6
3590 East Columbia St Tucson, AZ 85714 520-628-7415 http://www.appliedenergetics.com/highvoltage.asp Page 1 of 6 A Modular PFN Marx with a Unique Charging System and Feedthrough Richard J. Adler, Joshua A. Gilbrech, David T. Price Applied Energetics Abstract We have developed a reliable, high power Marx generator system capable of operation at up to 550 kV/11 kA with an output flat top of 1 μs nominal. The unit can operate at up to 5 Hz for 2 seconds (>30 kW output) from a power supply wi1th a "built-in" HV section. The Marx section consists of 14 series type C PFNs each with an impedance of approximately 3.5 ohms. The individual PFNs are built so that they are easily removed and modified. With this flexibility, the unit can be adapted to changes in the load such as those due to diode impedance collapse. The HV section of the HV power supply is not only in the Marx oil bath but it is "connected" to the semiconductor portion of the power supply using "air core magnetic flux coupling". This technique is as far as we know unique, and it eliminates the requirement for a high voltage or high current wired feedthrough in the system. It also dramatically reduces EMI coupling from the Marx to the semiconductor power supply modulator and the control system. The efficiency of this supply is approximately 80 %. In the paper we will discuss the theory of the coupling and the experimental results of operation of the power supply and Marx. I. INTRODUCTION The Marx Generator is a popular means of generating high voltage, high peak power output pulses and many Marx generators have been built over the years 1, 2 . Our goal was to build a Marx Generator that was reliable, easy to service, and capable of handling high power burst mode operation. Size constraints were traded for the ability to design the system in such a way that it was entirely modular and easy to assemble, build, test, and service. The second main design point was to be able to handle operation above 30kW for short burst periods. This required engineering design in order to provide a charging, control, and triggering scheme that would be resilient to the EMI generated by the repetitive erection of the Marx bank. In particular the EMI from the pulse can cause switching power supply control circuits to shut down so that only one pulse is generated. This is often not noticed in single pulse systems. The high voltage charging system is often an awkward part of the system since it involves mating a rack mount power supply through an oil bulkhead to the Marx. The cable connections are effectively a “highway” for noise generated by the Marx to enter the control and power system which are often in the same rack. We have often expended significant effort in protecting the HVDC power supplies used. We decided to develop a charging technique that would eliminate the technical difficulties previously experienced with the charging of Marx Generators. We have named this technique “Reslink” (patent pending), and we will discuss it in detail in this paper. Other design issues include the design of a low jitter reliable trigger and computer/FPGA control for the system that can reliably operate in the Marx EMI environment. II. SYSTEM DESIGN The parameters for the system are listed in Table 1. There are four main driving parameters: total energy, flat top pulse duration, recharge time and the need for a relatively fast rise time and fall time. Output Voltage 500kV into a 50-ohm load Maximum Voltage Droop <+/- 5% Maximum Pulse Width >1uS 90%-90% Rise Time 10%-90% rise time < 200ns Fall Time 90%-10% < 300ns Modes of Operation Single Shot (Repetitive at .1 Hz) Repetitive Burst [10 pulses at 5 Hz] Table 1. Design Parameters. These parameters drove the principal design choices as follows. The 1μs flat top pulse with ± 5% voltage variation requires a means of creating a flat top. We chose a type C PFN network 3 as the Marx Unit Cell in order to create a flat output pulse for 1μs, as shown in Figure 1. Figure 1. Marx Cell Schematic

A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

  • Upload
    volien

  • View
    252

  • Download
    5

Embed Size (px)

Citation preview

Page 1: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 1 of 6

A Modular PFN Marx with a Unique Charging System and Feedthrough

Richard J. Adler, Joshua A. Gilbrech, David T. Price

Applied Energetics

Abstract

We have developed a reliable, high power Marx generator system capable of operation at up to 550 kV/11 kA with an output flat top of 1 µs nominal. The unit can operate at up to 5 Hz for 2 seconds (>30 kW output) from a power supply wi1th a "built-in" HV section. The Marx section consists of 14 series type C PFNs each with an impedance of approximately 3.5 ohms. The individual PFNs are built so that they are easily removed and modified. With this flexibility, the unit can be adapted to changes in the load such as those due to diode impedance collapse. The HV section of the HV power supply is not only in the Marx oil bath but it is "connected" to the semiconductor portion of the power supply using "air core magnetic flux coupling". This technique is as far as we know unique, and it eliminates the requirement for a high voltage or high current wired feedthrough in the system. It also dramatically reduces EMI coupling from the Marx to the semiconductor power supply modulator and the control system. The efficiency of this supply is approximately 80 %. In the paper we will discuss the theory of the coupling and the experimental results of operation of the power supply and Marx.

I. INTRODUCTION

The Marx Generator is a popular means of generating high voltage, high peak power output pulses and many Marx generators have been built over the years1, 2. Our goal was to build a Marx Generator that was reliable, easy to service, and capable of handling high power burst mode operation. Size constraints were traded for the ability to design the system in such a way that it was entirely modular and easy to assemble, build, test, and service.

The second main design point was to be able to handle operation above 30kW for short burst periods. This required engineering design in order to provide a charging, control, and triggering scheme that would be resilient to the EMI generated by the repetitive erection of the Marx bank. In particular the EMI from the pulse can cause switching power supply control circuits to shut down so that only one pulse is generated. This is often not noticed in single pulse systems.

The high voltage charging system is often an awkward part of the system since it involves mating a rack mount power supply through an oil bulkhead to the Marx. The

cable connections are effectively a “highway” for noise generated by the Marx to enter the control and power system which are often in the same rack. We have often expended significant effort in protecting the HVDC power supplies used. We decided to develop a charging technique that would eliminate the technical difficulties previously experienced with the charging of Marx Generators. We have named this technique “Reslink” (patent pending), and we will discuss it in detail in this paper.

Other design issues include the design of a low jitter reliable trigger and computer/FPGA control for the system that can reliably operate in the Marx EMI environment.

II. SYSTEM DESIGN

The parameters for the system are listed in Table 1. There are four main driving parameters: total energy, flat top pulse duration, recharge time and the need for a relatively fast rise time and fall time.

Output Voltage 500kV into a 50-ohm load

Maximum Voltage Droop <+/- 5%

Maximum Pulse Width >1uS 90%-90%

Rise Time 10%-90% rise time < 200ns

Fall Time 90%-10% < 300ns

Modes of Operation Single Shot (Repetitive at .1 Hz)

Repetitive Burst [10 pulses at 5 Hz]

Table 1. Design Parameters.

These parameters drove the principal design choices as follows. The 1µs flat top pulse with ± 5% voltage variation requires a means of creating a flat top. We chose a type C PFN network3 as the Marx Unit Cell in order to create a flat output pulse for 1µs, as shown in Figure 1.

Figure 1. Marx Cell Schematic

Page 2: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 2 of 6

The Marx equivalent circuit is very close to a simple series stack of 14 of these 3.5 ohm units.

As shown in Figures 6 and 7 each of the LC branches of the network discharges in such a way that the total output of the cell is sufficiently flat for our requirements.

In order to meet the requirements the design had to minimize unnecessary series inductances. This requirement rippled through our design of current bus bars, inductor wire selection, and capacitor specification. Due to this inductance budget the current carrying bars were large (see figure 6). The distance from the bus bars to the spark gaps needed to be minimized while providing physical flexibility. Capacitors were selected for their peak current rating as well as lead inductance.

A. Initial Calculations

We note that the minimum energy requirement (no rise and fall time or other losses) is 5 kJ per pulse, but considering that 20% of the energy will be in the rise and fall, and that ≈ 20% will be in other losses, we require about 7 kJ per pulse. The voltage of operation was selected so that we could use 100 kV caps with a reasonable design margin (80 kV operating voltage). The open circuit voltage must be twice the operating voltage for a PFN, so the net open circuit voltage required was 1 MV. Given 1 MV and 80 kV operating voltage, the number of stages is 1 MV/80 kV = 12.5. 14 stages was our final selection (actually 13.5 stages). Ideally the impedance of a stage is 50 ohms/14 = 3.5 ohms so that was also a design requirement. We used a +/- charge configuration since this eliminates the need for trigger biasing, and because the deviation from a symmetric trigger bias is minimized with +/- charge. Our charging technique was designed to make the positive and negative voltages nearly perfectly equal. Since a single bipolar supply was used, we did not need to equalize two separate supplies. We had a “half-stage” because the first stage was grounded on one side and it was only charged to half voltage.

The charge power required is 7 kJ * 5 Hz for a constant power charge. One advantage of our charging technique is that unlike many supplies it is near constant power (within about 20% except for the first few milliseconds) We allow a 20 millisecond spark gap recovery time so the actual charge time duty cycle is only 90 %. Thus we need a power input of approximately 7 kJ* 5 Hz/0.9/0.8 ≈ 50 kW.

B. Charging system

One of the novel aspects of this Marx Generator is its unique method of charging. Typical Marx Generators make use of off the shelf capacitor charging supplies. These supplies must be protected from both the conducted and radiated noise of the generator and also any and all voltage reversals, ground bounces, and other problems introduced by the Marx Generator. The use of loosely coupled air core transformers has been utilized for years by the team at

Applied Energetics5. This experience was used to help design the Marx Generator charging supply. Figure 2 shows the 3d model for the primary, secondary, and output rectifier for the charging system.

A series resonant H-bridge IGBT inverter was used to drive the primary, which was air insulated. The secondary coil and a bipolar doubling voltage multiplier were housed on the interior of the Marx Tank in oil. The PVC that the windings were wrapped on served as both the feedthrough and the coil form.

Figure 2. Primary/Secondary Link 3D model. When in operation the primary on the right is inside the secondary spool (on the left). The secondary spool is the air/oil barrier.

There is no actual electrical penetration of the oil tank. Power transfer was done via magnetic coupling of the primary air insulated winding to the secondary oil insulated winding. PVC was chosen due to its easy of machining, dielectric strength, and low material cost. The low coupling from primary to secondary in conjunction with a ground shield on the inside of the secondary spool make the tank a nearly complete Faraday cage.

This design is based on numerous other Applied Energetics/Northstar Power Engineering designs that incorporate highly ruggedized power supplies by means of loosely coupled resonant drives. It has similarities to our “Nested High Voltage” technology.

C. Triggering System

The triggering system was composed of an Applied Energetics standard design for low repetition rate medium jitter Marx triggering. This system has been used at voltages of up to 400kV with rise times in the 10’s of nanoseconds. For the purposes of this Marx the output sharpening stage has been omitted. As such the output was 250kV (max) and most of our work was done at 160 kV output.

Page 3: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 3 of 6

The system is composed of an OEM HV charge power supply that charges two S series General Atomics capacitors. These capacitors are discharged by a three pin E2V spark gap through a low inductance single winding on an amorphous core transformer. The secondary of the transformer outputs the fast rise time trigger pulse. Figure 3 illustrates the base design of the trigger unit.

Figure 3. Trigger 3D model.

The switching column was designed such that it could be

easily removed from the oil and cleaned or adjusted. The rationale for utilizing a switching column was twofold: the primary reason was to minimize gas connections which are often problematic and the secondary reason was to allow for UV pre-ionization of later stage gaps. This secondary reason also allows for a possible wave erection4.

Figure 4. Switching Column 3D model.

D. System Assembly

The system was assembled as shown in figure 5. The design built has the 14 Marx cells linearly inserted into PVC “combs”, a single cylindrical gas insulated spark gap column, and charging inductors (which are also resistive) between each stage. The system was designed to allow for minimum work needed in order to remove stages for tuning and or maintenance.

Figure 5. System Assembly.

Voltage and output current monitoring were done through optically isolated voltage probes, V dot, and B dot sensors. All signal, command, and control lines were optically isolated via voltage to frequency conversion, transmission, and interpretation by a LabVIEW control computer. This control computer was designed to live within a double shell faraday cage. All power and control signals were isolated via transformers or optical isolation. As such the control computer was impervious to the EMI of the Marx erection at full voltage and maximum burst rate, even within close proximity (3 m).

One of the first parts of the system to be fully assembled was the Marx platters, shown in Figure 6. The platter inductors were wound to the simulated design and then discharged into a scaled load. The output current of each branch of the PFN was monitored. This allowed for the fine tuning of the discharge inductors to the precise values needed to create the flat top output pulse.

As shown in the schematic, our design differs from a conventional type C PFN by splitting the main capacitance into two values with different time constants.

Page 4: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 4 of 6

Figure 6. Marx Cell Assembly

Figure 7. Simulated output waveform from single platter

III. PERFORMANCE

The modular approach taken during the design phase worked out extremely well. Modifications to the system were able to be completed with very little down time due to the ability to pull Marx cells out with very little modification to the full system. Thus to make any sort of modification to the system a full tear down of the system was avoided. This allowed for a very painless tuning on the Marx output to precise specification. Figure 8 shows a picture of the full system constructed and in operating condition.

Figure 8. Marx Generator Assembly

The PFN Marx Generator performed well under full output testing. The following graph, Figure 9, shows an example scope shot of a 500kV single shot. The channel two trace is the integrated E-Dot output, channel one trace the Iout MON signal and the channel three trace is the Charge Voltage, Vset. The charge voltage, Vset, trace should be multiplied by 10 for an accurate reading. Note that the channel two voltage trace has a 10% instrumented droop and is in fact flat to within about 5% once that droop is corrected.

Figure 9. Output waveforms showing the output voltage (CH2), current (CH1) and half of the charge voltage (CH3).

Page 5: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 5 of 6

The PFN Marx Generator has a Pearson 101 current transformer and an E-Dot voltage probe within the tank. The E-Dot Probe waveform output is monitored after going through an RC integrator box into a 50 Ohm load. As shown in Figure 9, the output current is 10.2kA, half the charge voltage is 40kV, and the scaled Vout is 506kV. The 50 ohm load used was a custom made water resistor for this PFN Marx. The total charge voltage for this shot was 80kV. The measured 40kV on the channel three trace was measuring the positive end of the bipolar doubling multiplier with respect to ground.

To obtain the required 80kV charge on the capacitor banks, the charging system required approximately 200msec. An end of charge circuit detects when the capacitor banks are fully charged at the end of the 200msec cycle. Once the banks are fully charged, the charging system is inhibited and the trigger unit fires the PFN Marx. Successful operation was shown for single shots as well as burst shots of up to 10 pulses and rep rates of up to 5 Hz.

There is indication that the wave erection process is taking place due to the very fast rise time observed to 70 % voltage. This cannot occur due to the platters themselves (due to their series inductances) but instead it occurs due to discharge of the stray capacitance from the platters to ground. The wave erection results because each stage acts like a “peaking switch” and the switches close sequentially every 5 to 10 ns.

The wave erection process puts anomalously high voltages (up to 400 kV) across the final 3 to 4 inductors. This was observed to cause a partial discharge between the inductor wires. These 4 final inductors were replaced with water resistors and the gaps were changed to reduce the wave erection. These changes eliminated charging component problems.

A blanking circuit was implemented to inhibit system operation if a set current threshold though the load was exceeded. The internal Pearson 101 current transformer was used for this. An EOC (end of charge) inhibit feature was also implemented so that the user could mandate the PFN Marx to trigger, even if the set charge voltage was not reached.

It should also be noted that the performance of this PFN Marx was extremely reliable. During full output voltage testing, prefires and misfires were eliminated altogether by correctly balancing the set charging voltage, switch gaps, and spark gap column SF6 pressure.

The Marx noise has been extremely low (a screen box is not really required for most signals) and the computer has never crashed or rebooted during a shot. The power supply and the unique feedthrough have performed perfectly.

Figure 10. Waveform from data after droop correction (upper) and waveform from simulation with wave erection (lower).

Figure 11. Marx Generator Tank and Final Assembly Photos.

Page 6: A Modular PFN Marx with a Unique Charging System and ... · A Modular PFN Marx with a Unique Charging System and Feedthrough ... We used a +/- charge configuration since this

3590 East Columbia St

Tucson, AZ 85714

520-628-7415

http://www.appliedenergetics.com/highvoltage.asp

Page 6 of 6

IV. CONCLUSION

A highly modular, PFN Marx Generator was developed by Applied Energetics. This Marx Generator employed novel charging techniques and extremely reliable triggering methods that enable it to operate up and beyond design specifications. This system was designed such that future systems will take forward the modularity of the core design architecture. Future possibilities for the design are, folded platter setup to reduce overall length, faster rise-times made possible through wave erection or an output peaking gap, and higher repetition rate operation through an upgraded charging system.

V. REFERENCES

[1] J. O’Loughlin, J. Lehr, D. Loree, “High repetition rate charging a Marx type generator”, presented at the 13th IEEE International Pulsed Power Conference, Las Vegas, Nevada, 2001.

[2] J.Lehr and C. Baum, “Charging of Marx Generators”, Circuit and Electromagnetic System Design Notes, Note 43, Air Force Research Laboratory, Directed Energy Directorate, 29 June 2000.

[3] Glasoe G. N. and Lebacqz J. V. “Pulse Generators”. Radiation Laboratory Series Vol. 5 McGraw Hill Book Company Inc. (1948).

[4] W. J. Carey, J. R. Mayes, A. P. Electron, and T. X. Austin, "Marx generator design and performance," Power Modulator Symposium, 2002 and 2002 High-Voltage Workshop. Conference Record of the Twenty-Fifth International, pp. 625-628.

[5] R. J. Adler and R. J Richter-Sand “Loosely Coupled Parallel Resonant Converter” US Patent 6934165, 2005.