Patent Grant
7446601
The field of the invention is that of high-frequency electronic amplifiers intended for electromagnetic radio frequency reception and generation, both tuned and broadband. Applications also include digital signal processing and general purpose computing.
The twentieth century opened with the discovery of radio wave transmission by Marconi. World War II heralded the emergence of radar. The 1960's witnessed the launching of satellites. The 1990's saw the proliferation of commercial wireless data communications. These four events signaled epochal moments in history, opening up entirely new ranges of the electromagnetic spectrum for revolutionary applications such as radio, television, long-range surveillance, satellite communications and computer networking. The key components that made these advances possible were the development of electronic components capable of detecting, amplifying and re-transmitting high-frequency electrical signals: the point contact diode, the vacuum tube triode, the semiconductor transistor, the traveling wave tube, the integrated circuit. Each had—or is having—its moment and was superceded by a newer technology as demand for higher performance increased.
Today, RF communications, radar and other applications are pushing well into the high gigahertz region, as much as 200 GHz or more. Even home wireless networking and simple cordless telephones are operating at over 5 GHz, a domain once reserved to only the military a few short decades ago.
The key components that made these advances possible are high-frequency devices: transistors with current-gain-bandwidth product fT>200 GHz, LNAs with high linearity (IIP3), emerging power transistors made of SiC and GaN, and the venerable traveling wave tube (TWT). Many applications such as digital radio and military surveillance today are limited by the power or bandwidth achievable in a conventional semiconductor, or by the size, weight, cost, power and distortion products of the TWT. Space electronics is also limited by the radiation hardness and reliability of semiconductors. Military applications also require greater bandwidth, with tuning ranges exceeding 10:1 at frequencies up to 100 GHz.
Semiconductor Amplifiers
Despite the ubiquity of modem semiconductors, they suffer several limitations for the highest frequency RF applications. First, transistor breakdown voltage must be reduced significantly to achieve the necessary bandwidth, often to a volt or two or less. This severely limits the power they can generate, especially when low distortion is required. More fundamentally, semiconductors have an upper bandwidth dictated by the physics of the semiconductors: the maximum carrier velocity, especially, the saturated electron velocity. Current art places a limitation of perhaps 400 GHz fT on III-V compound devices such in InP, GaAs, InAs, and a theoretical limit of approximately 1 THz is dictated by the velocity of current-conducting carriers (electrons) in any semiconductor crystal. Practical applications such as an RF low-noise amplifier (LNA) usually can only operate at no more than 1/10 of the fT. Furthermore, to operate at speeds of 100 GHz or more (as in an RF LNA) requires considerable power. At this time, there are almost no semiconductor power amplifiers capable of operating much above 10 GHz, leaving the entire field of high-power antennas to the field of vacuum electronic devices, such as the TWT, which are orders of magnitude more expensive and bulky. Semiconductor amplifiers are also extremely sensitive to radiation induced degradation and failure in space environments.
TWTs and Other Traditional Vacuum Electronic Devices
TWT's offer direct RF amplification with power gains exceeding 40 dB, frequency of amplification over 100 GHz, and bandwidth of more than 2 octaves in specialized devices. The drawback is they are large, very expensive, power consumptive, noisy and introduce significant signal distortion. Size can vary from 10 cubic inches in very high frequency devices (˜100 GHz). Cost can be $10,000 in a typical device to as much as $100 k in a space-rated device. Minimum power consumption can be hundreds of watts even in a low power device. Noise figures are typically 40 dB, compared to as little as 1 dB in a semiconductor LNA. Distortion products for wideband operation can be similarly oppressive, restricting their use to power amplification. TWTs can in principle operate at frequencies approaching or exceeding 1 THz, but become extremely inefficient at these frequencies (as little as a few percent), and very hard to build because of the micron-sized dimensions. Machining tolerances of a few nanometers become necessary, and waveguide losses become dominant, since a long waveguide (such as a helix, serpentine, or many coupled cavities) has unavoidable ohmic sidewall losses.
Many applications today are severely constrained by the lack of high-frequency performance in available amplifiers. For example, an emerging application is wireless networking in dense urban environments. The demand for communication bandwidth on network channels is already exceeding 1 Gbps, yet the limits of present-day carrier frequencies is only about 5-10 GHz. As is known in the art, the carrier frequency must normally be much higher than the data rate—100 times higher or more. For example, 2.4 Ghz carriers typically provide 10 Mbps data rates or less in the well-known “Bluetooth” system (sometimes called “802.11b”). 1 Gbps data rates imply a carrier of at least 100 GHz or more.
The problem is exacerbated in dense urban environments, especially around large office buildings. Current technology increases the spectrum capacity by limiting the range of a limited number of sub-channels (which may be spectrally broad in spread spectrum or Ultra Wideband (UWB) systems). No more than a few hundred low-bandwidth (10 Mbps) channels can typically be made available within a short geographic radius of a few hundred meters. In an urban environment with thousands of network connections within a single building and other buildings in close proximity, it can be seen that there is a hard limit, indeed, on the number of network connections and the aggregate data transfer rate that is possible per cubic mile.
Hard-wired networks traditionally overcome this density limitation, but they are difficult to install and very expensive to retrofit an existing structure. Wireless systems have recently proliferated (based on the 802.11b standard, among others) using higher carrier frequencies, but for higher bandwidths and link densities, few or no solutions exist today.
As mentioned, semiconductor amplifiers cannot operate much above 100 GHz with any gain at all, and are very power inefficient. TWT amplifiers also cannot operate efficiently much above 100 GHz (though they are much better), but are prohibitively expensive for most applications. What is needed is a solution that offers the size and economies of scale of semiconductors, and the gain and frequency performance of TWTs, with power efficiency and linearity greater than both. Thus, it can be appreciated that there is a real demand for a low cost, efficient millimeter wave to sub-millimeter wave RF technology.
Related Art
As will become apparent, the present invention relates to microminiature electron beam devices applied to RF amplification and signaling, particularly those that operate in the millimeter to sub-millimeter wave region (50 GHz to 2 THz). Similar inventions have claimed advances that might operate in this region. For example, Manohara et al (ref. 11) have published work on sub-millimeter “nano-klystrons” based on many of the elements described herein for the present invention: semiconductor fabrication, MEMS and electron gun construction. An impressive development, it nonetheless suffers many deficiencies, including narrowband tuning, and relatively slow response to signal modulation, because of the resonant cavities inherent in the method. The nano-klystron also lacks integral phase and polarization control, which are highly desirable features of any RF power device intended for transmission purposes, yet expensive and bulky to provide as separate elements.
U.S. Pat. No. 5,497,053 issued to Tang, et al shows a deflection amplifier (or “deflectron”) that purports to offer wideband amplification, but suffers low gain, relative to the invention here, because the detrimental effects of space charge repulsion limit the maximum beam current. Furthermore, such beam current as Tang et al. can generate creates significant heating losses. Tang et al. also does not offer integral solutions to antenna coupling, phase and polarization control.
U.S. Pat. No. 3,725,803 issued to Yoder predates Tang et al., and teaches an electron beam driven P-N junction in a push-pull detector arrangement. Yoder does not suggest his method provides extra gain through the beam interaction with the semiconductor diodes, though it may be inferred. However, such extra gain as may be provided will be modest, and the apparatus does not lend itself well to microfabrication. Further, Yoder does not adequately elaborate on how his method will provide linear gain, and it may be inferred from the description that high linearity will not be achievable. For example, Yoder does not describe means for achieving a substantially uniform electron beam. Yoder does not indicate how the detection apparatus can be constructed so as to achieve a linear output from a uniform beam, and in fact, it achieves just the opposite. Thus, Yoder's arrangement is seriously deficient in regard to actual construction of a deflectron having linear response.
Chang, Muray, Lee, MacDonald (see references) have described “microcolumn arrays” of miniature electron guns and elements thereof for the purpose of improved electron beam lithography in semiconductor fabrication, yet they have not explored the potential of employing microcolumn arrays in amplifiers, RF generators or computing.
U.S. Pat. No. 3,922,616 issued to Weiner describes one way to provide gain from an electron beam, by means of an electron bombarded semiconductor. This is commonly called an “EBS” amplifier. The method is based on a p+-i-n+ diode with an intrinsic “i” layer. Kitamura et al (1993, ref 12) explicitly describes an EBS amplifier based on a silicon Schottky diode, but do not employ deflection means. U.S. Pat. No. 4,410,903 issued to Weider describes a heterojunction EBS amplifier based on InGaAs and InP compounds to improve the speed and bandwidth, but these suffer from lack of compatibility with low-cost silicon microfabrication. All three disclosures provide means to improve the gain of an electron beam deflectron amplifier over that of Yoder or Tang et al.
U.S. Pat. No. 5,592,053 issued to Fox et al. describes a variation on the EBS amplifier that provides gain via an electron-beam activated diamond conductor. U.S. Pat. No. 5,355,380 issued to Lin describes a related e-beam excited diamond switch for millimeter wave generation that depends on modulating the current of an electron beam. The principle disadvantage in either is that high beam energies are required with a diamond detector material. This causes extra heating losses, reduced efficiency, and severely limits the deflection gain. Another disadvantage is that Fox does not employ a precision e-beam forming device, such as a microcolumn. Another disadvantage is the difficulty of fabricating high-quality diamond films. Again, beam deflection is not incorporated in the gain mechanism.
A principle disadvantage of following Tang et al., Yoder, or Weiner is that they rely on high current electron beams, which are difficult to focus in low-energy beam systems because of the space charge effect. Lack of focus reduces amplifier gain, decreases bandwidth and increases amplifier distortion. Fox overcomes this with a high energy beam. High current and high energy beams are antithetical to microfabricated electron beam systems. High current and high energy beams dissipate excess anode heating power. High voltage beam circuitry is susceptible to destructive arcing and requires high voltage power supplies, which are difficult to build, bulky and power consumptive, and not amenable to microfabrication.
U.S. Pat. No. 4,328,466 issued to Norris et al describes an EBS amplifier that operates with a sheet beam to disperse the space charge and permit higher beam current, but sheet beams still suffer substantial space charge effects, thereby limiting the beam current and amplifier gain. Norris' amplifier suffers from the complexity of a distributed architecture to achieve high frequency broadband and high power operation, making it unsuitable for low-cost microfabrication.
Low current beams are desirable, yet they reduce amplifier gain. It may be appreciated that there is a need for higher current, but low energy electron beam systems for microfabricated high speed amplifiers.
U.S. Pat. No. 5,041,069 issued to Seiler, U.S. Pat. No. 6,177,909 issued to Reid, and Froberg (ref. 8) have constructed photoconductive antennas which employ semiconductor antenna excitation to generate THz radiation, yet they suffer from uncontrolled wideband transmission, no phase or polarization control, and require complex laser activation with slow pulse repetition rates. As will be seen, the present invention advances the art over all these examples of prior art, simultaneously providing, in different embodiments, controlled wideband modulation, high gain, RF transmission, phase and polarization control.
It will be appreciated in the following description and appended claims that the present invention combines many of the advantages of prior art while overcoming the deficiencies in a novel arrangement, to thereby achieve RF amplifier embodiments possessing higher gain, faster operation, less distortion and lower power consumption. These benefits accrue in almost any RF receiver or transmitter application including wireless networking and antenna beamforming, frequency multiplication, high-speed digital logic and computing.
The disclosure to follow provides method and apparatus for wideband RF amplification that solves the shortcomings of both semiconductor and conventional vacuum electronic amplifiers. It can simultaneously provide high frequency of operation (exceeding 1 THz), wide bandwidth (up to 10:1 frequency range or more), high power gain (60 dB or more), linear operation and low noise in a size comparable to an integrated circuit (several cubic millimeters) with similar cost and lower power consumption. What is disclosed is a hybrid of semiconductor and vacuum electronics. It can be constructed using standard semiconductor fabrication techniques. There are many embodiments of the same basic principle:
A first embodiment, amplifies a voltage signal and generates a highly linear current output by exciting a detector with a deflection modulated electron beam. The method includes a two-dimensional array of electron guns to generate beamlets, a distributed beam deflection apparatus in each electron gun array to provide high deflection gain to re-direct the electron beam in response to a voltage signal, and an electrostatic lens system to create a shaped electron beam spot where the beam strikes a current amplifying detector. The detector in one form comprises dual segments to differentially collect the beam in proportion to the deflection. Each segment converts a collected proportion of the beam to an electrical current, amplifies it, and couples it to an output network.
In the most linear configurations, the dual detector segments are triangular and oriented in opposition to respond to a narrow rectangular beam spot; for the highest linearity, the space separating the segments distorts the shape of the segments from pure triangularity. In the fastest configuration, the segments are rectangular and the beam spot is rectangular to give a configuration that has the smallest detector.
One construction is by semiconductor manufacturing processes including wafer bonding.
In another embodiment the detector is a Schottky diode made of a germanium-silicon heterostructure. In another, the detector is Schottky diode made from a low-ionization material such as InAs or InSb. In either case, the detector provides beam-generated cascade gain and avalanche multiplication by a sandwich of semiconductor between a beam contact and an output contact.
In another embodiment, the beam shaping is achieved with a shaped array of electron guns that are imaged on the detector by the electrostatic lens system.
In another embodiment, the lens system is a doublet of a retarding and accelerating lens constructed from planar electrodes in the drift cavity. One configuration comprises a circular disc electrode enclosing the electron gun array to generate the retarding lens, and a circular electrode enclosing the detector to generate the accelerating lens. The drift cavity is enclosed by a cylindrical drift can with the electron gun array centered in one end, and the detector centered in the other. Planar donut electrodes may enclose the first and second disc electrodes in their respective planes.
A variation achieves beam shaping with an astigmatic electron lens system comprising multiple shaping electrodes disposed around the exit plane of the electron gun array, and the electrodes are subject to different applied voltage potentials.
All embodiments employ electron gun construction comprising field emission cathodes, cathode gating, a plurality of focusing and aperture electrodes, and deflection plates. In one variation, the plurality of focusing and aperture electrodes is increased in number to reduce the diameter of the gun column (relative to the beam axis). In another a beam blanking deflector is incorporated for pulsed operation.
Another embodiment incorporates current control in every electron gun, comprising a ballast resistor to sense the cathode current and an amplifier to compare the ballast voltage against a reference, thereby generating an error signal that is applied to the cathode gate electrode.
In another embodiment, offset centering apparatus keeps the beam centered on the detector with a control loop comprising an integrator generating an offset correction signal in response to the beam offset as measured at the detector. A variation employs independent detector segments to measure the offset.
Another embodiment provides true time delay shifting by means of apparatus to adjust the energy of the electron beam and thereby the drift time through the drift cavity. One variation adjusts the potential of the detector plane, and in a configuration that improves the focusing, augments the cylindrical drift can electrode with a consecutive series of ring electrodes to approximate the fields potentials generated by a much larger drift cavity. In another variation the acceleration energy of the electron gun achieves the time delay control by augmenting the construction with a plurality of DACs coupled to deliver precise electrode focusing voltages for every time delay command. A further variation augments this arrangement with an analog-to-digital converter to couple a digitized measurement of the control gate with the time delay command, to generate electron gun focusing electrode potentials that are corrected for varying gate voltages in response to a current control loop.
Yet another embodiment achieves frequency multiplication. One configuration uses a multiplicity of detector segments in a linear array that provides programmable multiplication. Another configuration achieves lower inharmonicity by using a circular detector in a two-dimensional arrangement of segments similar to the slices of a pie, and uses horizontal and vertical electron gun deflection.
Another embodiment of frequency multiplication employs a single shaped detector segments and a shaped beam spot. The sweep of the shaped beam spot across the edge of the segment generates strong harmonics. The variations include triangular beam spots on rectangular detectors, rectangular beam spots on triangular detectors, rectangular beam spots on quadratically shaped detectors, and so forth, to generate second, third, fourth and so on harmonics.
Another embodiment, is a mixing device comprising a square detector made of four equal square segments arranged symmetrically around axes X and Y, a square beam spot disposed to sweep in X and Y directions in response to a first signal applied to an X deflection apparatus and a second signal applied to a Y deflection apparatus.
Another embodiment is a combinational logic device comprising a plurality of N deflectors X1, X2, . . . XN, a corresponding plurality of deflection signals V1, V2, . . . VN, and detectors D1, D2, . . . DM, each individually positioned to correspond to a logic state of the deflection vector V1 . . . VN. Some of the deflectors XN are oriented for horizontal beam deflection and some of the deflectors are oriented for vertical beam deflection to improve the degeneracy of states and the compaction of the system. A further extension of the concept employs deflectors of different geometries to achieve gray coding for a further reduction in the state degeneracy.
Another embodiment, is a method of exciting electromagnetic radiation by incorporating an antenna, such as a dipole, patch or horn. Some variations provide a selectable polarization dipole or patch by means of X and Y deflection, multiple detector segments and/or multiple addressable feedpoints.
Another radiating embodiment, excites a waveguide. The waveguide may be rectangular or circular. The excitation can be single or dual polarization to excite desired waveguide modes. The dual polarization device consists of four segments, with two opposing segments connected across a diameter of the waveguide, and the other two opposing segments connected across an orthogonal diameter of the waveguide. This may be augmented with a selectably shaped beam spot for selectable polarization, with a rectangular spot shape spanning two opposing detectors and a motion that sweeps between the two detectors. Any of the waveguide embodiments may be coupled to the feed of an antenna horn.
Another embodiment merges the detector and antenna in a single structure to make a novel radiator that can simultaneously generate harmonics and controlled phase and polarization. In a variation, multiple, independently steerable beams are employed to enhance the diversity of the output radiation.
Another embodiment, is constructed as an array of amplifiers according to any of the other embodiments, thereby achieving transmit antenna arrays, receive antenna arrays, T-R arrays and signal combining networks.
Another embodiment, is a crossbar matrix comprising a plurality of N independent electron guns, a plurality of M detectors and crossbar addressing means. Each electron gun includes independent X and Y deflectors, and receives N digital input signals and N X and Y offset control signals for addressably configuring the matrix. The crossbar addressing means comprises a plurality of DACs under the control of a processor or ROM.
An extension of the crossbar matrix further includes free-space photonic I/O comprising a photonic input array, an input lens system, a photodetector array, a laser diode array, an output lens system, and an output photonic coupling array. The lens system images the photonic input array on the photodiode array. The photodiode array electrically couples individual photodiodes to individual electron guns to transmit the signals to addressed detector outputs. The laser diode array electrically couples individual laser diodes to individual detectors. The photonic I/O can be provided by fiber optic bundles
Another embodiment, is a multiprocessing compute engine comprised of a crossbar matrix coupled to a plurality of processor elements.
Overview
Amplifier 10(1) operates by (1) emitting a composite electron beam (“e-beam”) 110(1) (consisting of electron beams 120 emitted from individual electron guns that are not shown in this figure), (2) deflecting composite beam 110(1) by applying voltage signal 140(1) to deflector apparatus 130(1), (3) generating output currents I1 180(1) and I2 180(2) through the action of composite beam 110(1) impinging upon detector segments 150(1), 150(2) at beam spot 170(1), and (4) transmitting output currents 180(1), 180(2) into output network 190(1). By deflecting composite beam 110(1) with voltage signal 140(1), a physical change in position of beam spot 170(1) impinging upon segments 150(1), 150(2) generates changes in output currents 180(1), 180(2) that can be coupled to an output load such as a resistor, a transmission line, a waveguide, or an antenna.
The principle of operation may be understood as follows. Composite beam 110(1) sweeps back and forth in sweep direction 210 from detector segment 150(1) to detector segment 150(2) in response to voltage signal 140(1). Electron beams 120, and thus composite beam 110(1), carry an electrical current equal to the well-known electronic charge q times a number of electrons emitted per unit time. Voltage signal 140(1), applied across a gap within beam deflection apparatus 130(1) establishes an electric field E that subjects electrons in e-beams 120 to a transverse force F as they travel through the deflector. The force is described by the well-known law F=qE. At a maximum positive beam deflection, detector segment 150(1) may collect all of the impinging beam current; at a maximum negative deflection, detector segment 150(2) may collect all of the impinging beam current. Between these extremes of positive and negative deflection, each of detector segments 150(1) and 150(2) collects a proportionate amount of the beam current. For example, when composite beam 110 is centered, each of detector segments 150(1) and 150(2) may collect 50% of the beam current. If beam 110(1) is positioned to 70% of maximum deflection in the positive sweep direction (i.e., the X direction of
Other factors being equal (as explained below), a deflection of composite beam 110(1) may be proportional to voltage signal 140(1), and a beam current collected by either of detector segments 150(1), 150(2) may be linear in response to the change in position of beam 110(1). As shown in
Current Multiplying Detector
A gain of electron-beam amplifier 10(1) may substantially increase when detector segments 150(1), 150(2) amplify collected beam currents so that output currents 180(1), 180(2) are much greater than the beam currents alone. For example, a gain of 1000 or more is possible with a Schottky diode detector. In the embodiment of
In certain semiconductor devices such as, for example, a Schottky diode, cascaded electrons can further multiply through the well-known avalanche multiplication effect. A key parameter for avalanche multiplication is thickness t2 of avalanche multiplication layer 250. With an appropriate reverse bias voltage between cathode and anode contacts, a thickness t2 of 250 to 1000 angstroms can create a sufficiently strong electric field within the diode to accelerate conduction electrons, generating even more hole-electron pairs (only exemplary electrons 260 are shown, for clarity of illustration). An avalanche gain of 10 or more is practical, and with a cascade gain of 100, an overall detector gain of 1000 is possible.
Alternative Detector Types
Many types of current multiplying detectors are possible, including Schottky diodes, junction diodes, photoconductors, and even micro-channel plates (MCPs, or micro-dynodes). Junction diodes operate similar to a Schottky diode, and may support higher voltage operation, but may have lower bandwidth. Photoconductors typically operate by generation of hole-electron pairs by photons to modulate the conductance of a resistor; a photoconductor can be designed to respond to electrons instead, generating conduction electrons by cascade excitation. A photoconductor may lack avalanche multiplication to supplement the cascade gain, and thus have lower gain than a diode; photoconductors also typically have a less linear response when coupled to a load. MCPs generate gain by a photomultiplier effect, but require high bias voltages (thousands of volts), complex construction, and have long response times.
It can be appreciated that a Schottky diode detector is preferred where high gain and fast response is desired.
Schottky Detector
The exemplary Schottky detector 150 of
However, in other Schottky diode embodiments, other contact metals and semiconductor materials (such as, for example, InAs) may be used; in such embodiments a beam contact may be a cathode and an output contact may be an anode. A beam contact may connect with a bias voltage and the Schottky diode may be reverse biased to establish a field gradient between the beam contact and an output contact. The field gradient (1) accelerates carriers to generate avalanche multiplication of current, and (2) sweeps carriers rapidly out of the diode. The output contact is coupled to a load, for example a terminating resistor or a transmission line. When a beam contact is an anode, the bias voltage may be negative with respect to a load.
In detector 150 of
Germanium is a desirable cascade layer material because it has a high cascade gain relative to other materials, such as silicon or diamond. In germanium, one cascade electron (and a corresponding hole) are generated for each 2.8 eV energy for each beam electron. The cascade energy of silicon is 3.5 eV; the cascade energy of diamond is 5.5 eV.
A cascade process generally occurs within approximately 50 angstroms of semiconductor depth for a beam energy of several hundred electron volts; for higher energy beams, the cascade may spread deeper. Because conduction electrons in germanium have lower saturation velocities than conduction electrons in silicon, thickness t1 of cascade gain layer 230 is optimally thick enough to allow completion of the cascade process, but not thicker, so that a transit time of conduction electrons to avalanche layer 250 is minimized.
Avalanche layer 250 of detector 150 optimally achieves two goals: (1) it supports a high saturated electron velocity, for fast detector response, and (2) it produces efficient, low-noise avalanche multiplication. Avalanche multiplication occurs when conduction electrons accelerate in a high-field region of avalanche layer 250. Accelerated electrons may impinge upon electrons in a crystal lattice of avalanche layer 250, generating more hole-electron pairs. Electrons thus generated accelerate again, and the process repeats, generating an avalanche current. The electrons are collected by output contact 270; holes thus generated travel through cascade gain layer 230 and are collected by beam contact 220. Avalanche multiplication can easily provide current amplification of 5, 10, 20 or more. Practical limits to avalanche multiplication are set by leakage current across a Schottky junction, and electrical noise generated by the avalanche multiplication. Silicon is a desirable avalanche layer material because leakage currents in Si are lower than in many other materials. Ge—Si epitaxy is desirable because a large body of experience in reliably and inexpensively fabricating this material system exists.
Thus, a Ge—Si Schottky diode may provide high cascade gain, high avalanche gain, high speed response, and low leakage. With a 280 eV beam, a cascade gain may approach 100, avalanche gain may be 10, and a total detector gain may be 1000.
III-V Detectors
Fast, high gain detectors may also be constructed with epitaxial systems other than Ge—Si, and such detectors may offer suitable performance for some embodiments of electron-beam amplifier 10. For example, all of Ge, Si and diamond are indirect bandgap semiconductors; in each, the cascade ionization energy is approximately ⅓ of the bandgap. Materials with direct, small bandgaps may have lower ionization energies. For example, Indium Arsenide (InAs) has a direct bandgap of 0.35 eV. Indium Antimonide (InSb) has a direct bandgap of 0.17 eV. These bandgaps compare with 0.66 eV for Ge and 1.12 eV for silicon. Either of these materials from groups III and V of the periodic table (the “III-V” group), or a ternary compound (such as for example InAs1-xSbx) may have a cascade ionization energy of 1 eV or less, and provide a cascade gain of three times or more the cascade gain of Ge.
III-V materials have a zincblende crystal structure; epitaxial growth of this structure on a diamond lattice of silicon may be problematic or impossible. In order to overcome this difficulty, InAs or InSb layers could instead be mated with another III-V avalanche layer, such as Indium Phosphide (InP).
For example, one drawback of a Ge—Si detector 150 is that its breakdown voltage is limited by a Si layer thickness (e.g., thickness t2 of
Detector Beam Contact
For electrons to penetrate a beam contact of a detector (e.g., beam contact 220 of detector 150) and enter an underlying semiconductor (i.e., Ge cascade gain layer 230, or another material), the contact metal must usually be thin. At beam energies of 100 eV to 300 eV, beam contact layer 220 may be around 10 angstroms, or thinner. However, a thin contact layer may have a high sheet resistance, for example about 10 ohms per square of metal. Contact layer 220 may conduct all of the detector current, which may be 100 mA or more, and an ohmic voltage drop across contact layer 220 may substantially de-bias a low-voltage detector 150. Such de-biasing may have consequences such as loss of detector gain, slower response, and signal distortion.
A width of each of beam conductors 280 and 285 may be much less than a space between adjacent beam conductors. For example, if a space between beam conductors is 1 um, the beam conductors' width may be less than 0.1 um. Thus, in each of detectors 150(3), 150(4), 150(5) and 150(6), the proportion of area that beam 110 cannot penetrate the thick beam conductors may be less than 10%.
Amplifier Gain
An overall electron-beam amplifier gain depends on deflection and detection gain and an output coupling impedance. Beam deflector, drift cavity and detector geometries can generally be chosen to (1) provide a given level of gain and frequency response, and (2) achieve 100% differential beam collection at a maximum deflector input voltage. That is, in the example of
The factor of 2 reflects the fact that the signaling is differential. For example, when a beam current is 100 μA, a maximum peak deflector voltage drive is 1V and a detector gain is 1000, the differential transconductance gain is 100 mA/volt.
When output network 190 has a differential impedance Z0=100 ohms, the amplifier voltage gain Gv=gm Z0 equals 10.
A power gain GP is given by a ratio of an AC input power, VIN2/2RIN, to an AC output power,
where RIN is an input impedance and ROUT is an output impedance. With equal input and output impedances (e.g., 50 ohms), power gain GP may be 20 db or more. For larger input impedances, the power gain will be larger. For instance, for an input impedance of 1 kohm, a differential output impedance of 100 ohm and a voltage gain of 10, GP is 1000, or 60 db. High frequency systems typically do not utilize high input source impedances, but specialized systems may.
Other Detector Shapes and Beam Spots
High Speed Detector
Beam spot 170(2) permits high beam current by dispersing beam charge over an area, rather than a line. Detector segments 150(9) and 150(10) are small, with a height of segments 150(9) and 150(10) matching the height of beam spot 170(2), resulting in lower parasitic capacitance and wider bandwidth into an output impedance. Vertical slot 160(2) enables linear differential beam collection, with some sacrifice of linearity because of the small dimensions.
In a preferred embodiment, a height of a beam spot is slightly greater than a height of corresponding detector segments, placing current density variation substantially outside the detector segments.
When a beam spot 170 is larger than a corresponding detector segment 150, most of a beam current density variation may fall outside detector segment 150, where it has no effect. Thus, the region of the most uniform spot current density (i.e., an interior of a beam spot 170) sweeps across a vertical slot 160, enabling high linearity of differential beam collection. Any portion of an beam spot 170 that falls outside a detector segment 150 is collected by a passive metallic anode and returned to ground.
The linearity of the detector of
Unipolar Detector
In certain embodiments, a unipolar detector for driving only one output load may be preferred. Two versions are shown in
Many detector configurations are possible for optimizing electron-beam amplifier operation and performance. Certain configurations will be described in the embodiments that follow, but others will be evident to those skilled in the art, as depending on the basic elements of a shaped e-beam spot and a high-gain detector consisting of one or more segments that are shaped.
Linearity Requirements
One attribute of many amplifiers is linearity of amplification. The linearity of RF amplifiers is characterized by a quantity known as a third order input intercept point (“IIP3”) that characterizes an input referred power of distortion products (i.e., an output distortion power divided by amplifier gain) in relation to an input signal power. IIP3 measures the most significant distortion product, a third harmonic, referred to an input of an amplifier. Fully differential operation of certain systems may eliminate the second and other even harmonics, or at least reduce them well below the third harmonic; thus the third harmonic is a useful measure of total non-linearity, including 5th, 7th, and higher orders, as well as intermodulation products.
IIP3 describes the concept that a ratio of third harmonic output power to signal output power may increase in direct proportion to a first harmonic input signal power (this ratio is the same when referred to the input). That is, small input signals may generate small distortion products, since the non-linearities present in an amplifier are less significant for the small input signals, while large input signals may generate proportionately larger distortion products. The output of an amplifier operating with large signals may “clip” peaks in an output waveform (i.e., the peaks of amplified signals may not achieve appropriate values, because such values would exceed the maximum voltages available). Generally, for a 3 dB increase in small-signal output power, third harmonic output power increases by 9 dB. Even if the third harmonic output power is much smaller than a linear output power under small-signal conditions, if the input increases sufficiently, the third harmonic output power may equal and even exceed it. The point where input signal power and the third harmonic output power are equal is called the third order intercept point. IIP3 is usually an extrapolated figure of merit since linear output power cannot usually reach this level of power because of gain compression (i.e., where amplifier gain starts to diminish at high signal levels).
IIP3 is a valid figure of merit for many amplifiers in a restricted range of actual operation. Higher IIP3 implies better amplifier performance in rejecting distortion, even if an amplifier cannot operate at an input signal level indicated by an IIP3 specification. A well-made low noise amplifier (“LNAs”) may achieve an IIP3 of +5 dbm. That is, 3 mW input signal power will generate 3 mW of distortion (referred back to the input). Certain amplifiers may achieve an IIP3 of +20 dbm or +40 dbm, but these performance figures may not be achieved at frequencies that exceed a few hundred MHz. Generally, the higher an operating frequency and the wider an operating bandwidth, the more difficult it is to achieve a high IIP3.
Electron-beam amplifier 10 may achieve an IIP3 as high or higher than typical solid-state amplifier, such as +40 dbm or better, at frequencies of many GHz, and potentially up to K band (40 GHz) or higher. This may be shown by considering an input-referred effect of third harmonic distortion as described by a transfer function of the form y=x+a3x3:
Vin=V1+a3V13=V1(1+a32Z0P1) (1.4)
where Vin is an input voltage, V1 is an input deflection voltage corresponding to a maximum beam deflection (e.g., a peak sinusoidal input cos(ωt)), a3 is a third harmonic distortion coefficient, Z0 is an input impedance, and P1 is an extrapolated input power. At a very high IIP3 of +50 dbm, P1 is 100 W from a 50 ohm source Z0. At an IIP3 intercept point, third harmonic power is the same as input power, so solving the above equation, the third harmonic distortion coefficient is
or 0.01%. This harmonic distortion coefficient is of the same order of magnitude as the manufacturing tolerances that may be achieved in a microfabricated embodiment of the electron-beam amplifier (for example, the reproducibility that may be achieved in the beam spot and the detector and slot geometries). For example, a detector of 10 um width may be made with segment tolerances of about 1 nm, about 10,000 times smaller than the width. Given the wide bandwidth of the electron-beam amplifier, it is possible to achieve high IIP3, and by the wideband nature of the amplifier, can achieve such high IIP3 at extremely high frequencies.
Distortion Compensation
Solid-state amplifiers have little flexibility in eliminating distortion. For example, low distortion requires high bias levels and amplifier bandwidth much wider than a signal bandwidth; reducing output signal level as a fraction of total bias level, in turn reducing the range and effect of non-linearities. The high bias levels lead to excessive power consumption in exchange for minor linearity improvement.
Non-linearity of electron-beam amplifier 10 is primarily related to non-ideal deflector apparatus 130, a shape and a current density of beam spot 170 and a shape of detector segment(s) 150. The most difficult linearity parameters to control are deflector apparatus 130 and beam current density. Though deflector apparatus 130 inherently has a linear response, fringing fields are unavoidable and difficult to compensate in a compact electron-beam amplifier 10. Beam current density is also difficult to control because of space-charge spreading effects and variations in currents among individual e-beams 120.
High linearity in electron-beam amplifier 10 can be achieved by optimizing the geometry of apparatus 130 and regulating beam currents of individual e-beams 120 with control loops to assure a uniform, controlled beam spot current density. Residual beam spot and deflection distortion can be compensated by appropriately shaping a geometry of beam spot(s) 170, and slot(s) 160 separating detector segment(s) 150.
As discussed above, beam spot 170 is an outline of a cross-sectional current density of e-beam 110 where it impinges upon detector(s) 150. This current density may be non-uniform, and a “spot shape” is simply a contour of some value of current density. For many configurations of electron-beam amplifier 10, it may be convenient to assume that this current density is essentially uniform within the spot, and zero outside. It can be appreciated that simply referring to the “beam spot” may facilitate understanding of the basic principles of electron-beam amplifier 10.
Non-uniform beam spots 170 may occur for many reasons, including imperfect electron gun focusing, thermal agitation of electrons, space charge spreading, imperfect focusing of multiple e-beams 120 into a single beam spot 170, and quantum effects. In electron-beam amplifier 10, the beam spot 170 and detector segments 150 may be shaped to effectively eliminate many distortion effects, substantially extending the linearity and utility of the amplifier.
Slot Deformation Linearity Correction
The principle of slot deformation can extend to other shapes of beam spots 170 and detector geometries 150. For example, in some configurations it may be convenient to utilize a circular spot shape rather than a line spot; others might employ a triangular shape. Other embodiments may unavoidably have beam spots 170 with non-uniform current density.
Detector Shaping Linearity Correction
Because slot 160(2) between high speed detector segments 150(9) and 150(10) of
Beam Centering
Proper operation of electron-beam amplifier 10 requires centering of e-beam 110 on detector segments 150, since a displacement of the beam with respect to a center position generates an amplifier output signal. Because of manufacturing tolerances in mechanical construction of the amplifier (including for example tolerances in geometries within beam deflection apparatus 130(1), and in axial alignment of deflection apparatus 130(1) to detector segments 150) the beam may be displaced from the center position when the voltage signal 140 is zero. For this reason, a feedback amplifier may be incorporated to center e-beam 110 through use of an offset control loop.
In beam offset control loop 375, a differential voltage ΔV develops when currents from detector segments 150 are applied to a load. Integrator 380 filters and amplifies ΔV over time to generate a correction signal VOS, which is a measure of a misalignment of beam 120 with respect to a center position 390 between detector segments 150. VOS is applied to deflection apparatus 130(2) as described below. Correction signal VOS acts to restore an average beam position so that it stays centered between detector segments 150. A static gain of electron-beam amplifier 10 may be high enough that a residual offset is negligible.
In beam offset control loop 375, the coupling from integrator output 384 to deflection apparatus 130(2) includes a summing circuit 400. Correction signal VOS is summed with an RF voltage input VIN being amplified, and the sum of these signals is applied to a single deflection apparatus 130(2). In a beam offset control loop 376 shown in
for frequencies f>>½πR2C2, where s is a Laplace frequency variable equal to j2πf. At high frequencies, the second term is near zero, and the device acts as an integrator with a time constant τ1=R1C1. At low frequency, the first term still dominates because
Thus, integrator 450 has feedback loop characteristics similar to those of integrator 410; both are suitable for beam centering in certain embodiments of electron-beam amplifier 10. In certain other embodiments of electron-beam amplifier 10, it may be advantageous to have dedicated detector segments, called “offset sense segments,” for measuring beam offset.
Microminiaturized Fabrication
Electron-beam amplifier 10 may be made with microminiaturized construction using wafer-based semiconductor fabrication technology. Microminiaturized deflectors may be as little as 1 μm long and may produce a frequency response greater than 1 THz. Single electron guns may have a cross-section of a few microns, and entire arrays of hundreds of guns may generate a precise beam with a diameter of 100 μm or less. Electron-beam detectors may be as small as a few microns, with femto-farad parasitic capacitance and THz bandwidth. An entire amplifier may have dimensions of only a few millimeters and thousands of amplifiers may be batch produced simultaneously with low cost, high yield and reliability characteristic of conventional integrated circuits.
Manufacturing of a microminiaturized electron-beam amplifier may include fabrication, alignment, and bonding of individual elements such as field emission cathodes, beam focusing electrodes, deflector plates and other components into electron gun assemblies called “microcolumns” or “electron gun microcolumns” herein. Techniques such as photolithography, etching, deposition, implantation, plating, multi-level metallization, wafer bonding, and possibly other methods may be used to assemble components such as microcolumns, drift cavities, detectors, output coupling networks and bias circuitry into a monolithic device.
Entire wafers may be constructed as arrays of amplifiers, for individual use or to work in concert. Silicon wafers are useful substrates for forming certain components because of silicon's low cost and because diverse fabrication techniques are available. For example, field emission cathodes on silicon wafers, including the molybdenum tips called Spindt cathodes disclosed in U.S. Pat. No. 3,665,241, have been especially successful. Wet etching may be employed for large drift cavities, and dry etching methods such as deep reactive ion etching can cut very small, precise, high-aspect ratio features such as the beam contact grid of the detector. Critical holes in electron guns can be fabricated with even more precise focused ion-beam and laser drilling. Multi-level planarized metallization processes using chemical and mechanical polishing (“CMP”) may form many of the electrodes, especially those in the microcolumn electron guns. Aluminum, gold, copper, nickel, tungsten and other metals are widely applied with both sputtering, vacuum deposition and plating techniques. Semiconductor devices (for example, bias circuits, output networks and other circuitry for use with electron-beam amplifier 10) may be formed concurrently with other electron-beam amplifier components on a silicon substrate, using similar, compatible techniques.
High aspect ratio etching technologies and waferbonding are characteristic of what is called “micromachining” or micro-electrical mechanical systems (“MEMS”) technology. Because of the complex three-dimensional geometries, different elements of the device may be constructed on separate substrates, and these substrates can be assembled into a single unit. Many methods of bonding wafers exist today, such as, for example, eutectic or fusion bonding. Techniques for wafer bonding have also been developed to create vacuum-encapsulated cavities, which are useful for electron beam devices, e.g., as shown in U.S. Pat. No. 5,842,680 issued to Davis and U.S. Pat. No. 6,479,320B1 issued to Gooch. Furthermore, SiO2 gettering materials are compatible with silicon semiconductor processing and have been demonstrated to sustain ultra-high vacuum and enhance cathode lifetime in electron guns, e.g., as shown in U.S. Pat. No. 4,771,214 issued to Takenaka et al.
Space Charge Spreading
A primary reason for limited beam current in any e-beam amplifier is an inherent, electric-field induced repulsion between beam electrons, which forces apart electrons in a focused beam, and is called “space charge spreading”. In high current beams, the forces are substantial, and as electrons travel through a drift cavity, these forces can spoil an initial focus that may exist just after a beam exits from an electron gun.
Coulomb's Law describes a force between two electrons:
F∝1/R2, (1.8)
where R is a distance between adjacent electrons. Since, for any two electrons at random positions within a beam, R is proportional to the radius r of the beam, so an average repulsive force between electrons decreases (to first order) quadratically with the total radius of a beam, for the same total beam current. Thus, a beam of 10 um diameter will have 100 times less repulsive force than a beam of 1 um diameter.
Electron Gun Arrays
An embodiment of an electron-beam amplifier minimizes space charge spreading by using a two-dimensional (“2-D”) array of electron guns. Like a linear (i.e., one-dimensional) array, a 2-D array of electron guns generates individual electron beams that are emitted as parallel beams from an emission plane (e.g., emission plane 20).
The aggregate sum of the individual e-beams is termed here the composite electron beam. The low Coulomb force interactions within individual e-beams reduces beam spreading in proportion to a cross-sectional area of the beam, permitting higher total beam current for a given amount of spreading force. For example, a linear array of electron guns emitting N e-beams of current I will have approximately the same spreading force as a circular two-dimensional electron gun array emitting N2 e-beams of current I. The circular array will have N times higher current for the same spreading force.
From this example, it may be appreciated that a 2-D arrays of electron guns provides a significant reduction in space charge spreading forces in a microminiaturized electron-beam amplifier 10. In combination with beam current amplification from an active detector 150, and optical focusing techniques described below, electron-beam amplifier 10 achieves higher gain and power, and requires no (large, heavy and costly) magnets. Thus, microminiaturized amplifier construction is possible, with attendant advantages including, for example, high bandwidth and low cost.
Distributed Deflector Array
To achieve high-gain deflection performance with a two-dimensional array of beams, it is not possible to simply pass all electron beams through a single large pair of deflection plates. A beam originating at an emission plane (e.g., emission plane 20) with a diameter corresponding to a 2-D electron gun array would require a deflector with a plate spacing that is too large to generate sufficient beam deflection at reasonable voltage drives. This reduces amplifier gain unacceptably, unless the plate lengths were made correspondingly longer; however, longer plates reduce bandwidth performance proportionately.
For example, if an electron gun array has a diameter of 100 μm at an emission plane, a deflector with 100 μm plate spacing would have 100 times less deflection force than a deflector with a plate spacing of only 1 μm. To get the same beam deflection as the deflector with 1 μm plate spacing, the deflector with 100 μm plate spacing would have to be 100 times longer.
Disadvantages of large deflectors include low bandwidth, and a physical size that is incompatible with microminiaturized construction. In the above example, bandwidth of the 100 μm long deflector is 100 times lower than bandwidth of a 1 μm deflector for a single e-beam. Large deflectors may also have uneven electric field gradients between deflector plates. For a large diameter beam, this causes uneven deflection for different parts of the beam; in an array of individual e-beams, it causes different deflections for different e-beams. In either case, beam misfocusing results, causing amplifier gain distortion.
One advantage of the instrumentalities described herein is the incorporation of independent, matched deflectors at the output of each individual electron gun in an array of electron guns. Each electron gun and a corresponding deflector is part of a single microcolumn.
In a microfabricated device, plate spacing and length may be less than 1 μm. Microfabricated plate tolerances may be controlled to under 1 nm, so that deflectors of all microcolumns are matched to 0.1% or better, so that all e-beams are deflected the same amount for the same drive signal. A set of deflectors (“ganged deflectors”) driven in this manner constitutes a distributed deflector structure that provides uniform deflection to an array of e-beams, with high gain and fast, wideband response.
Microcolumn with X-Y Deflectors.
X-Y deflection is required for certain embodiments of electron-beam amplifier 10. This is enabled by adding a second beam deflector to each electron gun. It will be appreciated that the use of “X” and “Y” is for reference only; actual beam sweep directions in an electron-beam amplifier 10 are a matter of design choice, but X and Y are meant to convey two orthogonal directions in which an electron beam may be swept.
Deflector Loading
Loading of an array of ganged deflectors is low. For example, if each deflector consists of two 1 um×1 um deflector plates with a spacing of 1 um between plates, a capacitance per deflector is only 8.85 aF (10−18 F). 100 deflectors in an array of 100 electron guns will have a total capacitance of only 0.9 fF (10−15 F). A 3 dB bandwidth (=½πZ0CLOAD) of a 50 ohm source driving the deflector array capacitance is 3.6 THz. The loading of an array of deflectors thus has little effect on the device performance, and enables a wide bandwidth that is compatible with that of the other system elements.
Electron Gun
The basic operation of the electron gun is as follows. A strong voltage between control gate 625 and cathode 620(2) (typically in the range of +10 to +50V) creates a strong electric field around cathode 620(2) that causes a release of electrons into free space. A current transported by the electrons may be described by the Fowler-Nordheim theory of electron flux over an energy barrier. Electrons may be released in the direction of the gate, with an angular distribution and an energy approximately equal to the potential difference between control gate 625 and cathode 620(2). By appropriate design, most of the electrons pass through the center of the gate electrode, and from there, they are focused within the electron microcolumn, as explained below.
Many electron gun microcolumn designs may be conceived as variations on the teachings herein to collimate electrons from a field emission tip into a narrow parallel beam.
Electron Gun Current
An electron gun 610 may be designed with a low enough beam current so that individual beam electrons are separated, on average, by a distance greater than the beam diameter. As a result, the electrons are far enough apart that mutual repulsion is minimized, so that space charge effects do not materially affect focusing.
Electron gun beams may have a diameter <1 μm and a maximum current of approximately 1 μA. This low current is consistent with negligible beam spreading because of a low density of electrons at beam energies typically used (around 100 eV to 300 eV). Generally, a lineal density λ that is a number n of electrons per unit beam length x, is given by
where IBEAM is a beam current, q is the electron charge, and vBEAM is a velocity of the electrons, given by
vBEAM=√{square root over (2qVBEAM/me)}. (1.9.1)
Here, VBEAM is a beam energy in volts, and me is the mass of an electron (9.11×10−31 kg). At 200V, vBEAM is 8.4×106 m/s, and at IBEAM=1 μA, the lineal electron density λ is 0.75 electrons per micron. A 1 μA beam current spaces the electrons apart by approximately the beam diameter, so that the electrons experience no significant lateral Coulomb force interactions or beam spreading.
Electron Optics
Focusing of electron beams 120 by electron optics can be understood by analogy to geometrical light optics. The advantage of the optical analogy is that it clearly predicts how focusing works for electron beams 120 from any direction, and provides insight into design of focusing fields.
If electron beams 120 exiting an emission plane (e.g., emission plane 20) are collimated into parallel beams they may be considered, by analogy, like light rays emitted from an object at an infinite distance from a lens. The lens is analogous to the electron optics. In geometrical optics, parallel rays can be focused to a point on an image plane on another other side of a lens, one focal length away.
where f=focal length, o=distance from object plane to lens, and i=distance from lens to image plane. (As in light optics, the lens “position” in this case is described in terms of “principal planes” 700(1) and 700(2), which are generally different for the object and image sides of a thick lens, but for purposes of this analogy the principal planes can be assumed coincident in position, which is the “thin lens” approximation from light optics.)
In electron optics, a “lens” consists of electrodes of appropriate sizes, shapes, and voltage potentials.
It can be understood from electron lenses 750(1) (FIG. 25) and 750(2) (
Electron Gun Focusing
In electron optics of an electron gun microcolumn, the “lens” may be constructed as a stack of electrodes perforated by circular holes (e.g., focusing electrodes 630). In the microcolumn, electrodes 630 may be metal layers (such as Al) separated by insulating layers (such as SiO2). Potential voltages applied to the electrodes create electric fields in the microcolumn that act on the emitted electrons to produce focusing action. In this way, electrons can be either accelerated or retarded in velocity.
Optical Aberrations
A limitation of optics, whether for light rays or electron beams, is focusing aberration. Two common aberrations that are relevant to electron-beam amplifiers are spherical and chromatic aberrations.
Spherical aberrations are characteristic of off-axis rays that meet the lens at a large angle. These rays are focused closer to the lens than rays that travel at angles close to the lens axis (called “paraxial” rays in optics). Correction of spherical aberrations can be accomplished in light optics through certain deviations of a lens shape from a spherical surface (“aspheric” lenses). In electron optics, analogous corrections can be made by shaping the electric fields via electrode sizes, shapes, spacings and potentials, although no “spherical” surface per se is being corrected.
Chromatic aberration is caused in light optics by different wavelengths being bent by different amounts within lenses. Chromatic aberration produces, in a given lens system, longer focal lengths for short wavelengths, and shorter focal lengths for long wavelengths. Correcting chromatic aberration in light optics can be done through certain combinations of lenses made from materials having different indices of refraction (for example, crown glass and flint glass), a combination referred to as an “achromat.” With the right combination of lens materials and curvatures, a lens system can balance chromatic variations in focal length for different lenses and can achieve approximately the same focal length for over a range of wavelengths.
In an electron optical system, chromatic effects arise from electrons of different energies. In an electron-beam amplifier, this may occur primarily at the point of emission from the field cathode. The general principle of correction through an achromat combination is analogous to an achromat in light optics; an electron achromat uses lenses of different field gradient densities to achieve the effect of different indices of refraction. However, it is difficult to combine separate lenses of different field densities because of the electrode structures required. An alternative to use of an achromat is to filter electrons of different energies with an aperture stop. This solution operates somewhat like a pinhole camera.
A disadvantage of filtering electron beams with apertures, as opposed to use of an electron achromat, is that some portion of beam current is blocked, reducing efficiency of an electron gun. An advantage is that a beam emerging from an aperture may be well focused and collimated. Spherical and chromatic aberrations may be corrected to produce an electron beam diameter of a few nanometers in a microcolumn that is several millimeters in length, at beam energies of 1 keV and currents up to 50 nA. Generally, higher energies, lower currents, longer columns and short drift distances achieve better focusing.
An electron-beam amplifier may require beam focusing on the order of a micron to ensure proper focusing across a drift cavity. Another way of looking at the beam focus requirement is that all e-beams emitted from a microcolumn array should act as if emitted from a single point source at infinite distance.
Electron Gun Fabrication
The components of an exemplary electron gun microcolumn include an FE tip cathode, a control gate (called a “wehnelt” in some literature), electrodes forming a first lens element, a first aperture plate, electrodes forming a second lens element, a second aperture plate, deflection plates, and a shield plate. The cathode may be a single field emitter tip; alternatively, a heated Schottky or other thermionic emitter may be used.
The microfabricated construction of an electron gun in an electron-beam amplifier may follow a sequence of fabricating components on individual silicon wafers, followed by alignment and wafer bonding of the wafers into a stack.
The wafers and assembly illustrated in
Multiple Focusing Electrodes
In electron-beam amplifier 10, multiple microcolumns are advantageously constructed concurrently in a compact array. Making a gun array as small as possible helps create high beam current density with good spot formation. For example, a single microcolumn may have a diameter of 5 μm or less to allow several hundred or more microcolumns to be fabricated in an array having a diameter of approximately 100 μm.
It is possible to use large electron lens electrodes achieve aberration-free focusing, as in light optics, in which large lenses improve image quality. In electron optics, as discussed above, perforated electrodes may act as lens elements (see
It can be appreciated that large lenses are not compatible with a small diameter microcolumn and a dense gun array. In an improved embodiment of an electron-beam amplifier, small microcolumns having a plurality of small electrodes approximate the focusing of a single large electrode.
Where the potential lines coincide with electrode surfaces, they have the same potential as the corresponding electrode. This is the principle of an improvement to electron-beam amplifier 10. The potential gradients near the centerline of the lens, within the radius of the perforation, can be preserved without a wide diameter lens by using a series of thin, small diameter electrodes. For example, each equipotential line 760 in lens 750(9) has the same spacing and shape as a corresponding equipotential line 760 in the small region between dashed lines 860, 860′ within lens 750(8). Thus, lens 750(9) may provide similar focusing action, within a smaller physical size, as lens 750(8). In the case of an infinite number of differential electrodes, the lens 750(9) performs exactly as lens 750(8). In practice, only a few extra electrodes are required to substantially approximate a large three electrode lens with a small, multi-electrode lens.
Beam Current Control
Formation of a useful beam spot 170 requires substantially uniform beam current from all electron guns that supply individual beams for the composite beam. Field emission cathode tips (“FE tips”) may have nonuniform current-voltage characteristics (“I-V characteristics”); applying a single potential to a gate electrode 625 of each gun in an electron gun array 100 may result in a beam spot 170 with large current density variations. For this reason, beam current from each electron gun may be individually regulated by a control loop so that each electron gun produces substantially equal current.
The gate electrode potential has a significant effect on the electron optical focusing of the microcolumn, and changes in gate potential may significantly defocus the electron gun beam unless compensated by changes in potentials of other electrodes. For this reason, an improved electron-beam amplifier 10 may include circuitry which adjusts certain electron gun focusing electrodes at the same time as the potential of a gate electrode is changed, to maintain constant focusing characteristics.
Focusing potentials are generally difficult to determine except by computer analysis. One method of adjusting electron gun focusing potentials in the presence of a current-regulated gate potential consists of an analog-to-digital converter (“ADC”), a digital-to-analog converter (“DAC”) and a read-only memory (“ROM”) that is programmable with digital values. The ADC may be coupled to the gate electrode, to develop a digital word representative of the gate potential. This word is transmitted to the ROM as an address. The ROM functions as a look-up table, and stores DAC codes representative of optimized electrode potentials for any given gate potential measured by the ADC. The DAC responds to the output of the ROM by generating a focusing potential, which may be applied to an electrode. Thus, one or more electrode potentials may be arranged to correlate directly to the gate potential. In alternative embodiments, it may be appreciated that the ROM can be replaced with other means of generating digital values, such as a processor element.
Focusing electrode controller 930 controls potentials of focusing electrodes 630(20), 630(21), 630(22) and 630(23) as follows. An ADC 940 connects with gate electrode 625 (2) and generates a digital gate word 950 which is transmitted to a ROM 960. ROM 960 accepts digital gate word 950 as input and generates electron gun focusing words 970(1), 970(2), 970(3) and 970(4) as output; the electron gun focusing words are transmitted to corresponding DACs 980(1), 980(2), 980(3) and 980(4) which generate gun focusing potentials corresponding to each electron gun focusing word, and transmit the gun focusing potentials to focusing electrodes 630(20), 630(21), 630(22) and 630(23).
In a focusing electrode controller (e.g., controller 930) each electrode driven by a ROM (e.g., ROM 960) increases a storage capacity required in the ROM by a number of input levels values resolved by a corresponding ADC (e.g., ADC 940), times the number of DACs, times the number bits of resolution required as input by the corresponding DACs. For example, in the case shown in FIG., if ADC 940 measures gate potential to 6 bit accuracy, the number of input levels resolved is 64; if each of DACs 980(1-4) requires a 7 bit word (e.g., electron gun focusing words 970(1-4)) as input, then the required ROM storage capacity is 64×4×7 bits (1792 bits).
The technique used in focusing electrode controller 930 may be extended to control all electrodes of an electron gun that are affected by a gate potential. Each electrode (e.g., electrodes 630) requires one DAC, and the required ROM storage capacity grows proportionately. There is no restriction on the number of bits in the electron gun focusing word supplied to a given DAC. Different DACs may resolve gun focusing potentials to different accuracy levels and may require correspondingly more or fewer bits per electron gun focusing word. For example, electrodes closest to the cathode may require high DAC accuracy and thus more ROM bits. Electrodes further from the cathode (in the microcolumn) may require less DAC accuracy and fewer ROM code bits. Generally, a first aperture plate of the electron gun (e.g., aperture plate 640(1) of
Typical Mechanical Parameters
Electron beam amplifier 10 may be designed or optimized for a parameter space of operation that may include gain, frequency response, bandwidth, power output, efficiency, noise, and drift time. Variables which may be manipulated as matters of design choice include electron gun energy, beam current, number of guns, number of deflection plates per electron gun (horizontal, vertical, cross-axis, blanking, offset centering), drift cavity acceleration, cavity length, detector size, shape and configuration, cascade and avalanche gain, diode material, voltage rating, bias, and output coupling method. Certain combinations of these parameters will result in amplifiers that may have vastly different mechanical dimensions and electrical specifications. For example, the mechanical dimensions shown in
Wideband Feedback
Certain systems require an amplifier with almost perfectly linear response such as, for example, a low noise amplifier (“LNA”) which may be used at the front-end of an RF receiver. High gain may not be required of an LNA, but distortion free response may be required to help detect small signals when a large interfering signal is present. For example, an interfering signal may have 1V peak-to-peak (“p-p”) amplitude, and a signal of interest may be 0.1 mV p-p (for example, when a jamming signal is present, or when a high-power transmitter is close to a receiver attempting to detect a distant signal).
In such applications, dynamic wideband feedback is often applied to a transistor amplifier to provide controlled gain with very low distortion. The transistor amplifier must be very wideband to operate with the feedback, since as is well known, this may be essential to achieving stable operation with the feedback. The wideband characteristic translates to a short delay through the amplifier; specifically, it is known that for feedback to be applied, the delay through the amplifier should normally be less than ½ cycle of a highest signal frequency for which the amplifier gain exceeds unity, or the feedback will be unstable and the amplifier will oscillate uncontrollably.
A delay time of an electron-beam amplifier may depend in part on a drift cavity length zdrift. For example, a 200 eV beam has a beam velocity of 8.4×106 m/s. With a 1 mm cavity, drift time is 119 ps. This is a short interval, but not short enough to use the amplifier with wideband feedback at frequencies for which it has useable gain. Since an electron-beam amplifier may offer significant gain at frequencies of 100 GHz or more (as described below), some embodiments may require a drift time of 5 ps or less. Based on this criterion, if stable feedback is to be applied at 100 GHz, a maximum drift cavity length zdrift is 40 um for a 200 eV beam.
A short drift cavity length zdrift has significant impact on parameters of an electron-beam amplifier. Short zdrift may mean that a smaller array of fewer electron guns may be used, since there is less distance over which to focus beams on a detector; but conversely, since less beam spreading occurs over the short zdrift, the guns may operate at a correspondingly higher current. For example, with zdrift of 40 μm, a gun array may have a diameter of 20 μm and may include only 16 guns. Individual beam currents may be on the order of 10 μA, since there will be less beam spreading over a short drift time, while a greater drift cavity length zdrift might only be compatible with beam currents on the order of 1 μA. Total beam current could therefore be 160 μA; not much different than in a long cavity, but higher beam energy may be used to reduce drift time and increase output current with higher cascade gain. For example, an 800 eV beam may provide a drift time of 2.4 ps. This is one-half the time of a 200 eV beam. Thus, feedback can be applied over 200 GHz bandwidth. With a 20 μm drift length, feedback bandwidth may be over 400 GHz.
Thus, it can be appreciated that many matters of design choice may be used to optimize an electron-beam amplifier 10 for a particular application, and that feedback may be applied to some electron-beam amplifier configurations to enable very low-distortion performance at high frequencies.
Typical Electrical Parameters
In typical configurations of an electron-beam amplifier 10, with or without feedback, beam energy may be 200-300 eV, individual electron beam currents may be on the order of 1 μA, detector gain may be 1000, and maximum deflector voltage drive may be 100 mV to 1V.
Like mechanical parameters, electrical parameters may range widely according to an intended application. Some parameters are related to mechanical dimensions, while others are more constrained by physics. For example, in most applications, one design objective is to generate an electron beam of maximum current without large spreading forces. At 300 eV energy, this translates to a maximum electron beam current of about 1 μA, based on electron density in the beam (though a shorter or longer drift cavity may increase or decrease the maximum electron beam current somewhat).
Another physical limitation is maximum beam energy. High beam energies at higher beam currents can cause excessive heating of a detector. High voltages (thousands of volts) which may be used to generate high beam energies can also cause arcing in a microminiature device, even at low beam currents. High energy also is not compatible with most integrated bias circuitry, which may withstand only a few hundred volts. Thus, a maximum beam energy in a range of 300 eV to 1000 eV is currently preferred.
Minimum beam energy is another limitation. If a beam energy is too low, cascade gain of a detector may be inadequate. As discussed above, low cascade gain cannot always be compensated by larger avalanche gain, since avalanche gain is limited by detector junction leakage and radiation sensitivity.
Another physical limitation is a minimum beam current which can produce a desired noise figure. Even with an ideal detector, electron beam shot noise (the effect of discrete electrons, rather than a smooth stream of current, striking a detector) is still amplified.
Many factors may drive deflector voltage drive range, including individual electron beam diameter, minimum plate spacing that can be manufactured reliably, drift cavity length zdrift, detector size, amplifier gain and input signal range. Since one application of an electron-beam amplifier 10 is as an antenna coupled LNA, its input signal may vary from microvolts to more than 1V. A maximum tolerable deflector voltage is set by the arc-limit of the plates, and may be around 10V per micron of space; if electron-beam amplifier 10 is fed from a solid-state amplifier, a lower limit of about 1V may be set by a voltage breakdown of high-frequency (GHz bandwidth) solid-state transistors.
These are not the only factors that constrain the electrical parameters, but illustrate some of the principles underlying the electrical parameter limitations.
Deflection Gain
Microminiaturization of e-beam dimensions and deflector plate spacing to micron or even submicron dimensions provides two benefits: high deflection gain and fast response. Thus, if plate spacing (e.g., spacing of deflector plates 600) is small, small signal voltages may generate strong electric fields for beam deflection, in turn creating large transverse beam displacement over very short transit times (of a beam through the deflector plates), permitting deflectors with short plate length LP. In practice, deflectors can be shorter than 1 μm, with transit times of much less than 1 ps.
A general relation for deflection force F is F=qE, where q is the electron charge and E is the electric field between two deflector plates, approximately
E=Vsig/WP, (1.11)
where Vsig is an instantaneous signal voltage applied across the plates separated by a spacing WP.
Deflector plates only approximate parallel plates, both in physical construction and in transfer function, but the parallel plate approximation may be used for most calculations. The essence of the approximation is that a one-dimensional, uniform electric field exists between two plates; from this, a basic relation may be derived for the deflection angle θ in response to an input signal ΔV. For a parallel plate deflector of plate spacing WP and plate length LP, a ratio of lateral transverse beam velocity vx (imparted by the deflection process) to a longitudinal beam velocity vz is
where beam energy is VBEAM (in volts). ΔX is lateral displacement of a beam after propagating across a drift region of length zdrift between the deflector and the detector plane.
Within an electron-beam amplifier, a ratio GBEAM of lateral beam displacement to a corresponding change in a deflection signal is
For example, with appropriate choice of WP and LP, a spot of a 100 eV beam may be deflected 71 μm per volt of signal at the detector when drift length zdrift is 1 mm. Longer drift lengths, longer deflectors and smaller plate spacings increase GBEAM; higher beam energies reduce GBEAM.
A collection gain Gcoll is a differential current collected by the detector with respect to a change ΔX in beam spot position. Gcoll may depend on width and geometry of a detector. As discussed above, an electron-beam amplifier may be constructed so that its detectors collect substantially all available beam current when a beam is fully deflected across a detector width XD:
Gcoll=IBEAM/XD. (1.14)
With kC and kA representing detector cascade and avalanche gain factors respectively, and kD=kC kA representing total detector gain, the above formula for Gcoll may be multiplied by kD to give the total amplifier transconductance gain gm, the change in differential output current between the detector segments, with respect to a change of input signal:
Parameters WP, LP, zdrift and VBEAM can be selected so that:
For example, if IBEAM=100 μA, ΔVin=1V p-p (i.e., +/−0.5V) and kD=1000, the transconductance gain is 100 mS (A/V). However, longer drift regions, smaller detectors and other parametric variations may allow an electron-beam amplifier to provide substantially higher gain from the amplifier, and ganging electron-beam amplifiers can provide even higher gain. Moreover, amplification may be very linear, so an electron-beam amplifier 10 may provide more usable gain than known amplifiers.
Deflector Frequency Response
Microfabrication also offers an advantage in terms of high frequency performance. When deflector plates (e.g., deflector plates 600) are shrunk to micron-scale dimensions, frequency response between input and output increases dramatically. Physically, the finite bandwidth of a deflector (e.g., deflector 130(1) consisting of matched deflector plates 600) can be understood as the time it takes a single electron to pass through the deflectors, since dynamic changes in deflector drive voltage will filter and average the deflection. For example, first define a transit time τ as the time it takes an electron to traverse the region between deflector plates. If a drive voltage is positive for half of τ and equally negative for the other half of τ, it can be appreciated that the net deflection will be zero. Thus, transit time τ should be designed as much less than a period of a maximum signal frequency. In a parallel plate deflector, a relation of 3 dB bandwidth to τ, or to beam velocity vz and plate length LP, may be derived as
When beam velocity is expressed in terms of the total electron gun accelerating potential VBEAM, the response is
where f3DB is the frequency at which the deflection gain is reduced to 0.707 (3 db) of the low frequency response.
Table 1 shows electron-beam amplifier physical and electrical parameters for selected values of VBEAM, Vin and LP. All entries in Table I assume WP is 1 μm and zdrift is 1000 μm. As shown in Table 1, frequency response of the deflector may exceed 1 THz. XDET The calculated values of f3DB, tan Θ, and
Dimensions and construction of detectors permit similar bandwidth, for example, where these bandwidths are where the gain is only down by 3 db compared to a low frequency response. Unity gain frequency response, or gain-bandwidth product, is another common measure of amplifier performance. With a voltage gain of 10, the gain-bandwidth product of an electron-beam amplifier may be 10 THz. Though an electron-beam amplifier has the potential for THz performance, gain-bandwidth product can be used as a figure of merit to assess usable gain at any frequency, or to determine the ultimate performance potential, or to make comparisons to other technologies. By way of comparison, a single-stage HEMT amplifiers may have gain-bandwidth products of about 400 GHz.
High Power Output.
Power output may be increased substantially by ganging amplifiers. A 100 gun array may have only 0.9 fF loading capacitance, so small that many amplifiers can be ganged and driven in parallel with little loss of bandwidth. For example, one electron-beam amplifier driven by a 50 ohm source may have an input bandwidth of 3.6 THz. Ten electron-beam amplifiers driven in parallel by a common 50 ohm source impedance may have an input bandwidth of 360 GHz, still high enough to pass most input frequencies, and the parallel gang provides 10 times the power output of a single amplifier. Similarly, a gang of 100 electron-beam amplifiers may have 100 times the power output, at 36 GHz.
With a hierarchical or “corporate” power input distribution system (see
By this means, a microfabricated electron-beam amplifier array may include as many as millions of amplifiers on a single silicon wafer, and the entire amplifier array may be driven from a single source. The total coherent power output of the amplifier array may exceed 10 kW, while preserving the wide bandwidth of individual amplifier elements. It can be appreciated that the ability to gang many amplifiers is one characteristic of electron-beam amplifiers for applications that require very high, wideband power output.
Efficiency
Another benefit is power-added efficiency (“PAE”). This is the RF power that is added to the output of an amplifier (i.e., POUT−PIN) as a percentage of total amplifier power PIN, including thermal losses:
Conventional semiconductor amplifiers can provide high power gain, but often have low efficiency, or somewhat higher efficiency over a narrow band of operation at relatively low frequencies (up to around 10 GHz). TWTs can provide much higher power output over an octave or more of bandwidth, with PAE approaching 50% in the best devices, but with a significant power overhead required to heat thermionic cathodes and generate a high-voltage collector bias (10 kV or more). For this reason, TWTs rarely operate with less than 100 watts of power, which is undesirable in many applications.
In contrast, an electron-beam amplifier 10 may provide high power gain (60 dB or more) in a miniature device dissipating as little as milliwatts of total power, or as much as many watts, at a PAE exceeding 50%.
A total amplifier power is approximately PTOT=PBEAM+Psupp, where PBEAM is the beam power and Psupp is the total detector power into the output power supply, Vsupp. The total beam power is PBEAM=IBEAMVBEAM., where VBEAM is the beam energy in electron-volts (i.e., the acceleration potential) and IBEAM is the beam current.
The supply power due to detector current is Psupp=I0Vsupp when a constant power supply absorbs a constant total current I0=kDETIBEAM from two detector segments (i.e., nearly 100% of the beam is over one detector segment or the other). If each detector segment terminates in a load resistor of value R, the optimum amplifier efficiency occurs for the largest output voltage swing within Vsupp. That is, if the signal is sinusoidal, the current output waveform from a single detector is
and the maximum voltage across the load resistor is Vsupp=I0R. Detector current causes an output voltage to swing between 0 to Vsupp across the load R (ignoring certain factors such as a minimum detector bias for generating detector gain kDET, but this is a reasonable approximation). Given these assumptions, supply power is
Psupp=I0Vsupp=I02R (1.21)
If all of the RF power from one detector segment is dissipated in the load, the RF power output is
averaged over one period T of the RF. Normalizing over an angle θ from 0 to 2π,
and finally the RF output power from one detector segment is
The total RF output load power PLOAD from both detector segments is twice P1, or
While certain RF amplifiers do not have a simple resistive load from which to calculate a transmitted POUT, a good first approximation is to use POUT=PLOAD.
Assuming an e-beam of 50 μA accelerated to a 200V potential, beam power PBEAM is 10 mW. If a detector has a gain of 2000, Io=100 mA. If an output load is 20 ohms, PLOAD=200 mW and POUT=150 mW. Using these assumptions, PTOT=110 mW. If the input is from a 50 ohm source with an amplitude of VIN=0.1V (peak), then
PIN=VIN2/(2×50)=0.1 mW. From these numbers,
This PAE compares favorably with solid-state or TWT amplifiers, but at higher frequencies and wider bandwidth.
Even higher PAE can be achieved in a specialized device that excites a resonant load with a non-sinusoidal pulsed current drive. If a detector is overdriven to operate as a photoconductive switch in such a case, the efficiency can approach 90% or more. Thus, it can be appreciated that electron-beam amplifier 10 may provide performance comparable to, or exceeding, that of known devices.
If amplifier power gain GP is high, the input power PIN is small with respect to the output power POUT. An electron-beam amplifier 10 may can achieve values of GP>106, so
This brings out the useful result that increasing output power supply voltage increases the efficiency (for example, by using a high breakdown strength detector material with high detector current), and decreasing the load resistance or the beam energy increases the efficiency, but the maximum efficiency can never be greater than 75%.
To understand the relation between detector gain, detector breakdown VBV and beam energy, let
Vsupp=VBV=IOR=kDETIBEAMR. (1.27)
As discussed above, detector gain is the product of the cascade and avalanche gain, kDET=kCkA, and the cascade gain kC is given approximately by kC=VBEAM/VCI, where VCI is the cascade ionization energy of the detector material. Solving for IBEAM,
Substituting into PBEAM=IBEAMVBEAM, the power added efficiency is
Notably, the beam energy and load resistance does not affect PAE. PAE is highest with a detector material that has the highest ratio of VBV/VCI, and a detector structure with a high avalanche gain. Table 2 gives material parameters VCI, EBV, and VBV for certain materials.
Thermal Heating.
High PAE corresponds to high thermal efficiency, which may be another benefit of electron-beam amplifier 10. With high detector gain and low beam current, little joule heating of the detector by a high energy beam occurs, so little power is wasted. For example, a 280 eV beam of 100 μA dissipates only 28 mW of power in detector heating, while generating 100 mA of diode current. Actual temperature rise of a detector is insignificant, on the order of a few degrees for typical semiconductor coefficients of thermal conductivity (eg, 100 degrees C. per watt).
Power Transformation
Electron-beam amplifier 10 is also an efficient power transformer, insofar as it converts a high-impedance, low-power input signal (a deflection voltage) to a low-impedance, high-power output signal (a detector current into a load network). This is another benefit of a high gain detector. A power-transforming advantage provided by electron-beam amplifier 10 is evident in radiating embodiments, as explained below.
Noise Figure
Noise in electron-beam amplifier 10 is predominantly shot noise. In an electron beam (e.g., composite electron beam 110), shot noise current iNB for a bandwidth Δf is spectrally white and is described by
iNB=√{square root over (2qIBEAMΔf)}(RMS,Amps/√{square root over (Hz)}) (1.30)
This is true because field emission obeys Poisson statistics, which are characteristic of current across a barrier potential. The detector introduces noise primarily through the avalanche gain. The cascade gain is essentially noise free, but the beam noise is amplified by the total detector gain. It can be shown that with sufficient cascade gain, the noise introduced by an avalanche process is negligible.
Shot noise is characteristic of a quantized current flow. The quantization in normal semiconductors arises from discrete charge quantities of electrons moving across a potential barrier, such as a P-N or Schottky junction. Shot noise in an e-beam is similar, since the charge quantities are still electrons. The effect of cascade gain on detector noise can be inferred from this. Each beam electron that penetrates the detector generates a cascade of kC electrons in only a few femtoseconds. The time frame of the cascade is so short that the effect is equivalent to a single particle of charge kCq (where kC is as defined above) striking a detector which has no cascade gain. Thus, the cascade-amplified beam current has a noise power iND that is still described by shot noise power:
where IBEAM is first rewritten as dQBEAM/dt and then as qdnBEAM/dt, with n being a number of electrons. In effect this can be rearranged as
iND2=2q(kC2IBEAM)Δf. (1.32)
This is exactly the noise of an ideal amplifier, showing that the cascade process introduces no excess noise. If a Noise figure NF is defined as
NF=10 log(1+NADDED/NIN), (1.33)
where NIN is an ideal minimum input noise and NADDED is the noise added by an amplifier, referred to the input, the cascade process is seen to have a noise figure near 0 dB. This can be understood by considering the noise added to a single beam electron—there is none, since the assumption is that each is exactly multiplied by the cascade factor kC. The total noise power, however, increases as the square of the gain because gain refers to current amplification, not power; hence the factor kC2. This is characteristic of any kind of amplifier.
By contrast, avalanche multiplication introduces noise through two mechanisms: multiplication of diode leakage current, and excess noise factor, which describes the statistical fluctuations in the multiplication arising from the sequence of hole or electron impact ionization events. Neglecting leakage, avalanche noise is given by
iNA2=2qFkA2ICΔf (1.34)
where IC is the beam current after multiplication by the cascade, kA is the avalanche gain, and F is the avalanche excess noise factor. F is a device specific parameter that is typically greater than 2, varying from 3 in silicon to 9 for germanium. The total noise at the output of the detector is
iND2=2qkA2kCIBEAM(kC+F)Δf. (1.35)
If kC>>F, this simplifies to
iND2=2qkD2IBEAMΔf (1.36)
where kD=kAkC, the total detector gain. Thus, a requirement for low noise detector operation is a cascade gain much higher than the excess avalanche noise. In one embodiment of the detector, the cascade occurs in a thin germanium layer and the avalanche takes place in a silicon layer. For example, a 280 eV beam will have a cascade gain of approximately 100 in germanium. A silicon avalanche diode can be optimized for F=3. Thus, it can be seen that the effect of avalanche excess noise is small, and for certain embodiments, the detector essentially operates as a noiseless amplifier (noise figure=0 dB). This is a key benefit of electron-beam amplifier 10.
Radiation Tolerance
Another benefit of electron-beam amplifier 10 is high radiation tolerance. An e-beam itself is inherently immune to radiation levels, and an energy flux of e-beams in electron-beam amplifier 10 is much greater than an energy flux of natural radiation (even in a low earth orbit of 700 km, where radiation is high). The primary effect of radiation on electron-beam amplifier 10 is leakage across diode junctions because of hole-electron pairs generated when high energy particles pass through semiconductors. High-energy electrons and protons are both significant, but the effect is similar. Under most natural conditions an effect of radiation may be a small increase in detector noise.
Beam Focusing in a Microminiaturized Amplifier
As discussed above, space charge induced beam spreading is mitigated by several means, including high detector gain to reduce beam current requirements, and by using electron gun arrays 100 to increase beam diameter. In a microminiaturized high-speed electron-beam amplifier 10 beam spreading may be significant, because small detector(s) 150 are necessary to achieve the high speed, and a beam spot 170 may be small, to match the detector. For operation above 100 GHz, a detector size of less then 10 μm is preferred. If a 100 μm diameter electron gun array is used, this means a 10:1 reduction in a diameter of a resulting composite beam 110 may be achieved by focusing action in a drift cavity 145. It can be appreciated that a means of overcoming space charge spreading forces to compress a composite beam diameter from approximately 100 um at an emission plane 20 (in a microminiaturized device) to a spot diameter that may be at least 10 times smaller at a detector plane 50 improves performance of an e-beam amplifier 10.
Improved Embodiment for Small Beam Spot
An improved electron beam amplifier 10 includes electron beam focusing in a drift cavity 145, providing higher beam current, higher power output, lower thermal heating, lower noise and higher efficiency.
Doublet Lens System
A second electron lens 1010, using an accelerating potential at the detector plane, creates a doublet lens arrangement of electrodes 1020, 1030 and 1040 to provide improved beam compression, cascade gain, and aberration correction. As also shown in
In electron-beam amplifier 10(4), electrodes 1020 and 1040 are circular discs surrounding the electron gun array and the detectors respectively. Electrode 1030 is an annular can or “drift can” partially closed at both ends by endplates, as shown in
Similarly to
In electron-beam amplifier 10(4), electrode 1030 may be at ground potential. Electron lenses 1000 and 1010 achieve focusing action through positive potentials on electrodes 1020 and 1040; the potential of electrode 1040 being substantially greater than the potential of electrode 1020, to provide acceleration through the drift cavity. For example, electrode 1020 might be at 50V and electrode 1040 might be at 300V.
The structure may be considered a doublet of two lenses. Both electron lenses 1000 and 1010 achieve lens action by the geometrical relationships of the sizes and the potential differences among electrodes 1020, 1030 and 1040, in a manner similar to that described above with respect to electron optics electron guns. The effect of using discs for electrodes 1020 and 1040, each in a common plane with electrode 1030, may be seen as making one of distances x13 or x23 in
According to the electromagnetic theory of superposition, the fields of electron lenses 1000 and 1010 may overlap, but the lenses may be treated as if they act independently. Both lenses 1000 and 1010 may be considered “immersion lenses,” since electron gun emission occurs inside lens 1000 and beam detection occurs inside lens 1010.
Since electron beam emission consists of parallel rays at emission plane 20, an optical “object” for the emission is virtually located at infinity behind the emission plane. The “image” of this “object” is a focal length away from a principal plane on an image side of a two lens system. The term “principal plane” from geometrical optics describes a point from which a focal length is measured in an optical system that has a non-zero thickness; there are two principal planes, one on an object side, and one on an image side (which in e-beam amplifier 10(4) is a region of drift cavity 145(4) towards detector plane 50).
An advantage of a doublet lens is that focusing and acceleration occur simultaneously. If only lens 1010 were used, the focusing action is not as strong because the short distance to detector 150 and the accelerating field reduce a transit time over which radial forces can act. If only lens 1 is used, the focusing action is strong because an inward momentum is imparted just past the emission plane, but a retarding field slows the beam, increasing transit time and reducing beam energy and detector cascade gain. A doublet lens provides the benefits of strong focusing and acceleration. Furthermore, a doublet lens provides extra degrees of freedom to correct for other well known optical phenomena such as spherical aberration, coma and field curvature.
Certain embodiments of an electron-beam amplifier may use only one electron lens. For example, in embodiments using single electron guns that are independently deflected by multiple signals, an electron lens like lens 1000 may be undesirable. In embodiments using multiple beams, an electron lens like lens 1010 may be undesirable. Several electron-beam amplifiers in which these considerations apply will be discussed below.
Parallel Beam Deflection and Focusing
In
ΔX=f sin Θ. (1.37)
For example, if Θ is 10 degrees and f is 1 mm, ΔX will be 174 μm.
Spot Formation
In a first method of spot formation, an electron gun array is arranged with an outline that is the same as an outline of an intended spot, and drift cavity optics image and demagnify electron beams from the array onto a detector. In a second method of spot formation, an array shape and astigmatic focusing optics are chosen to create a desired spot image.
Many spot shapes are possible, ranging from simple points, line spots and rectangles to circles, triangles and more complex shapes.
Placement of a detector at a focal point of a composite electron beam is undesirable in embodiments of an electron beam amplifier 10 wherein correct operation of the amplifier uses a shaped beam spot by design. To create a shaped spot, a detector may be placed ahead of, or behind, an image plane.
In
Astigmatic Optics
An electron beam amplifier 10 may generate a desired focused beam spot 170 with an electron gun array shape 101 that differs from the shape of the beam spot through use of astigmatic focusing optics. Astigmatic focusing optics are asymmetrical about an axis, and have different focal lengths in different axial planes.
Dynamic alteration of beam spot shape by electrical control of astigmatic electrodes is useful in other embodiments of an electron-beam amplifier, as explained below.
EBRX
From the foregoing, it can be appreciated that an electron-beam amplifier may include various combinations of the following elements: a two-dimensional electron gun array, low-current electron beams, composite electron beams, single or distributed beam deflectors, a drift cavity, drift cavity electron optics that provide focusing and/or beam acceleration, one or more high gain detectors, and one or more output networks; any of these elements may be made through microfabricated construction. Combinations of these elements may be termed here an “EBRX” for Electron Beam RF Amplifier (“X” being a common abbreviation for “amplifier”). As discussed below, certain of these elements are common to many embodiments of an electron-beam amplifier.
Time Delay Control
One embodiment of electron-beam amplifier 10 provides time delay control. Variable time delay is a feature of many RF systems such as, for example, phased array antennas and wideband electronic beam steering. In such systems, radio waves radiated by antenna(s) are timed to adjust a directionality and gain of receiving or transmitting antenna(s). True time delay shifting (“TTDS”) has an advantage over simple phase shifting (“PS”) in that control is broadband, rather than narrowband. Therefore TTDS is preferred, but traditionally both TTDS and PS have been expensive and complex to implement. Thus, a low cost time delay control of electron-beam amplifier 10 may provide a useful means of antenna beamforming.
In one embodiment, an output signal (e.g., output currents 180) from electron-beam amplifier 10 is variably time delayed by adjusting electron beam energy, thus adjusting electron velocity and transit time of electrons across a drift cavity to a detector. Variable time delay control is an almost free feature of electron-beam amplifier 10, since little extra power is required and physical elements of the amplifier (i.e., electron guns, drift cavity, focusing electrodes, detectors and so on) are not altered. A microfabricated electron-beam amplifier 10 may implement time delay control over a usable range of hundreds of picoseconds, which may support electronically steered antennas for narrow steering angles at millimeter and submillimeter wavelengths. For larger antennas or longer wavelengths, which may require total time delay control on the order of nanoseconds, specialized electron-beam amplifiers 10 may be used. For the largest antennas, multiple electron-beam amplifiers 10 may be cascaded for a control range of tens of nanoseconds, or an electron-beam amplifier 10 may be used as a delay fine-tuning mechanism in a hybrid arrangement, with large delays provided by other means, such as switchable delay lines.
Generally, the velocity of electrons in a beam is given by
ve=√{square root over (2qVb/me)}, (1.38)
where q is the electronic charge (8.85×10'19 C), Vb is a beam accelerating potential, and me is mass of an electron (9.11×10−31 kg). Transit time of a beam through a drift cavity of length zdrift is simply tDELAY=zdrift/ve, and a change in delay is
Thus, by adjusting a beam accelerating potential VBEAM, the transit time may be adjusted, and a signal at an output of a detector may be delayed. For example, if Zdrift=10 mm, VBEAM=50 v, and ΔVBEAM=+/−10 v,
tDELAY(min)=2.67 ns
tDELAY(max)=2.18 ns
ΔtDELAY=490 ps.
A ΔtDELAY of 490 ps may be expressed as a phase shift Δφ of a period T of certain RF frequencies:
Δφ=49 T @ 100 GHz
Δφ=4.9 T @ 10 GHz
Δφ=0.49 T @ 1 GHz.
Typical phase shifting applications delay a signal for a significant fraction of a period of an RF frequency. It can be seen that the time delay mechanism is suitable for the RF applications that operate above 1 GHz. Furthermore, electron-beam amplifier 10 introduces no dispersion (filtering) effects when a broadband signal is amplified, since electron-beam amplifier 10 is broadband, so all frequency components are delayed by the same amount. Thus, it can be appreciated that electron-beam amplifier 10 achieves true time delay control. Detector Plane
Adjustments for Time Delay Control
One effect of changing a potential in detector plane 50 is to alter the focusing properties of electron focusing optics. For example, in electron-beam amplifier 10(4) of
This can be understood by recalling that electron-beam amplifier 10(4) has a retarding lens 1000 and an accelerating lens 1010. The retarding effect of lens 1000 occurs because electrode 1020 is more positive than electrode 1030; the accelerating effect of lens 1010 occurs because electrode 1040 is more positive than electrode 1030. Thus, if the potential of electrode 1030 is constant, making the potential of electrode 1040 more positive increases the focusing power of lens 1010. By increasing the potential of electrode 1030 as some fraction of the change in potential of electrode 1040, the focusing power of both lenses 1000 and 1010 can be decreased, offsetting the increased power of lens 1010 in the absence of a potential change on electrode 1040.
Because changes in electron acceleration accompany adjustments of time delay, changes in deflection gain may also occur, even when a lens system is adjusted to maintain focal length. Even when transverse momentum imparted to beam electrons by a signal deflector is constant (since as-emitted beam energy of electron beams 120 remains constant), when transit time is reduced by increasing acceleration, lateral displacement less time to accumulate. Accordingly, deflection of electron beams 120 is reduced by increased acceleration.
A change in deflection gain caused by acceleration is independent of lensing action of a detector plane electrode (e.g., electrode 1180 of
One means of biasing ring electrodes 1290(1-4) includes potentials derived from a set of resistors 1330(1-5) with respective values RA, RB, RC, RD and RE, connected in series. As shown in
Electron Gun Adjustments for Time Delay Control
Adjusting potential of an electrode in detector plane 50 has advantages over adjusting an electron gun acceleration potential; adjusting potentials in an electron gun may affect deflection gain, and beam energy adjustments to a electron gun may be difficult due to complex electron gun electrode structure. Thus it is preferred, for most applications, to keep electron gun beam energy constant. Nonetheless, some applications of electron-beam amplifier 10 may benefit from a constant detector plane potential, such as for example applications which employ multiple independent e-beams, as discussed below. In these applications, time delay control may be achieved by adjusting electron gun acceleration potential.
Once electron beams 120 exit electron guns at an emission plane and enter a drift cavity, changes in beam energy affect beam focusing in this method, unless otherwise compensated. The reason is that the potentials of electrodes in a doublet lens system (e.g., electrodes 1160, 1170 and 1180 forming lenses 1190 and 1200 in
Gain Stabilized Time Delay Control
Time delay changes effected by altering the beam energy, either by electron gun adjustments or detector plane acceleration adjustments, may be accompanied by changes in both deflection gain of the beam and cascade gain of the detector. Thus, the overall amplifier gain is changed. As described earlier, amplifier transconductance is given by
This calculation assumes that one detector segment receives all available beam current at a maximum deflection signal voltage. Altering deflection gain is effectively the same as changing detector width XD. For example, increasing beam energy reduces transit time of beams through a cavity; XD decreases correspondingly. At the same time, increasing beam energy increases detector gain kD. The changes in XD and kD both increase gm when beam energy increases Likewise, decreasing beam energy decreases gm.
For this reason, amplifier gain may be stabilized by adjusting e-beam current. From the preceding equation, it is clear that changes in XD and kD can be compensated by changing the beam current. As beam energy is increased, beam current is decreased, and vice versa. For each change in detector plane potential, the electron gun currents are adjusted to maintain constant average output current.
Gain Controlled Amplifier
From the preceding, it can be appreciated that an electron-beam amplifier 10 may use a gain controlled amplifier. One method by which this can be accomplished is by implementing any of the methods of time delay control, but without current controlled gain stabilization. Another method is by a current controlled beam without beam energy adjustments. Finally, amplifier gain can be adjusted via beam energy adjustments working in concert with a current controlled beam, a difference being that current control works in the opposite sense of gain stabilization, so that it enhances the gain variation induced by the time delay control.
Pulsed Operation
Electron gun beam blanking is easily implemented in an electron beam amplifier 10. One application of electron gun beam blanking is an RF transmit amplifier that generates pulsed beams. This is beneficial for applications like radar and Ultra-Wideband (UWB) communications. With beam blanking, a continuous RF signal can be applied to deflection plates, and the amplifier output can be turned rapidly on and off with pulse widths as short as 10 picoseconds, without interrupting the RF signal.
Pulsing can be achieved by various means, for example, through gate electrode control, and through the inclusion of an extra deflector in each electron gun, called here a “blanking deflector.” Cathode control may involve a high loading capacitance and a slow response time. In many applications, such as radar and UWB, sub-nanosecond switching is desirable and cathode controlled gating is too slow. A blanking deflector has high-speed characteristics like other deflectors described above (e.g., deflector 130(1)) including very low loading of a driving source.
As in other electron beam amplifiers 10, electron guns with blanking capability can be arrayed to create a composite e-beam from many individual beams, and all such blanking deflectors may be coupled together under control of a single blanking signal.
Frequency Multiplication
Some high frequency applications utilize both frequency multiplication and amplification; for example, high-frequency oscillators, high-frequency references for TWTs and other high-power amplifiers, and RF carriers for radar transmitters and communications systems.
Frequency multiplication at RF frequencies is sometimes achieved by driving a non-linear element with a sinusoidal signal and filtering a resulting waveform with a tuned filter to extract a higher order harmonic. The principle can easily be grasped by considering simple second order non-linearity, y=x2. If the value x=cos ωt, the value y=(1+cos 2ωt)/2, so the frequency has been doubled. Higher order non-linearities can generate higher frequency multiples. However, extra filtering is required to extract the desired harmonic, and the process may be inefficient, since harmonics have energy that diminishes roughly in proportion to the order of the harmonic. For example, a 5th harmonic normally has much less energy than the 3rd harmonic.
A frequency multiplying electron beam amplifier 10 may provides efficient harmonic generation, even for higher orders. The method employs a detector with a multiplicity of segments greater than two, and may use one or two deflectors arranged for deflection in two orthogonal directions (e.g., directions X and Y of
By increasing a number of detector segments, higher order frequency multiplication may also be achieved. With a linear row arrangement, 6 segments achieves frequency tripling, 8 segments achieves quadrupling, and so forth; furthermore, frequency multiplication can be controlled by controlling the amplitude of an input voltage.
There are two limitations of the simple linear array. First, high orders of multiplication may require a wide layout of detector segments, and require a correspondingly large scan angle which may exceed the range of a deflector and voltage signal. Second, it may be difficult to achieve exactly periodic spacing of zero-crossings of a multiplied frequency output with a linear array of segments. The effect of aperiodic zero-crossings may depend on an application. In an RF mixer, spurious tones may be generated that can limit the sensitivity of a receiver. If an application is as a frequency reference for an analog-digital-converter (ADC), the aperiodic crossings may create sampling errors and limit conversion accuracy.
One method of achieving periodic zero-crossings is to adjust detector segment geometry to balance dwell times of a beam over all segments to lower the undesired harmonic content in the output. Detector segments 150(47-54) are arranged to compensate for the effect of a sinusoidal sweep pattern that spends more time on outermost regions of a sweep and less time on inner regions of the sweep. A beam spot (not shown) may scan all of segments 150(47-54), but the beam spot will spend more time on wider segments 150(50) and 150(51) due to their width, will spend less time on narrower segments 150(49) and 150(51), and so on.
Circular Frequency Multiplier
Another method of achieving periodic zero-crossings employs a circular detector with “pie-slice” segmentation and two-dimensional scanning that sweeps a beam in a circular pattern (for example, forming traces known as “lissajous figures” in the field of electron beam oscilloscopes).
A circular detector used with a beam swept in a lissajous pattern has an inherent tolerance with respect to variations in input signal amplitude. As long as a lissajous pattern formed by beam spot 170(17) stays centered on and within segments 150(56-59), the amplitude of Vx and Vy may vary without affecting an amplitude or duty cycle of an output waveform on output lines 183(1) and 183(2). Centering of the lissajous pattern on the detector may be ensured by means of beam centering arrangements, as described above. Nonetheless, there may be an optimum amplitude of Vx and Vy for a given beam spot shape that will minimize harmonic distortion in the output waveform.
The phase offset between voltage signals Vx and Vy may also be useful where phase offsets other than 90 degrees may lead to aperiodic zero crossings, which are equivalent to skews in duty cycle from the 50% duty cycle characterizing a sinusoidal output centered about a value of zero. Altering a phase offset between voltage signals Vx and Vy may be used to tune the duty cycle of an output waveform.
Detector 151(11) includes six output segments 150(61) through 150(66), with alternating segments connected to positive and negative output terminals as shown by the + or − sign within each segment. Detector 151(11) generates an output waveform with a frequency that is triple an input frequency applied to X and Y deflectors used to steer beam spot 170(18).
Other embodiments of an electron-beam amplifier 10 using X-Y deflection may optimize detector shape for low distortion or high frequency operation, such as, for example through use of an elliptical detector, or a segmented ring detector.
Other Frequency Multipliers
A multiply segmented detector is only one means of achieving frequency doubling. For example, in another electron-beam amplifier 10, frequency multiplication is achieved with a single detector segment. By appropriately shaping a detector and/or a beam spot, harmonic components may be emphasized as the beam spot sweeps across an edge of the detector. Emphasis of harmonic components results from a non-linear change in beam current collection with respect to beam spot position. An electron-beam amplifier 10 that multiplies an input frequency through shaped, single beam spots and detectors may generate output frequency tones that are not as pure (i.e., free of harmonics) as in multiple segment embodiments, but smaller, faster detectors and simpler microcolumns (i.e., with only one deflector instead of two) may be used.
It is also possible to make a beam spot 170(21) triangular and a detector segment 150(68) rectangular, as shown in configuration 151(14). Again, collected current changes quadratically in relation to a sinusoidal beam sweep. The triangular shape of beam spot 170(21) may be generated by the methods discussed above, including use of a triangular shaped microcolumn array imaged onto a detector plane. Configuration 151(14) may offer a somewhat smaller, faster detector, and illustrates the principle that it is the relation of beam spot to detector shape that is useful in generating a desired output.
Other shapes may be used to generate even higher frequency multiplication factors.
Mixer
RF mixing is another application of an electron-beam amplifier 10 that may multiply a frequency and generate intermodulation products of two frequencies.
Beam spot 170(23) will move in the X and Y directions across detector segments 150(71-74) so as to cause a differential current ΔIout across detector outputs 183(3) and 183(4) to have a fundamental frequency component at a frequency difference f1−f2. Harmonics that may exist in ΔIout may be filtered according to means known in the art.
In configuration 151(17), detector segments 150(71-74) each have a width and height of 2 W; square beam spot 170(23) is also of width and height 2 W, and has a uniform cross-sectional current density J. Beam spot 170(23) is deflected in an X direction in response to Vx and in a Y direction in response to Vy, instantaneous deflections in these directions are called Δx and Δy respectively, and Δx and Δy are linearly proportional to signals Vx and Vy. Currents generated from each of detector segments 150(71-74) are I1, I2, I3 and I4, respectively. These currents vary in response to beam spot deflections Δx and Δy, as shown below
I1=J(W+Δx)(W+Δy)
I2=J(W−Δx)(W−Δy)
I3=J(W−Δx)(W+Δy)
I4=J(W+Δx)(W−Δy) (1.42):
When the beam spot is centered, each segment receives a current J W2. Currents I1 and I2 are coupled to drive terminal 183(3) to form current IB and segment currents I3 and I4 are coupled to drive terminal 184(4) to form current IA. Net output currents IB and IA to terminals 183(3) and 184(4), respectively, are
IB=I1+I2=2J(W2+ΔxΔy)
IA=I3+I4=2J(W2−ΔxΔy) (1.43):
Differential output current ΔIout is given by
ΔIout=IB−IA=4JΔxΔy (1.44)
Thus, the action is that of a multiplier.
As known in the art of RF receivers, a multiplier is a basic element of many mixers. This may be seen when Δx and Δy are proportional, respectively, to sinusoids of amplitudes X0 and Y0, and frequencies f1 and f2:
Δx=X0 sin(2πf1t)
Δy=Y0 sin(2πf2t) (1.44):
As may be derived using the Law of Cosines,
This shows the sum and difference frequencies characteristic of a mixer. In certain RF applications, the sum frequency is removed by filtering, leaving a difference frequency (f1−f2) representative of an intermediate (IF) or modulation frequency.
It may be appreciated that e-beam spot deflections Δx and Δy are generated according to the basic principles of electron-beam amplifier 10. When scan deflections Δx and Δy are small with respect to the dimensions 2W of the spot, a linear multiplication is effected. When the scan deflections are large such that Δx and Δy approach or exceed the spot half dimension W, then a “bang-bang” rectifying type mixer is achieved, operating similar to known circuits which employ active switches, such as MOS transistors, or diodes.
Combinational Logic
Combinational logic is an application for an electron-beam amplifier 10 that resembles the mixing and frequency multiplying embodiments discussed above, but which operates in a different parameter space and for a different purpose. A combinational logic embodiment may include a short drift cavity and multiple deflectors, and may have only one electron gun per logic element. Detectors in combinational logic embodiments may have two or more segments. Voltage signals for Deflectors may be logic signals of binary or multiple quantized voltage levels. Combinations of quantized voltage input states correspond to quantized beam deflections, each quantized beam deflection being representative of a logic state formed by the combination of input states. By positioning detector segments at locations corresponding to quantized beam positions, the detector outputs may be representative of respective logic states. By this means, logic operations, such as AND, OR, XOR, and even complete functions (such as, for example, a full adder) may be constructed. With the inherent advantages, including high-frequency operation and microfabrication, it can be appreciated that combinations of logic elements can be incorporated as complex arithmetic units, digital multipliers or memory elements that operate at picosecond speeds.
The basic principle of a combinational logic embodiment is that if a signal representing a quantized logic value, for example a signal that may be −1V or +1V, is applied to an e-beam deflector, then the corresponding beam may be deflected to one of two states, corresponding to deflection angles, for example θ1 or θ2. If a second deflector that is likewise responsive to a signal representing a quantized logic value is incorporated, the number of possible states increases to four, such as beam angles θ1, θ2, θ3, θ4. With three deflectors, the number of possible states is 8, and so on. The principle may also be extended to multi-valued logic; for example, if 4-level logic signals are applied to two deflectors, the beam angle may have 16 states.
Only one detector is activated for each state, but this shows that two of the binary states have the same deflection angle (0). This is reflected in
A linear arrangement of deflectors and detectors may require a large deflection range when multiple inputs are used. For example, a binary deflection state corresponding to identical deflection angles applied to three successive deflectors may involve three times the deflection angle of a state in which only one deflector is active. Accommodating the deflection range necessary for all logic states may be difficult; this can be mitigated by use of a long drift region, but this increases the drift time of the beam, thus slowing the maximum switching speed and the latency of associated logic operations.
An electron gun microcolumn 610 may have multiple X and Y deflectors for logic involving more than two inputs. For example, for three logic inputs, a microcolumn may have two X deflectors and one Y deflector. For four logic inputs, a microcolumn may have two X deflectors and two Y deflectors. With X and Y deflection, the logic states are described by a two-dimensional set of beam states, detected with a two dimensional array of detector segments. The result is similar to creating a physical Carnaugh map, as known in the art of logic devices.
For the case of four logic inputs described above, the corresponding 16 logic output states are detected with a matrix of three rows and three columns of detector segments.
An examination of this table for particular detectors segments shows that degenerate states correspond to some form of exclusive—or combination; for example, detector segments 150(83), 150(86) and 150(89) correspond to A ⊕C, while detector segments 150(85), 150(86) and 150(87) correspond to B ⊕D.
Despite the degeneracy observed, orthogonal deflection drive is a preferred construction; it still minimizes degeneracy as compared to a linear array configuration, and a deflection required in each of the X and Y directions is smaller than would be required in a linear detector array configuration. Smaller deflection allows a proportionately shorter drift region, shorter drift time and smaller deflection drive voltages. For example, with only two deflectors, one in X and the other in Y, drift distance and time may be reduced by one-half when compared to a pair of X deflectors; correspondingly, logic switching operations occur twice as fast. Alternatively, for a given drift distance, a deflection voltage may be smaller (for example, 0.5V versus 1V) so that power consumption may be reduced or switching speed may be increased.
It may be appreciated that degenerate states are not the only way to combine logic states. In the case of
By “wire-oring” (as it is termed) deflector inputs and/or detector outputs using electrical connections, other logic functions may be implemented, providing great flexibility in a simple structure, since any of these means may switch almost as fast as any other. This is unlike conventional logic gates made from transistors, where certain gate types are much slower than others. For example, a CMOS NOR gate is slower than a CMOS NAND gate; also, conventional static CMOS logic lacks an inherent complement output, which must be generated with a second inversion gate, adding to switching delays. An ECL or current mode gate suffers loss in performance because multiple transistors are required for complex functions, and due to having a limited power supply range. In contrast, logic embodiments of e-beam amplifier 10 may be fast in almost any logic combination, because the logic function is encoded as a beam position (or state), rather than as a combination of switches.
Other combinations are possible. For example, deflectors may be physically designed to achieve more or less deflection for a given input voltage (“deflection gain”). One deflector might have a deflection gain of 10 degrees beam deflection per volt of deflection drive, while another deflector might have a deflection gain of 5 degrees per volt. As described above, longer or shorter deflector plates will alternately increase or decrease deflector gain; spacing deflector plates more closely or further apart will also increase or decrease deflector gain, respectively. By using deflectors with varying amounts of deflection gain, beam deflection states may be gray-coded to eliminate degeneracies and make detection more resistant to errors. These two goals follow directly from use of multiple deflection gains.
Gray coding is a well-known method of digital word encoding whereby single bit errors in the word cause only one bit of error in a digital count represented by a word. Gray-coded operation is useful for specialized functions often found in communication systems, where robust signaling that is tolerant of small errors is necessary. In electron-beam amplifiers 10, gray-coded beam states make detection resistant to single bit errors in beam displacement.
For example, if logic states A and B represent a binary number with A the most significant bit (“MSB”) and B the least significant bit (“LSB”), it can be seen that a maximum error in the output generated by a single logic state error (perhaps due to a noise glitch at an earlier stage of digital processing) may be 1 LSB. In contrast, the previous 2-input gate could exhibit a 1 MSB error. Gray-coding may be extended to more bits, as is known in the art.
One aspect of a logic gate may be that logic levels are compatible between gate inputs and outputs. In certain embodiments of an electron-beam amplifier, a difference in potential between detectors and deflectors may be up to several hundred volts. If the logic switching is dynamic enough, this potential difference may be accommodated with capacitive coupling.
Another means of logic level compatibility is to ensure that detector output levels are the same as deflector input levels. One method of keeping these potentials compatible is to use a zero bias drift cavity in which an exit plane of an electron gun is at the same potential as a beam contact and a detector plane (i.e., allowing electrons to drift from deflector to detector through a field-free region). Since a deflector is inherently a differential input device, a common mode level can be rejected to some degree, and detector output can be directly coupled to the deflector.
For logic operation, a suitable detector bias is less than 1V. This is consistent with an extremely high-speed device. Logic devices may use faster, lower bias detectors than amplifiers, since power is not required or desired. Operation at less than 0.5V is possible when detectors are Schottky diodes with turn-on potentials of around 0.2 to 0.3V.
A detector may be terminated in either a resistor or an active load, such as a resonant tunnel diode (RTD). When a resistor is used, beam current may pull down the output potential of the detector to the beam contact potential; this is a logic “0.” Without beam current, the resistor acts to pull up the output potential to the power supply voltage, representing a logic “1.” An RTD load behaves similarly, except that an RTD has a negative differential resistance, so the pull-up and pull-down are speeded up for faster operation.
As mentioned previously, it is desirable to operate e-beam logic elements with a single electron gun per gate. Because a very short drift region is required for low gate delay (a few microns), a single gun can tolerate higher beam current without space charge spreading causing beam defocusing during the drift time.
Nonetheless, a low beam current is still preferred to reduce detector heating. For this reason, detector gain should be as high as possible, but this conflicts somewhat with the requirement of high deflection gain. On one hand, high deflection gain is achieved with a low-energy electron gun; on the other hand, high detector gain is achieved with a drift cavity field that accelerates beam electrons to achieve high cascade gain. If the drift cavity is field free, all the cascade gain may come from the electron gun acceleration. One solution is to accept the lower cascade gain and compensate with higher avalanche gain in the detector. For example, photonic detectors with avalanche gains exceeding 1000 are relatively common. The downside is less radiation tolerance, which might be acceptable for many applications, and might be offset by a slightly higher beam current. For example, an electron-beam amplifier 10′ for an amplifying application might have a beam current of 1 μA, a cascade gain of 100, an avalanche gain of 10 and an overall detector gain of 1000; an electron-beam amplifier 10″ for a logic application might have a beam current of 2 μA, a cascade gain of 20, an avalanche gain of 25 and an overall detector gain of 500. The higher beam current of electron-beam amplifier 10″ provides the same detector output current, and almost entirely compensates for an increased radiation sensitivity due to the 2.5× higher avalanche gain.
In electron-beam amplifiers 10′ and 10″ above, the detector current is 1 mA; this may be inadequate for the highest speed operation, so even higher beam current and avalanche gain may be required. For example, a 50 ohm load, 500 mV switching application may require at least 10 mA detector current; avalanche gain may be increased by a factor of 10, or beam current may be somewhat (which may be tolerated because of a very short drift cavity). Beam current might be increased to 4 μA and avalanche gain increased by a factor of 5, or the beam current increased by a factor of 3× and the avalanche gain increased by a factor of 3.3. An advantage of sharing the gain increase between beam and detector is, again, to reduce radiation sensitivity.
As mentioned, a drift cavity of an e-beam amplifier 10 in a logic application may be very short, to minimize transit time of a beam. Beam delay directly affects a maximum cycle time that the logic can operate at. For example, if two deflectors are 1 μm long each, with a 1 μm drift cavity, the total drift distance is approximately 3 μm. For a 50V beam (with a velocity of 4×106 m/s), transit time from the input of a first deflector to a detector is 750 femtoseconds (10−15). This suggests an upper switching rate limit of around 1 THz.
Gate loading delays can also be estimated, by way of example. With a 1 um drift cavity, detectors may be on the order of 0.25 μm×0.25 μm in size. Junction devices such as Schottky diodes typically have capacitances on the order of 1 fF/μm2. Thus, a detector capacitance may be approximately 0.125 fF. The loading of a single deflector with plate spacing of 1 μm, a plate length of 1 μm and a plate height of 1 μm is 0.009 fF. For a 50 ohm load, capacitance is very dependent on construction, but may be well under 1 fF, so a value of 0.5 fF will be conservatively assumed here. Thus, a total loading capacitance may be 0.125 fF+0.009 fF+0.5 fF, or approximately 0.75 fF. The fall time when a detector turns on is dominated by pull-down current times into the total loading capactance, given by dv/dt=I/C. With a 500 mV power supply and a 1 mA beam current, a fall time may be 375 fs. A rise time when the detector turns off is approximately the RC time constant of the load resistor and capacitance, or, 50 ohms×0.75 fF=37.5 fs. These figures are approximate and will depend strongly on the application, but they demonstrate rise/fall times on the same order as the gate delay, thus an e-beam amplifier 10 used in a logic application may have switching speeds on the order of 1 THz.
As with other embodiments of an electron-beam amplifier 10, detectors provide gain with respect to collected beam current. This gain is essential if a single electron gun is to be used, which may be a preferred construction when many logic elements are combined in an integrated processor or other complex logic system.
Since detector gain is not precise, diode means may be used to limit detector output voltage to controlled binary logic levels. Schottky diodes are preferred, since they are readily available from the detector construction, and they are among the fastest clamping devices known.
A beam 120 is configured by an electron gun and focusing optics (not shown) to strike detector segment 150(97) or 150(98). A detector 150 that is not struck by beam 120 isolates a corresponding output 1560 from power supply 1550(2), allowing the corresponding resistor 1570 to pass a current IR so that the corresponding output 1560 reaches the potential of power supply 1550(1). In this illustration, detector 150(97) is not struck by beam 120, current IR passes through resistor 1570(1), and output 1560(1) reaches the potential of power supply 1550(1), but it will be appreciated that the circuit symmetry is designed to produce an equal effect on detector 150(98), resistor 1570(2) and output 1570(2) if the beam strikes detector 150(97).
A detector 150 that is struck by beam 120 emits an output current ID that drives the potential of a corresponding output 1560 until the corresponding output 1560 reaches a clamp potential of the corresponding clamping diode 1540. When current ID changes the potential of output 1560 to exceed the clamp potential Vclamp, clamping diode 1540 passes a current IC that prevents any further change to the potential of output 1560.
Thus the potential of an output 1560, corresponding to a detector 150 struck by a beam 120, will achieve the potential of power supply 1550(1) offset by the clamp potential Vclamp. It should be noted that the potential of power supply 1550(2) may be positive or negative with respect to power supply 1550(1) as a matter of design choice, for implementing suitable logic levels and choices of detectors 150 and clamping diodes 1540. The diode symbols used in
Power Combining Arrays
Ganging amplifiers is one way to increase the power output of amplifier embodiments while maintaining a wide signal bandwidth. Ganging may exploit a high input impedance of the deflector apparatus, such that many amplifiers may be driven from a common low-impedance source, for example, a 50 ohm transmission line.
The principle obstacle to ganging amplifiers is not input loading, but power-combining many outputs. In conventional technologies, such as solid state amplifiers, this type of combining may present a formidable problem. Simple electrical networks made of transmission lines or waveguides have significant ohmic losses that can drastically reduce the efficiency of the power summing, especially in large arrays. Efficient power combiners generally take two forms: waveguide combiners and free-space summing of electromagnetic waves. Waveguide power combiners suffer from ohmic losses, and are difficult to construct in a microfabricated form. The hierarchical structure of combiners, such as the Wilkinson type, also makes them suffer from wave reflections at the many summing nodes, resulting in high standing wave ratio and more lost efficiency.
As described below, free-space summing of electromagnetic waves is a preferred method of power-combining since there are no ohmic losses or standing waves. With free-space summing, amplifiers are coupled to radiating antenna elements, and the radiated fields naturally combine by coherent superposition. It is only desirable that the amplifiers be driven from a common signal input or sources that have the same frequency and similar phase. In many applications, these free-space fields may be used directly, as in a radar or communications transmitter. The effect of the phasing may, for example, create a directional RF beam. In other applications where RF radiation is not desired, the coherent sum can be collected in another, larger antenna, such as a horn or parabolic dish.
Thus, a radiating EBTX embodiment 4000 shown in
These radiating embodiments are termed here the EBTX (Electron Beam Transmit Amplifier) since they may amplify, as in a receiver mode, as well as transmit an electromagnetic field. Thus, free-space fields may be efficiently summed in large power generating arrays such as a phased array antenna.
Since EBTX amplifiers can be microfabricated the loading of many elements can be distributed by a hierarchical input feed constructed from EBRX amplifiers (EBTX sans antenna). By this method thousands or even millions of power combining elements can be constructed as entire wafer-based assemblies.
One radiating embodiment couples the detector of an EBRX to a separate antenna element via a short transmission line. In this case, the e-beam detector sees the network impedance of the transmission line, and the antenna accomplishes the impedance transform to free space. The antenna may be placed as closely as possible to the e-beam driven detector and uses integrated microfabrication technology to achieve a proximity of microns. Given the small dimensions of a microfabricated element, this may limit the antenna to a maximum size of some millimeters. Thus, radiating embodiments are most suitable for millimeter wave and sub-millimeter wave applications, which corresponds to a frequency spectrum of approximately 40 GHz and to 1 THz (K-band and above).
The nature of the microfabricated construction makes various types of strip and slot antennas compatible for coupling to the detector in forming an EBTX. These can be formed, for example, using multi-level metallization processes that are found in many microfabrication technologies. The most common types of strip and slot antennas are resonant structures such as the dipole and patch antenna, but there are also many broadband types, including the log-periodic, various forms of wideband spiral antenna, the wideband vivaldi flared type, and ultra-wideband structures.
Dipole
The power output of a single dipole can be estimated from P=V02/2Z0, where V0 is the peak sinusoidal voltage fed to the antenna, and Z0 is the theoretical feed impedance of the dipole. V0 is approximately ½ the detector reverse bias voltage since voltage excursions outside this range will de-bias the detector. For a 2V reverse bias, V0=1V. From these quantities, the power output of a dipole is approximately 7 mW.
Selectable Dipole Polarization
A dipole provides a single plane of polarized electromagnetic radiation. Many applications require selectable polarization.
As for other embodiments, the X and Y sweeps may be achieved by arrays of electron guns that each have X and Y deflectors
In another arrangement, a square beam spot 4118 is employed for both polarizations, as shown in
Broadband Antenna
Patch Antenna
A patch antenna 4172 is shown in
Selectable Patch Polarization
As with the dipole, a selectable polarization is possible with a patch antenna 4172, but in this case, by moving the feedpoint 4180.
The aiming may be accomplished as shown in
In another arrangement, two beams may be employed to achieve the selectable polarization, as shown in
Strip and Slot Antennas
In any antenna embodiment, the antenna can be constructed as either a strip of metal or a slot in a ground plane. These two configurations are based on swapping the conducting and non-conducting materials of the antenna geometries. Thus, a “slot” dipole antenna may look like strip, except it is mostly ground plane with two narrow slots in the shape of the antenna. Feeding arrangements between strips and slots are somewhat different due to the need to have a conductive contact, but performance is similar, though in some applications the slot can provide slightly better bandwidth and cross-polarization performance. In the literature, the strip and slots are known as “duals” of each other because of the geometrical similarity. Thus, it can be appreciated that the invention is not constrained to use one type or the other.
Integrated Detector/antenna
In another embodiment as shown in
The operational modes of EBTX 4206 include antenna radiation, polarization control, and harmonic generation. The basis for these modes is the fact that the beam spot can be deflected over a large area of the antenna. The beam deflection may span up to a half wavelength or more of the highest signal frequency and move the full length of the antenna, or the spot can simply be repositioned anywhere along the antenna and modulated with a small signal amplitude. Large amplitudes generate harmonics, while small amplitudes at particular positions can generate different polarizations and phases. By way of example, where the antenna 4210 is in the form of a strip-patch antenna,
The operation can be understood as follows. Where the e-beam 4214 strikes the beam contact 4216, relatively strong current flow between beam contact 4216 and the output contact (anode and cathode) because of the gain of the detector 4208 (see
The traveling wave is such that the edges of the patch look something like a transmission line terminated by the radiation impedance. Any mismatch in the impedance of the transmission line and free-space causes the traveling waves to be reflected. The waves therefore propagate back and forth through the patch detector 4208 establishing complex standing wave patterns. If the beam spot moves very little, the wave patterns are modulated at the frequency of the spot movement, and the patch will radiate at the same frequency. If the spot moves over a larger area, non-linear effects emerge because of interactions between waves generated at different positions of the patch, and the patch radiates harmonics as well.
The patch is generally a unique two-dimensional shape that may be adapted for a particular environment of use, though
The patch/detector concept may assume any geometry, including novel geometries or shapes. For example, as shown in
Generally speaking, efficient excitation of a diode detector/antenna structure requires an e-beam scan pattern that closely approximates the surface current density pattern of the antenna when radiating in a desired mode. This is one reason why the embodiment may use multiple e-beam spots with complex excitation, or may employ unusual antenna/detector shapes.
Because of the complexity of the device operation, the types of antenna shapes and scan patterns can only generally be indicated here. In practice, the exact construction may benefit from computer simulation and experimentation to determine the exact number of independent beams, together with the amplitude, position and scan pattern of each beam sweep for an intended environment of use. This may in turn determine the other parameters of the amplifier, including the number of electron guns, deflector drive, drift cavity dimensions, and focusing requirements, among others. It can be appreciated, however, from the general principles exposited here that the embodiment can combine the functions of antenna, frequency multiplier, phase shifter and selectable polarizer in a single device and thus offers an unusual flexibility.
Horn
In another embodiment as shown in
As shown in
Waveguide Coupling
TE10 is not the only mode that can be excited in a guide, but it the easiest mode to implement and describe, and so is shown by way of example. Besides the relative ease of guidewall excitation, a TE10 mode also has the lowest cutoff frequency of any rectangular waveguide mode, therefore offering the widest bandwidth. This bandwidth can span many octaves, making the waveguide fed horn much more useful than resonant dipoles or patch antennas for many applications.
A circular waveguide 4302 can also be used in place of waveguide 4270 (shown in
Dual Polarization Circular Waveguide
Beam spots 4324, 4326, 4328, 4330 excite the four detector segments 4314, 4316, 4318, 4320 with two independent deflections. Spots 4324, 4328 are moved vertically in unison to excite segments 4314, 4316. Spots 4330, 4326 move horizontally in unison to excite segments 4318, 4320. Spots 4324, 4328 move independently of spots 4326, 4330 to excite the waveguide 4302, and in this manner simultaneous dual polarization is achieved.
A central gap 4332 between segments prevents segments 4314 and 4316 from coupling to segments 4324 and 4326. A separation distance gap between beam spots 4324, 4328 matches the gap dimension between segments 4314, 4316, for example, so that as beam spots 4324, 4328 move up and down, the excitation of the segments 4314, 4316 changes in a uniform manner. The same considerations apply to the horizontal motion of spots 4326, 4330 exciting segments 4320, 4324. A diagonal polarization occurs when X and Y sweeps are driven in phase. A circular polarization occurs when X and Y sweeps are driven 90° out of phase at the same amplitude. An elliptical polarization occurs when X and Y sweeps are driven 90° out of phase at different amplitudes.
Capacitively Coupled Circular Waveguide
Aperture Antenna
A waveguide may also be used directly as an aperture antenna, without a horn. Though the directivity of a simple aperture is lower than a horn, in large arrays of apertures, free-space power combining improves the directivity substantially. In this kind of application the lesser directivity of the aperture is actually a benefit, since it permits beamsteering over a wider angle.
An aperture radiator also has one advantage of being much smaller than a horn, and therefore a high density of apertures can be used in large arrays for greater power output. Generally, horns are more appropriate for achieving high directivity from small arrays. Finally, the aperture retains the broad bandwidth of a horn, which far exceeds a dipole or patch.
That a waveguide has a broad bandwidth can be understood from the relation for group wave velocity of a guide:
Generally, the shorter the wavelength (or higher the frequency f relative to cutoff fc=c/2a), the more closely the group velocity approaches the free-space velocity of light, c. Thus, short wavelengths propagate at almost the same velocity and over short guide lengths there will be little dispersion. When the guide couples a detector on one side of a thin silicon wafer substrate (˜300 um thick) to an aperture on the other side of the same wafer, the dispersion will be negligible even at 1 THz.
Waveguides offer significant power advantage per element over simple antennas such as a dipole or patch radiator. The reason is that dipoles and patches have a relatively high feed impedance relative to the area of the antennas, in the range of 50 to 100 ohms. This limits the maximum current drive for a given detector bias voltage. Higher electromagnetic power feed can be achieved in a waveguide because the driving impedance can be lower for the same area. If the transmission impedance for a TE10 mode of a rectangular waveguide is ZT, the electrical impedance is
for a guide of width a and height b. The TE10 mode propagates down the waveguide by reflecting back and forth off the two sidewalls separated by the width a. ZT is given by
where the free-space radiation impedance η=377 ohms, the cutoff frequency fc=c/2a , and the speed of light c=3×108 m/s. The angle of reflection normal to the guidewall is given by θ. For guides of width a>>λ/2, the wave propagates nearly with the speed of light and the transmission and electrical impedances are minimum. For example, a guide that is a=2λ wide and b=λ/10 high will have ZT=389 ohms and Z0=20 ohms. At 100 GHz a=6 mm and b=0.3 mm. At 1 THz, a=600 um and b=60 um.
Thus, the lower electrical impedance of a wide guide permits more power to be transmitted from a low voltage source, such as an e-beam detector. This is one advantage of a waveguide over an antenna. Generally, the power down a guide as a function of the peak driving voltage is given by
V0 is approximately one-half the detector reverse bias voltage, since voltage excursions outside this range will de-bias the detector. For example, if the detector bias is 2V and Z0=20 ohms, the power output will be approximately 25 mW. This is over three times more power than the power from a half-wave dipole (Z0=73 ohms).
Since the long dimension of the waveguide is approximately the same as a dipole antenna (a˜λ, b<<λ), but the short dimension can be considerably less, arrays of guide-coupled EBTXs can have many more elements per unit area as arrays of dipole-coupled EBTXs, which are normally restricted to a one-halfwave separation in both directions. For example, a small array of 4 dipoles will be approximately λ×λ in area. This same area can have 10 waveguides of dimension λ×λ/10, and each guide will generate 40% more power than a dipole. The total array power will be 3.5 times more than the array of dipoles on a λ/2 element spacing. Thus, even a relatively narrow guide can generate higher power in an array.
Transmit Arrays
As discussed previously, antenna coupled amplifiers provide means for coherent power combining via arrayed embodiments.
One way to provide an efficient power combiner is as a dense array 4370 of microcolumn subarrays 4372, 4374 with integral local focusing optics over each microcolumn subarrays 4372, 4374. This is shown in
Arrays of RF emitters can be packed more densely than λ/2, as shown schematically in
With microfabrication, very large arrays and high radiated power are possible. A single wafer-fabricated transmit array might have more than 1 million elements. This is achievable at submillimeter wavelengths if standard 200 mm diameter silicon wafers are employed in the construction. This many elements cannot be driven directly, but as shown in
Transmit Beamformer
Transmit arrays can be extended to beamforming by employing time delay control of each amplifier element. The concept of a beamformer is an array of antenna elements that are independently controlled for time delay or phase to generate a beam or beams in designated directions. As mentioned before, phase control works for narrowband signals, and time control works for broadband signals. Time control is the more general concept, and the principle is shown in
where c is the speed of light and Δx is the element spacing, as shown in
Frequency Multiplying Radiating Beamformer
By constructing a detector according to the frequency multiplying embodiments described previously, the input frequency to the transmit beamformer can be a sub-multiple of the output frequency. One advantage is that very high frequency radiation can be generated from a low-frequency reference. Generally, a stable reference of pure tonal quality is more easily constructed if it is low-frequency, and is therefore preferred. In a large beamformer, there is the further advantage that a lower frequency signal can be distributed with lower losses through a corporate network of amplifiers and transmission lines.
Receive Arrays
EBRX amplifiers may be constructed in arrays to improve the performance of an RF receiver, in the same manner as EBTX amplifiers can be used to make transmit arrays. The same principles of beamforming apply, but in reverse.
According to one embodiment, a large antenna is constructed from an array of smaller unit antennas such as dipoles, patches or horns. Each unit antenna is coupled to the input of an EBRX and the combination comprises an element of the array. As shown in
b(t)=Σrn(t−Δtn). (1.51)
This function can be realized by many methods. One employs mechanical switching of transmission lines to generate the elemental delays Δtn, and electrical power combining to generate the summation. For example, one kind of power combiner 4460 is a corporate-fed Wilkinson combiner.
One embodiment generates a beam signal b(t) by quantizing the signals rn(t) with an analog-to-digital converter (ADC) coupled to the output of each element. The delays of each element and the power combining of all elements are generated with digital signal processing. This method can re-process the rn(t) signals M times with different sets of delays to generate M beams. Furthermore, the digital signal processing can selectively filter the resultant beams.
Another embodiment incorporates time delay control means in each EBRX to receive time delay control signals Δtn. Each output rn(t) is summed in an electrical power combiner to generate the beam signal b(t). The limitation of this approach is that only a single beam can be generated, but the benefit is the simplicity of the time delay construction and the beam generation.
Another embodiment achieves multiple beam formation by incorporating multiple EBRX amplifiers in each antenna element. As shown in
bm(t)=Σrnm(t−Δtnm). (1.52)
In a further improvement on this embodiment, an extra EBRX (not shown) may incorporated in each element to isolate the antenna from the loading of the M beamforming EBRXs. In this manner, the signal power can be further amplified before power combining, thereby overcoming losses in the combiner and improving the signal level.
Analog Beamforming Mixer
A related improvement integrates mixing action into the receiver array. One variant of an EBRX includes a mixer element (e.g., including beam spot configuration 151(17) shown in
Electron Beam Power Combiner
Another embodiment is an improved power combiner. The embodiment comprises k microcolumn arrays having independent deflectors, k beam offset means coupled to each deflector, a drift cavity, and a single detector. Each deflector of the kth microcolumn array receives a signal sk(t) that modulates the kth beam. Beam offset means keeps each average position of the beam centered on the detector according to embodiments described previously. The modulation then generates a detector signal. Since each beam excites the detector simultaneously, the detector output is the sum of all amplified signal components. Thus, power combining is achieved.
In another embodiment, the k beam offset means are achieved with electron optics. As shown in
TR Arrays
It can be appreciated from the microminiaturized nature of the construction that the foregoing benefits of a transmit beamformer can be combined with a receive beamformer in a single integrated bidirectional transmit-receive or “TR” unit.
Beamform Processor
In systems that employ digital signal processing to form RF beams, a plurality of signals rnm(kT) (received or to be transmitted) at successive times k of a sampling interval T are delayed by storing them in random access memory and selectively re-accessing them for beamform summation.
In some applications, the samples rnm(T) are multiplied by constants cnm so that each signal is not only delayed but scaled. Yet other applications may not use a simple progressive time-delay algorithm for beamforming, but may rely on specialized algorithms similar to the Fast Fourier Transform (FFT), which employs matrix mathematics to determine optimum time delays and scaling coefficients to achieve multiple beams with the low sidelobes. Even more complex beamforming algorithms are supplemented by adaptive nulling algorithms to suppress signals in certain directions where there may be interference (as in a receiver) or where interference must not be generated (as in a transmitter). In any of these examples, the beamforming might also have to form cross-polarization levels, which doubles the processing required. These are not the only types of processing, but are illustrative of the complexity of the processing that might be involved.
It can be appreciated that a beamform processor may have to accomplish many functions and require considerable computing power. In high performance systems, this is often achieved with multiple digital signal processors operating in parallel. These processors may have to access a common memory as well as the plurality of signals rnm(kT), and often have to transfer data between processors at very high rates.
Conventionally, data transfer between processors is via a shared input/output (“I/O”) bus, sometimes termed a “backplane”. Data is transferred between processors under the control of an arbitration arrangement, but since data transfer can only take place between one pair of processors at a time, the data transfer is necessarily sequential, and each processor waits its turn to transmit data to, or receive data from, another processor. The result is that processing slows significantly. As a number of parallel processors increase, the overall processing often improves no better than the logarithm of the number of processors. This limits multiprocessor computers, because the cost of parallel processing goes up dramatically with only minor performance improvements. Many real-time applications (such as, for example, synthetic aperture radar image processing or fast-fourier signal transforms) are severely constrained by data transfer delays.
Various methods have been employed to increase the performance of multi-processor systems. One method uses multiple buses between processors. Other methods use dedicated high-speed communication channels between each pair of processors. In general, the large number of data path combinations makes a full set of physical electrical paths prohibitively large, costly, power consumptive, slow and inefficient. Since for a number N of processors there are (N2−N)/2 processor pairs, even a subset of the datapaths becomes prohibitively expensive to implement using conventional printed circuit boards and cables, for large N (e.g., N>1024). Another difficulty is that each processor must drive N buses or channels, and the loading becomes prohibitive for high-speed operation.
Some sophisticated systems use active circuitry to create a device that attempts to exchange signal paths such as digital data streams across a “crossbar switch matrix” or “crossbar.” For example, a crossbar may dynamically reconfigure a fixed number of communication paths between processors on a demand basis, eliminating the loading effect by creating point-to-point connections between certain pairs of processors at one time. For instance, a crossbar may create a communication path between a processor A and some of any of N other processors, and a communication path between a processor B and some of any of N−1 other processors, and a communication path between a processor C and some of any of N−2 other processors, and so on.
This is only one application for a crossbar. The very nature of the device makes it of great utility for other applications as well. For instance, some types of crossbars can also be used as a switching element in reconfigurable computers and multiplexed data acquisition systems, among others.
Crossbar switches have historically had only a relatively few number of inputs and outputs, such as, for example, the 16 inputs and 16 outputs shown in
A traditional solution for dense interconnection has been to construct an array of many small crossbar switches. With appropriate cross-interconnection of small crossbar switches, the array can appear to be a much larger crossbar switch. One form of this is called an “active backplane”. A “passive backplane” consists simply of wiring among multiple processors, or processors and peripheral systems such as disk drives. In contrast, an active backplane incorporates active switching elements such as small crossbars to dynamically configure point-to-point connections among processors. Generally, some kind of crossbar switch elements are preferred and configured for duplex signalling.
However, even an active backplane may not allow simultaneous transfer between all processor pairs. In this case, it is termed “blocking,” to reflect the fact that communication paths between certain processor pairs will “block” simultaneous communication between some other processor pairs. When an active backplane can achieve simultaneous transfers between all processor pairs, it is termed “non-blocking”. The disadvantage of a “blocking” active backplane is that the transfer of data between processor pairs must be performed sequentially (i.e., certain transfers must wait for other transfers to be completed). This slows the overall data transfer rate among all the processors and reduces the computing throughput.
Similar considerations apply for a typical transmit beamformer.
It may be appreciated that N-element antenna RX array 3520 of
Some crossbar switches developed for active backplanes to date have used both electrical and optical means; many of these have limitations with respect to bandwidth, cost, power, complexity, and heat generation.
E-beam Crossbar Switch
In some EBXs, the number M of microcolumns may equal the number of detectors N, while other EBXs may have M≠N.
Programming (or re-programming) offset signal 3100 for any of electron beams 120 is achieved by delivering a matrix configuration command 3060 to control circuit 3055 that redirects a channel m coupling between a corresponding input signal sm and a detector DN. Each signal sm modulates one of the M deflectors, thereby causing the signal sm to excite one of the N detectors. This causes a current output to be generated from detector DN, thereby transmitting (and possibly amplifying) signal sm through a dynamic channel MN corresponding to targeting mth e-beam 120 onto detector DN.
A data signal corresponding to signal sm may be a small proportion of each deflector voltage signal 3130, as the data signal need only deflect the corresponding beam 120 by an angle subtended by a single detector element DN. Each detector DN may be formed, for example of one or two segments for digital signalling, but other arrangements are possible. Saturation means (e.g., high speed Schottky diodes) may be provided in the output networks Zn to clamp the output voltage levels, as discussed above with respect to
The mechanical dimensions of an EBX may be appreciated from an example. For a 5 μm wide detector, a 100×100 array of detectors has dimensions of 500 μm×500 μm. Similarly, a 5 μm diameter electron gun permits a 100×100 array of electron guns with the same dimension. (However, as mentioned above, the detector and gun arrays do not have to have the same size or dimensional number.) Assuming a maximum beamsteering tangent of 0.2 (corresponding to a deflection angle of 11.3 degrees), a minimum drift cavity length is approximately 2500 μm if an e-beam from one corner of electron gun array 3150 is to be steered to an opposite corner of detector array 3180. These dimensions are consistent with the fabrication techniques discussed above.
The electrical parameters of an EBX may be appreciated from an example. It is assumed for this example that input signals sm have a peak-to-peak amplitude of 100 millivolts, and are to be reproduced at detector outputs Zn that are terminated in 50 ohm loads. A 2 mA peak-to-peak current is thus required from the detector. With a beam acceleration of 280 eV and a detector gain of 1000, a beam current of 2 μA is required to excite each detector. From the previous description of the effects of space charge spreading, it can be seen that this is within the range of acceptable parameters, and a 2 μA beam is low enough in current that a single electron gun may be employed for each of the M input channels.
Crossbar Array Construction
Many arrangements of microcolumn arrays and detector arrays are possible. In the simplest, the microcolumns and the detectors can be arranged in a line; however, in this configuration, large numbers of channels result in excessive beamsteering angles.
In another arrangement, each of the microcolumn array and the detector array is arranged in a two-dimensional matrix.
Generally, the diameter of the microcolumn and detector matrices should be as small as possible for a compact construction, but these matrices need not be the same size. For example, if each microcolumn has a diameter of 5 μm, an array of 100 microcolumns could be a circular matrix about 70 μm in diameter. A detector size might be as small as 2 μm in diameter, so a detector matrix could be a circle about 20 μm in diameter.
For a given microcolumn array diameter, a smaller detector array size reduces a maximum beam steering angle, allowing for more channels and a shorter drift cavity. Maximum beamsteering angle is primarily limited by the maximum beamsteering deflection voltage that can be delivered by circuitry such as a DAC. A short cavity is consistent with a compact device, and simplifies wafer-based mechanical construction.
By way of example, a maximum beamsteering voltage may be estimated. From previous discussion, the deflection tangent is tan Θ=√{square root over (ΔV/2VBEAM)}. For a beam energy VBEAM of 50V at an exit of an electron gun Oust before deflection) and a maximum tangent of 0.2, a the maximum beamsteering voltage ΔV=4V. This is consistent with circuitry that may be used to generate beamsteering voltages.
By way of example, a modulation amplitude may also be estimated. For a 5 μm detector and a 2500 μm drift cavity, the maximum tangent of the digital deflection is approximately 5/2500=0.002 (0.11°). Again, from the previous formula, the deflection modulation voltage for a 50V beam (at the emission plane) is 400 μV.
Crossbar Signalling Rate
A signalling rate of each channel of an EBX can be estimated from these considerations. From prior discussion, it can be appreciated that a frequency response of deflectors in an EBX may exceed 1 THz. For example, a 1 μm long plate with a beam velocity of 4×106 m/s (beam energy of 50V) may support a bandwidth of 1.7 THz. If a corresponding detector has segments that are 2.5 μm×5 μm, detector junction capacitance may be on the order of 10 fF. If a load is 50 ohms and other circuit parasitics are of similar magnitude, (for example, 10 fF parasitic capacitance), then the bandwidth of the detector will be 160 GHz. Non-Return to Zero (“NRZ”) binary signalling may require a bandwidth that is 70% of the bit-rate, so a maximum bit-rate per channel may be over 200 Gbps.
Beam-steering
As discussed above, e-beams from a microcolumn array may be individually steered to a detector matrix by beamsteering signals applied to deflectors in a microcolumn array. In the case of a one-dimensional microcolumn array and a one-dimensional detector array, a single voltage applied to a deflector of a single microcolumn may position a beam from the microcolumn on a single detector. For a two-dimensional microcolumn matrix and/or a two-dimensional detector matrix, two voltages applied to an X deflector and a Y deflector in each microcolumn direct an e-beam from that microcolumn to a single detector. One of the X-Y deflectors may also be used for signal modulation, or a separate signal deflector may be provided.
With two-dimensional beam steering in an EBX with M input channels, there are 2M analog beam steering signals. Each pair of analog signals corresponding to an X-Y deflector pair is set to voltage levels corresponding to a physical offset (fixed by the mechanical design) between a particular microcolumn and a particular detector. Thus, for N detectors, each microcolumn will have associated with it N pairs of voltage levels. For example, if there are 100 detectors in a square detector matrix, each of an X and Y deflection voltage level may be chosen from 10 possible levels. A round or rectangular detector matrix may require more possible levels than a square matrix; additional range may be provided for channels near the ends of a microcolumn or detector array, since the corresponding e-beams may be deflected by greater angles than e-beams from microcolumns substantially within the matrix.
In one variant of an EBX, each beam steering voltage is generated by a DAC array 3090 controlled by an addressable memory 3070 and a matrix configuration command 3060 of X-Y matrix positioning signals (see
Crossbar Beam Centering Loops
Even after calibration, steering accuracy may be difficult to maintain in some EBXs. For example, high speed in each crossbar channel is achieved with a correspondingly small detector. It may be desirable to use a 1 μm wide detector, but it may be difficult to maintain beamsteering accuracy to a 1 μm tolerance, even with calibration. For example, temperature changes or vibration may cause beamsteering accuracy drifts which may be corrected to improve performance of an EBX.
One embodiment of an EBX includes a beam offset centering loop between each deflector and detector, which may operate the same as described for a simple amplifier (
Extracting, for example, X direction beam offset information from averaging is undesirable in a digital signalling context, because it may constrain bit patterns to have, on average, a same number of ones and zeros (for binary signalling), requiring special channel coding which may detract from signal throughput. However, if an averaging interval is very long relative to a signal bit rate, no special channel coding is required (for example, if the channel rate is 100 Gbps, and the averaging interval is 1 second). For long time intervals, averaging may be accomplished with a digital filter and a DAC for each channel; the DAC might be shared with a coarse “open-loop” beam-steering DAC.
Other arrangements are possible. For example, in configuration 151(20) of
X direction beam offset detector segments 150(108) and 150(111) of configuration 151(20) may operate in one of at least two ways. (It will be appreciated that in this discussion, the signal beam sweeps in the X direction; the same principles apply in other directions that are the same as a sweep direction.) In one method, a differential signal is averaged in an integrator of a control loop so that an average excitation of segments 150(108) and 150(111) is the same; this assumes the beam spot 170 is somewhat larger than segments 150(109) and 150(110) so that a one or a zero digital level will always excite segments 150(108) and 150(111). This requires a digital bit pattern with the same number of ones and zeros, on average, as in the previous detector embodiment.
In another arrangement, beam spot 170 may be made somewhat smaller than the segments 150(109) and 150(110). In this case, the digital modulation is designed so that with perfect spot centering, 150(108) and 150(111) are never excited, but if beam spot 170 is offset to the left (e.g.
If a width of beam spot 170 is somewhat less than the width of 150(109) and 150(110), and a spot deflection is approximately equal to the width of 150(109) and 150(110), then configuration 151 (20) does not require the same number of ones and zeros in a digital bit stream, on average or otherwise; this eliminates any need for special channel coding or long integrator time constants.
Beam centering loops may slow the rate at which a crossbar can be reconfigured. If an integrator time constant is long, transmission through the crossbar may have to wait for the integrator to settle so that signalling is reliably transmitted to the digital detectors.
Nonetheless, some applications may find beam centering loops advantageous, particularly when interconnection of many channels is required, since interconnection of many channels may only be achievable with very small (perhaps sub-micron sized) detectors. Such applications may tolerate a significant settling time delay. For instance, routing switches (e.g., for computer networking), may tolerate delays of tenths of a second or more. In applications requiring somewhat faster reconfiguration, it can be appreciated that a quadrature offset measurement detector is desirable, since it can have fast integrator time constants to quickly center a beam on appropriate detector segments.
Beam Centering Loop Reconfiguration Matrix
In a crossbar, beam centering loops may be dynamically reconfigured along with the connection that they support, so that they couple the correct offset measurements for a detector n back to an e-beam deflector steering a beam m.
For instance,
Thus, some kind of secondary crossbar matrix 3220 is necessary to connect the offset control signals 3210 back to the appropriate deflectors 130. Secondary crossbar matrix 3220 may be another e-beam crossbar, but since the beam centering loops may be much slower in operation than signals being transmitted, matrix 3220 may also be transistors integrated into an e-beam crossbar assembly.
A secondary crossbar matrix (e.g., matrix 3220) may be implemented by sequentially sampling the N detector offsets one at a time through a first multi-pole-single-throw switch, and then back through a second multi-pole-single-throw switch to the M input deflectors, calibrating the centering of each beam one at a time in a slow cyclic process. At any one time, a feedback signal may update a voltage on a storage capacitor coupled to a deflector of an input channel. This arrangement requires only a simple switching matrix, and works well when a slow loop update is preferred. Alternatively, a single ADC may measure beam offset at the detectors, and a sequential switching arrangement may transmit the ADC output as a digital correction through a bus structure to be stored in a register that controls a DAC coupled to an appropriate input channel. By way of additional examples, one or more ADCs may feed a processor which performs digital filtering, and may accelerate the initial error correction by non-linear means, or a ROM may be inserted between ADC and each DAC.
A number of ways of using offset corrections are also contemplated. For example, a memory which receives matrix configuration commands (e.g., memory 3070 of
Photonic I/O Coupling
Coupling a large number of I/O channels between an EBX of microfabricated construction and external circuitry may present challenges. For example, an EBX with 10,000 channels may occupy a package of only (5 mm)3 in size.
Direct electrical coupling is not easily achieved with such a large number of high-speed channels. While it is possible to electrically mate packages using technologies such as ball-grid arrays (“BGA”) or other high-density interconnect, coupling effects at speeds of 100 GHz or more may produce unacceptable signal distortion.
One embodiment of an EBX couples its inputs and outputs to external inputs and outputs (such as a computer bus) by means of optical interconnect.
Outputs 3270(1-4) of e-beam detector configurations 151(25-28) couple to laser diodes 3280(1-4); this coupling may also be direct or indirect, for example laser diodes 3280(1-4) may receive a DC bias current from a bias current source (not shown), with outputs 3270(1-4) capacitively coupled thereto. Light 3290(1-4) emitted by laser diodes 3280(1-4) is coupled to optical fibers 3240(5-8). Thus, in the e-beam configuration of
A photonic I/O coupled EBX preferably couples photodetectors in close proximity to deflectors, and couples laser diodes in close proximity to detectors, to minimize wiring-induced delays, and parasitic capacitance- and resistance-induced signal distortion.
A lens system may make a reducing image of light from an input optical fiber bundle onto a photodetector array.
Thus one embodiment of an EBX with photonic I/O coupling may operate as follows: a modulated input optical signal from an input fiber IFm is transmitted optically to a single photodetector PDm, wherein the input optical signal is converted to an electrical current and a voltage (by driving a resistive termination), and applied directly or indirectly to a deflector Pm of an electron gun EGm. The EBX directs an electron beam from gun EGm to a detector Dn, and an electrical current excited in detector Dn by the beam drives a laser diode LDn. The laser diode LDn generates an output optical signal with the same modulation as fiber IFm. This output optical signal is magnified and imaged onto a single fiber OFn of an output fiber bundle. This sequence of steps is performed in parallel across M potential input fibers and N potential output fibers so that optical signals in any given input fiber may be coupled to any given output fiber.
Advantages of this arrangement include leveraging known methods of manipulating fiber bundles for making reliable physical interconnects of high bandwidth. Fiber bundles may have a very high density of fibers, permitting a large number of channels. The optical imaging arrangement may couple thousands of channels to an EBX, which may have physical dimensions as small as a few millimeters. Furthermore, optical I/O provides level-shifting and high voltage isolation, which may allow a high common mode voltage difference between electrical input and output levels of the EBX. Flexibility with respect to high common mode voltage difference may permit high beam acceleration in an EBX drift cavity, high gain, and a high signalling rate for a given EBX electron gun current.
EBX Size
From the foregoing it may be appreciated that many configurations and applications of a crossbar are possible other than digital signalling applications. By the nature of the deflection process and the many variants of the EBTX and EBRX, functions such as analog amplification, time delay control, mixing, pulsing, frequency multiplication and combinational logic may be incorporated in crossbar channels. Thus, both highly integrated and highly specialized functions may be constructed in a single device.
For example, a Combinational Crossbar Logic (“CXL”) embodiment may be used as a reconfigurable computer that changes its functionality by forming specialized electron beams and addressing specialized detector configurations, as opposed to a computer that runs new software or firmware routines. In a CXL, extra deflection plates may be incorporated in the electron guns of a electron gun matrix, and specialized detector arrangements are incorporated in a detector matrix. By way of analogy, the electron guns and detector arrangements may be addressably configured in much the same way that logic cells are addressably configured in a field-programmable gate array (“FPGA”). A CXL may allow complex and reconfigurable logic processing in a very small, high speed device.
An Analog Crossbar Matrix (“AXM”) is an embodiment whereby, as previously discussed, each e-beam in a crossbar matrix modulates with continuous voltage levels, and each detector is a pair of segments as in an EBRX. Thus, steerable analog channels can be amplified. In an AXM, low noise operation may require higher beam currents for each channel, and sub-arrays of multiple electron guns per beam, as in prior embodiments (e.g.,
An Analog Crossbar Beamformer (“AXB”) is another embodiment for applications that can employ analog summation of multiple signals, as from antenna elements. This is similar to the power combiner of
Unterminated Waveguide Coupled Beam Deflection
Any RF amplifier is generally coupled to a signal source via some kind of wave-guiding structure, such as a transmission line or more generally, a waveguide. Usually the coupling requires terminating load resistors, or a more general matching network of resistors and reactive elements such as capacitors, inductors, waveguide stubs, etc, to provide a low-impedance match (say, 50 ohms) to the waveguide, and a simultaneous match to the input impedance of the amplifier. The match causes the transmission line to see a load with the same real impedance as the waveguide and the amplifier to see a reactive impedance that cancels any reactance at the input port of the amplifier.
Advantages of a terminating matched network between the waveguide and an amplifier are two-fold: First, the matched termination maximizes the power transfer from the waveguide to the amplifier. A load impedance that is the complex conjugate match of the same real part impedance or negative reactive impedance of the transmission line (or waveguide) absorbs the maximum signal energy in the real part of the load, e.g., a resistor. Likewise, when the matching network is the complex conjugate of the amplifier impedance, the maximum power is transferred from the network to the amplifier.
When the amplifier has no significant reactive input impedance the match can be accomplished with simple resistors. More often, however, the amplifier has a strong reactive impedance, and the matching network must incorporate reactive elements to cancel the amplifier reactance (within a frequency band of interest). This prevents the reactive part of the amplifier load from distorting the frequency response to the amplifier.
Generally, the matching network must transform the waveguide impedance of perhaps 50 ohms to a finite and fairly small amplifier impedance of a few kohms at most. Solid-state semiconductor amplifiers generally have a low amplifier impedance as an unavoidable consequence of the technology. For example, bipolar amplifiers are generally limited by the input resistance to the base of a transistor. This is often in the range of 1 kohm or less, dictated by the design requirements at higher frequencies of operation. Amplifiers made in FET technology (MOS, Schottky gate, etc.) may have a very high gate resistance, but a very low capacitive impedance from the large gate structure that is usually required to achieve significant gain.
The second advantage of a matching network is that it eliminates (or reduces, depending on the quality of the match) the back-wave reflection of the signal from the load onto the waveguide. This is a corollary to maximum power transfer. Thus, with a match termination, no forward-traveling wave energy is reflected back to the signal source at the input end of the waveguide. All the signal power is thus available to the amplifier (if the transmission line couples the signal to an amplifier), and the source does not have to absorb any reflected power.
Generally, the reflection is described by what is termed a “reflection coefficient”, usually denoted by the symbol a factor which is multiplied by the incident wave to determine the amplitude of the reflected wave. The general formula is
where ZL is the load impedance seen by the line, and Z0 is the line impedance (e.g., 50 ohms). Thus, a load open (ZL=high impedance) has Γ=+1, while a load short (ZL=0) has Γ=−1. In the case of a short, the reflected wave is inverted in amplitude, and the total voltage seen at the short is zero.
The case of a high impedance load is the one of interest. In this case, the reflected wave has the same polarity and amplitude as the incident wave, and the total voltage seen at the open is twice the incident voltage wave.
Backward reflected power is undesirable in some applications if the RF source is impedance mismatched to the transmission line (or waveguide). This is because the reflected wave can in turn get re-reflected at the source if the source is not matched well to the line. Thus, the backward wave is re-reflected towards the load, causing signal distortion. That is, the re-reflected wave reaches the load after the round-trip delay time of the transmission line (twice the line length divided by the velocity of the wave) and the load sees the signal plus a delayed version of the signal from an earlier time—albeit an attenuated, possibly inverted version, depending on the losses of the transmission line and the kind of source and load mismatch. If there is a strong mismatch at both ends and only weak attenuation along the transmission line, the successive reflections can seriously corrupt the signal being amplified with delayed representations thereof.
The advantage of a load matching network can thus be seen: for if the load match achieves a small ΓL that attenuates the reflection by x, and if the source match achieves a small ΓS that attenuates the reflection by y, then the total attenuation achieved is xy. For example, if ΓL=0.1 and ΓS=0.1, the total attenuation is 0.01. On the other hand, if the load was an open with ΓL=1, and if the source ΓS=0.1, the total attenuation is only 0.1—ten times worse.
Thus, a matching network at the load mitigates the non-ideal characteristics of the amplifier itself, improving the power transfer, frequency response and signal integrity. The signal VS is reflected with twice the voltage amplitude and four times the power gain.
Unterminated Waveguide Coupling
Though the EBTX or EBRX can be coupled to a waveguide in the conventional manner using a matched load termination, a reflective amplifier 5000 as shown in
A transmission line and/or waveguide 5008, 5008′ forms a circuit connecting antenna 5010 with deflectors 5004. Incoming RF 5012 strikes antenna 5010 to produce a voltage signal VS, which drives the deflectors 5004 in the usual manner; however, due to the large nature of RL, there is a reflected voltage signal VR which is approximately equal to or equal to VS. The reflected voltage signal VR communicates on transmission line and/or waveguide 5008, 5008′ to antenna 5010 for emission of re-radiated RF field 5014.
Because the total input capacitance of an EBTX or EBRX array may be as low as 100 fF, the bandwidth when coupled to a low-impedance waveguide can be very high. For example, 100 fF coupled to a 50 ohm line has a bandwidth of 60 GHz.
The key to using an unterminated line is to have a source impedance match. If the coupling at the source is a match of high quality, the reflection there can be made small enough to tolerate a load mismatch. The re-reflected wave will be much smaller in amplitude than the incident wave, and the effect on the signal at the load will be small.
This is often difficult to achieve in practical circuits if the source of signal power is another amplifier. Amplifiers usually have complex reactances in their output port that will create a poor match in the absence of a source-matching network.
There is one special case where the source can be well matched: an antenna. If the EBTX is directly coupled to its antenna with a very short transmission line (or no transmission line at all), the source match can be excellent. The antenna match can generally be well controlled, and the effect of the reflected energy is to simply be re-radiated without being re-reflected.
Two basic approaches may realize the unterminated coupling. In one approach, the transmission line or waveguide 5002 may end at the deflectors 5004. Alternatively, a transmission line may continue past the EBRX 5002, which merely taps off or “samples” the signal propagating down the guide. In this second case, the deflector 5004 can be the waveguide 5002 itself or the deflector 5004 can sample the voltage VS on a waveguide or transmission line by a wired connection to points of greatest voltage potential. In context of equation 1.53, it may be preferable for Γ=+1 where ZWN is the impedance or EBRX 5002 and Z0 is the impedance of a waveguide.
Direct Waveguide-Electron Beam Coupling
Although the input capacitance an EBRX or EBTX may be quite small, the loading effect may sill be significant if the frequency of operation is very high, e.g., 100 GHz or more. As shown in
The key is to make the e-beam travel at approximately right angles to the RF wave motion, because then the beam is subjected to approximately the same amplitude of the RF wave as it passes through. The waveguide must be constructed to ensure a single mode of operation, preferably TE or TEM, so that the electric field vector of the wave is perpendicular to both the e-beam and RF wave motions. This way, the e-beam is deflected uniformly in one direction, the direction of the electric field.
For this case the deflector does not really load the waveguide 5016 at all—it is the waveguide 5016 and has an impedance of Z0. according to Equation 1.53. That is, the capacitance of the deflector is just part of the natural distributed capacitance of the waveguide. There is no loading beyond a miniscule coupling to the electron beam itself, and the signal wave can propagate along the line without reflective obstruction or attenuation, and without distortion. The electron beam deflects directly in response to the propagating wave field of the signal VS without the need for a terminating load resistor to generate a voltage.
Solid-state amplifiers are not able to directly amplify a wave field. Transistors require the electric and magnetic field of a signal in a waveguide to first be converted to a voltage and current. Direct wave amplification is normally only possible to amplifiers such as TWTs and klystrons which couple the electromagnetic field of a signal to an electron beam by means of a special mechanical waveguiding structure or resonant cavities.
In principle, the signal power in a waveguide can generate an electric field of equal magnitude to that of a voltage across a deflector, so long as the wave can be guided into a constricted region having the dimensions of the deflector. In practice, this is not usually possible if the deflector has spacing and length dimensions of a few microns. The reason is that for most frequencies of operation a waveguide of such small cross-section will not sustain the propagation of a traveling RF wave. The maximum dimension for a closed waveguide (width or height) should be at least one-half wavelength. A 100 GHz frequency has a wavelength of 3 mm in free-space. Even a 1 THz frequency has a wavelength of 300 microns.
Nonetheless, there are specialized applications at extremely high frequency (100 GHz to 1 THz or more) where this might be done. If the waveguide is filled with a dielectric, for instance, the wavelength is much shorter, in inverse proportion to the relative permittivity of the dielectric. For example, SiO2, which has a relative permittivity of 3.9 would have a wavelength approximately ½ the free-space wavelength. A 1 THz frequency would have a minimum guide dimension of 75 um.
Thus, a direct coupling of the electron beam to the signal, by directing the beam through a waveguide, is one embodiment as shown.
Waveguide Voltage Sampling
Most applications of the EBTX or EBRX include deflectors coupled to a transmission line, which is a special case of a two-wire waveguide. The advantage of the transmission line is that each wire can have a different potential, and therefore the wire spacing is not constrained to be a minimum of one-half wavelength. Unlike the closed waveguide which can only sustain TE (transverse electric) or TM (transverse magnetic) modes of propagation (where waves bounce off the interior walls of a closed waveguide), the transmission line can sustain a TEM mode. Thus, the preferred embodiment couples an unterminated transmission line to the deflection apparatus.
Two advantages accrue to the unterminated load. The first is that the reflected wave doubles the signal voltage received by the amplifier. This has the same effect as 4 times the signal power in a conventional terminated connection.
The second advantage is an improvement in input noise. Solid-state amplifiers are normally used at the front-end of RF receivers to amplify the signal from an antenna, because they offer very low-noise amplification (1 to 5 dB noise figure). TWTs and other traditional electron beam amplifiers are normally used where large signal power of many watts is required, because they have only been practical to construct for high power operation, which is usually an extremely noisy process. A typical TWT might have a noise figure of 40 dB. In contrast, low-noise solid-state amplifiers often operate with signal levels that can be equal to or less than the noise power of a simple resistor, which is given by the well know formula PR=4 kTB. This low-noise amplifier (LNA) characteristic is extremely important in any RF receiver.
In an RF receiver coupled to an antenna, the LNA must normally have a wide bandwidth. For the reasons cited above, the amplifier coupling normally employs a matching network between the transmission line and the LNA. This terminating resistor is an unavoidable source of noise power diminishing the ultimate sensitivity and dynamic range of an RF receiver. In thermal equilibrium, the RF noise power is a simple result of the brownian motion of electrons in the resistor causing a varying resistor voltage that radiates RF; an equal amount of power is absorbed and re-radiated, and the radiated power is random broadband noise.
The EBTX or EBRX, therefore, when employed as a LNA, can improve the sensitivity of an RF receiver over prior art by eliminating the terminating resistor. The RF signal from, say, an antenna, can be amplified prior to being subject to other circuit noise. If the amplifier gain is high enough the added noise of the amplifier referred back to the input (i.e., divided by the amplifier gain) can be much less than the noise power of a simple terminating resistance. In the amplifier embodiment, the gain can be as much as 40 dB, or more. This makes it is possible to have an equivalent input referred noise power that is 1/10 or less of a simple resistor noise power at an ambient temperature of, for example, 300 K.
In this sense the effect of eliminating the resistor termination is like supercooling an input termination resistor to a temperature of only a few degrees Kelvin. The difference is that it can be done without any refrigeration, which is desirable in many applications such as spaceborne electronics, where the weight, power consumption, reliability and expense of cryogenic operation is unacceptable.
To achieve the noise reduction, however, it is desirable that the RF in the guide not be absorbed in any kind of resistance, either a load or losses in the waveguide walls. Any resistive power absorption will generate random RF noise that look just like a resistor, no matter where it is generated in the guide, since it will propagate back to the amplifier input.
In any of these embodiments, the goal is the same: to prevent remove the signal energy once it has been detected by the amplifier, without absorbing it in a noise-generating load. Otherwise this would eliminate the key advantage of the unterminated coupling: the reduction of input noise and the improvement of output signal-to-noise ratio (SNR).
Step-tapered Drift Cavity for Short Focal Length Electron Lens
For an EBTX or EBRX to operate with high gain, a high current beam is needed. This requires a large initial beam diameter, e.g., or several hundred microns or more, so that the beam can be propagated across a long drift cavity of up to 5 mm or even more without severe beam spreading from space charge forces, and then the beam must be focused down to a small beam spot at the detector to provide a useful output signal with wide bandwidth.
Focusing a large diameter beam to a small beam spot requires strong electron optical elements. Many schemes are possible, but one common approach employs what is called an “Einzel lens”. This consists of two annular ring electrodes with a gap between them, similar to a cylindrical soup can cut in half. Each electrode has a different potential applied to it, and the effect is to create the electron optical equivalent of a spherical lens, as in normal light optics.
As shown in
Because the electrons are traveling from a region of lower to higher potential, they are also accelerated as they pass through the equipotentials. The velocity of the electrons is therefore lower on the focusing side of the lens (the near-side), and higher on the defocusing of the lens (the far-side). The far-side equipotentials exert a strong defocusing force away from the axis of the same magnitude as the focusing forces, but because the electrons are traveling faster in this region, they are exposed to the defocusing action for a shorter period of time. Thus, the focusing action is not entirely cancelled by the defocusing and the lens exhibits a net focusing action. It can be appreciated, however, that the strong defocusing significantly diminishes the overall focusing power that might otherwise be achieved if the electrons were only subject to the focusing action on the near-side of the lens.
The essence of the problem with the conventional Einzel lens is that the equipotentials on either side of the gap are symmetrical. Even though the electron beam transit time through the defocusing region is shorter, it is not sufficiently shorter that the defocusing action does not cancel most of the initial focusing action. However, in the symmetrical can structure of an Einzel lens, it is not possible to make the equipotentials asymmetrical to any significant degree. This stems from the physics of static fields described by Maxwell's formula for a potential field in a charge free region of space.
The embodiment shown in
A variation on this theme is possible by electrically decoupling the flange from the first and second electrodes. In this arrangement, the flange acts as a third electrode to shape the equipotentials of the lens, such as to correct for lens aberrations and improve the focusing.
It may be noted that the electron beam 5030 should stay focused on a detector, meaning the beam 5030 is never deflected a great distance away from the optical axis. Since the beam stays close to the axis, it is possible to narrow down the initial drift can radius (which is required for a large diameter beam) to a smaller radius drift can (which receives a smaller beam diameter as a result of the focusing action). Thus, it may be appreciated that the stepped radius of the modified Einzel lens structure not only achieves stronger focusing, but is well suited to the electron beam amplifier concept in particular.
RF Cavity Detector
As shown in
The basic principle of the RF cavity detector 5040 is to receive the high energy swept beam energy 5042 at a porous beam contact, such as a gridded or slotted beam contact or wall 5050 that may act as an electron permeable RF shield. Wall 5050 permits the beam energy 5042 to be transmitted through, generally unimpeded. In this case, however, the beam electrons do not directly enter a semiconductor, but an RF cavity 5052 including conducting detector-waveguide 5054, 5056. The walls 5054, 5056 are generally at a different electrical potential from the potential of the beam contact or wall 5050, and the relation of the wall 5050 to the cavity walls 5054, 5056 creates an electron lens 5058, as has been described. In this, cause, a decelerating lens is preferred. When the beam energy 5042 enters the RF cavity 5052, it is immediately slowed down. Preferably, the speed of the electrons is reduced almost to zero. This is accomplished by having a cavity potential on detector waveguides 5054, 5056 that is negative with respect to the beam contact wall 5050 by the potential of the beam energy 5042. For example, if the energy of the beam entering the cavity is 1 keV, the cavity walls may be 1000V after the beam contact wall 5050.
The effect of the decelerating beam is to impart energy back into the cavity walls 5054, 5056 as a wall current on the wall surface. If the beam remained focused on one position, this would deliver a DC energy back to the power supply coupled to the cavity walls, less losses. However, the one feature is to convert this energy into RF field in the cavity 5052 by sweeping action along spots 5059, 5060 where the beam energy 5042 is steered by the action of field 5058. This modulates the spatial position of the beam energy 5042, moving the beam spot across the cavity walls, from left to right and back again, for instance. Many methods of spatial modulation are possible to achieve a desired signal or efficiency, but this one is illustrative as shown. In general, the goal is to mimic, to the extent possible, the wall current which would be present if an RF were already present in the cavity.
Thus, in this embodiment, the detector is a region of the cavity walls where the beam spot strikes it. The “detector” is simply a region of the metal guidewall in the cavity 5052. The detector may or may not provide current gain.
The beam contact wall 5050 in this embodiment is a gridded screen or slotted aperture to allow the electron beam to pass through unimpeded, and is actually spatially separated from the region on the cavity wall where the beam spot forms. The gridding of the beam contact is small enough relative to the RF field being generated (ie, the grid spacing is much less than a half-wavelength) that little RF can penetrate back into the beam drift cavity, where it would otherwise cause fields that would defocus the electron beam. The gridding isolates the RF in the cavity detector from the drift cavity.
In operation, the beam spot sweeps back and forth across screen grid (ie, beam contact) and back and forth inside the cavity, where the spot may be defocused or not, but where it will be “bent” in trajectory by the lensing action therein, causing the beam spot to sweep from one wall to the other (ie, the “segments” of the detector regions), with firehose action. If the spatial motion of the beam and the other factors are properly controlled, this can efficiently generate RF energy directly, which can be coupled out of the cavity by a waveguide, antenna horn or other RF guiding structure.
Crossbar Sequencing Control
In the case of the feedback loops, it can be difficult and complex to perform the all the feedback loops simultaneously, because for many channels of electron beams, just as many channels of feedback would be required. Moreover, some kind of secondary crossbar switch would be required to select each given detector and couple a feedback path back to each given electron gun, since these paths are different for every configuration of the main e-beam crossbar. Speed in the feedback paths can be orders of magnitude slower, though, since once the beamlets are properly centered they will not change except from thermal cycling, and so forth, so the secondary crossbar could be made of transistors, but even that would be excessively complicated if the e-beam crossbar had a lot of channels.
A solution is to simply achieve the feedback loops sequentially. In this case, a single detector output from detector output signals 5058 is selected, as may be coupled to a single filter 5067, and the single filter 5067 may couple a single error correction to a selected one of the electron guns (not shown). The selection of the detector can be a simple N to 1 multiplexor switch 5070, and the selection of the electron gun may be a simple 1 to N demultiplexor switch 5072, both made compactly and efficiently from conventional transistor technology. The filter 5067 may be analog but is preferably a digital filter so that the “state variables” of the filter 5067 can be stored and recalled each time a channel is updated, since otherwise the filter 5067 would retain the history of the error of the previous channel. This would slow the convergence of the feedback loops considerably and introduce undesirable transient settling errors into the beam steering. If a digital filter 5067 is employed, then the detector error transmitted from the multiplexor 5070 may be sampled by an analog-to-digital converter (not shown) before it is received by the digital filter 5067.
With the sequential update, the output of the filter 5067 is stored for each detector channel 5069. In an all-analog loop, this can be by means of capacitive storage (not shown), for example, a sample-hold on each deflector. In an digital loop, the storage can be a register 5074 s coupled to a DAC 5076, with the DAC 5076 driving the deflector (not shown) with the refined offset correction. In either case, the refined offset correction is summed with the coarse steering command in either digital or analog form, at any point after the refined offset is generated: either before the DAC or after it. The summing can be digital, analog, or even by means of a supplementary set of deflectors in each e-gun to drive the beamlets independently. A sequencer 5078 sequentially repeats this process for each detector in an array.
Though some embodiments it might be desirable to focus electron gun subarrays in that manner to achieve power combining, the concept has more general application. For instance, one problem is the maximum current of an electron gun. If the current is too high, the electron gun might focus it, but then the beamlet will spread out from space charge forces within the drift cavity. The doublet lensing of the drift cavity depends on the beamlets staying substantially focused during the drift time to the detector. This means the beamlet current should be quite low. Yet to obtain substantial overall beam current, a large array of electron guns is employed so that the beamlet currents combine additively.
Yet the problem of a large array is that if there too many electron guns, the input impedance seen from a signal source will be excessive, and the bandwidth of the amplifier will be reduced. Thus, fewer electron guns having higher beamlet current are desirable. This might be possible with a large diameter beamlet, but the problem is that as the beamlet diameter increases, the deflector plate spacing within the electron gun must increase also. This reduces the gain and the amplifier performance.
In the previous embodiments, the electron gun was described as generating a substantially parallel beamlet of electrons as they passed through the deflector and exited into the drift cavity. To increase the beamlet diameter while still maintaining a small deflector plate spacing, the beamlet can be brought to a tight focus near the deflector, then allowed to de-focus quickly so that the space charge forces have little time to cause repulsive effects. As the beamlet enters the drift cavity, the beamlet can be allowed to increase to a much larger diameter than the deflector plate spacing, and this would reduce the space charge forces, but uncorrected would still leave unresolved the problem of beamlet spreading as the beamlet travels to the detector.
The solution is microlensing where a series of successively larger lensing electrodes provide successively larger lensing fields 5080, 5046, as shown in
The first doublet lens 4806 is as shown before in
The idea can be extended any time more current is effectively required from an electron gun without increasing the number of guns, or to couple more signals into the deflector array. The key concept here is the idea of electron lenses inside electron lenses inside electron lenses, which has ever been done before. For example, single microlensed electron guns, then bigger microlenses for subgroups of electron guns, then groups of guns in a doublet lens of the drift cavity is a real possibility that is practical and useful.
Multiple Deflector Load Compensation
Depending on the application, the electron beam amplifier may require up to several hundred deflectors to be coupled to a waveguide or transmission line. Multiple deflector coupling can be accomplished in the same manner as a single deflector so long as the total capacitance of the multiple deflectors is small relative to the waveguide impedance and the bandwidth required, and the area encompassed by the multiple deflectors is small enough that transmission line delays do not cause substantial differences in the electron beam deflection between any two deflectors in the array.
One problem of coupling multiple deflectors to a transmission line is the additional capacitive loading. As indicated previously, the capacitance of the array (CARRAY) might be greater than 100 fF. This is large enough that it can cause enough mismatch on the transmission line for destructive signal reflections to occur.
One further embodiment therefore mitigates these reflections by compensating the waveguide structure so that the loading of the deflector array creates a constant waveguide impedance. The general principle is to transform the waveguide impedance from an initial value Z0, where the guide does not couple to the CARRAY, to a larger value Z1 in the region where the guide couples to CARRAY. As known in the art, a waveguide can be viewed as a distributed ladder of series inductors and grounded capacitors per unit length (
The magnitudes of L0 and C0 are determined by the physical structure of the guide, but in general it can be appreciated that if L0 is constant, then increasing C0 reduces Z0 and decreasing C0 increases Z0. Thus, excess load capacitance decreases Z0, and by the previous formula for □, there will be reflections generated.
The formula therefore suggests another embodiment: If the capacitance of the deflector array is enough to induce undesired reflections, the waveguide structure can be modified across a section to reduce the distributed capacitance of the guide, thereby raising the impedance to a different value Z1. Then the deflector capacitance can be coupled in distributed fashion along the modified section so that the average distributed capacitance is the same as the unmodified guide. Thus, the effective impedance along the modified section of guide will equal Z0, the magnitude in the unmodified sections of guide. This can substantially eliminate any reflections from the deflector array loading.
Modifying a section of the guide can be quite simple in principle though details must be carefully determined in practice. For a simple two-wire transmission line, the wire spacing can be increased for the distance of the modified section. For a closed waveguide, the guide walls on which the electric field lines terminate (as in a TEmn mode) can be spaced further apart. This is illustrated schematically in
One problem with a diode detector is achieving sufficient current gain without incurring distortion in the output waveform. The cascade gain mechanism multiplies beam current without sensitivity to the voltage of the load, since it depends only on the beam energy and the semiconductor material. But the gain from this mechanism is limited to perhaps a few hundred, even with high beam energies. For this reason, a detector might be supplemented with avalanche gain, to further multiply the diode current by a second gain factor of 5—perhaps 20 or more. Thus, overall detector gain, which is the multiple of the cascade and avalanche effects can exceed several thousand, thereby providing significantly greater output drive and output power.
Avalanche gain is inherently voltage sensitive. Avalanche operates by creates a strong field across a reverse biased diode junction that is near breakdown; as electrical carriers (electrons and holes) drift into the internal field of the diode junction, they are accelerated sufficient velocity to impact with atoms in the crystal lattice, breaking free more electrons. These electrons are themselves then accelerated in the field, breaking free more electrons, and so on in a chain reaction the grows until the electrons leave the high field region.
The problem is that the intensity of the high field region is very sensitive to the external voltage across the diode. Even small changes in the voltage can cause large changes in the avalanche gain.
When an avalanche diode is connected directly to a load, the large current modulates the load voltage and hence the avalanche gain. Thus, if the avalanche diode is a detector, the beam current generates a cascade current in the diode, and the cascade current is multiplied by the avalanche gain, generating a diode output current which drives the load—but as the load voltage changes in response to the diode output current the voltage across the detector changes, and hence the detector gain changes, thereby modifying the output current. This makes it impossible for the load voltage to linearly follow the collected beam current, and hence, the output voltage becomes distorted by harmonics. While this might be desirable in a frequency multiplier, it is very undesirable in a linear amplifier.
Thus, one option is to isolate the detector from the load voltage, as shown in
In effect, the high transconductance of the bipolar isolates the detector from the load. According to bipolar physics, large changes in the bipolar emitter-collector current are caused by very small changes of only a few millivolts in the base-emitter voltage, or vice versa. Thus, if the base contact of the bipolar is fixed to a bias supple, large changes in the avalanche current transmitted to the bipolar emitter cause very little change in the voltage across the avalanche diode. The bipolar in effect behaves as an impedance transformer so that the avalanche diode sees a small “AC” resistance, while the bipolar sees the high resistance of the load.
HBT Detector
Another option is to make a detector supplementary gain without using the avalanche effect, as shown in
If minority carriers (in this case, the electrons of the beam as multiplied by the cascade action) enter a base region, they generate bipolar gain described by a current gain factor “beta”, or {tilde over (β)} Beta is also often called “hFE”, and is the ratio of the collector current to the base current. Typical values are β=100. For example, if the base current is 1 uA and beta=100, the collector current is 100 uA. Generally, the emitter current is very nearly equal to the collector current by (1+β)/β, so the two can be assumed the same value here for convenience.
The method of operation may depend on the ratio of the carrier mobilities μn and μp between base and emitter, the thickness of the emitter and base layers XE and XB, and the doping concentration of emitter and base layers, NE and NB. according to a formula
Controlling these parameters in a suitable device structure can thus create a detector of very high gain. To use this as a detector, the base is simply coupled to a fixed bias supply, and the emitter is coupled to a beam contact of suitable thin construction so as to permit beam electrons to pass through, and the collector is coupled to the load.
Injecting a beam current into the base of a detector so constructed multiplies the beam current, first by cascade, and then by the bipolar β factor. In this manner, extremely high detector output current can be achieved at the bipolar collector. For example, if the cascade gain is 100 and the bipolar gain is 100, an overall gain of 10,000 is possible. It works. Moreover, the bipolar gain mechanism is not nearly so sensitive to voltage excursions of the output voltage on the collector. Thus, it achieve improvement of the detector linearity in the manner of the aforementioned cascode structure.
Nonetheless, the bipolar detector is not completely immune to gain non-linearity. As is well-known, bipolar devices suffer a second-order modulation of their current gain as the collector-base voltage. This is not expressed in the previous equation, but the effect can be as much as tens of percent or as little as a few percent. Compared to the voltage sensitivity of an avalanche diode, which might vary the gain from 1 to 1000 for a change in voltage of a few volts, this is not much, but it can still be significant.
A second problem with the bipolar detector is AC feedback from the collector voltage to the base region. This is due to the junction capacitance between these two point, and the effect is to substantially reduce the bandwidth of the detector, by approximately the factor β. In high frequency RF circuits this is generally (almost always) avoided by using a cascode(common base) transistor to achieve AC isolation.
Thus, it may be appreciated that the bipolar detector could, in some circumstances, profit from isolating the collector of the detector from the load voltage, in the same manner as the avalanche diode detector can: with a cascode transistor. The method can, in fact, be the same: a bipolar or HBT transistor.
The following documents are incorporated by reference:
1 T. H. P. Chang et al, “Electron-beam microcolumns for lithography and related applications”, J. Vac Sci. Technol. B 14(6), November/December 1996, pp. 3774-3781
2 M. G. R. Thomson et al., “Lens and deflector design for microcolumns”, J. Vac Sci. Technol. B 13(6), November/December 1995 American Vacuum Society, pp. 2445-2449.
3 E. Kratschmer et al., “Experimental evaluation of a 20×20 mm footprint microcolumn”, J. Vac Sci. Technol. B 14(6), November/December 1996 American Vacuum Society, pp. 3792-3796.
4 T. H. P. Chang et al., “Electron beam microcolumn technology and applications”, Electron-Beam Sources and Charged-Particle Optics, SPIE vol. 2522, 1995, 10 pgs.
5 T. H. P. Chang et al., “Arrayed miniature electron beam columns for high throughput sub-100 nm lithography”, J. Vac Sci. Technol. B 10(6), November/December 1992 American Vacuum Society, pp. 2743-2748
6 T. H. P Chang et al., “Electron beam technology—SEM to microcolumn”. Microelectronic Engineering 32, (1996), pp. 113-130.
7 H. S. Kim et al., “Miniature Schottky electron source”, J. Vac. Sci. Technol. B 13(6), November/December 1995, pp. 2468-2472.
8 N. M. Froberg et al, “TeraHertz Radiation from a Photoconducting Antenna Array”, IEEE J. Quantum Electronics, vol. 28, No. 10, pp. 2291-2301 (1992)
9 Sang-Gyu Park et al, “High-Power Narrow-Band Terahertz Generation Using Large-Aperture Photoconductors”, IEEE J. Quantum Electronics, vol 35, No. 8, pp. 1257-1268 (1999).
10 Cha-Mei Tang et al, “Deflection microwave and millimeter-wave amplifiers”, J. Vac Sci. Technol. B 12(2), March/April 1994, pp. 790-794.
11 Manohara et al, “Design and fabrication of a THz nanoklystron”, Far-IR, Sub-mm & MM Detector Technology Workshop, Monterey Calif.; Apr. 1-3, 2002. www.sofia.usra.edu/det_workshop/papers/session6/3-43manohara_rev020911.pdf: www.sofia.usra.edu/det_workshop/posters/session3/3-43manohara_Poster.pdf
12 Kitamura et al, “Microfield emitter array triodes with electron bombarded semiconductor anode”, J. Vac. Sci. Technol. B 11(2), March/April 1993.
This application claims priority to U.S. provisional application Ser. No. 60/482,106 filed 23 Jun. 2003 and hereby incorporated by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 3221207 | Kaufman et al. | Nov 1965 | A |
| 3665241 | Spindt et al. | May 1972 | A |
| 3725803 | Yoder | Apr 1973 | A |
| 3749961 | Bates et al. | Jul 1973 | A |
| 3922616 | Weiner | Nov 1975 | A |
| 4263531 | Yoder | Apr 1981 | A |
| 4328466 | Norris, Jr. et al. | May 1982 | A |
| 4410903 | Wieder | Oct 1983 | A |
| 4498952 | Christensen | Feb 1985 | A |
| 4771214 | Takenaka et al. | Sep 1988 | A |
| 4902898 | Jones et al. | Feb 1990 | A |
| 4980567 | Yasuda et al. | Dec 1990 | A |
| 5041069 | Horst | Aug 1991 | A |
| 5122663 | Chang et al. | Jun 1992 | A |
| 5186670 | Doan et al. | Feb 1993 | A |
| 5191217 | Kane et al. | Mar 1993 | A |
| 5319322 | O'Loughlin | Jun 1994 | A |
| 5355380 | Lin et al. | Oct 1994 | A |
| 5363021 | MacDonald | Nov 1994 | A |
| 5430347 | Kane et al. | Jul 1995 | A |
| 5489818 | Naff et al. | Feb 1996 | A |
| 5497053 | Tang et al. | Mar 1996 | A |
| 5514847 | Makishima et al. | May 1996 | A |
| 5592053 | Fox et al. | Jan 1997 | A |
| 5773933 | Yoder | Jun 1998 | A |
| 5842680 | Bustamante et al. | Dec 1998 | A |
| 5955849 | Tang et al. | Sep 1999 | A |
| 6023060 | Chang et al. | Feb 2000 | A |
| 6145438 | Berglund et al. | Nov 2000 | A |
| 6177909 | Reid et al. | Jan 2001 | B1 |
| 6281508 | Lee et al. | Aug 2001 | B1 |
| 6356221 | LeChevalier | Mar 2002 | B1 |
| 6369385 | Muray et al. | Apr 2002 | B1 |
| 6384519 | Beetz, Jr. et al. | May 2002 | B1 |
| 6433303 | Liu et al. | Aug 2002 | B1 |
| 6479320 | Gooch | Nov 2002 | B1 |
| 6498349 | Thomas et al. | Dec 2002 | B1 |
| 6518923 | Barquist et al. | Feb 2003 | B2 |
| 6545425 | Victor | Apr 2003 | B2 |
| 6617587 | Parker et al. | Sep 2003 | B2 |
| 7157703 | Nakasuji et al. | Jan 2007 | B2 |
| Number | Date | Country | |
|---|---|---|---|
| 20050285541 A1 | Dec 2005 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60482106 | Jun 2003 | US |
