加速器高压电源在紧凑型加速器与研究装置中的高压链

  A compact accelerator is a chain of potentials rather than a single voltage. The source sits at one end, the target or the extraction point at the other, and between them the beam passes through a series of electrodes whose voltages must be right relative to each other and not merely right in absolute terms.

  The chain begins at the supply. A single high-voltage unit feeds a resistor divider that sets every intermediate electrode, and the beam sees the sum of the gaps. If the supply is accurate but the divider drifts, the beam sees a distorted field even though the meter at the top reads the correct value.
  Resistor dividers drift for two reasons: self-heating and moisture. A chain of ten resistors carrying a milliampere each dissipates a few watts in a small volume, and the resistors at the top of the stack run hotter than the ones at the bottom, so the division is not the same as the one measured cold.
  The remedy is to specify the chain for the power it actually dissipates rather than for the voltage it holds. Resistors with a low temperature coefficient, mounted on a common heatsink, and measured at operating temperature, will hold a division to a part in a thousand where a casual chain holds it to five parts.
  The second element of the chain is the cable. In a compact machine the supply is often a separate cabinet and the accelerator is a tank, connected by a cable of a few metres. That cable carries the direct-current voltage plus whatever ripple the supply produces, and it also carries the displacement current of every spark.
  Cable capacitance changes the dynamic behaviour of the whole chain. A cable of five hundred picofarads per metre, run for four metres, presents two nanofarads to the supply, and that capacitance stores energy that must be counted in the spark budget of the accelerator tube.
  The third element is the tube itself. The accelerating tube is a stack of ceramic insulators and metal rings, and each ring is at a potential set by the chain. The tube's own capacitance is small, but the ring-to-ring capacitance and the surface leakage together determine how the field distributes when the beam pulses.
  Surface leakage is the failure that appears as a beam that drifts. A ring surface that has absorbed moisture will leak, the local field will shift, and the beam will steer by a fraction of a milliradian, which is enough to miss a small target.
  The fourth element is the target or the extraction. In an industrial electron machine the beam exits through a thin window and the window's potential is ground; in an ion machine the beam strikes a target at a potential that may float. Either way, the return path for the beam current must be defined, and it must not pass through the supply's measurement leads.
  Protection is distributed along the chain rather than concentrated in the supply. Each gap should have a spark gap or a voltage-dependent resistor that limits the energy that a discharge can deliver to the tube, and each of those devices must be rated for the voltage at its own position rather than for the total.
  A coordinated protection design uses the supply's current limit as the first line, a series resistor in the cable as the second, and a spark gap at each critical gap as the third. The three together keep a discharge below the energy that damages the ceramic.
  Commissioning the chain takes a defined sequence. The divider is measured cold, then the supply is run into a resistive load, then the cable is connected with the tube isolated, then the tube is connected with the beam off, then the beam is brought up at low current.
  Each step isolates a different element. Skipping the step in which the cable is connected without the tube turns a cable fault into a tube fault, and the tube is the expensive part.
  An industrial case illustrates the divider problem. A compact electron machine produced a beam that was uniform at low energy and asymmetric at high energy. The chain's upper resistors were running at sixty degrees while the lower ones ran at thirty, and the division had shifted by two percent, which tilted the field at the exit.
  The cure was thermal rather than electrical: the resistors were mounted on a common aluminium bar, the bar was tied to the cabinet wall, and the temperature spread fell to five degrees. The asymmetry disappeared without changing any resistor value.
  A second case involved a cable. The machine tripped on overcurrent every few days, always within a minute of start-up. The cable had a pinhole in its insulation near a bend, and the leakage appeared only when the insulation was cold and the capacitance was charging. Replacing the cable and supporting it at the bend removed the trips.
  A third case involved a spark gap that had been set for the total voltage rather than for the gap's own position. The gap fired at a voltage far above the local field, the ceramic between two rings was punctured, and the tube had to be replaced. Recalculating each gap for its own position prevented a recurrence.
  Instrumentation along the chain is worth the cost. A divider at the top, a divider at the middle and a divider at the bottom give three readings that can be compared; a single reading at the top tells only that the supply is behaving, not that the chain is.
  The comparison is the diagnostic. If the top and the middle agree and the bottom does not, the fault is at the bottom. If all three drift together, the fault is in the supply's reference. If all three read correctly and the beam is still wrong, the fault is in the tube or in the beam optics.
  Maintenance should follow the same three points. Each divider is checked quarterly against the top reading, the cable's insulation resistance is checked monthly, and the spark gaps are checked annually for their firing voltage.
  Spares for a chain are modest: one set of divider resistors, one cable assembly, one set of spark gaps and one window. The tube is not a spare; it is a rebuild.
  Cost accounting should include the diagnostics. A machine with three dividers costs perhaps ten percent more than one with a single divider, and the difference is recovered the first time a fault is found in an hour rather than in a week of substituting parts.
  Limits should be stated. No supply can correct a chain whose resistors have different temperature coefficients, and no protection can save a tube that is operated above its rated gradient. The supply sets the top of the chain, and the chain must be designed as a chain.
  The design intent is a sentence. Put a known voltage at the top, divide it with a chain that is thermally uniform, carry it to the tube with a cable whose energy is counted, and protect every gap against the energy that its position can deliver.
  A compact machine is compact because every element of that sentence has been made small; it is reliable because none of them has been made careless.
  Gas insulation inside the tank is part of the chain, not an accessory. The tank is filled with a dielectric gas at a pressure chosen so that the field at the most stressed electrode stays below the gas's limit, and that pressure must be monitored because a tank that loses a fraction of a bar loses a large fraction of its withstand.
  A pressure switch that only alarms is insufficient. It should inhibit the high voltage rather than merely sounding a horn, because a de-rated tank will hold until it does not, and the failure is a flashover that damages the tube.
  Grounding of the tank is the reference for the whole chain. The tank's earth must be a dedicated conductor to the building earth rather than a connection to a nearby frame, since the spark current in a fault flows through that conductor and its impedance sets the potential rise that the supply's control board will see.
  Control architecture should separate the fast loop from the slow supervision. The fast loop regulates voltage and current inside the supply; the slow supervision watches the chain's three dividers and the tank's pressure, and it is the layer that decides to stop the machine.
  A recording function is the cheapest diagnostic in the chain. Storing the three divider readings once per minute for a month gives a trend that shows a resistor aging before the beam shows it, and the cost of that function is a few kilobytes of memory.
  The beam current itself is the last signal worth watching. A chain that is correct but whose beam current falls at constant energy indicates a change in the source or in the gas, not in the high-voltage side, and knowing that saves the effort of opening the tank.
  Training should follow the chain. An operator who understands that the top reading is the supply, the middle reading is the divider and the bottom reading is the tube will report a fault with the right words, and the engineer will arrive with the right instrument.