Operating the PVX-4110, PVX-4130, PVX-4141 and PVX-4151 with common-polarity pulse and bias supplies

1Two rails and one output

A direct-coupled half-bridge high voltage pulse generator places two solid-state switches in series between a high rail and a low rail that the user supplies; the output is their junction. A logic-level gate decides which switch conducts: gate high connects the high rail to the output, gate low connects the low rail. The pulse amplitude is therefore the difference between the two rail voltages, and the pulse width and repetition rate follow the gate. Figure 1 draws the arrangement.

The two rails give the topology its polarity options. With the low rail at ground the output pulses from ground up to the high rail; with the high rail at ground it pulses from ground down to the low rail. The third case is the subject of this note: a pulse whose baseline is not ground. A detector gate that rests at a bias and pulses further in the same direction, or a deflection plate that pulses between two potentials of one sign, is set up with both rails on supplies of the same polarity, for example -HV In at +500 V and +HV In at +1,500 V, or -HV In at -1,000 V and +HV In at -500 V. The output then rests at the lower voltage and pulses to the higher one. The output stage accepts that; its only requirement is that the high rail is never below the low rail. The difficulty is on the supply side.

2What a DC supply can and cannot do

A regulated DC supply holds its output voltage by controlling the current it passes, and most high voltage supplies pass that current in one direction only. A positive supply draws electrons from ground; in conventional terms it sources current out of its positive terminal and takes it back at ground. A negative supply does the opposite. Neither can absorb current pushed back into its output terminal, because the regulating element conducts one way: if something external raises a positive supply's output above its setpoint, the supply's current falls to zero and its terminal voltage is set by whatever is pushing. With supplies of opposite polarity, or with one rail grounded, this limit is never tested, because every current returns through ground in each supply's natural direction. With two supplies of the same polarity the lower-voltage supply is connected to a node into which the pulse generator delivers current, and it has no way to take that current away.

3How the bias node drifts toward the pulse level

Two things deliver current into the low rail node. The first is leakage inside the pulse generator from the high rail toward the output while the generator is not gated: small, continuous, and with no return path it can only flow into the low rail's supply terminal. The second is the load. A capacitive load is charged to the high rail during each pulse and discharged to the low rail at each falling edge, and the charge that went in at the top comes out through the lower switch into the low rail node, more often as the repetition rate rises.

If the low rail's supply cannot sink that current, the node charges. The output baseline, which should rest at the -HV In potential whenever the generator is not gated, drifts toward the +HV In potential (Figure 2). The symptom is most visible when the load draws no DC current of its own and as the pulse repetition frequency is increased. At low rates with small loads it may not be seen, because the DC supply's own internal paths to ground absorb a small current; that is a property of the particular supply, not something to design around.

The result is a pulse of the wrong amplitude from the wrong baseline, and a bias supply whose current reads zero while its terminal voltage reads high. Neither instrument is faulty; the circuit is missing a return path.

4The remedy: a current path to ground

The low rail needs a path to ground that carries, in the supply's natural direction, at least as much current as the generator pushes into the node. In low power applications a resistor from the -HV In input to ground does this: it draws a steady current from the bias supply, which stays in regulation because it is always sourcing, and the current the generator delivers into the node flows out through the resistor instead of charging the node (Figure 3). In high power applications, where the resistor's dissipation would be large, an active shunt regulator from the bias input to ground does the same job with less heat, conducting only as much as it must to hold the node at the bias voltage.

The next two sections size the resistor by the method published with these pulse generators. It is deliberately conservative: a shunt too large in value leaves the node uncontrolled, while one too small costs only heat and supply capacity.

5Sizing the shunt resistor

Three quantities are needed, all of them from the application rather than from the pulse generator:

• the maximum pulse repetition frequency, F;
• the maximum differential voltage, V = V_HIGH minus V_LOW, which is the pulse amplitude;
• the total load capacitance, C, including the load itself, the interconnect cable and strays.

Because the load is capacitive, the charge that goes into it must also come out, and the power associated with switching it is the quantity the whole procedure is built on. The minimum power that the resistor or shunt regulator must be able to dissipate from switching the capacitive load is

From that power the discharge current through the shunt is

and the shunt resistor value follows from the bias voltage and that current:

A regulated supply holds the bias, and it stays in control only while it is sourcing current. If the current pushed into the node ever exceeds the current of equation (2), the supply's output current falls to zero and the bias is no longer controlled. To keep a margin, the value from equation (3) is halved, so the shunt always draws more current than the generator can deliver into the node and the supply is always sourcing.

DC current from the bias supply flows continuously through the shunt, and the resistor dissipates

For reliability the resistor is chosen with at least twice that power rating. It also stands at the full bias voltage continuously, so its voltage rating must cover V_LOW with margin: above a few hundred volts that means a high-voltage rated part or several resistors in series with the creepage between them respected.

6Worked example

The example is the one published with the original note. Suppose the load, the pulse amplitude and the repetition rate give a switching power of 1 W from equation (1), and the bias supply at -HV In is set to 100 V. Table 1 carries the steps: equation (2) gives 10 mA, equation (3) gives 10 kΩ, halved to 5 kΩ, and equation (4) gives 2 W, so a 5 kΩ resistor rated 5 W is chosen.

Figure 4 plots the procedure over a range of bias voltages. The halved resistance is V_LOW² / 2P, so it rises with the square of the bias voltage, while the continuous dissipation is 2P at any bias: at 1 W the shunt dissipates 2 W whether the bias is 50 V (1.25 kΩ) or 500 V (125 kΩ).

7What each DC supply must deliver

Once the shunt is fitted the two supplies carry different loads. The bias supply at -HV In must deliver the shunt power, 2 W in the example, which at 100 V is 20 mA of continuous current. The pulse supply at +HV In must deliver the C V² F switching power of equation (1), 1 W in the example, plus the generator's quiescent losses. Both must be regulated, and the bias supply's current rating must exceed the shunt current with margin, because a bias supply in current limit has lost control of its voltage.

8When the problem does not arise

Three common configurations never meet it, and need no shunt:

• Supplies of opposite polarity, for example -HV In at -1,000 V and +HV In at +500 V. Each supply passes current in its natural direction and the return path exists through ground.
• One input grounded, for example -HV In at ground and +HV In at +1,500 V, or the reverse. Ground absorbs whatever the generator returns.
• Small loads at low repetition rates, where the current delivered into the bias node is below what the supply's own internal paths to ground absorb. This depends on the supply, and raising the rate or the load capacitance later can cross the threshold: a setup that works at 100 Hz into a test load is not evidence that it will work at 10 kHz into the real one. If both rails share a polarity, fit the shunt.

9Confirming the fix on the bench

The front-panel voltage monitor is the safe way to watch the baseline: it reproduces the output waveform at a fixed ratio into a 1 MΩ oscilloscope input and needs no high voltage probe. The ratio is 2000:1 on the PVX-4110 and PVX-4130 and 1000:1 on the PVX-4141 and PVX-4151, from the published datasheets, so a 100 V bias reads 50 mV on the PVX-4110 monitor and 100 mV on the PVX-4141 monitor.

1. Set both supplies to their operating voltages with the gate held low and the generator enabled. Confirm at the voltage monitor that the output rests at the -HV In voltage.
2. Gate the generator at the lowest repetition rate the application uses, watch the baseline between pulses for several seconds, then raise the rate in steps to the maximum. A baseline that holds at every rate confirms the return path; one that climbs with rate, or over time at a fixed rate, means the path is missing or too small.
3. Read the bias supply's output current. With the shunt fitted it should show a steady V_LOW / R, 20 mA at 100 V into 5 kΩ, and never fall to zero while the generator runs. Zero current with the terminal voltage above the setpoint is the signature of section §§mechanism§§.
4. After several minutes at full rate, check the shunt resistor's temperature with the supplies off. A resistor dissipating 2 W is warm; one that is hot points to a switching power above the one assumed, and the sizing should be repeated with the measured load capacitance.

The shunt carries the full bias voltage to ground; treat it, and any resistor string, with the spacing and creepage of any other high voltage component.

10Applicability across the four models

The four pulse generators in scope share the half-bridge output stage and the pair of user-supplied inputs, so the mechanism and the procedure apply to all of them; they differ in voltage, power and rate, which set the switching power of equation (1). Table 2 collects the published values. The PVM-4210 is out of scope: it carries its own integrated high voltage supplies and has no +HV In or -HV In inputs, so there is no external bias supply to shunt.

11Limitations

The procedure is a sizing rule, not a model of the generator's leakage. It takes the full switching power as the quantity the shunt must handle, which gives a resistor that is safe at every operating point but larger in dissipation than a measurement of the actual return current would justify. Where that shunt power is inconvenient, the active shunt regulator of section §§remedy§§ is the alternative; this note does not design one.

The procedure assumes a capacitive load; the datasheets state that C V² F does not apply to resistive or inductive loads. The worked example uses a round 1 W and 100 V; a real application computes equation (1) from its measured load capacitance, cable included. And the behavior of a bias supply pushed toward zero output current depends on that supply's design, which is why the fix is a return path.

12Summary and selection

A pulse that starts and ends away from ground needs two supplies, and when both carry the same polarity the lower-voltage supply cannot sink the current the generator delivers into its node, so the baseline drifts toward the pulse level. A shunt resistor from the -HV In input to ground, sized by section §§sizing§§ and worked in section §§example§§, keeps the bias supply sourcing current and the baseline where it was set. Confirm it at the voltage monitor, and size the bias supply for the shunt current.

Choosing among the four models is a matter of Table 2: the differential voltage the load needs and the switching power at the intended rate. Within any of them the shunt is sized the same way.

13Definitions

Half-bridge (totem-pole) output stage

Two switches in series between a high rail and a low rail, with the output at their junction; one switch conducts at a time.

Pulse supply (+HV In, VHIGH)

The user-supplied DC input connected to the output while the gate is high. It sets the top of the pulse.

Bias supply (-HV In, VLOW)

The user-supplied DC input connected to the output while the gate is low. It sets the pulse baseline and is needed only when the output is biased away from ground.

Shunt resistor

A resistor from the -HV In input to ground: a continuous current for the bias supply to source, and a path to ground for returned charge.

Acronyms

14Further reading and one route onward

• Capacitive-Load Pulsing, Directed Energy technical note: the physics of I = C dV/dt and the C V² F power budget behind equation (1). https://www.berkeleynucleonics.com/docs/article-dei-capacitive-load-pulsing.html
• Selecting an HV Pulse Generator and How Do I Slow the Rise Time of My Voltage Pulser?, Directed Energy technical notes on model selection and on resistor voltage ratings and strings, which apply to the shunt as well. Both are listed at https://www.berkeleynucleonics.com/docs/dei-technical-papers.html
• The four model datasheets, the authority for Table 2: https://www.berkeleynucleonics.com/docs/bnc-dei-pvx-4110-datasheet.html and the matching pages for the PVX-4130, PVX-4141 and PVX-4151.

If you are setting up a common-polarity pulse on one of these generators and would like the shunt sized against your actual load capacitance, cable and repetition rate, our applications engineers would be glad to work through it with you. Write to info@berkeleynucleonics.com or call 800-234-7858.