RF design rules

Size: 10-inch pole diameter ... Dee voltage: 10,000 to 14,000 volts dee-to-dee; R.F. power: 2,000 watts; R.F. frequency: 25.68 megacycles ... Magnetic field strength: 17,000 gauss

McGuire, The Iowa State University 1.5 MeV Undergraduate Cyclotron (1961) — p. 9

Tabletop: This is the closest historical analogue to a next machine's target: same pole diameter as the reference machine, ~3x their field and ~10x their dee voltage buy ~10x the energy.

Cyclotron magnet design should always consider interaction with subsystems: RF system, vacuum pumping, ion source or injection system, extraction system or internal target, diagnostic probes.

Zaremba, Magnets for Cyclotrons (2005) — p. 3, 45

Tabletop: The most common amateur failure mode is a magnet that works but leaves no port for the probe or pump; run this five-item checklist on every layout iteration for a next machine.

For best RF efficiency, k=0.5 BUT to decrease machine dimensions k >0.5 (more hill, thus more field) CHOICE : k=0.67 (60 deg hills)... CHOICE : nu_z = 0.2

Zaremba, Magnets for Cyclotrons (2005) — p. 32-33

Tabletop: If a next machine goes AVF to escape the weak-focusing energy ceiling, these are proven starting numbers: 3 or 4 sectors, half-open valleys, and a modest nu_z ~ 0.2 target.

The cyclotron circuit was originally tuned to a frequency of 13.56 MHz due to the requirements of the commercial RF generator in use ... a magnetic field of 0.889 Tesla is required.

Chun, The Cyclotron Magnet and RF Oscillator (2003) — p. 10-11

Tabletop: Directly actionable option for a next machine: targeting 0.89 T instead of 0.59 T puts the machine on the 13.56 MHz ISM band where used generators, amplifiers, and matchboxes are plentiful and legal.

B (Tesla) 1 ... 1.6 ... f (MHz) 15.23 ... 24.36

Baumgartner & Heuer, The Cyclotron Kids 14-Inch Accelerator (2010) — p. 12

Tabletop: The reference machine's 0.59 T machine resonates at ~9.0 MHz; any next machine's field choice instantly fixes the oscillator/tank tuning range via this 15.23 MHz/T constant.

Polepieces of meehanite steel facemilled to a profile that gave appropriate field index... 1.6kVpp on Ds, 400Wpk. Base pressure 0.01mTorr

Niell, Effective Scientific Equipment Procurement Strategies: Building on a Budget (2010) — p. 11-17

Tabletop: An existence proof at exactly the reference machine's energy: modest dee voltage (1-2 kVpp), 1e-5 torr, and a machined pole profile suffice below ~300 keV; heroic RF and UHV are not prerequisites.

12 inch diameter poles pieces forming a 4-inch gap to which upper and lower pole tips up to 1-inch thick can be easily attached and removed.

Koeth, Undergraduate Education with the Rutgers 12-Inch Cyclotron (2015) — p. 2

Tabletop: A complete cross-check machine for a next machine's sizing; note the removable-pole-tip trick that lets one magnet host many field profiles.

For best RF efficiency, k=0.5 BUT to decrease machine dimensions k >0.5 (more hill, thus more field) CHOICE : k=0.67 (60 deg hills)

Beeckman, Cyclotron Magnets — ECPM37 lecture, Groningen (2009) — p. 69

Tabletop: Shows the exact arithmetic used to go from a required <B> to hill/valley fields and sector angle - reusable at any scale.

A position of suitably low field intensity, < 60 oersteds, was located by mapping the stray field about the magnet.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 59

Tabletop: Directly applicable: map the reference machine's H-frame fringe field with a hall probe and keep the LDMOS amplifier, its magnetics, and instrumentation outside the ~60 G contour.

Increasing the dee voltage, however, requires increasing the required voltage breakdown gap and thus the magnet hill gap, so that some compromise must be reached.

Livingston & Howard (eds.), The Oak Ridge Relativistic Isochronous Cyclotron — ORNL-2648, OSTI 4275955 (1958) — p. 65

Tabletop: The coupled optimization transfers: for a next machine, pick dee voltage and magnet gap together, since every kV of dee needs clearance that costs ampere-turns.

The conductivity is maintained at 10 micromhos or less, with a pH of about seven. Dee system water temperature stability of 1 F or better is required for steady operation.

Ramler & Parker, The Argonne 60-Inch Cyclotron — ANL-5907 (1959) — p. 6

Tabletop: Two directly portable specs: DI-water loop quality for any hollow-conductor coil, and tight dee-water temperature control if a next machine water-cools the dee.

It is necessary for the secondary coil to be cooled with oil or deionized water... because even minute thermal expansion of the copper can change the inductor's value.

Heuer & Baumgartner, Design of a 2 MeV Cyclotron (2009) — p. 23

Tabletop: Explains RF drift during long runs at the reference machine's power levels; cooling the tank coil stabilizes tune.

The Dee capacitance is approximately 76 pF. The secondary coil inductance of 0.87 uH or more in parallel with the Dee capacitance yields a resonance as high at 19.5 MHz

Yuly, The Houghton College Cyclotron: a Tool for Educating Undergraduates — Cyclotrons 2013, WE1PB01 (2013) — p. 3

Tabletop: Concrete LC numbers for a next machine's tank at the same scale; the swappable-primary approach lets you retune coupling without rebuilding the tank.

It is important for the coils should be in the air stream of one of the cooling fans (they will run hot if not).

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 28

Tabletop: Directly applicable to the homemade dee-tank coil, which sees circulating RF current far above the dee's DC feed current.

This operated normally at 65 C but when an identical coil, which had been nickel plated, was substituted, the operating temperature rose to 350 C.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 4

Tabletop: Reject nickel-plated hardware (and nickel underlays beneath chrome or silver) anywhere RF current flows in the resonator, coil, or ground-return path.

Most commercially available copper tube contains 0.015% to 0.08% phosphorus as a de-oxidising agent, so that its conductivity may range from 60% to 90% I.A.C.S.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 6

Tabletop: Buy the tank-coil tubing as electrolytic/electrical-grade (C10100/C11000) copper, not generic plumbing tube, for up to ~20% lower RF resistance.

typically have true Q values of up to about 800. Very few practical circuits require a Q above 900. Attempting to design a coil with a True Q much over 1,000 usually results in a coil with abnormal physical dimensions

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 1

Tabletop: Budget the resonant step-up assuming coil Q of a few hundred (loaded lower still), not textbook thousands, when sizing the amplifier for 5-13 kV dees.

the presence of the coil's distributed capacity causes the Q observed by the Q meter to be lower than the true Q of the coil

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 1

Tabletop: When characterizing the dee resonator with a VNA or Q meter, treat the reading as a lower bound and keep leads/fixture capacitance minimal.

Q increases with coil diameter (see figs. 1-4). Q increases with coil length, rapidly when the L/d ratio is small ... Q increases with frequency

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 1-2

Tabletop: At 9 MHz a 3-4 inch diameter tank coil can reach Q well over 1000, directly multiplying dee voltage per watt of drive.

The conductor diameter must be within the range of 0.45 and 0.70 times the center-to-center distance between adjacent turns (not all commercial stock coils meet this condition).

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 2

Tabletop: For a 9 MHz matching/tank inductor, space the turns so the wire fills 45-70% of the pitch; close-winding bare tubing throws away Q to proximity effect.

Maximum Q occurs at a coil L/d ratio of between (depending on other coil design parameters) 0.35 and 0.45, decreasing rapidly below that ratio and more slowly above

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 2-3

Tabletop: Make the resonator coil short and fat (roughly half as long as its diameter), not the long skinny solenoid that fits most easily in a corner.

The equations do not hold for coils with a length-to-diameter ratio of less than 0.35:1, coils with less than about 4 turns, or coils with conductor diameter-to-turn spacing ratios of less than 0.45:1 or greater than 0.70:1.

Murphy, The Elusive Q of Single-Layer Air-Core Coils — CQ Magazine, May 1999 (1999) — p. 3

Tabletop: A 2-3 turn link or coupling loop at 9 MHz is outside the formulas; measure it rather than calculate it.

Measurement of the capacitance with an L-C meter yields a value of 78.1pF. Nice agreement seen!

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 1

Tabletop: Directly usable on the reference machine's 8-inch dee: sum simple parallel-plate terms for top/bottom/edge and verify with a cheap L-C meter before winding the tank coil.

the trend of DEE voltage to follow the square root law of the input RF power is accurate over all measured power ranges

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 2-3

Tabletop: This is the sizing equation for the reference machine's LDMOS upgrade: doubling dee voltage costs 4x power, so going from 1.3 kV to 5-13 kV needs a 15-100x power increase unless L/C or Rs improves.

as if Rs had the value of 800mOhm - sixteen times that of the expected coil Rs ... take into account the stainless steel vacuum chamber return, the stainless steel Conflat DEE stem support and RF feed throughs.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 2-3

Tabletop: When predicting a next machine's dee voltage, don't use the coil resistance alone; the stainless chamber and stem return path dominates losses, so use copper return paths where possible and expect ~1 ohm scale Rs.

the induced voltage on the capacitive pickup facing the DEE was calibrated against forward power at lower levels. Extrapolation allowed us to measure forward power levels up to 1300 watts

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 3-4

Tabletop: Exactly the measurement chain the builder needs for the LDMOS upgrade: calibrate their pickup at 5-50 W against a scope probe, then trust the pickup alone at 100-500 W.

the P6015 probe introduced 3.0pF of capacitance; the tank circuit was indeed reduced in frequency corresponding to 3 pF

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 4

Tabletop: With the reference machine's ~78 pF-class dee, 3 pF is a ~2% frequency pull - enough to detune a high-Q tank, so calibrate with the probe on, then remove it and retune.

The connections to one of the coils are then interchanged and the equivalent inductance is measured again. The difference between the two measured inductances is then 4M.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 5-6

Tabletop: A bench L-C meter trick the builder can use to characterize their coupling loop; M of tens of nH is the expected scale for a matched half-turn loop.

is exactly 1/2 of Qo when the primary is critically coupled corresponding to the value giving maximum response ... a simple reflected RF power meter will suffice to show a perfect match - indicating critical coupling.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 6-7

Tabletop: The builder can tune their coupling loop with just an SWR bridge: adjust loop position/taps until reflected power nulls, and check Qloaded = Q0/2 with a NanoVNA S21 sweep.

From Fig.12 we determine Qmeasured at a distance of 11mm to be 460. This implies a Qo of 920.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 6-8

Tabletop: A benchmark for the reference machine's ~9 MHz tank: if their measured Q0 is far below ~900, there is excess loss (bad joints, steel in the return path) worth hunting down.

Empirically it was found for a given forward power into each of the (50+j0) Ohm points, the peak capacitor voltage was always the same.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 7

Tabletop: The builder need not agonize over loop position vs tap point: any combination that nulls reflected power delivers identical dee voltage, so pick the mechanically stable one.

minimizing Rs, or increasing L2 while simultaneously decreasing C2 (to maintain the resonant frequency) are the only parameters that can be adjusted to increase the DEE voltage for a given amount of RF power.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 8

Tabletop: For a next machine, shrinking dee-to-liner capacitance (larger dee-to-lid spacing) and using a bigger low-loss coil buys dee voltage for free before spending on amplifier watts.

It is not necessary to operate at 2000 watts, as shown above 1000 watts produces a peak-to-peak DEE voltage of approximately 15kV.

Koeth, Theoretical Calculations and Measurements of the DEE Voltage in the Rutgers 12 Inch Cyclotron (2005) — p. 8

Tabletop: Scales the reference machine's plan: with a similar tank, a 500 W LDMOS amp should land near 10 kVp-p - comfortably in their 5-13 kV target - and thermal management of stem and coil becomes the real issue.

the original purpose of the shield was to prevent RF from coupling to the pickup ... We agreed a large series inductance (an RF Choke) should mitigate this concern.

Koeth, Rutgers 12 Inch Cyclotron Ion Source Studies: Part I (2006) — p. 1

Tabletop: Lets the builder use an unshielded collector at nA levels: a ~100 uH-mH choke at the feedthrough kills 9 MHz pickup without blocking DC beam current.

an input RF power level of 50 watts is too low, and 500 watts should be sufficient. The first ions are expected to clear the chimney at approximately 200 watts.

Koeth, Rutgers 12 Inch Cyclotron Ion Source Studies: Part I (2006) — p. 5

Tabletop: A geometry trap for a next machine: any chimney or source structure must be smaller than the first half-turn diameter set by the dee voltage, or beam dies before the first gap crossing.

Protons: f (megacycles) = 1.52B (kilogauss) ... The actual technique used to control resonance in a cyclotron is to vary the magnetic field, with the applied frequency held constant.

Livingston & Blewett, Particle Accelerators (1962) — p. 156

Tabletop: The reference machine's 0.59 T (5.9 kG) gives 8.97 MHz, confirming their ~9 MHz choice; for a next machine pick B first, then f = 1.52*B.

there will be a somewhat lower potential at the ends of the D faces nearest the lines ... measured in some cyclotrons to be as great as 5 per cent ... a displacement of the ion source of over 2 in. has been necessary.

Livingston & Blewett, Particle Accelerators (1962) — p. 164

Tabletop: Make the source mount adjustable by a few mm in both directions and tune position for beam, not for geometric center.

the practical maximum migration in phase will be from zero to -pi/2 and back to zero, a total phase migration of pi radians or one half-cycle.

Livingston & Blewett, Particle Accelerators (1962) — p. 166-171

Tabletop: With a 3-4% field droop and 160 turns-scale acceleration, the reference machine's dee voltage sets how much phase slip they can afford: higher V = fewer turns = more field-shape tolerance.

Increasing the D voltage requires fewer turns for acceleration to maximum energy and will compensate for a larger phase shift. However, D voltage is usually limited by ... power and spark breakdown.

Livingston & Blewett, Particle Accelerators (1962) — p. 172

Tabletop: At 1.3 kV and 160 keV the reference machine's ions make ~60+ turns; doubling dee voltage halves turns and dramatically relaxes both field-uniformity and vacuum (scattering) requirements.

The gap between chamber lids was chosen to be 5 in., leaving 1 1/4-in. clearance between D's and lids ... resulting in a D-voltage limit of about 70 kv due to breakdown.

Livingston & Blewett, Particle Accelerators (1962) — p. 175

Tabletop: At 1.3 kV the builder has enormous margin; for a next machine at several kV, ~50 kV/in of clearance in vacuum with rounded edges is a comfortable design gradient.

it has been found necessary to have these tubes spaced as closely as 2 to 3 in. to prevent local heating and warping of the D's under power.

Livingston & Blewett, Particle Accelerators (1962) — p. 175

Tabletop: At tens of watts the builder needs no water, but the warping lesson stands: dee thermal drift detunes the resonator, so keep dee structures stiff and thermally anchored.

At MIT, with an internal D aperture of 1.6 in. the optimum length of ionization column was 5/8 in. For 4-in.-wide D's in the Carnegie Institution 60-in. machine it was 1 3/8-in.

Livingston & Blewett, Particle Accelerators (1962) — p. 178

Tabletop: Hood or collimate the reference machine's source so only ~1/3 of the dee aperture height of plasma column is exposed; more column means RF load, not more beam.

The only general rule is: The simpler the physical structure and the shorter the leads and connections, the less subject is the oscillator to parasitics.

Livingston & Blewett, Particle Accelerators (1962) — p. 185-187

Tabletop: If the next machine goes two-dee push-pull, watch for the push-push mode (no accelerating voltage, oscillator happily locked); a single-dee-plus-dummy design sidesteps it.

This loading of the D circuit by discharge currents holds the D potentials down to a few hundred volts ... Unless the loading is removed, the chamber will continue to operate in the low-voltage, blue-glow discharge condition indefinitely.

Livingston & Blewett, Particle Accelerators (1962) — p. 188

Tabletop: The reference machine's ~1.3 kV dee sits right in classic multipactor territory; surface conditioning, low pressure, and the ability to snap the drive up fast are the standard escapes.

Such a variable capacitance can be provided by a movable plate on the side wall of the chamber facing one edge of the D ... a range of motion sufficient to tune over about 1 per cent in frequency.

Livingston & Blewett, Particle Accelerators (1962) — p. 188

Tabletop: A bellows-actuated plate near the dee gives the builder live resonance trim without opening the chamber - invaluable when thermal drift walks the dee frequency.

dust should be controlled and all grease removed (even fingerprints), and under no circumstances should steel wool or coarse abrasives be used in cleaning.

Livingston & Blewett, Particle Accelerators (1962) — p. 189

Tabletop: After every chamber opening, budget an hour of gradually raised dee voltage for conditioning before expecting stable beam.

The most significant advantage of this circuit is its simplicity and compactness along with the freedom from delicate tuning requirements or precise construction.

Livingston & Blewett, Particle Accelerators (1962) — p. 190-193

Tabletop: The same logic favors the reference machine's self-excited or PLL-followed drive: let the dee resonator define frequency so mechanical drift retunes the drive instead of killing the beam.

about 50 per cent thicker than the depth of the groove to allow for compression ... 1/4-in. gaskets have proved adequate for even the largest seals ... Conductivity for rf currents through such a seal can be assured by half-covering the gasket with a thin copper-foil strip.

Livingston & Blewett, Particle Accelerators (1962) — p. 199-201

Tabletop: Directly usable rules for the reference machine's lid and port seals; the copper-foil RF bridge over elastomer joints prevents mysterious Q loss and local heating.

apply a radio frequency oscillating voltage to the electrode that is slightly less than the cyclotron frequency given at the center of the field, but greater than [that] near the edges.

King, A Preliminary Design for a Small Permanent Magnet Cyclotron — Houghton College thesis (2002) — p. 20

Tabletop: A concrete tuning rule for the builder: don't tune RF to the central field value; split the difference toward the outer-radius field.

It would only take a few cycles, on the order of 12, for most cyclotrons to have reached this velocity.

King, A Preliminary Design for a Small Permanent Magnet Cyclotron — Houghton College thesis (2002) — p. 20

Tabletop: Quantifies how little phase budget a mistuned tabletop machine has; explains failed runs where beam dies at small radius.

it has one dee-shaped copper electrode, and the grounded vacuum chamber functions as the other electrode

Dewan, Design and Construction of a Cyclotron Capable of Accelerating Protons to 2 MeV — MIT thesis (2007) — p. 13

Tabletop: The reference machine already does this; it remains the right choice for a next machine unless push-pull two-dee RF is needed for higher energy gain per turn.

The dees, made of thin sheet copper, arc 22.5 cm in diameter, 2.4 cm high, and they are separated by a gap of 1.5 cm.

McGuire, The Iowa State University 1.5 MeV Undergraduate Cyclotron (1961) — p. 7

Tabletop: A directly copyable dee geometry for an 8-10 inch pole; note the dees must be water cooled once RF power reaches ~kW.

This value is about 10% for 120 uamp of deflected deuterons, increasing to 15% for a 200 uamp beam. About 30% of the internal beam at the exit radius is extracted.

Ramler & Parker, The Argonne 60-Inch Cyclotron — ANL-5907 (1959) — p. 19

Tabletop: Sets expectations if a next machine attempts a deflector: losing two-thirds of the beam at the septum is normal, not failure.

The voltage breakdown limit determines the necessary gap width. The empirically determined limit (valid for clean, flat surfaces) is d[mm] >= 1.41 x 10^-2 * phi[kV]^(3/2).

Wolf (ed.), Handbook of Ion Sources (1995) — p. 379

Tabletop: Direct rule for the reference machine's source-to-puller spacing and any dee-to-ground clearance: a few-kV dee needs sub-mm minimum, but leave margin because sputtered metal films spoil the 'clean surface' assumption fast.

Dielectric strength / Thickness: 240 kV/mm at 0.025 mm; 70 kV/mm at 5 mm.

Wolf (ed.), Handbook of Ion Sources (1995) — p. 523

Tabletop: When insulating the reference machine's extraction or dee leads with PTFE sheet or heat-shrink, use the bulk (70 kV/mm-class) figure with a 5-10x safety factor, not the datasheet film value.

Compared to T or Pi networks, the L network uses only one combination of inductance and capacitance. This simplifies the microcontroller tuning algorithm.

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 11

Tabletop: If the builder automates their dee match at 9 MHz, a stepper-driven L-network is the simplest topology whose tuning can't get lost in redundant solutions.

The coupling factor is high, ~1000 or 30 dB, to minimize main line power loss ... enables 200 W power measurements using the AD8307 logarithmic detector IC

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 18

Tabletop: Exactly sized for the reference machine's 100-500 W upgrade: a homebrew 30 dB coupler plus AD8307 boards gives continuous forward/reflected monitoring across their whole power range.

Ferrite toroids wound with AWG 26 wire and surrounding two 2 inch sections of RG-8 50 Ohm coaxial cable form the coupling transformers.

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 18-19

Tabletop: A ~$5 coupler build that brackets the reference machine's 9 MHz band; the shield-through-toroid trick is the detail that makes homebrew directivity respectable.

the SWR measured using this directional coupler is 1.08 and 1.03 for 28 dB and 35 dB of directivity, respectively

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 19-20

Tabletop: Tells the builder not to chase SWR below ~1.1 on a homebrew bridge - that residual is the instrument, not the match.

Figure 42 - 30 W to 100 W Power Calibration Setup using Bird 43 Wattmeter

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 57-58

Tabletop: The builder already lives in this instrument ecosystem; a Bird 43 (or borrowed one) transfers absolute power calibration to permanently installed cheap sensors.

actuates stepper motors to alternately adjust a variable capacitor and a variable inductor to reduce VSWR to less than 1.5:1

Blodgett, 3.5 to 30 MHz Automatic Antenna Impedance Matching System (2012) — p. 6-8, 40

Tabletop: SWR 1.5:1 is the protection threshold for the reference machine's LDMOS amp too; coordinate-descent on C then L converges fine for the single-resonance dee load.

fo = qBo/2pi mi = (1.52x10^7) Bo(tesla)/A

Humphries, Principles of Charged Particle Acceleration (1986) — p. 524

Tabletop: One-line check of the reference machine's operating point: 0.59 T -> ~9.0 MHz for protons; sets the next machine's RF band for any target field.

the final kinetic energy is maximized by taking Vo large... a high gap voltage accelerates particles in fewer revolutions so that there is less opportunity... to get out of synchronization.

Humphries, Principles of Charged Particle Acceleration (1986) — p. 530-531

Tabletop: At sub-MeV this limit is distant (10 kV dee -> ~3 MeV proton ceiling), but the same physics governs field-flatness tolerance: fewer turns forgives more field error.

Low energy proton in 1 T field: 15.23 MHz

Antaya, Cyclotron Basics — MIT 8.277, Unit 10, Lecture 14 (2010) — p. 29

Tabletop: The single most-used number in the reference machine's notebook: 0.59 T -> 9.0 MHz; a 1.2 T higher-field successor -> 18.3 MHz, still comfortable amateur-radio-technique territory.

250 MeV protons; 17 KeV/turn: N~15,000... 250 MeV protons r=0.3m: dr/dN ~ 20 microns!

Antaya, Cyclotron Basics — MIT 8.277, Unit 10, Lecture 14 (2010) — p. 43

Tabletop: For the builder: 1 MeV at 2 kV/gap (2 gaps) is ~250 turns with final-orbit spacing ~0.2 mm at r=12 cm - explaining why higher dee voltage directly eases both extraction and vacuum requirements.

dphi/dn=360 [gamma-1] -> 8 deg. An ion on peak phase is lost in 11 revolutions. Only solution- very high energy gain per turn - 360kV

Antaya, Cyclotron Basics — MIT 8.277, Unit 10, Lecture 14 (2010) — p. 51

Tabletop: Reassurance and ceiling in one number: at 1 MeV gamma-1 = 0.001, ~0.4 deg/turn - a next machine is nowhere near the relativistic limit, and the classical (non-AVF) architecture is fine to several MeV.

The second DEE has been faked using the image of the real DEE on a grounded conductor ... By twisting the inductor, we can change the inductance to match our inductance requirements.

Chun, The Cyclotron Magnet and RF Oscillator (2003) — p. 11

Tabletop: This is the reference machine's exact topology, validated on a comparable machine; the deformable-inductor trim is a simple tuning mechanism for a next machine.

high precision is necessary for a coil or circuit with a high Q value ... The Q of this cyclotron was measured at 1600, under no loading.

Chun, The Cyclotron Magnet and RF Oscillator (2003) — p. 11-12

Tabletop: Direct benchmark for the reference machine's resonator: if measured Q is far below ~1000, hunt for lossy joints; and thermal drift of a Q~1600 circuit needs active or frequent retuning at 9 MHz.

For typical cases of interest, 'vacuum' pressure is lower than 10-5 torr, and 'gas' pressure higher than 10-4 torr.

Werner, Probing and Modeling Voltage Breakdown in Vacuum — Cornell dissertation (2004) — p. 23

Tabletop: Cyclotrons run 1e-5 to 1e-4 torr with gas feed - squarely in the gray zone - so the reference machine's spark limit will move with operating pressure, and tests at base pressure overstate what they can hold with gas flowing.

r.f. power should never be left on for prolonged periods under these circumstances, else the risk is run of cracking the glass dee insulators. The power and length of application should be gradually increased

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 10

Tabletop: Directly applicable startup ritual at the reference machine's 1.3 kV dee level; persistent glow during conditioning signals organic contamination (grease, oil, rubber) in the tank.

the 'single-dee' construction; this has many advantages ... Better ion focussing can be obtained by installing a 'dummy' grounded dee edge symmetric to the insulated dee, but this is a refinement

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 8

Tabletop: Exactly the reference machine's architecture; the dummy-dee edge is a proven low-cost next-machine upgrade for a cleaner accelerating gap.

The dee-to-ground capacity appears as the major portion of the capacitance in the tank circuit, which must be calculated taking this into account

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 8

Tabletop: Even with a modern solid-state chain, the builder must treat dee capacitance as the resonator's dominant C when designing the matching network; the confine-the-RF-current lesson is timeless.

it is important to provide short, broad paths for current flow, especially in the ground circuits ... While the tube should be placed as close to the tank as possible, it must yet be kept away from the magnetic field

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 8-9

Tabletop: Directly applicable to the reference machine's amplifier: wide copper sheet/strap grounds and a short feed run, with magnetically sensitive parts (and LDMOS heat sinks) out of the fringe field.

All circuits connected to tank elements should have adequate choking and bypassing to prevent r.f. from reaching the meters and supply lines.

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 9

Tabletop: Directly applicable; the reference machine's beam-current, bias, and gauge lines all need feedthrough RC/choke filtering at 9 MHz.

The voltages employed on the various cyclotron components are deadly; proper precautions must be taken, even during preliminary testing ... Interlock switches on the power supply covers and grounding hooks

Wouters, General Recommendations for Design of Small Cyclotrons — UCRL-476 (1949) — p. 9

Tabletop: Directly applicable home-lab safety baseline for a next machine.

the top and bottom of the chamber to bow in, which affected the capacitance of the dee and reduced the maximum voltage that the dee could withstand before flashing over.

Baumgartner, The Cyclotron Kids' 2 MeV Proton Cyclotron — Cyclotrons 2013, WE1PB05 (2013) — p. 2

Tabletop: Directly relevant to any thin-lid chamber on a next machine squeezed into a small magnet gap: plan support posts (clear of the beam spiral) from the start.

Having only one dee rather than two doubles the voltage requirement, but reduces the cost and complexity of having two RF feedthroughs in the vacuum chamber.

Baumgartner, The Cyclotron Kids' 2 MeV Proton Cyclotron — Cyclotrons 2013, WE1PB05 (2013) — p. 2

Tabletop: Confirms the single-dee choice for a next machine unless dee voltage becomes the binding constraint.

Mammoflex M-1 HV cable has C of 56 pF per foot ... ~20 feet total ~0.4 Joules at 30 kV ... Removed excess cable. Run is now ~ 5 feet total

Ponter, Beam Energy Measurements with a New HV Deflection System and Ion Source Upgrades on the Rutgers 12-Inch Cyclotron (2010) — p. 43-49

Tabletop: For any HV feed on a next machine (deflector, source bias): keep cable runs minimal - stored cable energy, not the supply, does the arc damage.

We've encased the resistor in a grounded shield, and the coax shields go through 68 Ohm, 2 watt resistors

Ponter, Beam Energy Measurements with a New HV Deflection System and Ion Source Upgrades on the Rutgers 12-Inch Cyclotron (2010) — p. 44-52

Tabletop: A ready-made HV-distribution recipe for the reference machine's deflector or PIG source bias; note even this shielding didn't stop arcs until the cable-energy fix - resistors limit damage, they don't prevent flashover.

HP 33120A Function Generator - ENI 155LCRH Power Amp - LDG AT-200PC Tuner - Bird 43A RF Power Meter - Dee

Yuly et al., Modifications on the Houghton College Cyclotron (2010) — p. 12-14

Tabletop: This is essentially the reference machine's current architecture, validated: a ham antenna tuner really can match a dee, at the cost of low Q and ~1-2 kV ceilings.

3.55 MHz 1700 Vpp (26 W) ... 3.48 MHz 800 V (10 W)

Yuly et al., Modifications on the Houghton College Cyclotron (2010) — p. 17-19

Tabletop: Benchmarks the reference machine's setup (1.3 kV from 5-50 W is right on this curve) and warns that the autotuner path plateaus in the low-kV range.

No peaks for magnetic fields larger than H2+ at 0.5 T -> 35 keV; 44 orbits at 800 Vpp

Yuly et al., Modifications on the Houghton College Cyclotron (2010) — p. 21

Tabletop: Quantifies why the reference machine's dee-voltage upgrade matters: at 1.3 kV their protons need ~hundreds of turns to reach interesting energies, and ~44 turns was already the survival ceiling at Houghton's pressures.

To develop high voltages with modest RF power. The highest voltage for given power occurs when: Qloaded = 1/2 Qo

Koeth et al., The Rutgers 12-Inch Cyclotron for Students (2010) — p. 18

Tabletop: The one-slide justification for the builder to move from an antenna-tuner match to a true high-Q tank circuit in a next machine.

Record Input Power 2kW: 8.4 kVpeak

Koeth et al., The Rutgers 12-Inch Cyclotron for Students (2010) — p. 19

Tabletop: Anchors the power budget: even a well-built 12-inch tank needs kW-class RF for ~10 kV dee voltage, so the reference machine's 500 W LDMOS should target ~4-8 kV peak.

Calculation showing first ions squeak by at 165 Watts ... Beam current abruptly dropped to zero at 170 watts !

Koeth et al., The Rutgers 12-Inch Cyclotron for Students (2010) — p. 20

Tabletop: A free end-to-end check for the builder: the RF power at which beam vanishes measures the true dee voltage through pure geometry, independent of every probe.

the DEE, chamber, tank coil, etc. heat up and slightly change the resonant frequency ... A DC 'error signal' is derived from comparing the phase of the driving RF to the Phase of the DEE pickup.

Koeth et al., The Rutgers 12-Inch Cyclotron for Students (2010) — p. 21

Tabletop: At 100-500 W the next machine will drift off resonance within minutes of turn-on; this phase-lock autotuner (or the equivalent PLL driving their signal source) is the fix.

Beam Inferred DEE Voltage

Koeth et al., The Rutgers 12-Inch Cyclotron for Students (2010) — p. 33

Tabletop: The builder can cross-check their 1.3 kV estimate by measuring where the first half-turn lands - the beam itself is the most honest voltmeter.

A large increase in current occurs only as the local field approaches 10^10 V per meter... After several minutes of current flow at the constant voltage, a remeasurement of the threshold voltage shows that it has increased. This phenomenon is called conditioning.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 111-113

Tabletop: Bring the reference machine's dee and extraction voltages up over tens of minutes on first pump-down, watching for micro-discharge pulses; a gap that arcs at 15 kV cold will often hold 20+ kV after patient conditioning.

They found that at a critical angle of 31.5 deg, the surface charge was zero... by screening the section of the insulation surface near the cathode... the breakdown voltage was raised by a factor of approximately 2.5.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 113-114

Tabletop: Free flashover margin for the next machine's source stalk and dee-stem insulators: cone the insulator ends at ~30 degrees toward the negative electrode and recess the triple junction behind a metal skirt.

Gleichauf also found that the breakdown voltage was strongly dependent on the material of the insulator but independent of the material of the electrodes.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 113-114

Tabletop: Budget ~2 kV per mm of insulator surface path in vacuum (before sputter contamination); a 20-kV extraction stalk wants >=10 mm of clean creepage plus corrugations.

Air spaces exist in solid and liquid dielectrics... the air will have the higher stress, possibly causing sparkover through the air space... The stress in the air gap can thus be k times that in the solid.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 116, 119

Tabletop: The classic failure of home-built HV feedthroughs: a loose PTFE sleeve over a rod arcs in the annular air film; pot it, oil-fill it, or evacuate the annulus so Paschen cannot be satisfied.

The optimum ratio as r_i/r_o = 1/e. This optimum ratio minimizes the stresses within the coaxial electrode arrangement, independent of the material of the dielectric used.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 117, 122-124

Tabletop: Sizes the reference machine's HV stalk directly: for a grounded 25-mm-bore chamber port, a ~9-mm center conductor minimizes field stress; and never leave a sharp-edged washer or nut on the HV end.

This phenomenon causes the cathode grid material from the IEC device to be deposited on the high-voltage (HV) stalk. That can in time cause premature breakdown at the stalk.

Miley & Murali, Inertial Electrostatic Confinement (IEC) Fusion: Fundamentals and Applications (2014) — p. 87, 105, 109

Tabletop: In the reference machine's small chamber everything sees the source; a simple washer-stack or skirt shielding the feedthrough ceramic from the chimney slit will multiply time-between-cleanings.

The axial component of the magnetic field can be decreased at larger radii in order to increase the radial (focusing) component, provided the ions only have a few revolutions left once they reach this portion of the field.

Morrow, Focusing in the Houghton College Cyclotron — Houghton College thesis (2015) — p. 27-28

Tabletop: Explains why the reference machine's next big win may be Dee voltage, not magnet shaping: at ~160 keV with a low Dee voltage the turn count is what kills the beam.

the location of current spikes at a given field strength always occur at or very near the theoretical resonances... at B/3, B/5, and so on.

Fuller, Exploring the Capabilities of the Houghton College Cyclotron — Houghton College thesis (2013) — p. 46-47

Tabletop: Essential for interpreting the reference machine's magnet scans: a peak at one-third field is a subharmonic, not a mystery species, and H2+ vs H+ vs They peaks can be disentangled this way.

It can be seen that in general, an increase in dee voltage results in a higher beam current.

Fuller, Exploring the Capabilities of the Houghton College Cyclotron — Houghton College thesis (2013) — p. 49-50

Tabletop: For a fill-gas machine like the reference machine's, dee volts are the strongest single knob on beam current; prioritize RF voltage over almost everything else.

the RF generator was adjusted in steps of 0.5 MHz ... adjusted in steps of 0.1 MHz near the maximum voltage ... a third sweep was performed using steps of 0.01 MHz

Haas, Characterizing the Performance of the Houghton College Cyclotron — Houghton College thesis (2009) — p. 63-64

Tabletop: A simple, scope-only resonance-finding recipe the builder can use after any mechanical change to a next machine's dee or stem.

the quality factor of the Houghton College cyclotron was determined to be Q=16.1. The previous chamber and dee constructed in 2006 had a quality factor of 22.

Haas, Characterizing the Performance of the Houghton College Cyclotron — Houghton College thesis (2009) — p. 64

Tabletop: Quantifies the reference machine's architecture choice: an antenna-tuner match (like their current setup) gives kV-class dee voltage; multi-kV needs a high-Q tank coil instead.

when the CT2591 HV probe was placed near the dee, the overall capacitance changed slightly. This would, of course, change the value of the resonant frequency.

Haas, Characterizing the Performance of the Houghton College Cyclotron — Houghton College thesis (2009) — p. 65

Tabletop: When the builder cross-checks NanoVNA SWR sweeps against probe measurements, a few-percent frequency disagreement is expected instrumentation loading, not a fault.

the real voltage was roughly 11,300 times the pickup voltage ... the probe had to be recalibrated every time the frequency was adjusted.

Haas, Characterizing the Performance of the Houghton College Cyclotron — Houghton College thesis (2009) — p. 65-66

Tabletop: The pickup scale factor is frequency-dependent - the builder must recalibrate their pickup whenever they retune, not assume one constant.

a SWR of 1:1 and forward power of 15.43 W were measured

Haas, Characterizing the Performance of the Houghton College Cyclotron — Houghton College thesis (2009) — p. 69

Tabletop: Reassurance for commissioning a next machine: hunt for first beam at tens of watts with a clean match before scaling power - beam detection, not power, is the bottleneck.

higher potential on the dees results in fewer orbits and a shorter time of acceleration, allowing for a higher maximum kinetic energy... gives a maximum of 15 MeV for protons with 50 kV peak-to-peak

Loucks, Initial Results from the Houghton College Cyclotron — Houghton College thesis (2007) — p. 24-26

Tabletop: At the reference machine's sub-MeV energies relativistic slip is negligible (~0.1%), so dee voltage matters mainly through path length and gas scattering - but this rule sets the fixed-frequency ceiling for any future MeV-class ambition.

A standard AEI hairpin electron microscope filament floating at approximately -90 V is heated by 2 A of current ... RF pickup on each filament lead is shorted through a 0.001 uF capacitor to ground.

Yuly, The Houghton College Cyclotron: a Tool for Educating Undergraduates — Cyclotrons 2013, WE1PB01 (2013) — p. 3

Tabletop: An off-the-shelf, cheap, replaceable filament choice plus the RF-bypass detail that keeps the filament supply alive next to a live Dee.

the Dee may be oscillated with voltage amplitudes of up to approximately 3000V relative to the grounded Dummy Dee ... For normal operation, 10-40 W of RF power are required

Yuly, The Houghton College Cyclotron: a Tool for Educating Undergraduates — Cyclotrons 2013, WE1PB01 (2013) — p. 3-4

Tabletop: Tells the builder that Dee voltage is a tank-Q problem, not a brute-force power problem: a modest amplifier plus a good resonator beats a big amplifier into a lossy one.

It was believed desirable to operate at as low a frequency as possible because of the larger amount of engineering information available for oscillators in the region below 15 megacycles/sec.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 15

Tabletop: The reference machine's 9 MHz sits in this sweet spot; for a next machine, avoid pushing frequency (i.e., field) past where cheap RF parts and simple technique still work.

This curve indicates that 96 kW of rf power is required for exciting the dees to 400 kv. ... The oscillator input was 162 kw.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 16, 25

Tabletop: Formula transfers directly: at 9 MHz, ~50 pF and Q~1000, 5 kV on the dee costs only tens of watts, telling the builder exactly how much amplifier a higher-voltage dee on a next machine needs.

It is possible to calculate the various effects quantitatively and to predict the minimum dee voltage required to obtain a given energy in a particular cyclotron.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 17-19

Tabletop: Directly applicable design step for a next machine: compute threshold voltage for the target energy and field taper before freezing the RF chain power budget.

momentary beam meter readings exceeded two milliamperes but operation at this level was very unsteady due to sparking; further increases were not attempted

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 24

Tabletop: Directly applicable test discipline for a next machine's dee-voltage conditioning: rate the machine at the level it holds quietly for minutes, not the level it touches.

As measured, efficiency tends to increase with dee-to-dee potential and with beam power.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 24-26

Tabletop: Order-of-magnitude expectation transfers: most RF power goes to resonator and ion-loading losses, so judge a next machine's RF sizing on resonator dissipation, not beam power.

two 12 in. split silver-plated, water-cooled copper rings which can be clamped securely around the stems; the dee stems are also silver plated over the adjustment range

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 53

Tabletop: Scaled down: any sliding or bolted joint in the reference machine's dee-stem/coil path should be a broad, clean, plated, firmly clamped contact - RF joints, not wires, set small-resonator Q.

The ceramic 'Lapp' coils originally used for introducing cooling water to the tube and the plate line failed whenever the oscillator voltage was increased to give 200 kv. They have since been replaced with choke coils wound from copper tubing.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 59

Tabletop: The principle (a conductive-liquid line into an RF-hot electrode must itself be an insulator or a choke) applies whenever a next machine adds cooling or bias plumbing to the dee.

a fused disconnect switch in the output of each supply permits the operator to remove a faulty unit from service without disturbing the remainder

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 61

Tabletop: Transferable architecture: paralleled small supply modules (or PA pallets) with individual protection give a home machine graceful degradation.

Oscillator starting difficulties due to 'ion loading' are avoided by the use of insulated negatively-biased dees.

Livingston & Boch, The Oak Ridge 86-Inch Cyclotron — ORNL-1196, OSTI 4357145 (1952) — p. 7, 47

Tabletop: Directly applicable if a next machine's RF start-up stutters or the dee glows at low voltage: insulate the dee for DC and add a few-hundred-volt negative bias through an RF choke.

The selected value of 100 kv peak voltage requires about 1.5-in. clearance from dee-to-liner ... Since the magnetic gap is so precious ... this minimum value is taken for design.

Livingston & Howard (eds.), The Oak Ridge Relativistic Isochronous Cyclotron — ORNL-2648, OSTI 4275955 (1958) — p. 100

Tabletop: Scales directly: the reference machine's 1.3 kV needs well under a millimeter electrically, so their clearances are set by beam aperture and tolerance, but a 20-50 kV dee on a next machine should keep several millimeters to grounded surfaces.

the maximum voltage gain/dee is Vd = 2*V0 sin(theta/2); for particles rotating on subharmonics of the dee frequency the angular width of the dee is n*theta to the particle

Livingston & Howard (eds.), The Oak Ridge Relativistic Isochronous Cyclotron — ORNL-2648, OSTI 4275955 (1958) — p. 100

Tabletop: Directly applicable when the builder trims a next machine's dee for probe or source clearance: keep the dee close to 180 degrees or account for the sin(theta/2) energy-gain penalty.

The high dee voltage required in cyclotrons can be achieved for practical driving power only by using a high-Q resonant circuit.

Livingston & Howard (eds.), The Oak Ridge Relativistic Isochronous Cyclotron — ORNL-2648, OSTI 4275955 (1958) — p. 17

Tabletop: Directly applicable framing for the reference machine's matching network: every dB of resonator Q lost to bad joints or lossy insulators is paid in amplifier watts.

it is possible to bias the dees to prevent multipactoring, and a more complex booster oscillator circuit is required

Livingston & Howard (eds.), The Oak Ridge Relativistic Isochronous Cyclotron — ORNL-2648, OSTI 4275955 (1958) — p. 99

Tabletop: Directly applicable: multipactor lives exactly in the few-hundred-volt, MHz regime of a starting tabletop dee; plan the DC-bias insulation into a next machine's dee stem from day one.

No solder to hold the board to the spreader, the screws are enough. Heat sink compound between copper spreader and aluminum heat sink.

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 10

Tabletop: Standard practice for any kW-class dee driver a home builder assembles from LDMOS boards.

the part description says '15 nH, connecting wires to R14 and R15, 1.5 cm each #20 AWG,' implying that they are just wires, not even coiled.

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 15

Tabletop: Useful if the builder builds a broadband solid-state dee driver: the devices have huge low-frequency gain and will oscillate without this.

The real issue was the third harmonic, though; in general, it was only down 10 dB down and on some bands only 8 dB down.

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 19

Tabletop: A cyclotron dee tank is narrowband, but the same rule holds: measure what the PA actually emits before designing filtering or worrying about RF interference from a garage machine.

reflecting all that energy back into the output of the FETs risked driving the oscillations I had worried about in the detailed design of the power deck... I chose the diplexer

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 19-20

Tabletop: Relevant if the builder drives the dee with a broadband solid-state PA instead of a tube: protect the FETs from the highly reactive dee load.

a fundamental flaw in my design settings had been that all the crossover and cutoff frequencies were too low, causing too much insertion loss and high VSWR in the passband.

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 20

Tabletop: Schedule tuning time (this cost the author three months); the same applies to a homemade dee tank and matching network.

I also added a limiting power resistor (20 ohms at 50w) in series with 50v to the TIP102 as a precaution...with this resistor in place, a short on the 12v line will limit the current and prevent a catastrophic failure.

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 31-32

Tabletop: Cheap fault-tolerance rules for any homebuilt high-voltage/high-current RF deck; voltage-derating the caps matters more with the reactive load a dee presents.

the maximum output power I've measured ... is about 1.4 kW. After going through T/R switches, filters and couplers, you can expect about 1.3 kW at saturation

Buckler, Solid-State, 2-Decade, 1.25 kW Linear Amplifier — Development Notebook (2015) — p. 4

Tabletop: Size the RF chain for the dee power you actually need plus ~10-15%; the same relay/coupler/filter tax applies to a cyclotron dee drive.

a plating having about half the conductivity of the copper base will cause the greatest increase in overall resistance ... the conductivity of much of the commercial silver plating is about half of that of pure copper

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 1

Tabletop: Skip decorative silver plating on the dee and coil; a jobbing-shop bright-silver finish would raise, not lower, resonator loss at 9 MHz.

a layer of high conductivity copper plating at least two skin depths in thickness, at the operating frequency, then protecting this against corrosion by a very thin layer of low conductivity plating or a layer of low-loss lacquer

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 10

Tabletop: For dees, stems, and tank coils at 9 MHz: bare electrical-grade copper plus thin lacquer beats commercial silver or nickel plate.

The resistance of the composite conductor reaches a maximum value when the thickness of the plating is approximately one and one half times the skin depth for the plated metal.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 3

Tabletop: A sub-micron corrosion-protection flash on copper is harmless at 9 MHz; a mid-thickness medium-conductivity coating is the worst case to avoid.

A gold flash (10 micro-inches) is often used although many workers have shown that the deposits are not pore-free and that at least 200 micro-inches of gold are necessary to provide adequate protection.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 7

Tabletop: For RF contact fingers and connectors on the resonator, distrust thin gold flash; specify thick gold or use bare copper with lacquer instead.

the contact resistance of two silver-plated wires rose from 6 milliohms to 200 milliohms after two hours' exposure to hydrogen sulphide.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 7

Tabletop: Any silver-plated RF joints in the shop atmosphere (or near vacuum-pump exhaust) need protection or periodic cleaning, or kV-level circulating currents will heat them.

This last problem has been investigated fully by Benson who recommends chemical or electrolytic polishing to produce a smooth surface and lower losses.

Fowler, Radio Frequency Performance of Electroplated Finishes — Proc. IREE Australia (1970) — p. 8

Tabletop: Polishing dee edges and stems serves double duty at 5-13 kV: lower RF resistance and higher voltage-breakdown threshold.

safety monitoring of temperature, load failure, and diplexer HPF outputs, and ALC feedback for driver

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 1

Tabletop: A directional coupler plus fast drive-cut on high reflected power is the single best defense when the cyclotron dee arcs or drifts off resonance mid-run.

regulated and temperature compensated bias, separate VDD feeds to the output transistor drains to keep high dc currents out of the RF transformers

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 2

Tabletop: Directly applicable to a homebrew 9 MHz LDMOS deck: thermal-tracking bias prevents runaway, and DC-free transformers avoid core saturation at high drain current.

the design underwent several changes along the way, including the addition of degenerative feedback after a 'smoke in the cockpit' incident after about 350 contacts had been made.

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 2-3

Tabletop: A dee resonator is a narrowband, sometimes-detuned load; build feedback in from day one rather than after the first blown transistor.

Rather than mounting the output transistors directly to a heat sink, they are first flow soldered to a thick copper heat spreader, which is then mounted to the heat sink.

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 3

Tabletop: At 100-500 W a copper spreader under the LDMOS pallet is cheap insurance against the die-temperature excursions that killed the reference machine's earlier MOSFET amps.

But the real issue was the third harmonic, which was only 10 dB down generally and on some bands only 8 dB down!

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 3

Tabletop: At 9 MHz the 27 MHz third harmonic can excite spurious dee-resonator modes and detune the match; filter it between amp and matching network.

favored the diplexer design for solid state amps in the HF range, because reflecting all that energy back into the output of the field effect transistors (FETs) risked driving the oscillations

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 3

Tabletop: The dee is a high-Q load that reflects everything off-resonance; an absorptive diplexer gives the LDMOS a resistive termination at harmonics and protects against the mismatch failures the builder has already had.

The position of the connection point at the drain trace is critical as it affects the match.

Buckler, A Solid State 1.25 kW Linear Amplifier — QST, January 2015 (2015) — p. 4

Tabletop: When copying an LDMOS pallet layout, reproduce the output-transformer connection geometry exactly; millimeter changes shift the match at hundreds of watts.

the maximum energy gain corresponds to a system in which the RF size of the dee is close to 180 degrees or is a multiple of 180 degrees with a factor of 3, 5, 7

Smirnov, The Cyclotron and Its Modeling — Phys. Part. Nuclei 52 (2021) — p. 11

Tabletop: The reference machine's single ~180-degree dee on h=1 is already the optimum; the formula lets the builder tool compute turns-to-energy for any future dee angle or harmonic choice.

By selecting the value of the generator frequency, it is possible to achieve that the point of changing the direction of the phase motion is near -90 degrees.

Smirnov, The Cyclotron and Its Modeling — Phys. Part. Nuclei 52 (2021) — p. 22

Tabletop: Directly applicable: at 100 keV-1 MeV the reference machine is far from the ceiling, but deliberately detuning the oscillator slightly low buys extra phase headroom against an imperfect field profile.

in the presence of one accelerating dee and a voltage of 60-70 kV, protons can be accelerated in a decreasing magnetic field to an energy of 9-10 MeV

Smirnov, The Cyclotron and Its Modeling — Phys. Part. Nuclei 52 (2021) — p. 51

Tabletop: Sets the scaling for a next machine: roughly, final energy in a classical machine tracks dee voltage; the reference machine's few-kV dee at ~160 keV is consistent, and ~1 MeV needs proportionally more volts per turn or a flatter field.

the common boundary of the maximum voltage in the accelerating gaps in the central region of the accelerator is traced, which is 1.3-1.4 times higher than the Kilpatrick criterion

Smirnov, The Cyclotron and Its Modeling — Phys. Part. Nuclei 52 (2021) — p. 57-58

Tabletop: At the reference machine's few-MHz, few-kV/cm gap fields this gives huge margin, but it is the correct sizing formula if a next machine pushes dee voltage up to tens of kV across small central-region gaps.

the relative radial increase is only half the relative energy increase. However, for a given cyclotron, the turn separation dr0 will double when the dee voltage is doubled.

Kleeven & Zaremba, Cyclotrons: Magnetic Design and Beam Dynamics — CAS 2015, arXiv:1804.08961 (2018) — p. 44

Tabletop: The reference machine (155 keV, 2.6 keV/turn, r=9.65 cm) gets only ~0.8 mm/turn; a 10 kV dee at the same radius gives ~3-6 mm.

This results in a phase acceptance of only a few degrees for the typical case of a few hundred turns.

Baartman, Cyclotrons: Why/How Are Their Dynamics Different? — JINST 18 T03005 (2023) — p. 10

Tabletop: Do not chase single-turn extraction: accept multi-turn with dE ~ 2*e*Vdee (~20 keV at 10 kV dee), which PIXE tolerates. (Spread and dB/B: botman p.11-14.)

it is mostly the squared ratio of extraction radius and pole gap at extraction which determines the maximal number of turns or the minimal energy gain

Baumgarten, Cyclotron Beam Extraction by Acceleration — arXiv:2205.04124 (2022) — p. 5-6

Tabletop: A next machine with R~10 cm and half-gap 1.27 cm allows only ~9 turns (needs ~17 keV/turn); shrinking the half-gap to 6-7 mm at the edge allows ~30 turns - borderline reachable with a 5-10 kV LDMOS dee.

For internal sources, the anode is usually grounded and 30-100 kV of rf voltage is used for extraction with a feeler or puller extending from the dee.

Clark, Ion Sources for Cyclotrons — Cyclotrons '81, Caen (1981) — p. 3

Tabletop: The reference machine extracts with its few-kV dee — 10x less voltage than any literature machine. Compensate with a small source-puller gap (1.5-2.5 mm, cf. Siemens 2.3 mm, K100 2.9 mm) since extracted current scales ~V^1.5/d^2, and expect proportionally lower beam than published uA figures.

It is supported on insulating columns, making it possible to provide a DC bias.

Creutz, Design and Construction of Synchro-Cyclotron — NYO-780 (1950) — p. 15

Tabletop: Directly applicable - a DC-isolated dee mount costs little at design time and gives the multipactor-suppression knob the Berkeley reports show is essential.

Numerous holes are drilled in them to facilitate vacuum pumping.

Creutz, Design and Construction of Synchro-Cyclotron — NYO-780 (1950) — p. 19

Tabletop: Directly applicable - virtual leaks behind liners and skins are a classic small-chamber trap; drill the next machine's liners generously.

Parasitic modes at higher frequencies than desired for proton acceleration were successfully eliminated with light-bulb loads inductively coupled to the tube lines, and the lower mode ... was avoided by means of a series resonant circuit from dee to vacuum chamber.

Creutz, Design and Construction of Synchro-Cyclotron — NYO-780 (1950) — p. 30

Tabletop: Oscillator/transmission-line practice, transferable - lossy coupled loads and mode traps work identically on a fixed-frequency dee resonator driven by an LDMOS chain.

results now indicate that four 880's will suffice, while the data from the three-fourths scale model had indicated that six would be necessary.

Creutz, Design and Construction of Synchro-Cyclotron — NYO-780 (1950) — p. 29

Tabletop: Transferable caution - stray C, proportions and device parameters do not scale cleanly; validate a next machine's dee voltage vs drive on the real geometry, treating models as guides.

good contact is established by the use of strips of thin soft copper sheet backed by foam rubber pads.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 10

Tabletop: Directly applicable - the cheap 1948 equivalent of RF finger stock for every removable panel, dee-stem clamp, and line cover on a next machine.

To minimize rf power losses, all steel surfaces exposed to rf fields are copper-plated.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 10

Tabletop: Directly applicable - on a next machine keep steel (chamber walls, bolts, pole faces) out of RF current paths or plate/line it with copper; Q and dee voltage per watt are at stake.

the vacuum gap be sufficient to withstand 50 kilovolts rf at the accelerating gap. Consequently, a minimum of 3" spacing was employed in the vicinity of the open front end of the dee.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 10

Tabletop: Directly applicable scaling - at 13 kV the same conservative ~17 kV/inch rule wants ~3/4" dee-to-liner clearance; tighter gaps must lean on the 0.080"/50 kV bench data with derating.

Over 50 kilovolts rf could be developed across the insulator at 13 mc by a 5 kilowatt oscillator. Under the most severe test conditions, air blast cooling of the insulators was found necessary.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 13

Tabletop: Directly applicable - a bench quarter-wave resonator lets the builder soak-test a next machine's dee-stem insulators at full 5-13 kV RF with only tens of watts of drive.

a .080" gap between copper or copper-plated surfaces having a reasonable polish would hold a maximum of 50 kilovolts at 13 mc at a pressure of about 5 x 10-6 mm.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 13

Tabletop: Directly applicable breakdown data for setting a next machine's dee-to-liner and puller gaps at 5-13 kV - and a warning that discharge-roughened surfaces lose ~40% of bench hold-off.

The water circuit is completed to ground potential by means of sets of flexible polyethylene tubing, each several feet long. Treated water of low conductivity is used.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 13

Tabletop: Directly applicable if the next machine's dee or stem is water-cooled while DC-biased - length of insulating hose plus DI water sets the leakage current.

For reasons of convenience, a quarter scale was chosen. The resonant frequency is then increased fourfold and all inductances and capacitances are reduced by a factor of four.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 14

Tabletop: Transferable method - a next machine's resonator can be prototyped at reduced scale with a VNA, remembering effective dee capacitance is not the static capacitance (500 vs 1000 pF on the 184-inch).

The evident discrepancy in transmission line length was eventually traced to a cumulative error of about two inches in various small errors in model dimensions.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 22

Tabletop: Directly applicable - give the next machine's resonant line/tank a deliberate tuning range (trombone section, tuning vane, trimmer C) instead of trusting calculated dimensions.

disclosed a re-entrant cavity resonator mode between the two pole pieces of the magnet with the vacuum tank walls as the return circuit resonant near the lower frequency limit. This was easily suppressed by strapping the pole pieces together.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 22

Tabletop: Directly applicable - the pole-chamber geometry of an 8-inch machine forms the same parasitic cavity; copper straps pole-to-pole (or liner-to-liner) are a one-hour fix worth doing preemptively.

All discharges were eliminated by cutting down the available volume by means of perforated shields around the sides of the dee, by adding a grounded dummy dee and by applying a negative bias to the dee.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 23

Tabletop: Directly applicable - the complete 1948 recipe for the RF discharges that plague small chambers at kV-level dee voltages; perforated shields keep pumping speed.

There is sufficient magnetic field at the rotary condenser to allow a Philips gauge discharge when positive bias is applied; a negative bias is therefore imperative.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 23

Tabletop: Directly applicable - the entire next machine's dee sits in 0.59 T, so if bias is used to kill discharges, start negative; positive bias risks a permanent Penning discharge.

The care taken in insuring that only pure tension and compression forces would be applied was thus well justified.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 24

Tabletop: Directly applicable to a next machine's dee-stem standoffs and feedthroughs - arrange the support geometry (threaded rods, spherical seats) so ceramics never see bending or shear.

this shielding was found sufficiently effective, the field being cut from 140 Gauss to less than 20 Gauss.

MacKenzie et al., Design of the Radio-Frequency System for the 184-inch Cyclotron — UCRL-64 (1948) — p. 20

Tabletop: Transferable - LDMOS amps, fans, and ferrite-cored parts near an 0.59 T magnet want a steel housing; remember the housing itself feels a large attractive force.

It would produce 15 kv peak volts on the dee at 10 me and 9 kv at 20 me with 6 kw input. It averages around 70%.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 4

Tabletop: Directly applicable benchmark near the reference machine's 9 MHz - scaled by its much smaller dee capacitance and Q, it frames how many LDMOS kW the 5-13 kV goal really needs.

The upper and lower frequency limits can be varied together about 6% by a tuning vane which varies the impedance of the transmission line. It is a movable copper sheet with flexible end connections.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 5

Tabletop: Directly applicable to a fixed-frequency machine - a vane gives the few-percent trim needed to land the dee resonance exactly on the magnet's cyclotron frequency.

The horizontal joints are also rubber backed, but no particular care is necessary to insure contact as the current flow is parallel to the joint.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 6

Tabletop: Directly applicable enclosure craft - plan a next machine's panel seams along the RF current direction and spend the contact-strip effort only on the seams that cross it.

It was tried this way at 20 megacycles, but it soon became necessary to copper plate most of the surfaces.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 6

Tabletop: Directly applicable at 9 MHz - bare steel may survive, but plating (or copper construction) is cheap insurance for Q and hot spots.

Polystyrene of this thickness used in this manner will stand over 15,000 volts DC and approximately 1500 volts r.f. provided the metal parts remain cool.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 8

Tabletop: Transferable construction recipe for homemade HV bypass/blocking capacitors in a next machine's RF chain (modern Kapton/PTFE substitutes upgrade the polystyrene).

Such oscillations will usually not occur if the time constant of the resonant system is more than ten times the time constant of the grid leak grid condenser network.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 8

Tabletop: Transferable oscillator rule - the same criterion governs bias-network time constants in any self-excited driver on a next machine's dee resonator (tube or transistor).

The loops are therefore made of large diameter tubing and the length of tubing which is not serving as mutual inductance in the dee stem circuit is kept at a minimum.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 9

Tabletop: Transferable to any self-excited or feedback-coupled driver in a next machine - minimum-plate-current (minimum-DC-input) trimming is a meterable, practical phasing procedure.

It has been predicted theoretically and demonstrated experimentally that most of the ions can be lost by such a variation if the ions reach their final radius at 19 megacycles.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 10

Tabletop: Directly applicable even at fixed frequency - sweep the next machine's system for modes near 2x and 3x of 9 MHz and detune any found; harmonic coincidences sap dee voltage and beam.

it has been used as an emission limiting device to protect the oscillator tubes when discharges occur in the tank and condenser.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 12

Tabletop: Directly applicable principle - fast current limiting/foldback in the LDMOS drain supply (plus VSWR trip) is the modern form of this arc protection.

the discharge is usually intense enough, even at 10-5 mm of Hg to prevent the voltage from building up to this extinction value. Such a discharge can be eliminated by a sweeping field obtained in any manner.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 12

Tabletop: The single most relevant discharge fact for the reference machine: its 1.3 kV dee lives just above this regime, and the 5-13 kV upgrade must punch through it - plan a DC bias supply on the dee from day one.

At 20 megacycles the space between electrodes which will allow an electron to reach an energy around 30 volts in 5 cm. There are very few paths, along the magnetic field, in the neighborhood of the dee that are greater than this.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 12

Tabletop: Directly applicable geometry rule - at 9 MHz electron oscillation amplitudes are larger still, so keep open RF-exposed volumes and along-field gaps in the next machine's chamber small or shielded.

Since the tickler oscillator does not derive its excitation from the load, it can drive the main oscillator over the critical voltage.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 12

Tabletop: Transferable start-up trick - with an LDMOS chain this becomes a driven start (external exciter) that rides through the multipactor band before handing over to normal operation.

For reasons which are not clearly understood this bias usually increases the size of the beam by a factor of two or more.

MacKenzie & Waithman, R.F. System for Frequency Modulated Cyclotron — MDDC-1045 (1946) — p. 12

Tabletop: Conditionally applicable - a bipolar, current-limited bias supply on the next machine's dee lets the builder try both polarities safely and keep whichever helps beam.

The measurements were made with G-R Impedance Bridge, Type 650-A Serial #1977, and are +/- 2 uuf. Lead capacity of 29 uuf has already been deducted.

Anderson, 184″ Cyclotron: Oscillator Capacitance Measurements — MDDC-964 (1947) — p. 3

Tabletop: A modern LCR meter with lead-nulling does the same job on the reference machine's dee stem; knowing C-to-ground before pump-down predicts the ~9 MHz resonance and flags assembly errors.

Dee-to-dee, kv 10 - 18 ... Internal Beam, Stable, ua 100

Howard, Cyclotrons and High-Energy Accelerators, 1958 — ORNL-2644 (1958) — p. 61

Tabletop: Proof that low dee voltage works at small radius: the 16-inch ISSP machine is the existence proof for a next machine's sub-MeV goal with a ~10 kV-class dee, provided the field profile keeps the many extra turns focused.

Oscillator type self-ex. Oscillator tube 5771 ... Osc. input, max 35 kw. Osc. output, max 20 kw.

Howard, Cyclotrons and High-Energy Accelerators, 1958 — ORNL-2644 (1958) — p. 106

Tabletop: These kilowatts bought 20-90 kV dees at high Q, not beam power; a few-kV tabletop dee needs only ~100 W-1 kW, and the census shows simple self-excited oscillators (no synthesizer, no feedback loop) ran every one of these machines.

Beam defining slits used on 1, 2, and 3rd revolutions to define center of beam rotation; 3rd turn slit is 1/2 mm wide. 100% extraction efficiency with low beams, requires better than 1/2 % stabilization of dee volts.

Howard, Cyclotrons and High-Energy Accelerators, 1958 — ORNL-2644 (1958) — p. 27

Tabletop: The cheapest extraction upgrade known: mechanical slits in the center region plus tight dee-amplitude regulation; for a next machine's turn-separation budget, orbit-center definition on turns 1-3 matters more than deflector finesse.

Magnetic field variable, by only changing the coil current, from 14 to 18 kg with 1 to 2.5% field drop-off at the exit (r = 16 cm).

Howard, Cyclotrons and High-Energy Accelerators, 1958 — ORNL-2644 (1958) — p. 61

Tabletop: The builder can trim B to match a fixed RF (or vice versa) and expect the shim profile to survive, as long as the iron is not driven into locally different saturation - measure n(r) at both ends of the intended current range.

As a result of these circuits improper tuning cannot damage the power tubes.

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 5-6

Tabletop: Directly transferable to the LDMOS upgrade in solid-state form - an ALC loop whose setpoint is overridden by drain-current/SWR limiters protects a 100-500 W pallet from mistuned-dee experiments exactly as it protected 50 kW tubes from inexperienced operators.

to let the reference voltage vary with the rf plate voltage so that plate dissipation would be limited to a constant value

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 6

Tabletop: The exact analogue for LDMOS is limiting device dissipation (drain current x voltage headroom) rather than a fixed current clamp, which either under-protects off-resonance or throttles available power on-resonance.

The dee voltage is depressed about 1% for each 10 uA of injected current, and because a balance is maintained at the input of the amplifier, the recovery time is limited only by the bandwidth

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 6

Tabletop: A tabletop ALC loop gets beam pulsing for activation or timing experiments for free - inject an offset into the amplitude setpoint; no mechanical or source-side hardware needed. The 1%/10 uA constant is specific to their circuit, not a scaling law.

below a dee voltage of 15 kV, the flip-flop will remain in one state and only if the dee circuit happens to be tuned to the low frequency side of resonance will the servo run in the proper direction

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 7

Tabletop: Any auto-tune loop on a next machine (phase comparison of PA drive vs dee pickup) needs the same amplitude gate plus a manual jog mode to walk the tuner into range before handing over - the failure mode is detector-technology-independent.

adjusting Cn for coincidence of maximum dee voltage and minimum plate current as the dee was tuned through resonance

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 4

Tabletop: Neutralization per se is a triode/tetrode issue, but the acceptance test transfers - on any amplifier-dee chain, dee-voltage peak and PA input-current dip should line up when sweeping through resonance; a skew flags feedback or coupling problems.

a Phillips-Ion-Gauge-type discharge can start in the magnetic field inside the vacuum tank near the positive transmission line

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 5

Tabletop: Very relevant at higher dee voltage on the reference machine or a next machine - crossed E and B in vacuum is exactly a PIG geometry (it is how the ion gauge and PIG source work), so any DC-carrying feedline, bias lead, or probe wire inside the field must be RF/DC separated; they melted an electron trap and a ceramic window learning this.

no discoloration or other indication of heating of the contacts, despite routine operation to 110 A/in and occasional operation to higher current densities

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 8

Tabletop: A tabletop coaxial resonator tuning short carries far less current, so 110 A/in is a generous ceiling - but the recipe (thick fingers, positive clamping, plated surfaces, cooling on both sides of the joint) is the proven pattern for any movable-short tuner.

vacuum capacitors have been used in various ways in its plate circuit. None has been found to stand up satisfactorily.

Osterlund & Smythe, A Cyclotron Power-Amplifier RF System Using a 4CW50,000C/8350 Tetrode — COO-535-543 (1963) — p. 8

Tabletop: Dated in absolute terms - modern vacuum caps are far better and fine at a next machine's kW-class levels - but the underlying point stands; put lumped variable capacitors where RF current is low and do coarse tuning with distributed elements.

It minimizes electric defocusing, which is ordinarily a major cause of beam loss, and actually provides some focusing during the usually defocusing part of the phase excursion.

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 10-11

Tabletop: The reference machine's beam losses in the first turns are exactly this mechanism at nA scale; a flat-topped dee is likely too much RF plumbing for a next machine, but the rule explains why phase excursion and gap-crossing timing dominate small-machine transmission.

the same bunching occurs even if a third harmonic is added to the r-f wave form to square the wave

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 10

Tabletop: Reassurance that waveform shaping and center-region bunching are separable problems; also a reminder that the bunching mechanism itself (Cohen) is what sets which ions survive the reference machine's center region. OCR note - theta prints as (c) in these appendix equations.

If this is done at the rate of 10 megacycles per second, the power requirement would be 20 megawatts!

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 13

Tabletop: The cleanest back-of-envelope argument in this collection for why dee voltage is bought with Q, not amplifier watts - scale it to a next machine (7-9.5 kV on tens of pF at 6.78 MHz) to show why a few hundred LDMOS watts suffice only through a good resonator.

The extra current element can, however, be a second transmission line

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 18

Tabletop: A lumped-plus-stub version is buildable at tabletop scale and the design tables (PDF 33-60) are precomputed; even unused, the method shows how to place a resonator's higher modes deliberately instead of discovering them by accident.

Tune the length of one of the lines for a null on the admittance meter.

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 21

Tabletop: The written five-step procedure is a model for documenting any coupled- adjustment RF tune-up (a next machine's coupling loop + trimmer interact the same way); interpolating from precomputed tables to know which way to tune is the transferable trick.

if the coupling line is one-quarter of the fundamental wave length it is three-quarters of the third harmonic wave length, and the impedances are simply inverted by the line at both frequencies

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 22

Tabletop: Handy odd-harmonic identity for any coax-fed system - it also warns that a quarter-wave feeder presents transformed impedances to your amplifier's harmonics, which matters for LDMOS stability even in a plain sine-wave system.

It was somewhat difficult to get both the phase and the amplitude of the third harmonic adjusted correctly; however no serious attempt was made to develop a good self-excited driving system.

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 25

Tabletop: Mirrors the next machine's decision already leaning MOPA - independent control of each degree of freedom beats a self-excited loop whenever more than amplitude must be set, and a DDS + LDMOS chain is the modern MOPA.

A visual means of monitoring the wave form and a remote tuning or an automatic servomechanism for tuning should be provided.

Goodman, A Square-Wave Cyclotron Oscillator — ORNL-2403 (1958) — p. 25

Tabletop: Even a plain sine system drifts (thermal detuning is logged on the reference machine); a calibrated capacitive pickup on a scope is the minimum instrument, and it is a prerequisite for any auto-tune servo on a next machine.

The loop gain should be such that one degree of phase error will apply full power to the servo motor.

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 7

Tabletop: Stated as easily-achieved perfect-performance values, not minima - a stepper-driven trimmer on a next machine's resonator can copy all three numbers directly; 2% range comfortably covers thermal drift on a machine holding ~5 G field tolerance.

The phase detector is quite sensitive and produces an output signal of 0.76 volt for a phase error of 1

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 13

Tabletop: A modern phase detector IC gives ~10-20 mV/deg, so their 0.76 V/deg shows how much detector gain a robust motor loop wants - budget amplification accordingly, and characterize K_d so loop gain is a number, not a knob.

the amplitude of the intermediate frequency is exactly equal to the magnitude of the local oscillator signal and entirely independent of the magnitude of the phase signals

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 12-13

Tabletop: The same requirement is met today by limiting amplifiers or digital phase detection - the principle (servo dynamics must not change between 10% and 100% dee voltage, and the source arc is an in-band noise generator) applies verbatim to a next machine's tuning loop.

The most serious objection to the dee-to-dee capacitance is the coupling between servo systems which it provides. The problem of servo stability becomes insuperable.

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 15

Tabletop: Confirmed independently on the machine side in ucrl-3187 p.5,11 (dees could not be servoed individually until neutralized). A single-dee next machine dodges this entirely - which is itself the design lesson - but it governs any future two-dee or dee+dummy-dee variant with separate tuners.

It was necessary to do this while the machine was down to air, in order to avoid multipactoring.

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 18

Tabletop: Two transferable habits - quantify RF isolation as a measured coefficient with a pass number, and remember that low-level RF tests in vacuum sit exactly in multipactor territory (the reference machine has seen multipactor-like loading); atmospheric-pressure RF checks avoid it.

if the line is excited from a constant current source the magnitude of the voltage appearing across the input of the line is constant but the phase of the voltage depends upon the load resistance

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 11

Tabletop: A DDS sets phase digitally today, but the lambda/8 trick remains a zero-active-parts phase adjuster for RF plumbing (e.g. trimming a pickup or reference arm) and a nice classroom demonstration of transmission-line properties for the educational-machine line.

Once the three dees are isolated electrically by adjusting the neutralizing loops the machine behaves like three separate single-phase systems.

Smith, A Three-Phase Radiofrequency System for Cloverleaf Cyclotrons — UCRL-3153 (1955) — p. 18

Tabletop: The architectural moral - decouple first, then control each loop as SISO - applies to any interacting set of tabletop loops (tuner vs coupling vs amplitude on a next machine); trying to servo a coupled system is how the programme burned months (ucrl-3187 p.5-6,11).

The beam load would cause a two- to threefold increase in the amplifier plate currents compared to the source-off condition.

Heusinkveld et al., Studies with a Three-Dee Three-Phase Proton Cyclotron — UCRL-3187 (1955) — p. 14

Tabletop: The linear beam-vs-dee-voltage check transfers to nA scale and is a good run-log plot for the reference machine; the 2-3x loading signature does NOT - milliampere beams absorb real RF power, whereas a nA beam is invisible in amplifier current, so use it only as an upper-bound sanity argument.

protected by an interlocked air-cooling system, spark gaps at both ends of the half-wave transmission lines leading to the dee stems, and by an rf-dc fault circuit

Heusinkveld et al., Studies with a Three-Dee Three-Phase Proton Cyclotron — UCRL-3187 (1955) — p. 11

Tabletop: The rf-dc comparison is the tube-era ancestor of modern SWR/output-detect foldback and ports directly to the LDMOS upgrade (DC applied but no RF developing = arc or detune, kill drive); spark gaps at the feedthrough remain cheap insurance at 5-13 kV dee voltage.

during this time the servos received spurious signals and tended to creep away

Heusinkveld et al., Studies with a Three-Dee Three-Phase Proton Cyclotron — UCRL-3187 (1955) — p. 12

Tabletop: A control-system rule that transfers verbatim to any next machine's auto-tune or ALC firmware - hold integrators and actuator positions during spark recovery/restart, and make both the recycle delay and the servo-hold adjustable; same lesson as coo-535-543's amplitude gate, learned independently.

thermal effects detuned the machine sufficiently so that ion lock prevented the rf from being restored

Heusinkveld et al., Studies with a Three-Dee Three-Phase Proton Cyclotron — UCRL-3187 (1955) — p. 12

Tabletop: The reference machine already shows warm-up drift; the transferable practice is a permanent low-level sweep capability (VNA or dip meter on a pickup loop) so resonance can be found cold without RF power, plus logging tune position vs temperature - cheap now, standard then.

peak efficiency did not correspond to 120 phase difference between the dees

Heusinkveld et al., Studies with a Three-Dee Three-Phase Proton Cyclotron — UCRL-3187 (1955) — p. 11

Tabletop: Same programme's RF paper (ucrl-3153 p.7) draws the identical moral - measure the quantity you care about at the electrode, not a correlate at the amplifier. For a next machine - derive tuning/phase feedback from the dee pickup, not from LDMOS drain current or forward power, which optimize at subtly wrong points.

the diode charges a capacitive load to the peak value, and the stored charge does not subtract (appreciably, at any rate) from the output voltage.

Smith, Deflector Power Supply for Sector-Focused Cyclotrons — UCRL-10655 (1963) — p. 10

Tabletop: Historical for parts choice (modern fast-recovery diodes moot the issue) but the design insight stands - CW stacks feeding capacitive deflector loads tolerate slow diodes; spend money on voltage rating and grading instead.

These tubes cannot be operated at plate voltages above 11,000 volts, even though rated at 16,000 volts, because, in operation, the plate voltage reversed in 0.3 us causing arcing between plate and grid.

Kerns et al., High Voltage Pulser for 184-inch Cyclotron Electric Deflector — UCRL-95 (1948) — p. 13

Tabletop: The derating discipline (waveform-specific stress, not catalog rating) transfers to every switching element an amateur uses - MOSFET/IGBT avalanche and dV/dt limits in a Marx or inverter play the exact role the 5C22 reversal limit played here.

These lines provide a 1000 volt trigger of 20 ohms impedance for approximately 0.20 microseconds. Applying this trigger to the grids results in positive ionization of all the tubes in 0.1 +/- 0.01 us.

Kerns et al., High Voltage Pulser for 184-inch Cyclotron Electric Deflector — UCRL-95 (1948) — p. 13

Tabletop: Classic paralleling technique still used in pulsed-power; relevant only if a next machine ever adds a pulsed element (fast chopper, kicker for time-of-flight work), but then it is the reference method.

Approximately ninety percent of the total power input to the system is eventually dissipated in the transformer cores as heat. At rated operating levels this loss is approximately 500 watts.

Kerns et al., High Voltage Pulser for 184-inch Cyclotron Electric Deflector — UCRL-95 (1948) — p. 12

Tabletop: Not needed for a CW deflector, but the de-aerated-oil corona cure, the 3x overvoltage proof test, and the interlaminar-voltage failure mode are transferable HV-construction craft for any oil-insulated amateur component.

a plate voltage of 11,000 volts on the thyratrons can be maintained with a power supply voltage of 2,750 volts, a step up ratio of four to one.

Kerns et al., High Voltage Pulser for 184-inch Cyclotron Electric Deflector — UCRL-95 (1948) — p. 15

Tabletop: Pulsed-modulator craft, not CW-deflector material; file under 'if a next machine ever needs a kicker' - it means the HV DC supply can be a quarter of the switch voltage.

It is an advantage to choose the smallest value of l, since the choice allows the least sensitivity in matching the perturbing frequency to the particle motion.

Stubbins, Radiofrequency System for Extracting Particles from a Cyclotron — UCRL-8578 (1958) — p. 7

Tabletop: A genuinely tabletop-compatible extraction assist - an electrode pair driven by a small independent oscillator. For a next machine (nu_r ~ 1) the required frequency is near the orbital frequency's sidebands; worth a tracker experiment before committing hardware.

This rf electrical-field gradient, less than 5 kv/cm, is believed to be easily obtainable by an oscillator which is independent of the cyclotron oscillator and which can be tuned to the optimum frequency and amplitude.

Stubbins, Radiofrequency System for Extracting Particles from a Cyclotron — UCRL-8578 (1958) — p. 8

Tabletop: kV/cm-class rf on a small electrode is trivially available at a next machine's scale (hundreds of volts across a few mm). The scheme was never demonstrated on hardware in this report - treat as a promising computed option, not proven practice.

In the weak vertical field component of the accelerating voltage in the 184-inch cyclotron the beam loss was impressively fast.

Stubbins, Radiofrequency System for Extracting Particles from a Cyclotron — UCRL-8578 (1958) — p. 5

Tabletop: A real design caution at any scale - dee misalignment or asymmetric liners give exactly the z-proportional E_z this resonance needs. Keep a next machine's dee/dummy-dee vertically symmetric and check whether nu_z resonates with any strong rf harmonic at operating field.

A dee voltage of 75 kv dee-to-dee was reached after some routine difficulties were overcome. The cyclotron now maintains this voltage for long periods of time without showing any tendencies to failure.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 31 March 1952 — ORNL-1339 (1952) — p. 15

Tabletop: Directly transferable acceptance test for the reference machine's LDMOS upgrade — run the dees 25% above the planned operating voltage for hours before calling the RF done; a margin that survives only seconds is not margin.

the vacuum tube rectifiers are unchanged by neutron bombardment and, unless the probe-to-dee distance is changed, the calibration remains constant.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 31 March 1952 — ORNL-1339 (1952) — p. 10

Tabletop: Two transferable halves: (1) semiconductor sensors near the chamber are a calibration-drift risk once neutrons appear; (2) a capacitive dee-voltage pickup is calibrated GEOMETRY — mechanically fix the probe-to-dee distance or every calibration is void. Bears directly on retiring the uncalibrated ~800 V nominal dee-voltage number on the reference machine.

the ion source arc slit was shortened from 2 1/2" to 1/2". Thereafter the ratio of accelerated beam power to ion loading power was increased, as predicted.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 31 March 1952 — ORNL-1339 (1952) — p. 16

Tabletop: On a tabletop machine where every watt of RF matters, an oversized source aperture wastes drive as ion loading; try a shorter emission slit on the reference machine's source and watch accepted beam per unit dee loading, not raw source output.

the net ion-loading efficiency was 40%, that is, 40% of the power expended in acceleration of ions was to the target. There was a two-fold increase in electrical efficiency as the beam was increased.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 30 June 1952 — ORNL-1345 (1952) — p. 7

Tabletop: On the reference machine at nA the beam power is invisible next to fixed RF losses — the transferable lesson is the metric, not the number: log P_beam/P_RF per run; on any small machine almost all RF power is overhead, so chase Q and coupling, not amplifier watts, for efficiency.

A negative voltage bias, 1 to 2 kv, is applied to the dees in order to sweep out any ions that may be formed while oscillation is being established.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 30 June 1952 — ORNL-1345 (1952) — p. 9

Tabletop: Transferable at proportional scale (tens of volts on the reference machine's ~1 kV dees, a few hundred on the LDMOS upgrade) — multipactor/ion loading during RF ramp-up is a classic small-machine failure mode and a bias supply is the classical cure.

Carbon lips of a new design were installed on the edges of the dees to reduce sparking at the increased dee-to-dee voltage, 400-500 kv, required for operation at the high energy level.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 30 June 1952 — ORNL-1345 (1952) — p. 7

Tabletop: The 400-500 kV is MW-era and does not transfer; the material practice does — if the reference machine's 5-13 kV upgrade sparks at the dee gap, graphite edge pieces are the period-proven remedy and are trivially machinable.

This additional pumping speed then can be considered as being in parallel with that through the opening at the mouth of the dee.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 129

Tabletop: Dees are pumping dead-ends by construction - drilling the dee back and any stem shrouds (small holes, below RF-significant size) is free conductance exactly where the ion source dumps its gas.

The frequency-sensitive circuit "gates" the phase-sensitive circuit, and the phase-sensitive circuit triggers the deflector.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 168

Tabletop: The architecture for any timed kick against an extraction gap, on a next machine or a small synchrotron - a coarse condition (frequency, turn count, integrated field) opens a window, and the RF itself supplies the fine phase. Two easy measurements replace one impossible one.

Block 4 is a local oscillator with a frequency about 1 megacycle below the desired deflection frequency P. When the dee-oscillator signal sweeps through point P, the 1-megacycle filter passes an a-c transient.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 168-169

Tabletop: Classic measurement doctrine - move precision to a fixed low frequency where stability is cheap. The same trick (mix dee RF down, detect at fixed IF) is how a modern amateur frequency/turn marker gets crystal accuracy with junk-box filters.

The crystal oscillators, Nos. 3 and 5, will maintain a constant frequency within one part in 10,000.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 169-170

Tabletop: The stabilization pattern transfers even where the parts are now silicon - put the reference AND the analog discrimination components in one controlled thermal box; the filter drifting is as fatal as the oscillator drifting.

Unfortunately it is theoretically impossible to filter a transient without introducing time delay. The time delay thus introduced was kept to a minimum.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 170-171

Tabletop: General fast-timing wisdom for beam-pulse and kick timing chains - every smoothing stage costs latency; calibrate the chain end-to-end and remove the constant part in the threshold or delay setting rather than chasing zero-delay filters.

the time at which the transient is at a peak is somewhat later than the time at which the dee-oscillator frequency is in the center of the filter pass band.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 171-172

Tabletop: Matters wherever a resonant pickup watches a changing frequency - including a synchrotron RF ramp or an FM-tuned marker on a cyclotron; if the ramp rate changes, the timing calibration silently moves.

That alternative was abandoned in view of the susceptibility of high-frequency switching circuits to unforeseen difficulties.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 172

Tabletop: A complexity-avoidance rule with 2026 force - blanking a known-bad time window (one line of firmware now) beats analog cleverness whenever the artifact's timing is predictable.

Probable error in firing time, <1 microsecond. Range of firing frequency, 19 to 21-1/2 megacycles, with provision for changing this.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 174

Tabletop: Calibrates ambition - microsecond-class event timing off a tiny RF sample needed no exotic parts in 1952; any modern comparator + MCU implementation should beat it by orders of magnitude, so the architecture (not the hardware) is the thing to copy.

The fairly low efficiency, 27 per cent, indicates that it would be desirable to go to six type-880 tubes in the final model, especially since power-supply capacity is available for the additional tubes.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 162

Tabletop: The transferable method rule - prototype the next machine's dee/stem/liner as a cheap scale model (or full-scale mockup, given the small size) and measure resonance, Q and parasitics before final fabrication; NYO-780 p.29ff records the same practice. Cite both.

equipped with a small coupling loop ... used to load the unwanted mode, which on the three-fourths scale model was about 50 megacycles, in which the oscillator stub forms a capacity-loaded half-wavelength line.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 162

Tabletop: Both suppression tools are amateur-accessible - a strap that shorts the parasitic mode's voltage pattern without disturbing the wanted mode, and a loop selectively coupling the parasite into a lossy load; relevant the moment the LDMOS upgrade raises the reference machine's gap voltages.

The operation of the oscillator determines the variation of capacity with time which will keep the ion phase constant. This variation is not very critical.

Foss et al., Cyclotron Component Design Technical Reports — TID-454 (1952) — p. 152

Tabletop: Synchro-only for cyclotrons, but the pattern maps onto a small synchrotron's RF ramp - program tolerance is loose if phase stability (not exactness) is the criterion, and cycle dead time is a direct beam-current tax.

From (5) the measurement of dr is essentially a determination of the phase.

Cohen, Spatial Distribution of Current on an Internal Cyclotron Target — ORNL-1348 (1952) — p. 9

Tabletop: Table verified on page image. Energy-independent physics: on the reference machine or a next machine a differential probe (shadowed double tip) or the sectioned-target map gives dr, and with the known dee voltage that is a measurement of ion RF phase — the quantity a next machine's 5-G field tolerance is protecting. A rare direct experimental handle on phase for machines with no beam-position monitors.

The cost of the magnet would be increased unreasonably if a gap suitably large for conventional dees were provided. For this reason, a system for acceleration with vertically oriented resonant cavities was adopted.

Oak Ridge National Laboratory, A Proposal for the Mc² Isochronous Cyclotron — ORNL-3540 (1963) — p. 145

Tabletop: The specific topology (vertical TEM cavities) is 810-MeV-only; the transfer is the coupling: every inch of dee clearance is bought with magnet cost, so dee-gap and magnet-gap must be traded as one system, as in a next machine's gap decision.

chosen for the final design since this is the ordinary coaxial cavity and is more amenable to design ... was chosen as a compromise between voltage-holding ability and transit-time effects.

Oak Ridge National Laboratory, A Proposal for the Mc² Isochronous Cyclotron — ORNL-3540 (1963) — p. 174-176

Tabletop: Fully scale-free method: a dee-stem system is likewise a transmission line with machine-fixed dimensions and a few free ones; name each spacing's compromise pair when sizing the next machine's 5-13 kV dee.

Checking of the model dimensions revealed a construction error of several percent in the spacing near the median plane ... a resonant frequency of 151.3 Mc/s was calculated for the model; this is within 4% of the measured value.

Oak Ridge National Laboratory, A Proposal for the Mc² Isochronous Cyclotron — ORNL-3540 (1963) — p. 177

Tabletop: Scale-free: their 1/12-scale model both certified the three-step-line calculation (4% agreement) and caught a build error; a bench mock-up of a next machine's dee/stem before the LDMOS amp arrives serves the same double duty.

Computed power loss in cavity 500 kW; Beam power 160 kW; Contingency 200 kW; total 860 kW

Oak Ridge National Laboratory, A Proposal for the Mc² Isochronous Cyclotron — ORNL-3540 (1963) — p. 180

Tabletop: The kilowatts are 810-MeV numbers; the budget structure — separate lines for copper loss, beam load, and contingency, plus >=4x drive reserve — sizes a next machine's 100-500 W LDMOS chain honestly.

Thus, when a spark occurs in the cavity a large resistive load is reflected to the plates of the power amplifier and the tube plate current would decrease.

Oak Ridge National Laboratory, A Proposal for the Mc² Isochronous Cyclotron — ORNL-3540 (1963) — p. 182-183

Tabletop: Scale-free fault-mode-first design: dees spark at every scale, so choose a next machine's amp coupling and protection for arc behavior, not just matched-condition efficiency — directly relevant to protecting an LDMOS pallet.

Since the dee voltage in the experimental unit was limited to 10 kv, application of injection potentials of over 10 kv resulted in deceleration of ions between the accelerating electrode and the dee.

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 31 December 1951 — ORNL-1269 (1952) — p. 47

Tabletop: Any reference-machine or next-machine source-bias or puller experiment must respect the same ordering - source extraction potential is bounded by the rf accelerating potential actually available at the first gap, or the first gap runs backward.

The dee voltage will undoubtedly be limited, though, by the spacing between the dee stems outside the vacuum tank, which is only 1.5" at one point

Howard (ed.), Electromagnetic Research Division Quarterly, period ending 30 September 1952 — ORNL-1383 (1953) — p. 23

Tabletop: For the reference machine's LDMOS upgrade toward 5-13 kV dees, walk the whole rf path in air - feedthroughs, stem gaps, coupling hardware - because atmospheric-side spark gaps, not vacuum gaps, set the first ceiling.

Since this circuit depends upon the electrical characteristics of the resonant dee system, it cannot be designed until these characteristics are determined.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 18

Tabletop: The template for the reference machine's LDMOS upgrade - freeze the amplifier and dee-resonator designs, but specify the matching/coupling network only after measuring the real dee system's f0, Q, and shunt impedance on the bench. Ordering the coupling parts first is the classic mistake this rule prevents.

The whole dee system is insulated from ground so that a bias potential may be applied to control ion loading.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 18

Tabletop: The reference machine already uses dee bias; the design rule for a next machine is to make bias a first-class requirement - insulate the dee-stem support (see the ornl-1884 cantilever-on-insulators execution) rather than retrofitting isolation later.

The removal of the flat shims from the tank increased the magnet gap to 13 1/2 in. and provides sufficient clearance to permit operation of the dees at a potential of 100 kv.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 18

Tabletop: For a next machine the same ledger applies at 5-13 kV: dee-to-liner spark distance plus dee aperture plus liner clearances must fit inside the gap, and every millimeter given to voltage clearance is field (B ~ 1/gap) taken from energy. Decide dee voltage and gap in the same trade study.

Beam radius, in. 11 / 20; Proton energy, Mev 1.5 / 4.9; Magnetic field, oersteds 6400; Magnet gap, in. 13.5; Maximum dee-to-dee potential, kv 100; Frequency, megacycles/sec 9.7 (spec table, condensed)

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 18

Tabletop: The nearest professional sibling to a next machine in this collection - same ~0.64 T field class and ~9.7 MHz as the reference machine's 0.59 T / 9 MHz, scaled up in radius and voltage. Use it to sanity-check B-f consistency and to see what 100 kV (vs the reference machine's ~0.8 kV) buys in radius terms.

The push-pull circuit for this test was constructed by using an 8-ft section of the electrical model of the 63-in. cyclotron dee stems as the resonant system.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 19

Tabletop: A bench-scale dee-stem mockup (pipe sections plus padding capacitors) lets the next machine's oscillator/coupling scheme be raced against alternatives for pocket change - the same measure-on-model philosophy as the deferred filament-coupling rule, one report earlier in hardware form.

the resonant impedance of the dee stems varies markedly with the frequency due to the comparatively low "Q" of the secondary circuit.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1953 — ORNL-1663 (1954) — p. 19

Tabletop: Relevant if a next machine ever adds a variable-frequency or remote trim element - any lossy tuning appendage coupled to the dee resonator drags its shunt impedance (hence dee voltage per watt) across the tuning range; keep trim elements high-Q or mechanical.

The error of the control in tuning the oscillator to the frequency of the secondary is of the order of 0.1%.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 March 1954 — ORNL-1670 (1954) — p. 24

Tabletop: A 1954 peak-hold autotune implementable today in a microcontroller for the rf chain of the reference machine or a next machine - sweep the exciter, record the dee pickup peak, re-sweep and lock; 0.1% at 9 MHz is ~9 kHz, comparable to the tuning precision the next machine's 5-G field budget implies.

at an r-f load of 100 amp per lineal inch, the temperature rise could be held to less than 10 C by a water flow of only 0.5 gpm.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 September 1954 — ORNL-1795 (1954) — p. 23

Tabletop: 100 A/in. is a design allowable for any sliding or clamped rf contact (shorting planes, tuning bars) in a next machine's resonator - and the material lesson (plate the stainless with copper; bare SS is an rf resistor) applies at any scale.

The whole dee system is supported by a cantilever mounting at the outer end of the dee stems. This mounting is supported on insulators in order to permit the application of a bias potential to the dees.

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 March 1955 — ORNL-1884 (1955) — p. 19

Tabletop: The mechanical execution of the 1663 insulate-for-bias rule - for a next machine, one stiff cantilevered dee-stem mount outside the field region, isolated by insulators, is simpler than distributed insulated supports and keeps the bias feed and rf geometry clean.

Dee-to-dee r-f voltage (design), kv 200; Threshold voltage for N5+, kv 110 (Table 3, condensed)

Howard (ed.), Electronuclear Research Division Semiannual, period ending 20 March 1955 — ORNL-1884 (1955) — p. 20

Tabletop: Same margin philosophy as the 63-inch 75-vs-60-kV acceptance hold (ornl-1339) - for the reference machine's LDMOS upgrade, compute the threshold dee voltage for the intended turn count and buy amplifier/resonator headroom for ~2x it, not 1.1x.

the pulses are obtained in a way which makes them insensitive to beam current changes, and b) all reference pulses are smooth and identical in shape.

Fulbright et al., A Fast Neutron Time of Flight System for Use with Cyclotrons — NYO-9360 (1962) — p. 6

Tabletop: The most directly transferable finding here: the machine's RF is a free timing fiducial at ANY scale. Reference-machine and next-machine experiments (beam-phase measurement, gated counting, TOF over cm-scale paths for keV protons) can clock everything off a capacitive sniff of the dee line; beam-derived triggers die exactly when you need them (low current).

connected to a short (10") antenna of No. 12 wire which extends inside the oscillator house to within a foot or two of the grid circuit of the oscillator.

Fulbright et al., A Fast Neutron Time of Flight System for Use with Cyclotrons — NYO-9360 (1962) — p. 8

Tabletop: DIRECT recipe, buildable on the reference machine in an afternoon - loose capacitive coupling (never galvanic) plus stub-tuned harmonic mixing is how you sharpen a 9 MHz sine (55 ns rise as a sinusoid) into a few-ns edge with zero active electronics at the pickup. The check-the-pulse-shape-after-retuning discipline (p.9) carries over verbatim.

stop trigger pulses emerge only after an event occurs in the neutron detector.

Fulbright et al., A Fast Neutron Time of Flight System for Use with Cyclotrons — NYO-9360 (1962) — p. 8

Tabletop: DIRECT - this reversed (common-stop) architecture is still how RF-referenced timing is done; it inverts the time axis but slashes dead time and pileup. With a modern TDC or digitizer the same logic applies: trigger on the detector, timestamp against the next RF zero crossing.

In our case reducing the frequency of beam pulses to about 3 mc can virtually eliminate the problem.

Fulbright et al., A Fast Neutron Time of Flight System for Use with Cyclotrons — NYO-9360 (1962) — p. 16

Tabletop: SCALE-HONEST: frame overlap needs multi-MeV spectra over meter paths, so sub-MeV machines rarely hit it - but the tool is general: a pair of deflection plates driven at an RF subharmonic is the cheapest beam chopper a cyclotron can have (single-bunch selection, duty-cycle control, background gating), and the odd/even zero-crossing subtlety is real circuit-level physics worth teaching.

In order to simplify tuning procedures the scaler system was designed with a minimum of tuned circuits; there is only one, the tank circuit associated with the beam deflection plates.

Fulbright et al., A Fast Neutron Time of Flight System for Use with Cyclotrons — NYO-9360 (1962) — p. 17

Tabletop: DIRECT pair for a next machine's auxiliary electronics: (1) every tuned circuit is a knob someone must retune at every frequency change - minimize them by design (a lesson the reference machine's self-excited oscillator experience already rhymes with); (2) use derived/divided signals for SELECTION logic but always take the precision timing edge from the primary RF - never from a divider chain.

Performance of the model is considered sufficiently satisfactory to proceed with the full scale design and construction based on the model dimensions.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 16

Tabletop: A next machine's dee/stem/tank is already benchtop-sized, so the transferable form is the mockup itself — a cheap RF-only copy (no vacuum) of the dee-liner geometry, swept with a VNA before the vacuum parts are machined. Same lineage as UCRL-64 and MDDC-1045 already in this collection.

Only the radio frequency circuit was simulated in the model, the vacuum system and purely mechanical equipment was not included.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 7

Tabletop: License to mock up the next machine's RF cavity in bare copper/aluminum on a bench plate — no chamber, no pumps — provided every conducting surface that carries RF current (liner included) is reproduced.

For 30 kv on the full scale system the above power input figures become 146 kw at 9.25 mc, 132 kw at 12 mc, 146 kw at 15 mc, and 98 kw at 23 mc for continuous operation.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 15

Tabletop: The V^2 term is the live part for a next machine's power budgeting — measured drive power at a safe low dee voltage extrapolates as (V_target/V_test)^2 on the same hardware, since Q is voltage-independent until multipactor/breakdown; a 1500-V measurement predicts the 5-13 kV LDMOS requirement.

These were the capacity presented to the dee by the dummy dee, the minimum capacity of the rotor, and the inductance at the throat of the dee.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 10

Tabletop: Direct checklist for why a tank on the reference machine or a next machine does not resonate where the lumped-element estimate says — dummy-dee proximity, feedthrough/trimmer minimum C, and stem-to-dee transition inductance are the three knobs.

the insertion of the dummy dee had dropped the upper frequency limit from 48.8 to 44.5 mc, and the lower limit from 19.9 to 18.8 mc

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 9

Tabletop: The reference machine's lore confirmed at lab scale — final RF tuning of a next machine's cavity must be done with dummy dee, source structure, and probes installed, or budget a multi-percent retune.

a shorted stub - a section of transmission line less than a quarter wave length at 50 mc - was connected to each side of the dee.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 9

Tabletop: The recovery plan if a next machine's fixed-frequency cavity misses 9-ish MHz after assembly; note both fixes tax drive power, so aim the design low in frequency and trim up with the cheaper capacitive side when possible.

This region is a current maximum point at the highest frequency and most sensitive to volume or inductance.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 10

Tabletop: On a small machine the dee-stem-to-chamber-wall clearance is the same critical region — it sets both the resonant frequency and where I^2R heating concentrates; machine it to drawing, do not shim it by eye.

The oscillator must be stable enough to sustain an arc drawn from the dee face (simulating discharges in that region).

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 12

Tabletop: The arc test transfers verbatim to the planned LDMOS amplifier — prove the driver (and its protection) rides through a real drawn arc at the dee before trusting it in vacuum, where sparking during conditioning is guaranteed.

A deviation of an inch or two one way or the other will cause this response to rise or fall at either end of the range.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 13

Tabletop: Even a fixed-frequency amateur system inherits this through the amp-to-dee coax and its stray resonances — sweep the assembled feed system, not just the cavity, and expect inch-level sensitivity at low VHF.

the original ones used were resonant at 18 mc, which may account for a sharp drop observed in the dee voltage as this frequency was approached.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 13

Tabletop: Filament, bias, meter, and interlock leads entering the reference machine's or a next machine's tank all need RF chokes whose self-resonance is measured, not assumed — a choke resonant near 9 MHz silently loads the dee.

Ratio Max-capacity/Min-capacity = 1370/50 = 27.6

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 14

Tabletop: Same differential technique as Koeth's Rutgers dee-capacitance note already in this collection — an LCR meter plus one disassembly step yields every lumped C in a next machine's tank model, feeding the resonance and Q predictions.

The excitation was removed by connecting the plate line to the oscillator housing so that no r.f. currents would flow and all the power input would appear in the plates of the triodes.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 8

Tabletop: The thermal-reference trick survives translation — calorimetry on the LDMOS heatsink (or dee cooling loop) calibrated at DC gives dissipated power without trusting directional couplers; and publish instrument error bars the way Anderson did.

the inductance inherent in the 304-TL triodes is large compared with that in the 9C21 triodes to be used in the final oscillator.

Anderson, Half-Scale Model Tests on the Three Quarter Wave R.F. System — UCRL-31 (1947) — p. 12

Tabletop: When bench-testing a next machine's RF with a stand-in amplifier or without the real chamber wall, write the fidelity caveats into the test log — the model predicts the cavity, not the parts that were substituted.

volts per turn substantially lower than the designed 280 kev/turn would result in sharp reduction of both extraction efficiency and optical quality.

Blosser & Gordon, Computational Study of a Resonant Extraction System for a 3-Sector Cyclotron — MSUCP-9 (1961) — p. 45

Tabletop: The quantitative ancestor of "dee volts buy extraction" — the reference machine's ~800 V nominal dee is why it is internal-beam-only, and the next machine's 5-13 kV target is what would make any future extraction scheme even thinkable.

The sinusoidal voltage, it is seen, shifts the final position of the beam spot but has almost no effect on the distortion.

Blosser & Gordon, Computational Study of a Resonant Extraction System for a 3-Sector Cyclotron — MSUCP-9 (1961) — p. 40

Tabletop: A permission slip for CYCLOPS-lite staging — start with constant energy gain per gap to get the radial dynamics right, then add cos(phi) gain and phase slip as a second-stage refinement, checking that conclusions survive.

Table 1. k/h = 0.1: at x/h = 0, E/(V0/h) = 0.99388 [values verified against page image]

Beal, Computation of Electric Field and Potential of an Idealized Dee Geometry — MSUCP-12 (1961) — p. 11

Tabletop: Sets the real ceiling on gap-field strength for any dee redesign — with a 1-inch aperture (h = 0.5 in) and 5 kV dee-to-dummy (V0 = 2.5 kV), peak field cannot exceed ~2 kV/cm no matter how tight the gap; widening the aperture for beam height costs peak field one-for-one.

Table 1, k/h = 0.1: V/V0 = 0.73760 at x/h = 1.0, 0.94468 at 2.0 [verified against page image]

Beal, Computation of Electric Field and Potential of an Idealized Dee Geometry — MSUCP-12 (1961) — p. 11

Tabletop: Transit-time factors and gap-crossing phase errors for the tiny machine and a next machine must be computed on this extended profile — at low first-turn velocities the particle spends a large RF phase interval inside a field region ~2h long, which a delta-kick model at the gap centerline gets wrong.

Table 1, k/h = 2.0: E/(V0/h) = 0.37823 at x/h = 0, maximum 0.38966 at x/h = 1.2 [verified against page image]

Beal, Computation of Electric Field and Potential of an Idealized Dee Geometry — MSUCP-12 (1961) — p. 16

Tabletop: Kills the tempting back-of-envelope E = V_dee/gap for both breakdown margin and energy-gain estimates on a next machine's geometry, where gap and aperture are the same order (k/h ~ 1) and neither limiting approximation holds — use the formulas or the tables.

Commercial mu metal shields, if properly used, will normally provide sufficient shielding against magnetic fields. To eliminate the effects of RF fields, aluminum foil or screening can be used.

Martin (ed.), Accelerator Radiation Protection — AD-755510 / USA-NLABS-TR-73-7, US Army Natick Laboratories (1972) — p. 128

Tabletop: Written for exactly such a bench — a scintillator PMT near a 0.6 T fringe field and a 9-MHz (soon LDMOS) transmitter. The Keithley 617 grounding lore in the reference machine's as-builts is this rule independently rediscovered.

The capacitor housings have separate vacuum systems using turbomolecular pumps. RF feed through insulators separate them from the main cyclotron vacuum system.

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 10

Tabletop: Scales down well — a differentially pumped appendage for any mechanism (the reference machine already does this in spirit with the diff-pumped source region) keeps ion-gauge-clean chamber pressure honest.

the rotors are at low RF due to their < 0.010 in. high capacitance face gaps to ground

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 10

Tabletop: General lesson for any rotating RF machinery (choppers, tuners) at any power level; the rotating-capacitor FM tuner itself is synchrocyclotron-specific and does NOT transfer to a fixed-frequency tabletop machine.

The basic variation of line Zo along the resonator tends to minimize the capacitor Cmax/Cmin ratio needed.

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 9

Tabletop: FM machinery itself does not transfer, but the impedance-profile trick applies to any tunable tank, and to any RF cavity that needs a swept or trimmed frequency range.

a half-wave resonator permits the rotating capacitors to be located outside the main vacuum chamber for good shielding from both the magnetic field and radiation

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 10

Tabletop: The placement principle transfers — keep variable capacitors, trimmers, and drive mechanisms of a next machine's tank outside the pole gap and chamber where field, beam spray, and pumpdown cannot reach them.

The design has used a 1/2 scale model, in conjunction with detailed computer calculations, to establish all parameters which can be varied

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 9

Tabletop: At tabletop size the "scale model" is the full-size mockup on the bench — cold-test a next machine's dee/stem with a VNA before power exists; the mode-spectrum audit transfers verbatim.

The dee resonator will be dc floating so a negative bias of amount sufficient to control multipacting can be applied.

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 10

Tabletop: Directly relevant at a next machine's planned 5-13 kV dees where multipactor bands are widest; corroborates the glow corner's multipactor-mechanism rules with a 1971 operating-lab remedy and a concrete bias magnitude to scale from.

a slow linear reduction (turn off) of the RF amplitude there will result in a near adiabatic spreading out of the phase angle

Rainwater et al., The Columbia University Nevis Synchrocyclotron Major Modification — NEVIS-189 / R-774 / CU-295 (1971) — p. 5

Tabletop: Meaningless for a fixed-frequency CW tabletop cyclotron; it belongs to swept-frequency machines, where adiabatic capture and slow parameter ramps are the RF-program design space.

The Q of the model will be 0.7 times the Q of the actual installation and so will require 1.4 times as much power for a given dee voltage.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 6

Tabletop: Bench-model a dee-stem or resonator geometry at reduced size before cutting full-size copper; apply the sqrt(scale) Q correction before comparing model power and coupling measurements to full-scale predictions.

It was not excited satisfactorily due to the fact that at 100 megacycles the transit time effects were quite noticeable on the fundamental mode.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 6

Tabletop: Cold measurements (network analyzer) scale exactly, but any POWERED model test needs a driver whose parasitics and transit time scale with the geometry, or the model will exhibit modes the real system never sees and vice versa.

dimensions can be calculated fairly exactly whereas in the system shown in Figure 5 one must depend on model tests (which are safer anyway).

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 6

Tabletop: Transmission-line formulas ignore end effects, bends, and support hardware; for any resonator whose geometry is not a textbook line, a cheap model measurement outranks the calculation it checks.

A total of 6 wavetraps sufficed to suppress all unwanted modes throughout the proton range. The parasitic modes were not completely eliminated, however, when the dee was shorted, as would occur in a discharge.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 7

Tabletop: Sweep the assembled dee/stem/liner system with a signal generator and probe before first power-up; every resonance within the amplifier's gain bandwidth is a candidate parasitic, and a dee spark momentarily retunes the system into modes the clean census missed.

two of them require that large currents flow in the rotor supports, which can be made fairly high resistance. These modes are therefore not excited in favor of the much higher Q correct mode.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 7

Tabletop: Mode-selective damping - resistance placed at a current maximum of the unwanted mode and a current null of the wanted one - is cheaper and more robust than trying to tune the parasite out of band.

the insulators will be subjected to about 3 times the rotor voltage to ground. This is because the long metal supports act as open transmission lines

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 8

Tabletop: Check the electrical length of every support, cooling line, and instrument stalk inside the RF volume; a mechanically convenient standoff can sit at a voltage antinode and flash over at dee voltages its rating should easily hold.

The oscillating circuit actually is a long folded transmission line consisting of the two rows of meshed teeth.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 16

Tabletop: The diagnostic transfers whole - a sharp, frequency-specific dead spot in dee voltage means some conductor assembly (screen, liner seam, feedthrough array) is resonant there; fix it by shortening or breaking up the structure, not by driving harder.

if the frequency of the transverse mode is still higher than the fundamental when the teeth are fully meshed, then the transverse frequency will always lie above the fundamental for any partially meshed position.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 16

Tabletop: For any tunable element (trimmer panel, movable shorting plane), verify mode separation at the extreme of travel where the wanted and unwanted modes converge; monotonic behavior between extremes lets one measurement clear the whole range.

over about a 2 to 1 frequency shift, the voltage on the 1/4 wave-shorted line (which will be referred to as the "stub" line) is practically the same as the dee voltage.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 3

Tabletop: Even a fixed-frequency machine drifts with thermal expansion and plasma loading; feeding at a voltage-stable point of the resonator keeps drive impedance roughly constant as the resonance moves.

start with the tap at the end of the stub line and then move toward the shorted end until the tube draws rated plate current at rated voltage.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 11

Tabletop: The same walk-the-tap procedure sets link or tap coupling on any dee tank - begin overcoupled-safe at high impedance and converge on rated loading, rather than computing a tap position and committing to it.

End effects and bends in the line can cause this much variation. The conclusion is that there should be some possibility of varying the line length by a small amount

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 11

Tabletop: Design connection lines and stubs with a sliding section or trombone worth a few percent of a wavelength; the calculation gets you to the right neighborhood and the adjustment does the rest.

the shift is reduced by making the filament grid capacity large, so that the out of phase r.f. current will be large compared with the in phase electron emission current.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 14

Tabletop: The mechanism matters for any self-excited tube oscillator on a dee - drive phase error costs efficiency quadratically; padding the input capacity is the one-component fix, and the graphical Shanklin incident/reflected-wave construction (Fig. 15) checks it.

The 1/2 scale model uses about 400 watts input to the oscillator at 48 megacycles and 600 watts input at 18 megacycles to produce 1500 volts on the dee.

MacKenzie, Preliminary Report on the “Three Quarter Wave” R.F. System for Frequency Modulated Cyclotrons — AECD-1850, University of California (1947) — p. 16

Tabletop: Dee power scales as voltage squared - measure watts-per-volt-squared on the bench, correct for Q, and the amplifier requirement for any target dee voltage falls out; joint quality and surface material shift Q enough to budget for.

the R. F. dee voltage was lowered to approximately 50% of its normal value which is less than the threshold voltage.

Cyclotron Staff, Report of Cyclotron Operation 1 July – 31 December 1969 — NRL Memorandum Report 2103, Naval Research Laboratory (1970) — p. 23

Tabletop: There is a dee-voltage threshold below which no ions survive to full radius; gating against it - by stepping the regulator reference, not by unkeying the RF - pulses beam for detector duty-cycle or background measurements without any retuning transient.

The filament voltage of the 6949V1 must be raised slowly from zero to the normal operating value to prevent filament current from exceeding a value of 1,700 amperes, even momentarily.

Cyclotron Staff, Report of Cyclotron Operation 1 July – 31 December 1969 — NRL Memorandum Report 2103, Naval Research Laboratory (1970) — p. 24

Tabletop: Scales straight down to any transmitting tube or big thoriated filament - a soft-start (variac ramp or NTC inrush limiter) plus an undercurrent trip covers both the cold-start and the loose-socket failure modes that shatter filaments.

The four PAA tuning capacitors are now coupled together by a heavy-duty chain driven by a single large servo motor with one translator. No trouble has been experienced since this modification.

Cyclotron Staff, Report of Cyclotron Operation 1 July – 31 December 1969 — NRL Memorandum Report 2103, Naval Research Laboratory (1970) — p. 24

Tabletop: Wherever two adjustments must stay matched (paired trimmers, symmetric shorting planes), a shaft, chain, or belt enforces the constraint by construction; independent actuators plus software matching is a standing failure mode.

The modulation on the dee voltage, defined as the peak-to-peak ripple riding on the RF voltage as a percentage of the peak RF value, has recently been reduced to about 0.5% from 1.5%.

Cyclotron Staff, Report of Cyclotron Operation 1 July – 31 December 1969 — NRL Memorandum Report 2103, Naval Research Laboratory (1970) — p. 25

Tabletop: Dee-voltage ripple modulates turn energy and orbit phase; put the envelope on a scope, log the percentage, and remember the excitation source (a cheap signal generator's spurs included) is a candidate cause before blaming the amplifier or resonator.

we increased the flow rate to the 6949 tube plates from 42 gpm to 60 gpm by the addition of a 4-inch return pipe to separate the high pressure and low pressure water loops

Cyclotron Staff, Report of Cyclotron Operation 1 July – 31 December 1969 — NRL Memorandum Report 2103, Naval Research Laboratory (1970) — p. 29

Tabletop: When an amplifier cannot hold rated dissipation, check hydraulic head losses in shared manifolds before derating the tube; and duplicating the single-point-of-failure pump is a reliability purchase the outage ledger justifies.

The holder ring was split to prevent eddy currents in the ring from rf (Ramsay, discussion of "Thick Targets for In-Beam Hyperfine Structure Study")

Fifth Annual Conference of the International Nuclear Target Development Society — LA-6850-C, Los Alamos Scientific Laboratory (1977) — p. 201

Tabletop: An internal target probe near the dee gap lives inside the machine's own RF field; a closed metal frame is a shorted turn that heats and perturbs — one saw cut fixes it.

in cyclotrons using a simple dee system as the tank circuit difficulties are encountered in getting the oscillator to break into full oscillation.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 2

Tabletop: DIRECT — this sentence names the exact configuration of the 8" machine (simple dee system as the tank circuit) and matches its documented multi-year pattern of RF amplifiers failing to bring the dee to voltage. Multipactor loading in the ~100 V band is a named, testable candidate cause for that history, distinct from amplifier inadequacy.

multipactor loading, which occurs with voltages of the order of a hundred, cannot build up sufficiently to prevent the rise of voltage through the multipactor region.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 2

Tabletop: The decision menu for any machine that stalls in the multipactor band: bias, drive-through, or (this report's contribution) impulse shock. On a small machine the driven start maps to an external exciter ahead of the power stage; the bias cure maps to a few-hundred-volt DC offset on the dee — pick by which fights the existing hardware least.

The dee biasing scheme was considered too awkward to apply, chiefly because the variable frequency requirement had already led to a rather complicated mechanical design.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 2

Tabletop: Transferable decision pattern: on a machine whose dee stem is grounded through the tank structure, retrofitting DC bias means rebuilding the stem insulation, while a shock starter touches nothing but a spare port. Choose the quench that is additive.

the dee circuit begins to oscillate with a dee voltage amplitude of several hundred volts. The oscillator then begins to carry the voltage on up to its full value.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 3

Tabletop: The cheapest cure in this collection for a stalled self-excited start: one loop, one capacitor, one spark gap, one HV supply — all amateur-stock parts. The key insight is that the kick need only clear the top of the loading band, not deliver operating power; everything above a few hundred volts is the oscillator's own job.

The spark gap is adjusted so that the sparking rate is roughly two per second. Ordinarily a single spark will cause oscillation to commence.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 3

Tabletop: Complete parts list (values on Fig.2, PDF p.8). A sub-joule impulse sufficed on a 27" machine; a smaller dee system needs proportionally less. The 30 kV supply is the only nontrivial part, and a NST or flyback-class supply covers it.

automatically turned on when the oscillator power is on and there is no dee voltage, but which is automatically turned off as soon as dee voltage appears.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 3

Tabletop: DIRECT automation pattern: gate the starter on (RF enabled) AND (dee pickup below threshold). The dee capacitive pickup already present for voltage monitoring is exactly the signal needed, and the same logic doubles as a stall alarm — sparks repeating at 2/s means the machine is failing to start.

Because the loop is oriented at right angles to the axis of the cavity, the RF voltage induced in the loop is very small, even with full dee voltage.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 3

Tabletop: General principle for any diagnostic or starter coupling added to a resonator: choose an orientation that is null for the operating mode. The impulse still couples because the spark's broadband ring excites wall currents, not the clean mode — an inefficient coupler is acceptable precisely because the required kick is small.

the envelope of the oscillations was found to fall smoothly until the dee voltage had fallen to a value roughly 1/3 its maximum, then for a short time to drop steeply, then afterward to decay slowly once again.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 4

Tabletop: A free diagnostic: ring the dee (impulse or drive-and-release), scope the pickup envelope, and look for a kink. A visible steep-decay segment localizes the multipactor band on YOUR machine and tells you whether nominal operating voltage sits inside it — the critical question for any dee running near a few hundred volts.

Changing the tuning of the grid circuit changed the amplitude of spark-induced dee oscillation severalfold, which seems indicates the flow of considerable capacitative current through the oscillator tube.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 4

Tabletop: Two practical consequences: the shock method transfers to solid-state drivers (nothing about it is tube-specific), and every branch circuit hanging on the resonator participates in the ring — tune-dependent shock amplitude means the starter should be commissioned at the operating tune, not on the bench.

The ringing frequency of the current in the sparker loop was found to be about 12 mc, independent of the oscillator conditions. The amplitude of ringing is found to decay 50% in about 3 cycles.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 4

Tabletop: Sets expectations honestly: the shock cures the multipactor stall, not bad vacuum or a detuned feedback network. If a spark does not start the machine, the fault list is pressure, tune, or outgassing state — a diagnostic branch in itself. The low-Q broadband ring means one fixed sparker covers a whole tuning range (trivially true at fixed frequency).

partly because it was more convenient in our case since the dee chamber is rather crowded, and partly because we thought that the RF pickup problem would cause less trouble with inductive coupling.

Fulbright, The Sparker, a Device to Overcome the Multipactor Difficulty in Starting the Oscillator of a Cyclotron — NYO-9359, University of Rochester (1961) — p. 4

Tabletop: Directly usable trade note: a loop near the dee stem (outside the beam region) beats a capacitive plate facing the dee in a small chamber where every square inch by the dee is contested. Do not optimize the coupler — "works with margin" arrived at the first, deliberately inefficient geometry.

The skin is of electrolytic high conducitvity copper with 1/4, 3/8, and 5/8" copper tubes silver soldered on the back side for water cooling.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 58

Tabletop: The construction vocabulary (EHC copper skin, silver-soldered cooling, silver-plated RF joints, high-pressure sliding contacts) is exactly what a 5-13 kV LDMOS-driven dee upgrade needs; split-for-repair is cheap foresight at any scale. Same contact-pressure concern as the nyo-9683/ornl-2648 sliding-contact rules.

upon considering the approximations necessarily made in this type of analysis, the figure of 150 kw maximum r-f power was selected.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 75

Tabletop: DIRECT scaling method for the LDMOS upgrade — measure the dee system's Q and C, compute watts per kV from P = V^2/(4QX) (equivalently V^2/(2R_shunt)), then buy the amplifier with a 2-3x factor for the terms the lumped model misses. The Koeth Rutgers dee-voltage note is the same math on an 8-12 inch machine.

This method should be easiest of the methods used to assure oscillation at the proper frequency with the dees operating 180 degrees out of phase.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 76

Tabletop: For a one-dee-plus-dummy machine the mode problem collapses, but the principle stands for any driven system — verify which resonance the amplifier is locking to (a network analyzer sweep distinguishes the modes), because the wrong one accelerates nothing. Also this collection's second explicit SELF-EXCITED architecture choice (see ucrl-9435 rule on the self-excited-vs-MOPA tension).

Electron oscillations in the vicinity of the dees and dee stems at low r-f voltages tend to absorb energy and prevent the oscillations from building up.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 75

Tabletop: The corpus's driven-start cure (mddc-1045 tickler; nyo-9359's catalog) as a 1951 DESIGN feature rather than a retrofit — including the elegant half-frequency/doubler isolation trick so the starter needs no changeover switch. Directly relevant to the reference machine's dee-voltage buildup pathology: any LDMOS drive chain is inherently a "driven" start, but only if it can push watts through the ~100-V multipactor band without foldback/protection tripping.

there are many uncertainties in the exact determination of the constants of the equivalent circuit ... the calculated values ... serve well as a guide

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 82

Tabletop: A tabletop resonator IS the scale model — build the dee/stem mockup on the bench, measure f and Q before committing to vacuum hardware, and trust lumped calculations as guides not gospel. Berkeley used the same quarter-scale method on the 88-inch (ucrl-9435), which also confirms the Q-degradation-at-joints lesson (ornl-2648).

the plate is operated at d-c ground potential so that no insulation is required in the water lines.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 85

Tabletop: Solid-state amps moot the HV plumbing, but two transferable doctrines survive — pick the grounding scheme that keeps coolant out of the HV problem, and use source impedance (here transformer reactance) as passive inrush protection instead of active circuitry where a component's own rating allows it.

measured by means of a saturable reactor ... The dee voltmeters are of the series type coupled ... by a small capacity between the probe and a copper paddle soldered to the dee edge.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 86

Tabletop: The capacitive-paddle dee voltmeter is the same instrument Koeth calibrated on the Rutgers 12-inch and the missing calibration on the reference machine's "~800 V nominal" — a soldered paddle + defined-gap probe + diode peak detector, calibrated once against a real HV probe, converts dee voltage from folklore to data. The saturable- reactor trick survives as the Hall-effect/isolated-shunt principle: never bring an HV node to the meter.

On the basis of cost it was decided to omit this refinement.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 90

Tabletop: The decision pattern (name the deleted protection, name the two things standing in for it, keep a commissioning-only resistor in the drawer) is directly reusable; contrast ucrl-9435, where the 88-inch — with 20x the stored energy — bought the full hard-tube-modulator protection instead. Scale decides.

a minimum of 80 kv dee to ground r.f. potential is necessary in order that the ions do not enter the region of decelerating phase at any time within the dees.

The University of Washington 60-Inch Cyclotron: Progress and Status Report of Design and Construction — AECU-1951, University of Washington (c. 1950) — p. 116

Tabletop: The RF-ground deflector-plate trick (steal the existing dee field, add DC only as a supplement) and the adjustable-channel flux-vs-resolution trade are scale-free ideas for any future extraction study on a next machine; the phase-floor analysis is the kind of computation this collection's deflector cluster (nyo-9360 etc.) expects before metal is cut.

(b) the multipactoring problem, common in the field of particle accelerators, rarely occurs in other industrial applications

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 7

Tabletop: The checklist for adapting ANY industrial/ham RF gear (including an LDMOS pallet) to a cyclotron: add spark protection, add a multipactor start plan, and only then worry about power. Fixed-frequency tabletop machines are spared only (c).

The threshold of secondary emission is about 150 electron volts for most surfaces; consequently, multipactoring becomes possible when the voltage across the dees reaches this value.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 4

Tabletop: DIRECT — the ~100-150 V band is exactly where a small machine's dee voltage must pass on every start. Completes this collection's cure set with the DC SWEEP variant: mddc-1045 (bias + tickler), nyo-9359 (impulse), ucrl-64 (volume reduction + bias), aecu-1951 (booster drive-through). Same author lineage as ucrl-3153/3187.

It usually involves permitting the dee to spark several hundred thousand times. Afterwards, it will usually hold many times the voltage that it would initially.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 4

Tabletop: DIRECT for the 5-13 kV dee upgrade: plan a conditioning campaign (auto-recycle protection makes it unattended) rather than interpreting early sparking as failure. A tabletop resonator stores millijoules, so conditioning is gentle — the count, not the violence, does the work. Corroborates the ornl-2648/nyo-9683 conditioning rules and quantifies them.

In this service it will open the anode circuit within 10 usec of a fault.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 6

Tabletop: The modern translation is exact: LDMOS drain supplies want a fast electronic disconnect (the hard-tube modulator's descendant), a slower breaker layer, snubbers, and a matched/terminated dc feed. The dual-use insight — the series regulator IS the fast protection switch — carries straight into a solid-state dee supply. Cross-check ornl-2403's protection chapter; Smith's is the self-excited counterpart.

In this type of system the resonator is the frequency-determining element of the system; hence it is called a self-excited oscillator.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 4

Tabletop: The live architecture decision for a next machine. An LDMOS chain driven by a synthesizer is a MOPA — it inherits ornl-2403's virtues (frequency authority, instrumentation) AND the self-excited literature's start-up disease (nyo-9359): the synthesizer holds frequency while multipactor holds the dee at zero. Smith's phase-discipline logic (feedback phase correct across the whole operating range) is the checklist item either way. High-SWR argument p.6.

At the maximum particle energy, the beam requires 60 kw of power.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 5

Tabletop: The four-line budget is the right form at any scale — a tabletop version is watts of copper loss, ~zero stray-ion, uW of beam, and a misc line that is mostly coupling/radiation. The named "stray-ion loss at the center" term is a reminder that source gas load steals RF power — one more reason the reference machine's beam and RF problems interlock.

The circuit elements of the rf system were adjusted so that the first two higher modes would not be excited by an oscillator harmonic.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 5

Tabletop: DIRECT for a fixed-frequency machine: sweep the dee system to a few hundred MHz on a VNA, list the modes, and check none sits at 2f or 3f of the drive — if one does, detune it with a strap or stub BEFORE blaming the amplifier for instability. Same mode-vs-harmonic discipline as the msucp-9 and ornl-2648 lines.

For accelerator applications, a tube should have a sufficiently heavy grid to absorb several joules of energy in an area determined by the cross section of the spark.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 6

Tabletop: The solid-state translation: LDMOS gates tolerate NO spark energy — the joule-absorbing ruggedness must move into the coupling network (series blocking, clamping, circulator/isolator) because it no longer lives inside the active device. Budget those parts as the modern "shield tees."

The rf-dc interlock compares the dee voltage with the amount of oscillator anode dc. If the ratio is too low, indicating the presence of an arc, the fault detector opens the anode circuit and recycles

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 7

Tabletop: DIRECT and cheap: a comparator on (dee pickup voltage vs forward power) with a ~1 s drop-and-retry turns dee sparks from session-enders into log entries, and is precisely the automation a conditioning campaign needs. The ratio form matters — absolute thresholds miss arcs that still draw full power.

If the circuit is made too fast, it takes too long to bake the resonators in. If it is made too slow, the spark damage to the dee and linear surfaces will be excessive.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 7

Tabletop: A protection spec you cannot derive from electronics alone — the dwell time is chosen so each spark finishes cleaning the spot that caused it. For a tabletop supply: let a dee spark burn ~1 ms before the drop-and-retry, but trip amplifier-device faults as fast as the electronics allow. Two speeds, two purposes.

the resonator has to be vacuum-tight, automatically making it rf-tight.

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 8

Tabletop: DIRECT and comforting for a residential machine: a metal chamber dee system radiates almost nothing — EMI escapes via feedthroughs, viewports and the amplifier side, so gasket those and shield the drive chain and the neighbors' radios stay quiet. (The same reasoning applies to viewport mesh in the fusor literature.)

Maximum operating conditions for the RCA 6949 for the 88-in. cyclotron

Smith, The RCA 6949 as a Self-Excited Cyclotron Oscillator — UCRL-9435, Lawrence Radiation Laboratory (1960) — p. 10

Tabletop: The ratio discipline transfers: a healthy class-C/E chain shows 70-90% final efficiency and 15-20 dB final-stage gain whether it is 319 kW of tube or 500 W of LDMOS — efficiency far below that in the reference machine's or a next machine's amplifier means mistuning, parasitics, or multipactor loading, not "small machines are just lossy."