Detectors & Nuclear Instrumentation

Detectors & Nuclear Instrumentation

A detector laboratory runs on more than one kind of high voltage. A photomultiplier draws a fraction of a milliampere through its divider at 1 to 2 kV. A silicon photomultiplier draws microamps at 25 to 70 V. An ion-source extraction gap carries the beam current it launches, milliamps to hundreds of milliamps at kilovolts. An electron gun or an electrostatic accelerator column carries its beam at tens of kilovolts. A feedthrough, cable or divider under test has to be held at voltage while its leakage is read. These loads differ in current by five orders of magnitude or more, and no single supply class serves all of them.

The deciding numbers are a supply's floors rather than its ceilings: the smallest current its current loop regulates, the smallest voltage step it sets, and the ripple it leaves on the output. The Berkeley Nucleonics HVX-Series is a 2U supply in 500 W, 2 kW and 3 kW classes with microcontroller and FPGA regulation, and each of its floors scales with the nominal rating of the model. Table 1 works them out from the datasheet: ripple is specified at 0.01 percent of nominal voltage plus 100 mV peak to peak, the setting range as 0.01 to 100 percent of nominal, and setting resolution as 16 bits.

Read against a silicon photomultiplier, those floors rule the HVX-Series out. A device with a 28 V breakdown run at 3 V of overvoltage is biased at 31 V. Its gain is proportional to the overvoltage, so the 0.25 V ripple floor of a 1.5 kV model modulates the gain by 8 percent peak to peak, and the 23 mV voltage step is 0.8 percent of the overvoltage. The microamp bias current is below the 140 µA at which the current loop begins to act. The right source is the bias generator built into the ScintIQ silicon-photomultiplier detectors and readout modules, which runs from a 5 to 16 V supply and corrects the bias for temperature.

A photomultiplier is a closer call and still the wrong fit. At 1,500 V the ripple floor is 170 ppm of the bias, and for a ten-stage tube whose dynode gain varies as the 0.7 power of voltage that moves the gain by about 0.1 percent peak to peak, which is small against the energy resolution of any scintillator. The mismatch is in current and in scale: a tube's divider draws 0.1 to 1 mA, below the 140 to 200 µA at which the current loop of a 1.5 kV model engages, and a 2 kW supply on a 1 W load has none of its range in use. For a single tube the ScintIQ bMCA carries its own 0 to 1,500 V supply in 4,096 steps; a counting room full of tubes is served by a multichannel detector bias supply built for microamp loads. Figure 1 places those loads against the HVX-Series bars.

Ion-source extraction and acceleration gaps. The electrode carries the beam current it launches plus the drain of the gap. A 10 kV extraction passing 100 mA of beam is a 1 kW load, within an HVX-10000-200 at half its current. A source that sparks during conditioning is what Arc Detection is for: it reports the flashover and can switch the output off, and Ramp Control brings the gap back up at a set gradient, from 1 V/s upward, rather than as a step. The external interlock input on the I/O terminal ties the output to the source enclosure.

Electron guns and electrostatic columns at 20 to 30 kV. The HVX-20000-25 and HVX-30000-17 deliver 500 W in fixed positive or negative polarity: a 20 kV gun at 15 mA is 300 W on the HVX-20000-25. Line regulation is better than 0.01 percent of nominal for a 10 percent mains change, and stability is 0.01 percent of nominal over 8 hours, which holds a 20 kV gun within 2 V over a shift.

A bulk bias bus for a large photomultiplier array. Four hundred tubes at 1,500 V and 1 mA each draw 0.4 A, 600 W, from one HVX-1500-1400, provided each tube's divider is fed through its own series resistor and capacitor so the bus ripple is filtered locally. A 10 kohm and 1 µF stage has a 16 Hz corner and attenuates a 20 kHz component by 62 dB, taking 0.25 V to 0.2 mV, at the cost of 10 V of drop at 1 mA and a per-tube trim where the tubes need one. The arithmetic depends on the ripple frequency, so measure the bus before sizing the filter.

High-voltage test of detector hardware. Feedthroughs, cables, dividers and insulators are held at 10 to 30 kV with the current limit set to protect the item under test and the actual current shown on the display, and the event log time-tags every trip. Discharge at no load takes 60 s typical and up to 120 s, which is the wait before the item is handled.

The DC source behind a high-voltage pulser. The DEI PVM-1001, which gates photomultipliers and microchannel plates with 0 to 950 V pulses into 50 ohm, takes an external supply of up to 975 V; a 1.5 kV HVX-Series model in the matching polarity supplies it.

Two options matter for these loads. Arc Detection detects a flashover in the output, reports it and can switch the output off. Ramp Control sets the gradient of every rise and fall from 1 V/s to ten times the nominal voltage per second. Both are field-retrofittable. Polarity on the 1.5 kV to 10 kV models is positive, negative or electrically reversible; the 20 kV and 30 kV models are ordered positive or negative. Ethernet and RS232 with SCPI, the external interlock and the time-tagged event log are standard.

Limitations. The smallest current the HVX-Series regulates is 0.01 percent of nominal, 140 µA on the 1.5 kV, 1.4 A model, so a load below that runs in constant-voltage mode with no current regulation. The ripple floor is 0.25 V to 3.1 V peak to peak by model and is filtered at the load where a quieter rail is needed. The datasheet does not state how far the FLO models may be referenced away from earth; confirm that figure for a stacked installation before ordering.

For a quick question, chat with an engineer at berkeleynucleonics.com.

Ask now → Chat with our assistant and get an answer in seconds.