微通道板探测器高压电源在成像设备与辐射环境中的增益

A microchannel plate multiplies electrons by a factor of thousands, and its gain is a function of the voltage across the plate in a way that is close to exponential. That makes the plate's supply a gain control rather than a bias source, and it means the supply's stability, its range, and its behavior during a discharge all show up directly in the image or the count rate. Imaging instruments that use plates in a radiation environment add one more variable: the plate's gain changes as it accumulates dose, so the supply has to be adjustable over a wide range to compensate for the plate's aging.

The gain's sensitivity to voltage is steep. A plate operating at 800 volts delivers one gain; at 900 volts the gain is several times higher. The practical consequence is that a supply specified as a percentage of full scale must be read as a percentage of the gain. A 0.1 percent voltage stability at 800 volts is a 0.1 percent gain stability only if the exponent is one; with the typical exponent of ten to fifteen, a 0.1 percent voltage change gives a one to one and a half percent gain change, which is enough to be visible in a quantitative imaging measurement.
The supply must therefore be specified in gain-referred terms rather than volts. In practice that means the noise and ripple should be measured and then multiplied by the local exponent, which is measured once by sweeping the voltage and recording the output current for a fixed input. The sweep takes twenty minutes and produces a curve that can be inverted to convert any voltage change into a gain change. Without that curve, the supply's contribution to the image's noise cannot be separated from the plate's own noise.
Plate aging is the slow variable. Under continuous exposure, the plate's gain falls, and the rate of the fall depends on the total charge that has passed through the channels. A supply that is fixed at one voltage will show a slowly decreasing signal over months, and the operator will attribute it to the source or the scintillator. Recording the voltage and the signal at a reference input, and computing the gain from the two, gives an aging curve that predicts when the plate needs to be replaced. In practice the curve's knee arrives without much warning, and the prediction is worth a few weeks of scheduling flexibility.
The supply's range matters as much as its stability. A plate that has aged by a factor of two needs the voltage raised, and a plate that is new may need the voltage lowered to avoid saturating the readout. A supply whose maximum is set just above the new plate's operating point will run out of range as the plate ages, and the instrument will be declared obsolete before the plate is actually worn out. Specifying the maximum at thirty percent above the new plate's nominal costs little and extends the instrument's life by a year or more.
Discharges are the recurring event in plate operation, and their energy matters. A discharge occurs when a channel's wall releases a burst of gas or when a high flux saturates the channel's current; the plate's stored energy is small, but the power supply's output capacitance adds to it. The supply should therefore have a low output capacitance at the plate's terminals, which is achieved by locating the filter close to the supply and feeding the plate through a series resistor. The resistor's value is chosen so that the discharge current is limited to a level that does not damage the channel's output end, and the resistor's thermal capacity must handle the discharge's energy without cracking.
The series resistor has a cost: it creates a voltage drop that depends on the plate's current, and the plate's current varies over several decades with the input flux. At high flux, the current through the resistor can be tens of microamps, which at a megohm is tens of volts, which is a significant gain change. A supply with a remote sense at the plate's terminals would correct for the drop, but remote sensing at the end of a resistor that is there for protection means the loop sees the discharge as a load step and fights it. The practical compromise is to use a smaller resistor with a faster current limit in the supply, accepting that the protection is electronic rather than resistive.
Radiation environments add a requirement that is easy to forget: the supply's own components are being irradiated. A supply located near the plate in a gamma field will see its reference and its divider degrade, and the degradation is cumulative and silent. Moving the supply outside the field, with a longer cable, is usually better than shielding it inside the field, because the cable's capacitance is a smaller problem than a drifting reference. If the supply must be inside, its reference should be a type with a characterized radiation tolerance, and its divider should be checked against an external meter at each maintenance interval.
Imaging with a plate is often photon-counting, and the counting mode has its own requirement. In counting mode, each event's charge is compared against a threshold, and the threshold's position relative to the pulse height distribution determines which events are counted. A supply whose voltage drifts moves the entire distribution relative to the threshold, so the count rate changes without any change in the source. For quantitative imaging, the distribution's position should be monitored, and the voltage adjusted to keep it at the same place, which is a form of closed-loop gain control that the supply should support.
The high-voltage cable between the supply and the plate is a part of the gain path. Its capacitance in parallel with the plate's means that the discharge current comes from the cable as well as from the supply, and a long cable therefore increases the discharge energy. Keeping the cable short, and using a low-capacitance construction, is worth more than adding protection downstream. Cables with a conductive layer over the dielectric, sometimes called a semiconductive layer, also suppress the microdischarges that occur within the cable itself under high voltage, and those microdischarges are a source of noise that appears in the image as sporadic single events.
The supply's ramp behavior matters when the instrument is started. A plate that is ramped quickly will emit a burst of charge as the channels' walls charge, and that burst can saturate the readout or, in the worst case, trigger a protection in the readout's front end. A ramp that takes ten to twenty seconds, with a rate limit that holds the current below the plate's maximum, avoids the burst. The rate limit should be settable, because a fresh plate and an aged plate have different burst behavior.
Grounding and the reference's cleanliness are the last items and the ones that are hardest to diagnose. A plate's output is a current into a transimpedance amplifier, and any ground current that flows through the amplifier's reference is indistinguishable from signal. The supply's return should go to the chassis at a point separate from the amplifier's return, and the two should meet only at the instrument's single ground point. A ground loop of a few milliohms is enough to inject a hum at the line frequency into the image, and that hum is often blamed on the plate.
For instruments that operate in a radiation field with pulsed sources, the supply's response to a pulsed load is worth measuring. The plate's current can step by orders of magnitude within a microsecond when a pulse arrives, and a supply with a slow loop will sag, recover, and sag again, producing a gain modulation that follows the pulse pattern. The measurement is a scope trace on the plate's voltage with a high-impedance divider, and the trace shows whether the sag is within the acceptable gain change. If it is not, more local capacitance at the plate and a faster loop are the two levers, and the capacitance's cost is the discharge energy discussed earlier.
The narrow conclusions: specify the supply in gain terms, not volts; keep the filter close and the cable short so the discharge energy stays small; provide range headroom for the plate's aging; and move the supply out of the radiation field rather than shielding it inside. Those four decisions account for most of the difference between an instrument that holds its calibration for a year and one that needs attention every month.
A field observation worth passing on: plates that are stored with the supply's voltage left on a test fixture age faster than plates stored at zero volts, even though neither has been exposed to input. The channels keep a small current flowing through surface leakage, and that current erodes the channel walls slowly. Training the operators to leave the plate at zero between runs extends the plate's life measurably, and it costs nothing but a line in the procedure.
One last habit that pays off: keep a log of the plate's voltage, the reference input's signal, and the room's humidity at every maintenance visit. Three numbers, taken four times a year, are enough to separate a supply's drift from a plate's aging from a humidity problem, and the separation is what keeps a spare plate from being spent on a fault that was never in the plate.