuit eciency 85% 6.17 9.48 10.67 11.22 Power dissipation on cathode (kW) 0.467 0.681 0.801 0.884 Power dissipation on anode (kW) 4.61 8.40 10.21 11.00 Overall eciency 50.6% 47.3% 46.3% 45.5% Vane voltage (kV) (amplitude) 3.76 4.65 4.95 5.07 Output voltage (kV) (amplitude) 0.84 1.04 1.10 1.13 Output current (A) (amplitude) 16.76 20.76 22.09 22.59 are consistent with those of the real magnetic eld Bmax = 0.418T , nut only at higher anode current levels, see Table 5-F. Anode currents are lower and the overall eciencies are higher than the experimental results. At high anode voltage levels of 5.90kV and 6.09kV, the anode currents are much lower than the measured values. It can be seen from Figure 5.13 that the computed performance with real magnetic eld Bmax = 0.405T is close to that of the uniform magnetic eld 0.405T when the anode voltage is higher than 5.7kV. At lower anode voltages, the computed performance at a given anode voltage with the uniform magnetic eld tends to yield more anode current than that with the real magnetic eld. The important eect of the real magnetic eld is not limited to this phenomenon, in Chapter 6, discussion of further eects due to the real magnetic eld
101
will be presented. In the above simulation power dissipation due to electron bombarding of the cathode and anode has also been obtained, see Table 5-D, 5-E and 5-F. As mentioned in Chapter 1, this is a great advantage of computer modelling since power dissipation inside a magnetron is dicult to measure experimentally. In all these simulations, energy balance is fully satised. Besides the power dissipation on both cathode and anode, the power dissipation on the other parts inside the magnetron such as the pole pieces, has also been investigated and is found to be negligible and therefore omitted from the tables. This also conrms the conning eect of the cathode end hats, which prevent the electrons from owing sideways and beyond the cathode (see Figure 5.4(b)). It can be summarised from the above that the current computer model cannot provide enough anode current, especially at high anode voltages, so that the overall eciency of the model is about 10% higher than in a real magnetron. Therefore there must be some factors missed out in our magnetron model that may induce more anode current and at the same time reduce the eciency. These factors may be the spurious current from the end hats, the copper surface losses, the cathode secondary emission, or some other features that have been neglected. In the following modelling work, all these factors will be investigated one by one.
5.4.2.3
Investigation of dierent emission current densities
In all the above modelling work, the cathode emission current density was assumed to be 15A/cm2 . In order to investigate the eect of dierent cathode emission current densities, modelling has been carried out with the cathode emission current density of 8A/cm2 , 10A/cm2 and 18A/cm2 respectively. The magnetic eld was set at real Bmax = 0.418T and the anode voltage at 5.75kV or 5.74kV (roughly the same). For the convenience of comparison, the simulation results with cathode emission current density 15A/cm2 and magnetic eld Bmax = 0.418T are presented again in Table 5-G. It can
102
be seen from the table that at given anode voltage when reducing the cathode emission current density from 18A/cm2 to 8A/cm2 , the overall eciency is slightly reduced and at the same time the anode current is also reduced. At a high emission density 18A/cm2 , the anode current is close to the measured value, but still the high eciency problem persists. Table 5-G: Investigation of beam emission model with varied cathode emission density Magnetic eld: real magnetic eld distribution Bmax = 0.418T Qext from cold test: 244 Cathode emission beam voltage: 5eV Cathode beam current density (A/cm2 ) 8 10 15 Cathode emission current (A) 3.35 4.19 6.26 Eective anode voltage (kV) 5.75 5.75 5.75 mode oscillation start time (ns) 9 9 7 mode frequency (GHz) 9.451 9.451 9.452 Back bombardment current (A) 1.45 2.24 4.19 Anode current (A) 1.89 1.94 2.08 Total system energy (J) 76.8e-6 78.5e-6 80.7e-6 DC energy (J) 53.6e-6 54.2e-6 54.2e-6 System RF energy (J) 23.2e-6 24.3e-6 26.5e-6 W Hot test Qe xt = 2f0 Po ut 225 220 221 Output power (kW) 6.12 6.56 7.12 Output power modied by circuit eciency 85% 5.20 5.58 6.05 Power dissipation on cathode (kW) 0.350 0.382 0.491 Power dissipation on anode (kW) 4.54 4.40 4.34 Overall eciency 47.8% 50.0% 50.6% Vane voltage (kV) (amplitude) 3.46 3.57 3.73 Output voltage (kV) (amplitude) 0.77 0.80 0.83 Output current (A) (amplitude) 15.42 15.94 16.63
18 7.54 5.74 8 9.451 5.25 2.30 83.7 54.2e-6 29.5 222 7.89 6.71 0.596 4.78 50.8% 3.92 0.88 17.50
5.4.3
The space-charge-limited emission model
The space-charge-limited emission model has been investigated with uniform and real magnetic eld respectively. Dierent from the beam emission model, which has the cath-
103
ode emitting at a xed current density, the space-charge-limited model has the cathode emission limited by the amount of the space charge facing it. The space charge eld combined with the DC eld keep the normal electric eld at the cathode surface equals to zero. In the space-charge-limited emission model, the magnetic eld distribution is another factor aecting the cathode emission. This issue will be discussed in detail in Chapter 6. Simulation has been carried out with uniform magnetic eld 0.405T and real magnetic elds Bmax = 0.418T and Bmax = 0.405T . For all magnetic eld distributions, modelling has been carried out with varied anode voltages. The results are summarised in Table 5H, Table 5-I and Table 5-J. Figure 5.14 compares the VI curves, the output power against the anode current and the overall eciency against the anode voltage for the computed and e2v measured results. Figures 5.14(b) and 5.14(c) use the same legends as Figure 5.14(a). The calculated performance with a beam emission model are again presented in these diagrams for comparison. The calculated performance with uniform magnetic eld 0.405T coincides with the calculated performance with real magnetic eld Bmax = 0.418T , as shown in Figure 5.14(a). At given anode currents, their output powers and overall eciencies are very close. Here, the output power was modied by including the circuit eciency of 85% to account for the anode copper surface losses. At given anode voltages, the calculated anode currents of the uniform magnetic eld 0.405T are about 0.2Amps higher than for the real magnetic eld Bmax = 0.418T . Comparing the simulation results of the two non-uniform magnetic eld distributions with the experimental results, one can see that at a given anode voltage the computer model tends to provide anode current about 50% higher than that of the experimental value. The simulated output power is also much higher than the experimental values at a given anode current, resulting in the calculated overall eciency more than 10% higher than the corresponding experimental value.
104
Ia
(a)
Ia
(b)
Figure 5.14: Comparison of the performance of space-charge-limited emission model with the experimental performance of MG5241P. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Anode voltage vs anode current. (b) Output power modied by circuit eciency of 85% vs anode current. (c) Overall eciency vs anode voltage.
! ! " " # # $ % & $
(c)

Va (kV)
Bmax= .425T Bmax= .413T Bmax= .398T Space charge limited emission; Bmax=.418T Space charge limited emission; Bmax=.405T Uniform emission current density: 15A/cm2; Bmax=.418T Uniform emission current density: 15A/cm2; Bmax=.405T Space charge limited emission; Uniform B=.405T

105
Table 5-H: Investigation of space-charge-limited emission model with uniform magnetic eld B=0.405T Magnetic eld: uniform B=0.405T Qext from cold test: 244 Cathode emission model: space-charge-limited Cathode emission current (A) 8.59 9.74 Eective anode voltage (kV) 5.61 5.70 mode oscillation start time (ns) 16 12 mode frequency (GHz) 9.456 9.455 Back bombardment current (A) 5.36 5.86 Anode current (A) 3.26 3.86 Total system energy (J) 99.3e-6 110.1e-6 DC energy (J) 52.2e-6 54.0e-6 System RF energy (J) 47.1e-6 56.0e-6 Hot test Qe xt = 2f0 PW 224 224 o ut Output power (kW) 12.51 14.87 Output power modied by circuit eciency 85% 10.63 12.64 Power dissipation on cathode (kW) 0.550 0.640 Power dissipation on anode (kW) 5.41 7.17 Overall eciency 58.1% 57.4% Vane voltage (kV) (amplitude) 4.90 5.36 Output voltage (kV) (amplitude) 1.09 1.13 Output current (A) (amplitude) 21.88 25.80
11.62 5.87 Va step up 9.453 6.21 5.40 134.3e-6 57.3e-6 77.0e-6 226 20.16 17.14 0.907 11.30 54.1% 6.26 1.40 27.92
Another important dierence between the calculation and measurement is that when varying the magnitude of the magnetic eld, the experimental eciency wont change much, but the calculated eciency varies signicantly, see Figure 5.14(c). When the anode voltage is between 5.4kV and 6.0kV, the experimental eciency varies about 2 to 3%; but the calculated eciency drops by about 5% when the magnetic eld is reduced from Bmax = 0.418T to Bmax = 0.405T . Comparing the two emission models, the space-charge-limited emission model can provide more than enough anode currents at given anode voltages, while the beam model cannot provide sucient anode current. The space-charge-limited model also operates
106
Table 5-I: Investigation of space-charge-limited emission model with real magnetic eld, Bmax = 0.418T Magnetic eld: real magnetic eld distribution Bmax = 0.418T Qext from cold test: 244 Cathode emission model: space-charge-limited Cathode emission current (A) 7.03 8.90 Eective anode voltage (kV) 5.45 5.61 mode oscillation start time (ns) Va step down Va step down mode frequency (GHz) 9.457 9.454 Back bombardment current (A) 4.95 5.74 Anode current (A) 1.96 3.00 Total system energy (J) 76.9e-6 95.8e-6 DC energy (J) 49.5e-6 52.4e-6 System RF energy (J) 27.4e-6 43.4e-6 W Hot test Qe xt = 2f0 Po ut 217 223 Output power (kW) 7.53 11.56 Output power modied by circuit eciency 85% 6.40 9.83 Power dissipation on cathode (kW) 0.457 0.665 Power dissipation on anode (kW) 2.80 4.91 Overall eciency 59.9% 58.4% Vane voltage (kV) (amplitude) 3.84 4.73 Output voltage (kV) (amplitude) 0.86 1.06 Output current (A) (amplitude) 17.15 21.13
9.97 5.70 14 9.453 6.19 3.60 106.1e-6 54.1e-6 52.0e-6 224 13.78 11.71 0.792 6.38 57.1% 5.19 1.16 23.18
12.53 5.87 Va step up 9.453 7.20 5.20 131.2e-6 57.4e-6 73.8e-6 226 19.36 16.46 1.180 10.76 53.9% 6.12 1.37 27.32
at a higher eciency than the beam model. For the beam model, when increasing the cathode emission density, both the anode current and the overall eciency tend to increase. This is reasonable because if the cathode emission density keeps increasing, the beam emission model will eventually approach the space-charge-limited model. From the above analysis, it can be concluded that rstly, the current computer model with both beam emission and space-charge-limited emission can represent the magnetron performance to some degree, the simulated performance curves of these two emission models agreeing with the experimental performance over a range of values; secondly, the space-charge-limited emission model should be used for further investigation, since the
107
Table 5-J: Investigation of space-charge-limited emission model with real magnetic eld, Bmax = 0.405T Magnetic eld: real magnetic eld distribution Bmax = 0.405T Qext from cold test: 244 Cathode emission model: space-charge-limited Cathode emission current (A) 6.72 8.60 9.79 Eective anode voltage (kV) 5.25 5.40 5.49 mode oscillation start time (ns) Va step down 16 13 mode frequency (GHz) 9.454 9.453 9.452 Back bombardment current (A) 4.57 5.42 5.92 Anode current (A) 2.07 3.07 3.82 Total system energy (J) 74.3e-6 90.9e-6 102.4e-6 DC energy (J) 46.0e-6 48.7e-6 50.5e-6 System RF energy (J) 28.3e-6 42.2e-6 51.9e-6 W Hot test Qe xt = 2f0 Po ut 217 222 224 Output power (kW) 7.74 11.27 13.76 Output power modied by circuit eciency 85% 6.58 9.58 11.70 Power dissipation on cathode (kW) 0.387 0.575 0.720 Power dissipation on anode (kW) 2.89 5.03 6.83 Overall eciency 60.5% 57.8% 55.8% Vane voltage (kV) (amplitude) 3.92 4.67 5.17 Output voltage (kV) (amplitude) 0.87 1.04 1.15 Output current (A) (amplitude) 17.49 20.83 23.07
12.53 5.66 Va step up 9.451 6.87 5.54 127.9e-6 53.7e-6 74.2e-6 225 19.59 16.65 1.120 11.54 53.1% 6.16 1.37 27.46
beam emission model will eventually approach the space-charge-limited emission model when its emission current density is suciently increased; thirdly, the current computer model shows a high overall eciency and the real magnetic eld distribution is not the reason. Possible reasons for this discrepancy will be investigated in the following sections.
5.5
Electric Field Distribution between Cathode and Anode
The radial and azimuthal electric eld distribution between the cathode and the anode when the magnetron is oscillating in a stable mode has been computed for both the
108
space-charge-limited cathode emission and the beam emission models. An initial objective of the computations was to conrm whether the electric eld at the cathode was zero for the space-charge-limited emission and to compare this with the eld distribution in the beam emission model for a current density of 15A/cm2 . However, the variation of the eld with radius has revealed some interesting features which justied further investigations.
5.5.1
The calculation paths
The Er and E components have been observed at three dierent points of an RF cycle, when the system was in stable mode oscillations. The calculations have been carried out along four radial paths: r (rc , ra ), = 300 , z = 0; which is a radial line in a radial plane through the centre of a cavity, at z=0. r (rc , ra ), = 200 , z = 0; which is a radial line in a radial plane through the centre of a vane, at z=0. r (rc , ra ), = 300 , z = 1.5mm;
(5.1)
(5.2)
(5.3)
which is a radial line in a radial plane through the centre of a cavity and near one of the end hats. r (rc , ra ), = 200 , z = 1.5mm; (5.4)
which is a radial line in a radial plane through the centre of a vane and near one of the end hats. Here rc = 1.9mm is the cathode radius and ra = 2.82mm is radius of the inner
109
anode vane surface. Figure 5.15 shows the angular position of the dierent paths used for calculations.
= 20
= 30
Figure 5.15: Electric eld is measured along the solid lines indicated in drawing
5.5.2
Electric eld distribution in the beam emission model
The electric eld distribution has rst been calculated in the beam emission model. The operating conditions are anode voltage 5.75kV, non-uniform magnetic eld Bmax = 0.418T , anode current 2.08Amps and cathode emission density 15A/cm2 (for detailed operation parameters, see Table 5-E). Figures 5.16 and 5.17 show the variation of Er and E along the four dierent paths. The position of the cathode and anode surfaces is indicated on the diagram. The electric elds are marked by and for Er and E respectively. In the computer model, the calculation grids are allocated so that the anode and cathode surfaces are just on the cell edge. As mentioned in Chapter 3, MAGIC is an FDTD code, in which the radial and azimuthal electric components are calculated alternately at the edge and in the middle of a cell. Therefore on the diagrams the E component is calculated at the cell edges and the Er component at half cell points. When a magnetron is oscillating in a stable mode, there is a core of space charge surrounded by rotating spokes of charge. In Figure 5.16(a), which is plotted along a radial line through the center of a cavity, the eld strength at the cathode surface
110
E field distribution along the centre plane using a beam emission model (=30degrees, the second cavity)
5
Er and E (kV/mm)
-5 cathode surface -10 Er (t=30.250ns) Er (t=30.276ns) Er (t=30.316ns) E (t=30.250ns) E (t=30.276ns) E (t=30.316ns)
-15
anode surface
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(a)
E field distribution along the centre plane using a beam emission model (=20degrees, the second vane)
5
Er and E (kV/mm)
-15
anode surface
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(b)
Figure 5.16: Electric eld components are measured along (a) r (rc , ra ), = 300 , z = 0; (b)r (rc , ra ), = 200 , z = 0; is not zero. The Er components increase gradually towards the anode and reach a maximum at the boundary of the core of the space charge at about 2.5mm, and then
111
E field distribution along the plane near the end hat using a space charge limited emission model (=30degrees, the second cavity)
5
Er and E (kV/mm)
-5 cathode surface
-10 Er (t=30.250ns) Er (t=30.276ns) Er (t=30.316ns) E (t=30.250ns) E (t=30.276ns) E (t=30.316ns)
anode surface
-15
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(a)
E field distribution along the plane near the end hat using a beam emission model (=20degrees, the second vane)
5
Er and E (kV/mm)
-15
-20 1.6 1.8 2
anode surface 2.6 2.8 3 3.2
2.2
2.4
r (mm)
(b)
Figure 5.17: Electric eld components are measured along (a)r (rc , ra ), = 300 , z = 1.5mm; (b)r (rc , ra ), = 200 , z = 1.5mm fall o, approaching zero at the back of the cavity. There is little variation with time, either because the radial electric eld due to the electron spokes is small or because the
112
spokes are opposite the vanes at the instant of measurement. The azimuthal electric eld, E , has a small variation in the charge core and then varies with the RF, the amplitude between the vanes being approximately 4kV/mm. In Figure 5.16(b), which shows conditions along a radial line through the center of a vane, there is only slight variation of E with time. This, of course, should be zero, except for space charge uctuations. It should be noted that there is an axial variation of the back bombardment current (for detailed discussion, see Chapter 6), therefore the radial electric eld variation at the end of the cathode near the end hats is interesting. Figures 5.17(a) and 5.17(b) show this eect, as measured along the lines through the cavity and the vane respectively. It can be seen from these two gures that the Er components are in general bigger near the anode surface and smaller near the cathode surface than in Figures 5.16(a) and 5.16(b) respectively. This is due to the fact that the gap between the cathode end hats and the anode is quite narrow.
5.5.3
Electric eld distribution in the space-charge-limited emission model
With the space-charge-limited emission model, the operating conditions are the anode voltage 5.70kV, the real magnetic eld Bmax = 0.418T and the anode current 3.6Amps (for detailed operation parameters, see Table 5-I). The computation results plotted in Figures 5.18(a) and 5.18(b) for space-chargelimited emission show that the electric eld at the cathode surface is near zero and then increases almost linearly until the boundary of the space charge core is reached at about 2.4mm. This conrms the space-charge-limited emission model discussed in Chapter3. E varies similarly to the beam emission model but its amplitude is greater. The eld variation near the end hats also exhibits similar pattern as the beam emission model but at a greater amplitude.
113
E field distribution along the centre plane using a space charge limited emission model (=30degrees, the second cavity)
5
Er and E (kV/mm)
-5 cathode surface
-10 anode surface -15

Er (t=30.250ns) Er (t=30.263ns) Er (t=30.316ns) E (t=30.250ns) E (t=30.263ns) E (t=30.316ns)
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(a)
E field distribution along the centre plane using a space charge limited emission model (=20degrees, the second vane)
5
Er and E (kV/mm)
-5
-10
cathode surface
-15
-20 1.6 1.8 2
anode surface
2.2
2.4
2.6
2.8
3.2
r (mm)
(b)
Figure 5.18: Electric elds are measured along (a) r (rc , ra ), = 300 , z = 0; (b)r (rc , ra ), = 200 , z = 0; The above investigations of the electric eld distribution with both beam emission model and space-charge-limited emission model show quantitatively the dierences
114
E field distribution along the plane near the end hat using a space charge limited emission model (=30degrees, the second cavity)
5
Er and E (kV/mm)
-5 cathode surface
-10
Er (t=30.250ns ) Er (t=30.263ns ) Er (t=30.316ns ) E (t=30.250ns) E (t=30.263ns) E (t=30.316ns)
anode surface
-15
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(a)
E field distribution along the plane near the end hat using a space charge limited emission model (=20degrees, the second vane)
5
Er and E (kV/mm)
-5 cathode surface -10

anode surface
-15
-20 1.6 1.8 2
2.2
2.4
2.6
2.8
3.2
r (mm)
(b)
Figure 5.19: Electric elds are measured along (a)r (rc , ra ), = 300 , z = 1.5mm; (b)r (rc , ra ), = 200 , z = 1.5mm between the two models. Also from the electric eld distribution between the cathode and anode, the radius of the electron cloud when in a stable mode oscillation can
115
be measured. Actually the electron cloud radius can always be obtained from an electron cloud snapshot, for example using Figure 5.4(a), which clearly shows the electron cloud in the case of a thick tapering vane model. Thus the method discussed here is a double check on the accuracy of our simulation results. An analysis of the electric eld distribution can also help in understanding the intrinsic operation mechanisms of the magnetron.
5.6
Varying the Output Coupling Loop
The output coupling circuit plays a very important role in magnetron operation. Varying the output coupling area, the magnetron performance such as the output power, the anode current and electronic eciency can change greatly. Figure 5.20 illustrates the geometry of the output coupling circuit. The intrusion length of the coupling loop into the anode is denoted as d in the picture and it will be varied in the following simulations.
Figure 5.20: The geometry of the output coupling circuit The initial objective of this investigation is to study the reason for a high overall eciency of the computer model. It is found that by tuning the output coupling circuit, the overall eciency is decreased at the expense of Qext being increased at the same time and a comprehensive set of results has been achieved by varying Qext , as shown
116
in Table 5-K. Hot test simulations have been carried out with the output coupling loop length being varied from 2.765mm to 0.78mm. The Qext varies correspondingly from 183 to 680 and the anode current increases from 3.36A to 6.25A. As mentioned before, the overall eciency is obtained from the value of output power as modied by the circuit eciency, which is determined by the Q0 and Qext (see section 5.7). cold = 1 1+
Qext Q0
With Qext = 183, cold = 85%, the Q0 is 1037. If the Q0 does not change, increasing the Qext from 183 to 680 reduces the circuit eciency from 85% to 60%. The overall eciency therefore decreases signicantly from 60.7% to 21.7%. Table 5-K: Investigation of the dierent output coupling with space-charge-limited emission model Magnetic eld: uniform 0.405T Cathode emission model: space-charge-limited Eective anode voltage (kV): 5.70 The output coupling loop length d (mm) 2.765 1.97 1.58 Qext from cold test 183 260 330 Circuit eciency 85% 80% 76% mode oscillation start time (ns) 8 7 7 mode frequency (GHz) 9.418 9.461 9.454 Anode current (A) 3.36 3.83 4.40 Output power (kW) 13.68 14.30 14.96 Output power modied by circuit eciency 11.63 11.44 11.37 Power dissipation on cathode (kW) 0.494 0.640 0.830 Power dissipation on anode (kW) 5.23 7.17 9.80 Overall eciency 60.7% 52.4% 45.3% Vane voltage (kV) 4.15 5.15 6.32
1.38 387 73% 7 9.451 4.84 14.95 10.91 0.972 12.11 39.6% 7.07
1.18 440 70% 8 9.450 5.10 14.70 10.29 1.100 13.90 35.4% 7.50
0.98 540 66% 8 9.448 5.60 14.00 9.24 1.200 17.20 29.0% 8.20
0.78 680 60% 7 9.446 6.25 12.86 7.72 1.410 21.66 21.7% 8.95
Table 5-K presents detailed results of such modelling, e.g. the vane voltage, anode current, output power and the anode and cathode electron power dissipation. The mag-
117
(a)
(b)
(c)
(d)
Figure 5.21: Magnetron performance as a function of the output coupling loop. (a) External Q vs output coupling loop length d (b) Anode current vs external Q (c) Overall eciency vs external Q (d) Vane voltage vs external external Q netic eld is uniform in the axial direction at 0.405T and the input voltage is 5.70kV. At this point MAGIC could not cope with the real magnetic eld distribution. The constrains on boundary condition of this version of MAGIC is also dierent from those of more recent versions. Therefore with the same geometry, ie., the coupling loop length d=1.97mm, the computed Qext s is somewhat dierent in these cases. In the old version of MAGIC, Qext with d=1.97mm is 260. In the new version of MAGIC, by tuning the absorbing boundary conditions of the output ports, the Qext with d=1.97mm is 244, which is close enough to the old value for a direct comparison of results. We have used this value of Qext in previous sections. In Figure 5.21 various relations between dierent parameters are presented. Figure 5.21(a) shows that the external Q varies with the output coupling loop length nearly

! " $ # %
! " #
# # $" !
! ! # " " $ % #
% & ' ) ( 0 1 2 3
118
parabolically. Figures 5.21(b) and 5.21(c) show that the anode current and the overall eciency vary roughly linearly with the external Q. With increasing Qext , the overall eciency decreases linearly. This observation is consistent with e2v measurements. Table 5-L presents one set of measurement of the output mismatch, where the eciency drops when the hot test Qext increases. The oscillation frequency has also changed, due to the frequency pulling eect. The dependence of RF voltage between adjacent anode vanes on Qext is shown in Figure 5.21(d). Basically, with increasing Qext , the anode current will increase and so does the RF vane voltage. In this section the values of Qext of dierent geometries are measured by cold tests, as described earlier in this chapter. Table 5-L: Measurement of the output mismatch (e2v)
Mismatch phase Mean current Anode voltage Pin Pout Eciency Hot Qext
No mismatch max. power m/m min. power m/m
(mA) 5.0 4.72 5.30
(kV) 5.8 5.85 5.75
(Watts) 29 27.6 30.5
(Watts) 11.1 12.6 9
(%) 38.3 45.6 29.5
187 125 280
Since the overall eciency is decreased only when Qext is increased well above its usual value, the conclusion is reached that the output coupling is not the responsible for the high eciency of the computer model of the MG5241 magnetron.
5.7
Copper Surface Losses
All previous computations have been carried out without considering the anode copper surface loss. Though the overall eciency can be obtained by multiplying the electronic eciency by the circuit eciency, a process which accounts for the copper surface losses, it is really necessary to simulate the copper surface losses when investigating their eect on the magnetron start-up of oscillations. Since during the start-up of process, competition between and non- modes depends directly on their respective loaded Qs. Although the denitions of Qext , Qload and Q0 are available in most microwave books, they are given in Appendix 2 as a convenient reference together with the calculation of
119
the circuit eciency. Two dierent lossy anode models have been investigated. First, the entire inside surface of the anode was dened as a lossy material with conductivity = 5.99e7mhos/m, this gives a loaded Q of 220, see Table 5-M. But with this model, the simulation time is unbearably long, about twice that for a normal model. Therefore another approach, which introduces lumped lossy elements into the anode system, has been developed in order to reduce the simulation time. Figure 5.22 shows the geometry of the lumped lossy elements. Now the simulation time is greatly reduced and is nearly the same as the ordinary simulation without any copper losses. By properly selecting the size and conductivity of the lossy elements, Qext , Qload and the cold eciency can all be tuned close to their experimental values.
Figure 5.22: The geometry of the lumped lossy elements is shown in dark blue (The conductivity of the two lossy rings is 5.99e5mhos/m) In the two models, Qext = 260. The corresponding calculation of Qload is based on the cold test energy decay process, which is the same as the calculation for Qext , but with the surface losses or lossy elements being added. Table 5-M presents Qext , Qload , Q0 and the circuit eciency for the two models. Three hot test simulation runs have then be carried out respectively without surface loss, with distributed copper surface losses and with lumped lossy elements. Table 5-N summaries the detailed results for these three simulation processes. The space-charge-
120
Table 5-M: Calculation of Q0 and circuit eciency Loss arrngements Distributed surface loss Lumped lossy elements Qext 260 260 Qload 220 212 Q0 1430 1148 Circuit eciency 84.6% 81.5% Table 5-N: Investigation of the copper surface losses Qext : 260 Magnetic eld: uniform 0.405T Cathode emission model: space-charge-limited Surface loss arrangement No surface loss Distributed surface loss = 5.99e7mhos/m Qload / 220 Oscillation start at (ns) 7 14 Eective anode voltage (kV) 5.84 5.85 mode frequency (GHz) 9.460 9.460 Anode current (A) 5.07 4.83 Output power (kW) 18.50 16.20 Power dissipation on cathode (kW) 0.839 0.740 Power dissipation on anode (kW) 10.87 9.45 Surface power loss (kW) / 2.50 Overall eciency 53.1% 57.3% Vane voltage (kV) 6.03 5.60 * The overall eciency is obtained by correcting the electronic eciency circuit eciency of 85%.
Lumped lossy element 212 14 5.82 9.462 4.58 15.17 0.685 8.64 2.84 56.9% 5.40 with the
limited emission model has been used for all these runs and the magnetic eld was kept uniform at 0.405T. In order to help the mode to dominate, all three simulations start at a lower anode voltage of 5.7kV. This is because at a higher anode voltage, competition between mode and non modes is very strong and sometimes may cause the magnetron to jump into non mode oscillations. When the magnetron settles into stable mode oscillations, the anode voltages can then be stepped up gradually to obtain a 5Amps anode current, which is the normal operating current in practice. It can be seen from Table 5-N that when without surface losses, the magnetron starts oscillating in the mode at 7ns; with surface losses or lumped lossy elements, the start of mode
121
oscillations is delayed to 14ns. This indicates that by introducing copper losses, the mode competition is getting stronger at the same anode voltage and makes it dicult for the mode to dominate. The overall eciency of the two models is still about 10% higher than the experimental value, which is consistent with our earlier investigations in the absence of the copper surface losses. Therefore, the conclusion can be reached that the copper surface losses are not the reason for the low eciency of the X band magnetron.
5.8
The End Hat Emission
From the previous investigation, both beam emission model and space-charge-limited emission model operate at overall eciency levels about 10% higher than the experimental eciency. Therefore the main objective of the following work is to nd out the reason for the lower eciency of the X band magnetron. It is found from the experimental plots of the output power against anode current as measured at e2v, that a leakage current of about 0.5mA exists in the magnetron. This indicates that a source of spurious current must exist, which has not been included in the current computer model. It is known that cathode end hats emission is a possible source of the spurious anode current and this could reduce eciency. Therefore in this section, simulation runs will be carried out to investigate the eect of the cathode end hat emission under various conditions. Both space-charge-limited model and beam emission model have been investigated, together with uniform and real magnetic eld distributions.
5.8.1
Emission models of end hats
Figure 5.23 illustrates the geometry of the cathode and end hats. The inside surface of both end hats, where possible spurious emission could occur, are divided into 6 cells labeled as 1-6 in our computer calculation. Initially the space-charge-limited emission
122
model has been used for both the cathode and the end hats. The copper surface losses have not at this point been included in order to reduce the simulation time. Instead, all the output power is corrected by 85% referring to circuit eciency.
Figure 5.23: The geometry of the cathode end hats First, preliminary simulation runs have been carried out to investigate possible end hats emission areas. Table 5-O summarises the results of those calculations. In the rst emission model all inside cells of the end hats are allowed to emit electrons. With this setup, the anode current rises to 17Amps within 2ns. Then the emission from one, two and three top layers of cells has been investigated. Still the anode current was unrealistically high. Therefore a conclusion must be drawn that the emission from the top three layers of cells of the end hats using space-charge-limited emission model take the system beyond the cuto and causes a dramatic increase in the anode current. This in itself is an interesting discovery. Table 5-O: Investigation of possible emission from dierent areas of the end hats Emission area Anode currents All cells 1 6 Rise up to 17A within 2ns Cells 1, 2 Rise up to 15A within 1ns Cell 1 Rise up to 9A within 2ns Cells 3, 4, 5, 6 Rise up to 3A within 2ns Cells 4, 5, 6 mode oscillation, 4.92A Cells 5, 6 mode oscillation, 4.1A In order to verify the above observation, analytical calculations of the cuto voltage corresponding to dierent end hat positions have been carried out. The radial coordinates
123
14
12
rc=2.5mm rc=1.9mm rc=2.2mm
10
Anode voltage (kV)
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Magnetic field (Tesla)
Figure 5.24: Cut o voltage corresponding to dierent cathode radii of the top three end hat cells are from 2.2mm to 2.5mm. Figure 5.24 shows the cuto curves with assumed cathode radii 1.9mm, 2.2mm and 2.5mm. It can be seen that when the magnetic eld is 0.405T, the cuto voltage is 8.50kV with cathode radius 1.9mm and 4.37kV for cathode radius of 2.2mm. The cuto voltage for cathode radius 2.5mm is even lower. Therefore, the cuto voltages corresponding to the top three cells of the cathode end hats are far below the operating anode voltage, which is about 5.7kV. This explains why the emission from these cells is so high. Stable mode oscillations occurr when end hats emissions are restricted to the bottom two or three layers of cells. Detailed conditions and simulation results are presented in Table 5-P. The magnetic eld is kept uniform at 0.405T. For the convenience of comparison, the simulation results without the end hats emission under the same magnetic eld and eective anode voltage 5.70kV, are also presented in the table. When including the bottom two layers (cells 5, 6) of the end hats in emission, with the anode voltage 5.69kV, the end hats emitted current is 0.35A from each; the anode current is slightly increased from 3.93A to 4.1A; the overall eciency is reduced from 55.7% to 53.0%. When including the bottom three layers (cells 4, 5, 6) in emission, the eective anode voltage drops to 5.66kV; the end hats emitted current is 1.15A from each; the anode
124
Table 5-P: Preliminary investigation of the end hats emission Qext : 260 magnetic eld: uniform 0.405T Cathode emission model: space-charge-limited End hats emission model: space-charge-limited End hats emission area / cells 4, 5, 6 Eective anode voltage (kV) 5.70 5.69 Oscillation start at (ns) 7 11 Cathode emission current (A) 9.74 8.76 end hats emission current (A) / 1.15*2 Total cathode back bombardment current (A) 5.86 5.72 Total anode current (A) 3.93 4.92 Output power (kW) 14.30 14.73 Output power modied by circuit eciency 85% 12.16 12.52 Power dissipation on cathode 0.640 0.705 Power dissipation on anode 7.17 12.79 Overall eciency 55.7% 45.0% Vane voltage (kV) 5.15 5.24
cells 5, 6 5.66 14 9.43 0.35*2 5.90 4.10 14.55 12.37 0.698 8.20 53.0% 5.19
Figure 5.25: Comparison of anode current evolution. (a) Anode current evolution without end hats emission. Anode voltage 5.7kV, magnetic eld uniform at 0.405T. (b) Anode current evolution with end hats emission from cells 4, 5, 6. Anode voltage 5.66kV, magnetic eld uniform at 0.405T, pre-oscillation anode current 1Amps.
125
current increases to 4.92Amps and the overall eciency decreases signicantly to 45.0%. Here, in all cases, the overall eciency is corrected by including the circuit eciency of 85%.
Figure 5.26: The electron cloud including the end hats emission from the bottom three layers of cells 4, 5, 6 at the beginning of the hot test simulation. Anode voltage 5.66kV, magnetic eld uniform at 0.405T. (a) t=0.525ns (b) t=1.275ns Figures 5.25(a) and 5.25(b) compare the anode current evolution with and without end hats emission. Without end hats emission, the waste current or pre-oscillation current is nearly zero. With end hats emission (cells 4, 5, 6), the pre-oscillation current is about 1Amp and the oscillations start at 11ns. Emission from the polepieces has also been observed, but the corresponding current is very weak, about 0.045Amps from each polepiece. The amplitude of vane voltage oscillations does not change signicantly. Figures 5.26(a) and 5.26(b) show snapshots of the electron cloud at the beginning of a simulation. It can be seen that the end hat electrons tend to move horizontally towards the centre of the cathode, but do not go straight to the anode.
126
From our preliminary investigations, it can be concluded that with a space-chargelimited emission model, the end hats emission from the bottom three layers of cells (4, 5, 6) can signicantly reduce the overall eciency and bring it close to the experimental eciency; this in itself justies further investigations.
5.8.2
End hats emission with space-charge-limited model and uniform magnetic eld of 0.405T
Including electrons emitted from the bottom three layers of cells of the cathode end hats, a number of simulation runs have been carried out with anode voltages varying from 5.70kV to 5.14kV and anode current from 5.11A to 2.24A, the applied magnetic eld being uniform at 0.4050T. Since this piece of work was completed before the MAGIC code was updated to include real magnetic eld distribution, the cold test external Q of the computer model is 260 in agreement with the early version of MAGIC. Table 5Q presents the details of the simulation results. In this table, the end hats emission electrons and the cathode emission electrons are treated separately. Therefore their separate contributions to the anode current, back bombardment current, anode power dissipation and back bombardment power are clearly shown. This is very important for the analysis and understanding of the interaction mechanism. In Figure 5.27, the relations between the anode voltage, output power and anode current are shown. In these diagrams, the results of space-charge-limited emission model without the end hats emission have also been presented for the purpose of comparison. It can be seen from gure 5.27(a) that at given anode current, the output power now agrees well with the experimental result. But at low anode current, i.e., below 3Amps, the output power is lower than the experimental result; extending the curve to zero output power, the intersection point occurs at about 1Amps. This means that the waste current of this computer model is about 1Amps. But the waste current from the measurements is only no more than 0.5Amps. We can see from the diagram of the VI curves (Figure 5.27(b)) that at a given anode voltage, the anode current is far
127
Table 5-Q: Investigation of the end hats emission with space-charge-limited emission model and uniform magnetic eld at 0.405T Qext : 260 magnetic eld: uniform 0.405T Cathode emission model: space-charge-limited End hats emission model: space-charge-limited; emitting from cells Eective anode voltage (kV) 5.70 5.66 5.49 5.40 Oscillation start at (ns) 14 11 5 2 Cathode emission current (A) 9.10 8.76 7.34 6.57 end hats emission current (A) 1.16*2 1.15*2 1.11*2 1.09*2 Anode current from cathode emission (A) 3.32 3.16 2.13 1.69 Anode current from end hats emission (A) 1.79 1.78 1.71 1.68 Total anode current (A) 5.11 4.92 3.84 3.37 Back bombardment current from cathode emission (A) 5.66 5.62 5.15 4.86 Back bombardment current from end hats emission (A) 0.122 0.128 0.177 0.200 Output power (kW) 15.25 14.73 11.12 9.40 Output power modied by circuit eciency 85% 12.96 12.52 9.45 7.99 Power dissipation on cathode by cathode emission (kW) 0.623 0.600 0.448 0.381 Power dissipation on cathode by end hats emission (kW) 0.102 0.104 0.145 0.166 Power dissipation on anode by cathode emission (kW) 6.41 5.92 3.67 2.74 Power dissipation on anode by end hats emission (kW) 7.04 6.87 6.02 5.59 Total power dissipation on polepieces (kW) 0.287 0.201 0.166 Overall eciency 44.5% 45.0% 44.8% 43.9% Vane voltage (kV) 5.40 5.24 4.64 4.25
4, 5, 6. 5.24 2 5.43 1.04*2 1.00 1.60 2.60 4.46 0.270 6.78 5.76 0.284 0.215 1.50 4.96 0.122 42.3% 3.61
5.14 3 4.81 1.00*2 0.70 1.54 2.24 4.10 0.316 5.30 4.51 0.237 0.237 1.00 4.74 0.089 39.2% 3.20
higher than the experimentally measured current. Figure 5.27(c) shows that the overall eciency is lower than the experimental eciency at low anode current and high at high anode current. The legends of Figure 5.27(a) and 5.27(c) are the same as those of Figure 5.27(b). It can be concluded that with this emission model, the spurious current being introduced is higher than expected, which aects the magnetron performance more signi-
128
Ia
(a)
Va (kV)
Bmax= .425T Bmax= .413T Bmax= .398T Space charge limited emission; Uniform B=.405T Space charge limited emission model. End hats emission from cells 4, 5, 6. Uniform B=0.405T
Ia
(b)
Ia
(c)
Figure 5.27: Investigation of the end hats emission from the bottom three cells 4, 5, 6. Uniform magnetic eld at 0.405T. Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Output power vs anode current (b) Anode voltage vs anode current (c) Overall eciency vs anode current

! " # !

129
cantly at low anode current levels. This is because at low anode current levels, the contribution of the end hats emission to the total anode current is proportionately greater than at higher anode current levels. For example, when the anode voltage is 5.7kV, the total anode current is 5.11Amps, of which, 3.32Amps is from the cathode emission and 1.79A is from the end hats emission. But when the anode voltage is reduced to 5.14kV, the total anode current is reduced to 2.24A, of which, only 0.70Amps is from the cathode emission and 1.54Amps is from the end hats emission. Although this model can reduce the overall eciency in general, it cannot give the right results at low anode current.
5.8.3
Reduced spurious current and the real magnetic eld distribution
The end hats emission current can be reduced in two ways. One is to reduce number of the end hats emitting cells; the other is to use the beam emission model and control the emitting current density. The rst method has already been investigated in section 5.8.1, where simulations have been carried out with dierent combinations of the end hat emitting cells. We have concluded in that section that with space-charge-limited emission from end hats, only emission from the bottom three layers of cells can give reasonable results and at the same time reduce the overall eciency signicantly. However this model was proved in the previous section to be unrealistic by generating too much spurious current. Therefore the only way to reduce the end hats current is to use the beam emission model and thus control the end hats emission articially. Since conditions regarding the end hats emission are uncertain and dicult to measure, though articial, this investigation is still deemed to be useful for studying possible reason for the lower eciency of the magnetron. Also at this time, the MAGIC code was updated to allow for a real magnetic eld distribution. Therefore in this section, all the simulation runs are carried out with a real magnetic eld distribution, in order to diminish possible discrepancy caused by the use of an unrealistic, uniform magnetic eld. With a beam emission model, the total end hats emission current is now reduced to 1.664Amps, with 0.832Amps from each
130
Table 5-R: Investigation of the end hats emission with beam emission model and real magnetic eld distribution Bmax = 0.418T Qext : 244 Non-uniform magnetic eld: Bmax = 0.418T Cathode emission model: beam emission with current density 15A/cm2 ; Total cathode emission current: 6.26A End hats emission model: beam emission with current density 10A/cm2 ; Total end hats emission current: 0.832A*2 Eective anode voltage (kV) 5.90 5.80 5.72 5.63 5.54 Oscillation start at (ns) 12 10 10 10 10 Anode current from cathode emission (A) 3.07 2.29 1.90 1.45 1.10 Anode current from end hats emission (A) 1.20 1.13 1.10 1.06 1.05 Total anode current (A) 4.27 3.42 3.00 2.51 2.15 Back bombardment current from cathode emission (A) 3.19 4.00 4.36 4.85 5.14 Back bombardment current from end hats emission (A) 0.171 0.222 0.261 0.286 0.295 Output power (kW) 12.24 9.58 8.15 6.62 5.71 Output power modied by circuit eciency 85% 10.40 8.14 6.93 5.63 4.85 Power dissipation on cathode by cathode emission (kW) 0.627 0.495 0.456 0.410 0.370 Power dissipation on cathode by end hats emission (kW) 0.088 0.078 0.075 0.067 0.054 Power dissipation on anode by cathode emission (kW) 7.32 4.95 4.00 2.72 1.75 Power dissipation on anode by end hats emission (kW) 5.26 5.10 4.70 4.49 4.29 Total power dissipation on polepieces (kW) 0.353 0.399 0.444 0.509 0.537 Overall eciency 41.3% 41.0% 40.4% 39.8% 40.7% Vane voltage (kV) 4.88 4.33 3.99 3.61 3.35
5.43 11 1.00 1.04 2.04 5.14 0.288 5.18 4.40 0.343 0.048 1.46 4.16 0.553 39.7% 3.17
end hat. Now the emission areas are the whole inside surfaces of the two end hats, that is cells 1-6 in Figure 5.23. The cathode also uses the beam emission model with emission current density of 15A/cm2 , which gives the total cathode emission current of 6.26Amps. A number of simulation runs have been carried out with anode voltage varying from 5.43kV to 5.98kV. Table 5-R presents detailed simulation results. The cold test Qext is 244 using this new version of MAGIC, as explained in Section 5.6.
131
Ia
(a)
Va (kV)
Bmax= .425T Bmax= .413T Bmax= .398T Uniform cathode emission current: 6.26A; Bmax=.418T Uniform cathode emission current: 6.26A; end hats emission: 0.832A*2; Bmax=.418T
Ia
(b)
Ia
(c)
Figure 5.28: Investigation of the end hats emission with the beam emission model for both the cathode and the end hats. Real magnetic eld distribution Bmax = 0.418T . Solid lines: MAGIC calculation; Dashed lines: e2v measurements. (a) Output power vs anode current (b) Anode voltage vs anode current (c) Overall eciency vs anode current

! " # !

132
Figures 5.28(a), 5.28(b) and 5.28(c) show the calculated performance curves together with the measured curves. These three gures share the same legends. For the convenience of comparison, the calculated performance without the end hats emission are also presented on these diagrams. At given anode currents, the dierence between the simulated output power and the experimental output power is very small; the overall eciencies are reduced greatly, especially at low anode currents. But the VI curve is still not right compared with the experimental results. When without end hats emission, at a given anode voltages, the calculated anode currents are always lower than the experimental anode currents. When the end hats emission is included, at a given anode voltage, the anode currents are consistently at least half amp higher than the experimental anode currents; especially when the anode current is as low as 2.04Amps, the calculated anode currents are about 0.8Amps higher than the measured values. Referring to Table 5-R, when the total anode current is 2.04Amps, the contribution from the end hats emission is 1.04Amps. This situation is similar to the previous investigation of the space-charge-limited emission model for the end hats, that is, the waste current is too high at low anode currents. Also from Table 5-R, when the anode voltage varies from 5.9kV to 5.43kV, the contribution of the cathode emission to the total anode current decreases from 3.07Amps to 1.00Amps, but the contribution of the end hats emission to the anode current changes very little, from 1.20Amps to 1.04Amps. Figure 5.29 presents the current contribution relationship between the cathode emission and end hats emission. It can be seen that the end hats emission current actually acts as that in a beyond cuto diode and therefore does not seem to contribute to the output power. From the above discussion, the conclusion can be drawn that on its own the end hat emission does not account for the lower eciency of the X band magnetron, though it can give good agreement with the experimental results for the output power and eciency, but only at certain anode currents and anode voltages.
133
Figure 5.29: The contribution of the cathode emission and end hats emission to the anode current as a function of the anode voltage
5.9
Slow Rate of Rise of Voltage
In all the previous modelling work, in order to save the simulation time, the rate of rise of anode voltage was set at 3000kV /s, which is 30 times bigger than the value used in practice. In order to investigate the eects of the real RRV on magnetron start-up, which is only 100kV /s, simulation runs have been done with anode voltage at 5.70kV and 5.87kV. Table 5-S presents the operation conditions. Figure 5.30 shows the transient evolution of the stored system energy, the vane voltage oscillations, the input anode voltage and the anode current while the eective anode voltage is 5.87kV. Figure 5.30(a) shows that the anode voltage rises up to 6.0kV within 60ns and drops to 5.87kV, which is an average value, when the magnetron gets into a stable mode oscillation. It can be seen from Figures 5.30(b), 5.30(c) and 5.30(d) that before the magnetron gets into a stable mode oscillation, the anode current is very low, about 0.7A; the RF energy stored in the magnetron is also very low; the voltage oscillation between anode vanes is very noisy due to the spurious oscillations of non- modes. The calculation of the DC energy and RF energy has been discussed in section 5.2.3. Figure 5.31 presents the spectrum of the vane voltage oscillation during the magnetron start-up between 55ns and 70ns. Apart from the mode oscillation around 9GHz, there are strong spurious
#! $" 9B Q F H8@RH@ FA 9B@ CA97 8QB H I GP G 8E 8 DB @8 9B Q F Q P IHH@ FA 9B@ CA97 8QB H @ G 9 G 8E 8 DB @8
% ' & ( ) 2 0 1 2 ) & 3 4 5 6 %
134
(a)
(b)
(c)
(d)
Figure 5.30: The simulated transient evolution of the main parameters obtained with the slow RRV of 100kV /s. (a) The eective anode voltage 5.87kV. (b) The anode current 5.26A. (c) The system stored energy, the DC energy is 58.5e-6J; the RF energy is 73.1e-6J. (d) The voltage oscillation between two adjacent vanes.
135
Spectrum of vane voltage oscillation between 55ns and 70ns 0
10
20 DB (Volts/GHz)
30
40
50
60
10 11 Frequency (GHz)
12
13
14
Figure 5.31: Oscillation spectrum of the vane voltage during the start-up of magnetron oscillations Table 5-S: Simulation runs with slow RRV of 100kV /s and space-charge-limited emission model Qext : 244 Non-uniform magnetic eld: Bmax = 0.418T Eective anode voltage (kV) 5.70 5.87 Oscillation frequency (GHz) 9.454 9.454 Anode current (A) 3.26 5.62 Output power (kW) 13.87 19.42 Overall eciency 57.1% 53.4% Vane voltage (kV) 5.20 6.13 modes at 10GHz, 11GHz, etc. Therefore by applying an experimental value of the RRV in our simulations, details of the transient oscillations in a magnetron have been revealed, such as the mode competition, the RF energy build-up and the anode current evolution.
136
5.10
Summary
In this chapter, an X band magnetron has been modelled in some detail. Investigations of dierent computer models show that the thick tapering vane model is the best to represent this particular X band magnetron. Dierent cathode emission models together with uniform or non-uniform magnetic elds have been investigated, in order to obtain a complete set of results for comparison with experimental results. The computer model generates an overall eciency which is about 10% higher than the experimental eciency. Possible reasons for the discrepancy have been investigated, such as the copper surface losses, the output coupling loop circuit, the end hats emission and the rate of rise of anode voltage. Though most of these factors have been proved not to be the reason for the low eciency of the magnetron, in the process, some important insights into the operation of the X band magnetron have been gained. The quantitative relations between the output coupling loop area, the overall eciency and the Qext have been obtained. Though the end hats emission can bring the simulated performance close to the experimental one, the eect appears to be articial, bring the tube beyond cuto condition. Simulation using a real RRV has revealed details of the start-up characteristics of a magnetron. So far, no simulation included the important cathode secondary emission because of some instability problems. The conditions for cathode secondary emission and its possible eects on magnetron operation will be discussed in Chapter 6 and 7.
References
[1] S.Y.Liao, Ed., Microwave devices and circuits. Prentice Hall Internationsl, 1990.
Chapter 6
The Eects of Non-Uniform Cathode Emission and Electron Back Bombardment

6.1 Introduction
In this chapter, the cathode emission and back bombardment distribution along the length of the cathode have been investigated using two dierent cathode emission models, the space-charge-limited emission and the beam emission. For each emission model, simulation runs have been conducted both with the uniform magnetic eld and the real magnetic eld distributions respectively. Their corresponding eect on magnetron performance is then carefully considered. Since a non-uniform electron back bombardment distribution along the cathode length would result in a complicated secondary emission distribution, its possible eect on magnetron performance was investigated by articially simulating cathode emission with a Gaussian prole.
137
Chapter 6. The Eects of Non-Uniform Cathode Emission and Electron Back Bombardment
138
6.2
Simulation Model
In our previous work the cathode was treated as a single unit, which emits electrons and absorbs back bombarding electrons. In this chapter, in order to investigate the distribution of the cathode emission and back bombardment along the cathode length, the cathode is divided into 14 ring shaped units, as shown in Figure 6.1.
Figure 6.1: The length of the cathode is divided into fourteen rings
Here cath1 is located adjacent to the left end hat, its axial coordinates being (1.75mm to -1.5mm); cath2: cath3: z (1.5mm, 1.25mm); z (1.25mm, 1.0mm); ...... cath7: cath8: z (0.25mm, 0); z (0, 0.25mm); ...... z (1.5mm, 1.75mm).
cath14:
Calculations of the cathode emission and back bombardment have been carried out individually on each ring.
139
6.3
6.3.1
Investigation of the space-charge-limited Emission Model

Characteristics of cathode emission and back bombardment distribution along the cathode length
With the space-charge-limited emission model, modelling has been carried out with different DC voltages and dierent values of both uniform and real magnetic eld. The results show that in all the cases, the cathode emission current is non-uniformly distributed along the cathode length, see Figure 6.2. When the real magnetic eld distribution is applied, the distribution of the emitted current density along the cathode length is approximately Gaussian. When using a uniform magnetic eld, the current density distribution becomes somewhat unexpected; there are two spikes in the distribution curves which are situated near the end hats. The magnitude of the DC voltage or the magnetic eld (uniform or real) do not aect the emission distribution curves signicantly.
Emission current density with space charge limited emission model
45
Cathode emission current density (A/cm2)
40 35 30 25 20 15 10 5 0 -2
Vdc=5.70kV;uniform B=.405T Vdc=5.50kV;uniform B=.395T Vdc=5.40kV;uniform B=.395T Vdc=5.50kV;Bm ax=.405T Vdc=5.62kV;Bm ax=.418T Vdc=5.62kV; Br=0, Bzm ax=.418T
-1.5
-1
-0.5
0.5
1.5
z (mm)
Figure 6.2: Cathode emission current density distribution along the cathode length when using the space-charge-limited emission model Tables 6-A and 6-B present the corresponding results and operating conditions. As introduced in Section 5.3, the real magnetic eld distribution is 2D with both Bz and Br components, while a uniform magnetic eld has only the Bz component. Is this then
140
the Br component that cause the dierence in the cathode emission distribution? In order to test this hypothesis, a simulation was carried out with zero Br component but retaining the Bz component as measured at e2v (Bmax =.418T). Now, the two spikes appear in the emission current density distribution appear again as they did in the case of a uniform magnetic eld. This conrms that the Br component aects the cathode emission distribution signicantly and in modelling the real magnetic eld distribution should never be replaced by a uniform magnetic eld. Table 6-A: Simulation results with the space-charge-limited emission model and real magnetic eld distribution Cathode emission model Eective input voltage (kV) Non-uniform magnetic eld (T) Anode current (A) Input power (kW) Output power (kW) Vane voltage (kV) Electronic eciency (%) space-charge-limited 5.62 5.50 Bmax : 0.418 Bmax : 0.405 3.0 17.05 11.56 4.73 67.8 3.82 21.26 13.78 5.17 67.8 5.62 Br=0 Bz(max)=0.418 2.78 15.60 10.94 4.60 70.1
Table 6-B: Simulation results with space-charge-limited emission model and uniform magnetic eld distribution Cathode emission model Eective input voltage (kV) Uniform magnetic eld (T) Anode current (A) Input power (kW) Output power (kW) Vane voltage (kV) Electronic eciency (%) space-charge-limited 5.62 5.50 5.62 0.405 0.395 0.395 3.86 3.78 3.05 22.40 20.64 16.56 14.88 13.40 11.41 5.15 5.20 4.71 66.4 64.9 68.9
Figure 6.3 and 6.4 show the back bombardment voltage and current density distributions along the cathode length. The back bombardment electron voltage (or electron power) has an approximately Gaussian distribution along the cathode length, somewhat surprisingly irrespective of a uniform or real magnetic eld distribution. However the current density distribution depends on the shape of the magnetic eld. When the magnetic eld is uniform, the back bombardment current density is roughly uniform in the
141
center region, but drops sharply to zero towards both ends of the cathode. When the real magnetic elds is used, back bombardment current gradually decreases towards the ends of the cathode. Thus again the Br component plays an important role in the back bombardment distribution. An explanation of this phenomenon is provided in the next section, using simple particle kinematics.
Back bombardment voltage with space charge limited emission model
400 350
Vdc=5.70kV;uniform B=.405T Vdc=5.50kV;uniform B=.395T Vdc=5.40kV;uniform B=.395T Vdc=5.50kV;Bm ax=.405T Vdc=5.62kV;Bm ax=.418T Vdc=5.62kV; Br=0, Bzm ax=.418T
Back bomdment voltage (V)
300 250 200 150 100 50 0 -2
-1.5
-1
-0.5
0.5
1.5
z (mm)
Figure 6.3: Back bombardment Voltage distribution along the cathode length
Back bombardment current density with space charge limited emission model
45
Back bombardment current density (A/cm2)
40 35 30 25 20 15 10 5 0 -2
Vdc=5.70kV;uniform B=.405T Vdc=5.50kV;uniform B=.395T Vdc=5.40kV;uniform B=.395T Vdc=5.50kV;Bm ax=.405T Vdc=5.62kV;Bm ax=.418T Vdc=5.62kV; Br=0, Bzmax=.418T
-1.5
-1
-0.5
0.5
1.5
z (mm)
Figure 6.4: Back bombardment current density distribution along the cathode length
142
6.3.2
Electric and magnetic eld variation along the cathode surface
In order to reveal how the Br component aects the cathode emission and back bombardment distribution along the cathode length, the electric and magnetic elds above the cathode surface and the electron motion under their inuence are investigated. As shown in Figure 6.1, the cathode is divided into fourteen rings along its length. Since the boundary conditions on a metal surface require that the tangential electric eld component must be zero, in the region near the inside surface of the end hat the electric eld is mainly in the Ez direction; moving towards the center of the cathode, the electric eld becomes mainly radial or Er . Figures 6.5 and 6.6 show the variation of Er , E and Ez along the cathode length at dierent azimuthal and radial positions, respectively when uniform and real magnetic elds are. With the space-charge-limited emission model, very few electrons are emitted from the cathode rings 1 and 14 because the Er component near the end hat (z = 1.75mm) is too weak; Er component distribution along the cathode length is roughly uniform between rings 2 and 13 (-1.5mm -1.5mm). In the presence of a uniform magnetic eld, the only force on the electrons in the longitudinal direction is eEz , so that electrons leaving rings 2 and 13 tend to move towards the cathode center under the inuence of the Ez component. This keeps the space charge density above those two rings quite low. When using the space-charge-limited emission model, this is no longer the case, since more electrons will be emitted from the two cells. This explains the two spikes in the emission current density distribution in the presence of a uniform magnetic eld. Figure 6.7 shows the electron cloud emitted from dierent cathode sections at the beginning of the emission and at the magnetron oscillates in a stable mode. It is obvious from these diagrams that electrons emitted from the sections near the end hats tend to move towards the cathode center.
=0degree; t=37.995ns; Vdc=5.7kV; Uniform B=.405T Solid line: Er Dashed line: E Dotted line: Ez
4
143
8 6
x 10
r=1.9mm r=2.0mm r=2.1mm r=2.2mm
Er, E and Ez (V/m)
2 0 -2 -4 -6 -8 -2
-1.5
-1
-0.5
0.5
1.5
z (mm)
(a)
8 6 4 x 10
6
=10degree; t=37.995ns; Vdc=5.7kV; Uniform B=.405T Solid line: Er Dashed line: E Dotted line: Ez r=1.9mm r=2.0mm r=2.1mm r=2.2mm
Er, E and Ez (V/m)
2 0 -2 -4 -6 -8 -2
-1.5
-1
-0.5
0.5
1.5
z (mm)
(b)
Figure 6.5: Electric eld distribution along the cathode length at dierent azimuthal positions. Vdc = 5.7kV , uniform magnetic eld 0.405T. (a) Opposite the vane, = 0 (b) Opposite the cavity, = 10 But the situation is dierent when the real magnetic eld is applied. Figure 6.8 shows the Br and Bz variation along the cathode length. It can be seen that the Bz component
=0degree; t=38.040ns; Vdc=5.62kV; Bmax=.418T Solid line: Er Dashed line: E Dotted line: Ez
4
144
8 6
x 10
r=1.9mm r=2.0mm r=2.1mm r=2.2mm
Er, E and Ez (V/m)
2 0 -2 -4 -6 -8 -2
-1.5
-1
-0.5
0.5
1.5
z (mm)
(a)
8 6 4 x 10
6
=10degree; t=38.040ns; Vdc=5.62kV; Bmax=.418T Solid line: Er Dashed line: E Dotted line: Ez r=1.9mm r=2.0mm r=2.1mm r=2.2mm
Er, E and Ez (V/m)
2 0 -2 -4 -6 -8 -2
-1.5
-1
-0.5
0.5
1.5
z (mm)
(b)
Figure 6.6: Electric eld distribution along the cathode length at dierent azimuthal positions. Vdc = 5.62kV , real magnetic eld Bmax 0.418T. (a) Opposite the vane, = 0 (b) Opposite the cavity, = 10 does not vary much, but the sign of the Br component changes along the cathode length, its magnitude increases from zero at the cathode center to about 0.1T near the end
145
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Figure 6.7: Electrons emitted from dierent cathode sections at the beginning of the emission (a, c, e and g) and when the magnetron is oscillating in a stable mode (b, d, f and h) when the magnetic eld is uniform. Vdc=5.70kV; Uniform B=.405T (a)(b) Electrons from rings 1 and 14 (c)(d) Electrons from rings 2 and 13 (e)(f) Electrons from rings 3, 4, 11 and 12 (g)(h) Electrons from rings 5 to 10.
146
hat.
0.5 0.4 0.3 Br and Bz (T) 0.2 0.1 0 -0.1 -0.2 -2
Br (r=1.9mm) Br (r=2.0mm) Br (r=2.2mm) Br (r=2.4mm) Bz (r=1.9mm) Bz (r=2.0mm) Bz (r=2.2mm) Bz (r=2.4mm)
Static magnetic field distribution along the cathode length
-1.5
-1
-0.5
0 z (mm)
0.5
1.5
Figure 6.8: Real magnetic eld (Bz and Br) distribution along the cathode length, Bmax =0.418T. The Br component exerts an axial Lorentz force on the electrons near the end hats, Fz = zev Br . Together with the eEz force, the net eect is for the electrons to move towards the end hats, causing the space charge density to grow there. With the space-charge-limited emission model, emission from these areas will hence be suppressed. Thats the reason of the cathode emission distribution being roughly in Gaussian shape when the real magnetic eld is used. Figure 6.9 illustrates the electron cloud emitted from dierent cathode sections at the beginning of the cathode emission and when the magnetron oscillates in a stable mode. It can be seen that the electrons tend to move towards the two ends of the cathode. In back bombardment the electron motion is more complicated, but according to the above analysis, the dierence in the back bombardment distributions along the cathode length when uniform and non-uniform magnetic elds are applied is also due to the axial Lorentz force generated by the Br component of the eld.
147
(a)
(b)
(c)
(d)
(e)
(f)
Figure 6.9: Electrons emitted from dierent cathode sections at the beginning of the cathode emission (a, c and e) and when the magnetron is oscillating in a stable mode (b, d and f) using the real magnetic eld distribution. Vdc=5.62kV; Bmax =0.418T. Very few electrons are emitted from the two rings 1 and 14, therefore no electron cloud diagrams are displayed for those two sections. (a)(b) Electrons from rings 2 and 13 (c)(d) Electrons from rings 3, 4, 11 and 12 (e)(f) Electrons from rings 5 to 10.
6.4
Investigation of the Beam Emission Model
In Chapter 5, the X band magnetron has been investigated using both space-chargelimited cathode emission and a beam emission model, the back bombardment distribution of the beam emission model will therefore be investigated here for comparison. The beam
148
emission model assumes an initial voltage of 5eV and we have chosen beam current density of 15A/cm2 . With the beam emission model, the cathode emission current density distribution is uniform along the whole cathode length. Here our investigations are focused on the back bombardment and its distribution. Modelling has been carried out with both uniform and real magnetic eld distributions. Figures 6.10 and 6.11 show the back bombardment current density and the back bombardment voltage distribution with uniform and real magnetic elds respectively. Table 6-C summarizes the simulation results. Table 6-C: Simulations of beam emission with uniform and non-uniform magnetic elds Cathode emission model Eective input voltage (kV) Magnetic eld (T) Anode current (A) Input power (kW) Output power (kW) Vane voltage (kV) Electronic eciency (%) Beam current density:15A/cm2 Beam voltage:5eV 5.75 5.62 Real B eld: Bmax =0.418 Uniform B=0.405T 2.08 3.21 12.0 18.28 7.12 10.58 3.73 4.54 59.3 57.9
It can be seen from Figure 6.10 that the back bombardment current density distributions are more or less uniform in the centeral region of the cathode irrespective of whether uniform or real magnetic eld is used. But the back bombardment voltage distribution presents a very dierent picture. When a real magnetic eld is applied, the back bombardment voltage has a Gaussian distribution, which is consistent with the results obtained in the case of a space-charge-limited emission; but when the uniform magnetic eld is applied, the distribution is roughly uniform except two spikes occurring at rings 3 and 12. A detailed explanation for the two spikes must be quite involved because of the complicated motion of the electrons. But the reason for the spikes again must lie in an unrealistically uniform magnetic eld distribution, as indicated in the previous section.
149
Back bombardment for beam emission model

45
Back bombardment curremt density (A/cm2)
40 35 30 25 20 15 10 5 0 -2
Vdc=5.62kV; Uniform B=.405T Vdc=5.75kV; Bmax=.418T
-1.5
-1
-0.5
0.5
1.5
z (mm)
Figure 6.10: Back bombardment current density distribution with beam emission model
Back bombardment for beam emission model
400 350
Back bombardment voltage (V)
300 250 200 150 100 50 0 -2
Vdc=5.62kV; Uniform B=.405T Vdc=5.75kV; Bmax=.418T

-1.5 -1 -0.5 0 0.5 1 1.5 2
z (mm)
Figure 6.11: Back bombardment voltage distribution with beam emission model
6.5
Inuence of a Non-Uniform Cathode Emission on the Operation of a Magnetron
The existence of a non-uniform back bombardment distribution has been demonstrated in the previous sections. This phenomenon conrms that the cathode emission must be non-uniformly distributed when the secondary emission process is included. Currently a complete model of performance including secondary emission is numerically unstable.
150
Therefore instead of introducing the secondary emission into the full magnetron model, the non-uniformity of the cathode emission has been represented in a simple way by applying dierent emission proles along the cathode surface. Although the actual cathode emission prole, including secondary emission, is yet not fully known, the eects of a non-uniform emission can be investigated. It is hoped that this is way some useful information can be revealed on the relation between magnetron operation and the non-uniform cathode emission.
Cathode emission current density along the cathode length
50
Cathode emission current density (A/cm2)
45 40 35 30 25 20 15 10 5 0 -2
Peak: 50A/cm2 Peak: 35A/cm2 Peak: 31.5A/cm2 Peak: 28A/cm2
-1.5
-1
-0.5
0.5
1.5
z (mm)
Figure 6.12: Cathode emission current density distributions (Total emission current is about 5.89A for all emission proles) All our cathode emission proles are Gaussian. Figure 6.12 shows several cathode emission current density distributions along the cathode length, each with a dierent peak value. Table 6-D: Simulation runs with dierent cathode emission current density distributions Peak cathode emission current density 28A/cm2 31.5A/cm2 35A/cm2 50A/cm2 Eective input voltage (kV) 5.73 5.73 5.73 5.72 Anode current (A) 2.50 2.65 2.80 3.17 Input power (kW) 14.5 15.3 16.3 18.3 Output power (kW) 8.0 8.4 8.7 9.4 Electronic eciency (%) 55.1 54.5 53.5 51.2 Following the modelling results of Chapter 5, here the total emission current has
151
been adjusted to about 5.89A, which is close to the cathode emission current with a uniform beam emission current density of 15A/cm2 . The peak current densities range from 28A/cm2 to 50A/cm2 . The simulation results are summarized in Table 6-D. It can be seen that the anode current and output power increase when the peak cathode emission density is increased and thus the shape of the Gaussian distribution is narrowed. However the anode current is increased more than the output power, so the eciency would decrease. Figure 6.13 illustrates the variations of the anode current and the overall eciency against the peak cathode emission current density. This shows that the variation of the cathode emission distribution aects the magnetron eciency by about 4%.
(a)
(b)
Figure 6.13: (a) Anode current vs Peak cathode emission current density. (b) Overall eciency vs Peak cathode emission current density.
' " " $ # ! )(& % ! $ " !
% "! ' &$ # "
0 1 2 3 4 5 5 2 6 8 6 7 7 2 8 @ 9
( ) 0 1 2 3 4 5 5 ) 2 6 7 8 9 (
152
6.6
Summary
In this chapter, the cathode emission and electron back bombardment distributions along the cathode length have been investigated for both space-charge-limited emission model and beam emission model. Their non-uniform distribution along the cathode length have been demonstrated for the rst time using computer modelling. It has been found that the axial Lorentz force caused by the Br component of the magnetic eld plays an important role in shaping the distributions. Since a non-uniform back bombardment distribution obviously would aect secondary emission, dierent Gaussian cathode emission proles have been used to check their eect on the magnetron performance. It has been shown that the cathode emission proles only aect the magnetron eciency to a small extent.
Chapter 7
The Eect of Cathode Emission Voltage

7.1 Introduction
One signicant dierence between our current computer model and the real magnetron is that only the primary emission is modelled, while in a real magnetron the cathode emission comprises both primary and secondary emission. In all previous modelling, the initial cathode emission voltage is very low, at most a few volts, whether we use a spacecharge-limited or a beam emission model; these values are typical of thermionic emission. However experimental results have shown that at some circumstances high energy secondaries are present in an oscillating magnetron. Figure 7.1 shows energy distributions of secondary electrons emitted by silver [1, 2]. There are three regions in the diagram, representing the true secondaries (below 50eV), inelastically reected or rediused primaries (above 50eV, but below the primary energy) and elastically reected primaries (near the primary energy). Will these high energy secondaries aect magnetron operation significantly? Our investigations carried out by using computer modelling have indicated, as described in this Chapter, that the energy distribution of the back bombardment elec-
153
Chapter 7. The Eect of Cathode Emission Voltage
154
Figure 7.1: Energy distribution of secondary electrons emitted by silver [1] trons is quite high, a signicant fraction having kinetic energy large enough to produce high energy secondary electrons. Since at present a MAGIC model of secondary emission becomes numerically unstable, we have articially increased initial cathode emission voltage in order to partially simulate the process.
7.2
Energy Distribution of Electron Back Bombarding
In order to estimate the energy of secondary electrons, the energy distribution of the back bombarding primary electrons has been investigated. Modelling has been carried out with anode currents of 3.60A, 4.13A and 5.14A respectively. Simulation results show that the back bombardment energy distribution ranges from 0eV to 500eV when the anode current is 3.60A; the upper limit of the back bombardment energy increases to 600eV when the anode current reaches 5.14A. Table 7-A presents simulation results with Ia = 3.60A, 4.30A and 5.14A; for average initial cathode emission voltages of 4.0eV, 5.0eV and 5.7eV, the average back bombardment voltages are respectively 130eV, 147eV and 163eV. Figure 7.2 shows the energy distribution at dierent time intervals when the magnetron is oscillating in a stable mode and the anode current is 5.14A. It can be seen
155
(a)
(b)
(c)
(d)
Figure 7.2: The back bombardment energy distribution at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a). t1=39.930ns (b). t2=t1+0.14RFperiod (c). t3=t1+0.28RFperiod (d). t4=t1+0.42RFperiod
156
Table 7-A: The average electron back bombardment energy at dierent anode currents Emission model: space-charge-limited Magnetic eld: Bmax = 0.418T Eective input voltage (kV) Anode current (A) Output power (kW) Overall eciency Average emission voltage (V) Average back bombarding voltage (V)
5.70 3.60 13.77 57.0% 4.0 130
5.79 4.35 16.35 55.2% 5.0 147
5.87 5.14 19.0 53.5% 5.7 163
that a signicant fraction of the back bombarding electrons exhibit kinetic energy higher than 100eV; this is large enough to produce high energy secondaries. Figures 7.3 and 7.4 provide a contour distribution along the cathode length of the electron back bombardment energy taken at the same time intervals as in Figure 7.2. Figure 7.5 presents the RF voltage between two adjacent anode vanes during one RF cycle. The red stars denote the instants when the electron back bombardment energy has been calculated. In Figures 7.2, 7.3 and 7.4, the electron back bombardment energy distribution has been obtained over a very short accumulation time of 75fs, which is about 1/700 of one RF cycle. It is desirable to evaluate the electron back bombardment energy distribution over a longer period. Figure 7.6 shows the electron back bombardment energy distribution measured over 10RF cycles when the magnetron is in stable mode oscillations. It can be noticed from this gure that maximum back bombardment occurs at 30eV and that 57% of the total back bombarding electrons have kinetic energy in excess of 100eV. This again suggests a possible mechanism for generating in large numbers high energy secondary electrons. The above results indicate that in the real operation of a magnetron, the initial velocity, or the initial voltage of a fraction of the cathode emitted electrons can be up to several hundred volts. But in our current modelling, the initial voltage of the emitted electrons is only a few volts for either space-charge-limited or beam emission model. This justies further investigation using high initial voltage of the emitted electrons.
157
700
Back bombarding energy distribution along the cathode length (dn/dke/dz) at t1=39.930ns
x 10 1.8
10
600
1.6
1.4 500 1.2
ke (eV)
400 1
300 0.8 200 0.6
100
0.4
1.5
0.5
0.5
1.5
0.2
z (mm)
(a)
Back bombarding energy distribution along the cathode length (dn/dke/dz) at t2=t1+0.14RFperiod
10
700
x 10 1.8
600
1.6
1.4 500 1.2
ke (eV)
400 1
300 0.8 200 0.6
100
0.4
1.5
0.5
0.5
1.5
0.2
z (mm)
(b)
Figure 7.3: The back bombardment energy distribution along the cathode at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a) t1=39.930ns (b) t2=t1+0.14RFperiod
158
700
x 10 1.8
10
600
1.6
1.4 500 1.2
ke (eV)
400 1 300 0.8 200 0.6
100
0.4
1.5
0.5
0.5
1.5
0.2
z (mm)
(a)
10
700
x 10 1.8
600
1.6
1.4 500 1.2
ke (eV)
400 1 300 0.8 200 0.6
100
0.4
1.5
0.5
z (mm)
0.5
1.5
0.2
(b)
Figure 7.4: The back bombardment energy distribution along the cathode at dierent time intervals. Vdc=5.87kV, Ia=5.14A. (a) t3=t1+0.28RFperiod (b) t4=t1+0.42RFperiod
159
Voltage oscillation between two adjacent anode vanes

8 6 4 2
Volt (kV)
0 -2 -4 -6 -8
39.89 39.9 39.91 39.92 39.93 39.94 39.95 39.96 39.97 39.98 39.99
time (ns)
Figure 7.5: The RF voltage between adjacent anode vanes within one RF cycle. Vdc=5.87kV, Ia=5.14A. Electron back bombardment energy distributions shown in Figures 7.2, 7.3 and 7.4 have been calculated at points indicated by a red star .
Figure 7.6: The electron back bombarding energy distribution measured over 10RF cycles.
7.3
High Initial Cathode Emission Velocity
Since a direct simulation of cathode secondary emission is not yet possible, an equivalent method has been developed to investigate the eect. The space-charge-limited emission model has been modied with an articially high initial velocity spread in order to simulate high energy secondary emission. Basically the velocity distribution of the emitted electrons is Maxwellian, multiplied by the sine-of-polar angle from vertical.
160
A number of simulation runs have been carried out using various peak values. For simplicity, the average emission voltage is used to evaluate the initial velocity of emission. Table 7-B: Magnetron performance as a function of the initial velocity of the emitted electrons Eective input voltage (kV) Anode current (A) System RF energy (1 106 J) Output power (kW) Overall eciency Back bombarding power (kW) Average cathode emission power (kW) Average cathode emission voltage (V) Average back bombarding voltage (V) Vane voltage (kV) 5.70 3.60 52.11 13.77 57.0% 0.807 0.040 4.0 130 5.18 5.70 3.86 49.25 13.18 50.9% 1.49 0.709 54.5 168 5.05 5.69 4.03 48.2 12.92 47.9% 1.95 1.25 87.3 197 5.02 5.68 4.29 47.32 12.72 44.4% 2.68 2.14 134.6 241 4.95 5.68 4.45 46.58 12.52 42.1% 3.24 2.81 167.1 273 4.92
Table 7-B summarizes the simulation results. The anode voltages is kept constant at around 5.70kV. When the average initial velocity is increased from 4eV to 167eV, the anode current grows from 3.60A to 4.45A; the output power decreases from 13.77kW to 12.52kW; therefore the overall eciency decreases from 57.0% to 42.1%. The explanation is that when electrons are emitted with high initial velocity, some synchronous electrons loss their excess kinetic energy by striking the anode. This reduces the overall eciency, even though there is more DC power being supplied. Figure 7.7 compares the electron cloud when the initial velocity is 4eV and 100eV. With the average initial velocity of 4eV, the electron spokes are well dened; when the initial velocity is increased to 100eV, the spokes become blurred and get bigger, indicating a wide energy spread among asynchronous electrons. Another phenomenon to note is that when the cathode emission velocity is increased, the cathode emission power signicantly changes from 40Watts to 2.81kW. In magnetron operation, the only power source is the DC power supply, so the extra emission power is thought to come from the back bombardment, which has been shown to produce high energy secondary emission. Thus, by increasing the cathode emission voltage, one can reduce the computed
161
(a)
(b)
Figure 7.7: Electron clouds in stable mode oscillations with dierent cathode emission initial velocities. (a). The average initial velocity of cathode emission is 4eV. (b) The average initial velocity of cathode emission is 167eV. eciency of the magnetron signicantly and bring it close to the experimental result.
7.4
Summary
By investigating the energy spread of the back bombarding electrons, it is shown that the X band magnetron is capable of producing high energy secondary electrons. Simulations have then been carried out with an articially increased emission voltage in order to investigate the eect of high energy secondary electrons. It is found that when the initial voltage is increased from 4eV to 167eV, the overall eciency of the X band magnetron
162
drops from 57.0% to 42.1%. It is concluded that the high energy secondary emission is the main reason for the low eciency of the X band magnetron.
References
[1] D.Chernin, A.Drobot, and M.Kress, A model of secondary emission for use in computer simulation of vacuum electronic devices, in Eelctron Devices Meeting, ser. Technical Digest., International, Dec 1993, pp. 773776. [2] J.Rodney and M.Vaughan, A new formula for secondary emission yield, IEEE Transactions on Electron Devices, vol. 36, no. 9, pp. 19631967, Sept. 1989.
Chapter 8
Conclusions and Future work

8.1 Conclusions
The research presented in this thesis comprises 3D computer modelling of magnetrons using an FDTD/PIC code called MAGIC. The modelling process has been vastly extended to include realistic performance details of two e2v technologies S and X magnetrons, all in close collaboration with e2v engineers. Considerable insight into the characteristics of magnetron operation have been gained through modelling. The following is a summary of the key items considered in this work. 1. Investigation of the melting of the output coupling rods in the S band MG5193 magnetron. It is found that when the glass dome is replaced by a ceramic one, the impedance match is improved, therefore the RF power reection cannot be the reason for the melting of the rods. Investigation of the RF current distribution along coupling rods shows that the RF current in the input coax is much higher than that on the coupling rods, therefore RF current overheating can also be eliminated as the cause for 163
Chapter 8. Conclusions and Future work
164
rod melting. Preliminary work on a multipactor eect indicates that electron bombarding of the coupling rods and of the central support could last long enough to cause melting. 2. Analysis of the frequency spectrum of an S band anode (e2v MG5193). A computer model for an S band anode has been optimised by considering the bare anode and the strapping system. When a tuner is included, simulation spectrum agrees well with experimental result. 3. Simplication and approximation rules in setting up a computer model for the whole magnetron. There are two important aspects in building a computer model. One is to t the model to the codes underlying symmetry in order to maximise computing eciency. Another is that when an approximation cannot be avoided, it should be kept away from the beam/wave interaction region. Based on these two rules, a thick tapering vane model is selected for the X band (MG5241) magnetron as being the best t. 4. Investigation of a high overall eciency of the X band magnetron model. In hot test modelling two dierent cathode emission models have been used, the space-charge-limited model and the beam emission model. Both models yield higher overall eciency than that of a real magnetron. Possible reasons for the eciency discrepancy have been investigated, including the end hats emission, the output coupling circuit, the copper surface losses and the rising rate of voltage, but all have failed in explaining the eect. However, the above investigations have revealed some interesting features of operation of magnetrons, such as the role of the end hat emission, the relationship between the output coupling loop area and the Qext , the oscillation start-up and its dependence on the rate of rise of anode
165
voltage. 5. Investigation of non-uniform cathode emission and back bombardment distribution in the X band magnetron. Both non-uniform cathode emission and back bombardment distribution have been demonstrated for the rst time by computer modelling; this indicates the inherent complexity of the cathode emission process when secondary emission is included. Investigation of uniform and non-uniform or real magnetic eld distribution clearly shows that the Br component plays an important role in the cathode emission and back bombardment distribution pattern along the cathode. Articially created cathode emission distribution of Gaussian shape has been used to show that different emission proles can cause marked dierences in the overall eciency. 6. Investigation of the eect of high initial velocity of emission. Investigation of the energy spread in back bombardment electrons hitting the cathode indicates that high energy secondary emission is present in the X band magnetron. Since a direct simulation of cathode secondary emission so far proved to be numerically unstable, simulation with articially increased initial electron velocity has been substituted. The results show that a high initial velocity of cathode emission reduces the overall eciency signicantly. A clear conclusion is that a high energy secondary emission is capable of substantially lowering the eciency of a magnetron. As mentioned in Chapter 1, it is hoped that our computer modelling should help to substantially reduce the time and cost of development of new devices. In our project, a range of modelling techniques has been developed towards this objective. Though there are still limitations in current magnetron modelling, such as proper simulation of secondary electrons and the modulator/magnetron interaction, at least we were able to show that computer modelling is an eective and reliable way to investigate magnetron operation.
166
8.2
Future Work
At this stage various details of operation of the X band magnetron have been investigated and the reason for the low operating eciency has been deduced. Future work should be concerned with magnetron/modulator interaction and with further development of secondary emission model. 1. Magnetron/Modulator interaction The start-up of oscillations is important in magnetron operation, it is then that mode competition and stability problems occur. In section 5.9, the magnetron start-up of oscillations has been investigated, preliminarily by simulating a slow RRV. But this investigation cannot provide adequate information because the inuence of the modulator is not included in the computer model, even though the modulator plays an important role in magnetron operation. The interaction between the magnetron and the modulator determines the rate of rise up of anode voltage. Investigation of the magnetron/modulator interaction process is therefore strongly recommended to be carried out in the future. 2. Cathode emission Our current investigation of cathode high energy secondary emission has shown that it reduces the magnetron eciency signicantly. To conrm this observation, a magnetron that does not depend on secondary emission from the cathode (e.g. BM75L) should be modelled to see if the computer predicted electronic eciency agrees with measured values. At this stage of work, the energy distribution of cathode emitted electrons is articial, whether it is maxwellian or uniform. It is necessary and of considerable interest to investigate the real energy distribution of the cathode emission and its possible eect on magnetron operation.
Appendix A
Denitions of Qs and Circuit Eciency

The denitions of the external Q, loaded Q and internal Q of a resonant cavity are given below [1, 2] Qext = 2f0 W Pout (A.1)
Qload = 2f0
W Pout + Ploss W Ploss
(A.2)
Q0 = 2f0
(A.3)
where W is the systems stored RF energy, f0 is the resonant frequency, Pout is the output power and Ploss is the cavity copper surfacer loss. Substituting equations A.1 and A.2 in equation A.3, the Q0 can be expressed in terms of Qext and Qload by writing Q0 = Qext Qload Qext Qload (A.4)
167
Appendix A. Denitions of Qs and Circuit Eciency
168
When Qext and Qload are obtained using computer simulation, Q0 can then be obtained from A.4. The circuit eciency of the cavity is < 100% due to copper surface losses and it can be expressed in terms of the Qs.
cold = or cold =
Qload Qext 1
(A.5)
1+
Qext Q0
(A.6)
References
[1] S.Y.Liao, Ed., Microwave devices and circuits. Prentice Hall Internationsl, 1990. [2] J.D.Kraus and D.A.Fleisch, Eds., Electromagnetics with Applications. WCB McGrawHill, 1999.
Appendix B
Authors Publications
Journal Papers
1. L. Ma, X. Chen, M. Esterson and P. Lindsay, The eects of the initial velocity of cathode emission in an X band magnetron, in preparation .
Conference Papers
1. L. Ma, X. Chen, M. Esterson and P.A. Lindsay, Computer Modelling of the Possible Multipactor Eect in the Output System of an Experimental S Band Magnetron, 3rd IEEE International Vacuum Electronics Conference, Monterey, California, USA, April 2002, pp. 246. 2. L. Ma, X. Chen, M. Esterson, P.A. Lindsay, P. Burleigh and M. Brady 3D Computer Modelling of an X band Magnetron, in 4th IEEE International Vacuum Electronics Conference, Seoul, Korea, April 2003, pp. 281. 3. L. Ma, X. Chen, M. Esterson, P.A. Lindsay, P. Burleigh and M. Brady, 3D cold test modelling of an S band magnetron, in 9th International Conference on Microwave & RF Heating, Loughborough University, UK, Sept 2003, pp. 233. 169
Appendix B. Authors Publications
170
4. L. Ma, X. Chen, M. Esterson, P.A. Lindsay, P. Burleigh, K. Saleem and D. Wilson, Investigation of an X band experimental magnetron through MAGIC modelling, in 31st IEEE International Conference on Plasma Science,Baltimore, Maryland, May 2004, pp. 177. 5. L. Ma, X. Chen, M. Esterson, P.A. Lindsay, P. Burleigh, K. Saleem and D. Wilson, Investigation of the non-uniform back bombardment of an X band magnetron, in MAGIC user group meeting, 31st IEEE International Conference on Plasma Science,Baltimore, Maryland, May 2004.
Project reports
1. L. Ma, X. Chen, P.A. Lindsay and M. Esterson, Computer Modelling of X Band Magnetron, 1st Oct. 2003 30th Oct. 2004. 2. L. Ma, X. Chen, P.A. Lindsay and M. Esterson, Computer Modelling of X Band Magnetrons, 1st Oct. 2002 30th Oct. 2003. 3. L. Ma, X. Chen, P.A. Lindsay and M. Esterson, Computer Modelling of S/X Band Magnetrons, 1st Oct. 2001 30th Oct. 2002.

Magnetron - Thesis - LiliMA

Uploaded by

Copyright:

Available Formats

Magnetron - Thesis - LiliMA

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Magnetron - Thesis - LiliMA

Uploaded by

Copyright:

Available Formats

What is the research work primarily based on?

What is the research work primarily based on?

What insights were obtained about the S band MG5193 magnetron?

What insights were obtained about the S band MG5193 magnetron?

3D Computer Modeling of Magnetrons

Department of Electronic Engineering Queen Mary, University of London United Kingdom

Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Operating Principle of Magnetrons

2.1 2.2 2.3

Emission Models in MAGIC Code . . . . . . . . . . . . . . . . . . . . . . 48

3.4.1 3.4.2 3.4.3

The space-charge-limited emission . . . . . . . . . . . . . . . . . . 48 The beam emission . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 The secondary emission . . . . . . . . . . . . . . . . . . . . . . . . 49

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 4 Modelling of an S band Magnetron 4.1 52

Investigation of the tuner . . . . . . . . . . . . . . . . . . . . . . 73

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 5 3D Modelling of an X band Magnetron 5.1 5.2 77

5.4.2 5.4.3 5.5

The beam emission model . . . . . . . . . . . . . . . . . . . . . . . 95 The space-charge-limited emission model . . . . . . . . . . . . . . 102

5.6 5.7 5.8

Slow Rate of Rise of Voltage . . . . . . . . . . . . . . . . . . . . . . . . . . 133

6.3.2 6.4 6.5

7 The Eect of Cathode Emission Voltage 7.1 7.2 7.3 7.4

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 8 Conclusions and Future work 8.1 8.2 163

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 167

Appendix A Denitions of Qs and Circuit Eciency

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 Appendix B Authors Publications 169

= 4 ; Red colour rep-

ABCS CFA CW FDTD FFT HPM PIC RF RRV TWT

Fundamentals of microwave tubes

Dierent microwave tubes and their applications

Motivation and Contributions

Contribution to industrial design

Organization of the Thesis

[4] S.Y.Liao, Ed., Microwave devices and circuits.

Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

Negative Resistance Magnetron

Chapter 2. Operating Principle of Magnetrons

Principle of Transit Time Oscillation

Chapter 2. Operating Principle of Magnetrons

Kinetic motion of electrons in a planar diode

Figure 2.3: Electron trajectory in a planar crossed eld diode

Chapter 2. Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

2 2 vy0 + (A2 vx0 )

Chapter 2. Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

Kinetic motion of electrons in a cylindrical diode

Rearranging Eq 2.18 results in the following equation d(r2 eB d )= rdr dt m

Chapter 2. Operating Principle of Magnetrons

Chapter 2. Operating Principle of Magnetrons

from the DC eld to the RF eld via the electrons.

Resonant Cavity Magnetron Operation

Chapter 2. Operating Principle of Magnetrons

Equivalent circuit and oscillation modes

Chapter 2. Operating Principle of Magnetrons

Figure 2.10: Equivalent circuit of an eight-cavity magnetron [2]