Optically isolated, 2 khz repetition rate, 4 kv solid-state pulse trigger generator

Transcription

1 REVIEW OF SCIENTIFIC INSTRUMENTS 86, (2015) Optically isolated, 2 khz repetition rate, 4 kv solid-state pulse trigger generator D. H. Barnett, 1 J. M. Parson, 1 C. F. Lynn, 1 P. M. Kelly, 1 M. Taylor, 1 S. Calico, 2 M. C. Scott, 2 J. C. Dickens, 1 A. A. Neuber, 1 and J. J. Mankowski 1 1 Center for Pulsed Power and Power Electronics, Texas Tech University, Lubbock, Texas 79409, USA 2 Lockheed Martin Missiles and Fire Control, 1701 W. Marshall Dr., Grand Prairie, Texas 75051, USA (Received 22 October 2014; accepted 18 February 2015; published online 5 March 2015) This paper presents the design and operation characteristics of a solid-state high voltage pulse generator. Its primary utilization is aimed at triggering a gaseous spark gap with high repeatability. Specifically, the trigger generator is designed to achieve a risetime on the order of 0.1 kv/ns to trigger the first stage, trigatron spark gap of a 10-stage, 500 kv Marx generator. The major design components are comprised of a 60 W constant current DC-DC converter for high voltage charging, a single 4 kv thyristor, a step-up pulse transformer, and magnetic switch for pulse steepening. A risetime of <30 ns and pulse magnitude of 4 kv is achieved matching the simulated performance of the design. C 2015 AIP Publishing LLC. [ I. INTRODUCTION Numerous high voltage trigger generator topologies exist utilizing various techniques from gas/vacuum to solid state switches, and combinations of the two. 1 8 Such trigger generators are used to trigger various pulse power systems driving a variety of loads. This paper presents a design used to trigger a trigatron spark gap used to erect a 10-stage, 500 kv, 42 J Marx generator. In the last few decades, advances in semiconductor devices have made it feasible to build complete solid-state pulse generators. Designs utilizing MOSFETs, IGBTs, thyristors, and more recently static-induction thyristor (SIThy), silicon carbide junction FET (SiC-JFET), 1 as well as semiconductor opening switches (SOS) have made a strong appearance in trigger generator designs. However, designs using gas/vacuum switches 2 5 are still frequently utilized due to their relative ease to operate while providing fast risetimes as well as large voltage hold off. While significant progress has been made in high power solid state switching over the past few decades, semiconductor switches still have slower risetimes, higher power loss, and lower operational voltages compared to gas/vacuum switches. 5 Therefore, when using semiconductor switches, the design often relies on utilizing large stacks in series and parallel to achieve the desired operation. This leads to physically large devices and large power dissipation. 5,6 Systems that operate with only a single semiconductor device typically employ various different spark gap combinations to achieve a risetime of <50 ns and are generally limited to currents of only a few amperes. 2 4 The systems utilizing spark gap switches in their designs are also limited in how fast they can operate; operation ranges from less than a Hertz repetition rate up to roughly 1 khz. 2 5 In contrast, trigger systems using only semiconductors have achieved rep-rates exceeding 2 MHz in short burst mode operation. 1,6 In order to steepen the trigger pulse front, magnetic pulse compression has been used successfully in various pulse generators designs. 5,8 In general, utilizing semiconductor switches rather than gas triggered switches has the tendency to produce a trigger generator with smaller jitter and longer lifetime. The presented trigger pulser design makes use of a single thyristor switch achieving rep-rates up to 2 khz with a pulse magnitude of 4 kv and a pulse risetime of <30 ns. Besides the 4 kv rated solid state switch, the trigger topology relies on passive magnetic switching and an output voltage step-up transformer. II. SYSTEM DESIGN This section will be a discussion of the system design and components used as shown with a simplified schematic of the control circuitry and high voltage components in Fig. 1. Figure 2 shows waveforms generated from the high voltage section. The trigger generator design is based on only a few critical components: a high voltage power supply for charging, thyristor for pulse generation, and MOSFET for command charging. For complete isolation, the board has fiber optic inputs, and the whole system is operable from a single 32 V battery including all circuitry and the high voltage power supply. With operation from a lithium ion polymer 2.8 A h battery, the system can produce approximately pulses before a recharge is needed. In Secs. II A II C, the characteristics of the high voltage operation will be discussed. A. Pulse generation The present design generates the high voltage pulse by creating a voltage differential across a single pulse transformer labeled as T3 in Fig. 1. The final pulse is ultimately determined by how fast this transformer is driven. Using a single thyristor without magnetic compression, the fastest risetime that may be obtained is determined by the closing time of the thyristor. In /2015/86(3)/034702/5/$ , AIP Publishing LLC

2 Barnett et al. Rev. Sci. Instrum. 86, (2015) FIG. 1. Simplified circuit schematic showing measurement points. Thick traces indicate the pulsed, high current path. Voltage and current measurement points are indicated as well, cf. Fig. 2. order to achieve a faster risetime, a magnetic switch, similar to Refs. 6 and 8, L1 in Fig. 1, is used in series with the thyristor to decrease the risetime across the transformer, T3, to less than 30 ns. This is done by having the primary inductance value of the transformer T3 much smaller than the inductance of L1 before saturation and much larger than the saturated inductance of L1. The resulting output waveforms for an initial charge of 4 kv across C5 reveal the much faster voltage collapse across the magnetic switch. The magnetic switch approaches saturation at t 0.95 µs with falltime of 30 ns. In comparison, the voltage collapses across the thyristor, onset at t 0.5 µs has a falltime of 200 ns. The secondary of the 1:1 transformer was left open circuited when these waveforms were taken. The transformer core is a Magnetics, Inc., tape wound core, part number A. This core matches the requirement of having a very square B-H curve with a high permeability. The high permeability allows for good coupling between primary and secondary windings, benefiting the circuit by having as few turns as possible on the primary. The transformer core operates optimally with 19 turns with an inductance of 350 µh. The magnetic switch core is a Toshiba, amorphous cobalt tape wound core, part number MS21x14x4.5W. This core also needs to have a very square B-H curve with a much smaller cross section area than the transformer. The core size and turns are adjusted to saturate right after the thyristor is fully closed. This core operates with 72 turns and an inductance of approximately 4 mh at high frequencies. FIG. 2. Comparison of the simulated (a) and experimental (b) pulse power section with an open circuit secondary, with (c) a zoomed in portion of notated portion of (b). The voltage waveforms displayed are across each element. thyristor, magnetic switch, capacitor, transformer. Measurement points for each waveform correspond to the shapes in Fig. 1. Time t = 0 marks the trigger input driving the top H-bridge in Fig. 1. B. Thyristor switching The thyristor used is rated for 10 ka maximum current, 4 kv hold-off voltage, and is made by Silicon Power, part number CSSC14N40A10. It is driven from a full H-bridge through a small pulse transformer labeled T1 in Fig. 1. The transformer and the diodes, rated at 1 A and 100 V blocking voltage, serve to isolate the high voltage portion of the circuit from the control section. On the secondary damping resistors R3 and R4 with values on the order of 100 Ω are used to dampen oscillations on the gate. An H-Bridge is used to drive the gate of the thyristor negative in an attempt to pull charge

3 Barnett et al. Rev. Sci. Instrum. 86, (2015) carriers out of the device to achieve a faster recovery time. However, in the case of this specific thyristor, this was found to have only a minimal effect toward speeding up the recovery of the device. The thyristor is a current controlled device and requires a low impedance path from gate to cathode to prevent erroneous switching. This is achieved by the gate to cathode resistor, labeled R7 in Fig. 1. The switching turn-on time depends heavily on how the gate is driven and the amount of gate to cathode resistance. A smaller resistance value decreases the recovery time (turn off time) of the thyristor, but more current is required to maintain the same turn on time. For the present design, a relatively small value of 0.5 Ω was chosen as the gate to cathode resistor to achieve a recovery of <100 µs. With this resistor being small and the current from the H- Bridge internally limited to 4 A output current, the thyristor turn-on time is approximately 200 ns. Turn-on times have been observed as low as 190 ns to over 500 ns for different gate currents. The longer turn-on times need to be avoided. Since, the magnetic switch was designed to saturate at approximately 300 ns. Hence, the maximum thyristor switching time needs to be smaller than 300 ns, otherwise, the transformer, T3, would not see the full voltage drop. C. Command charging In the present design, command charging of the main trigger capacitor C5, is implemented for several reasons. The first is to decrease charge time of the capacitor, the second is to reset the cores used for the pulse transformer and magnetic switch, and third is to aid in the recovery of the thyristor allowing for faster operation. Before command charging was utilized, the trigger capacitor was charged through a large series resistance of 75 kω. This resistance allows rep-rates up to 1 khz, using a 1 nf capacitance value for C5. The downfall in this approach is the large power dissipation in the series resistance during operation which requires a 250 W power supply to achieve continuous operation. In addition, the charge time of the trigger capacitor severely limits the frequency of operation due to the charge time of almost 800 µs. A single MOSFET is utilized to control the command charging cycle, labeled M1 in Figure 1. The MOSFET is an IXTL2N450 rated for 8 A at 4.5 kv pulsed. The signal to the MOSFET gate is provided by the H-bridge to supply both a positive and negative pulse which insures that the device is turned off before the thyristor is switched. The H-bridge is isolated from the MOSFET by a 1:1:1 pulse transformer, with both negative polarity windings tied together to form a 1:1 center tapped transformer, see T2 in Fig. 1. The transformer is center tapped on the source of the MOSFET, with the secondary windings on the gate. This essentially shorts the gate-source, only allowing the voltage that develops on the primary to be across the secondary, insuring that the gatesource voltage cannot exceed that placed on the primary (24 V). This is also used with D3 and D4, rated at 1 A and 100 V and 100 Ω damping resistors R5 and R6 to prevent oscillations created by the inductance from the transformer and the capacitance from the gate. To allow the command charge operation to be controlled from a single input signal, a FIG. 3. Command charging of the trigger capacitor: voltage across capacitor C5. simple RC circuit is established with logic ICs to switch the signals on the direction pins of the H-bridge at a predetermined time. For testing purposes, the circuit is setup to switch at 60 µs, which is right before T2 saturates. Fig. 3 shows the trigger capacitor being command charged. During operation, a large reservoir capacitor C3, 500 nf 1 µf, is charged first. Then, once this capacitor has reached the desired charging voltage, M1 is triggered with a 70 µs pulse to give the MOSFET a 60 µs positive and 10 µs negative pulse; the thyristor is then triggered. The large charging current, flowing from C3 through M1 to C5, enables an increase of the trigger capacitance, C5, to a few nanofarads. In testing, it is observed that the current integral is now sufficiently large to saturate the magnetic switch, L1, in the direction opposite to C5 discharging. It should be noted that the associated large dv change causes the thyristor to switch erroneously past 500 Hz operation. To resolve this issue, the resistance of the current limiting resistors R10 and R8 were initially 300 Ω but were increased to 1000 Ω, and an RC snubber R9 and C4, values 8 Ω and 1 nf, respectively, was added to keep the dv in check, allowing operation up to 2 khz with no further problems. Past 2 khz, the required snubber values consume more energy than the available energy, making testing past 2 khz impractical. Both the main pulse transformer T3 and the magnetic switch L1 need to be reset for every pulse. The required voltseconds (V-s) to reset a core, if completely saturated, is the area cross section times the delta saturation flux density; because of this, the required volt-seconds for the magnetic switch are small compared to that required for the transformer core at 433 V µs. The delivered volt-seconds are derived from measured voltage data using equation one; t 1 and t 2 are chosen to be the area just before and after the pulse, λ = t2 t1 v (t). (1) Before command charging was implemented, charging only provided 56 V µs for resetting the transformer which is lower than the required amount of volt-seconds to reset the core for every pulse. With command charging, the charge time is under

4 Barnett et al. Rev. Sci. Instrum. 86, (2015) 60 µs for 2 nf and provides 5500 V µs which is more than an order of magnitude higher than what is needed to reset the core completely. Overall, command charging allows for a faster recharge time but also lowers the thyristor recovery time to below 100 µs. With the snubber circuit added, the circuit consumes 58 mj/shot, allowing continuous operation of 1 khz with a 60 W power supply. III. SYSTEM RESULTS In this section, the system s operation results are discussed and compared to a simulated model. The simulation is helpful initially to verify circuit components and to optimize them accordingly. Operation at different repetition frequencies will be detailed for burst mode with up to 1000 pulses. Most testing is done with the secondary of the pulse transformer left open. This was done to initially see the pulse at an open circuit voltage because the system was originally designed to trigger spark gaps. The spark gaps used are covered in more detail in Ref. 8. A. Numerical simulation A simulation of the high voltage section, bold traces in Fig. 1, is carried out using P-Spice. The simulation of the circuit is done to verify the main characteristics of the pulse power section. This yields insight on the overshoot of the transformer and voltage decay on the capacitor. The simulation is a basic model with the magnetic switch inductance being reduced in a prescribed fashion after the thyristor voltage collapses to zero. In Fig. 2, the initial switching command is sent at t = 0, and the magnetic switching takes place at 0.95 µs. The model also neglects detailed thyristor switching characteristics; this is noticeable by the slope difference between the red, upside down triangle waveform in the two graphs. B. Experimental results The following experimental results are obtained by the previously described setup and components. The waveforms are measured with an open secondary to determine the maximum magnitude and the risetime. Total available power is not a concern since the trigger generator triggers a trigatron spark gap where the high voltage pulse needs a large magnitude and dv for field distortion. The high voltage section is charged using a single 60 W power supply. The single 60 W power supply is only able to provide enough energy for continuous operation at 1 khz. At frequencies above this, voltage droop is very evident. The voltage droop at 2 khz is almost 700 V from the first to the 25th shot across the transformer. In Figure 4, the voltage droop is visibly evident at 2 khz, along with the stability and repeatability of operation at 1 khz. At 1 khz operation, only 25 waveforms have been plotted out of a 1000 shot pulse train for comparison to the 2 khz waveforms. Because the voltage bus droops 20 V/shot at 2 khz, resulting in over 500 V total, only FIG. 4. Overlaid voltage signals across T3 primary. (a) 1 khz operation, 25 traces, every 40th pulse; (b) 2 khz operation, first 25 shots only. 25 shots are done at 2 khz. The resulting jitter measurements are made at 500 Hz, 1 khz, and 2 khz, yielding <1 ns, 4 ns, and >6 ns, respectively. The 500 Hz and 1 khz are calculated on a 1000 shot pulse train while the 2 khz pulse train is from a 25 burst. The jitter starts to increase with the voltage droop as the volt-seconds across the cores change from shot to shot. This change affects the delay and the time it takes the cores to saturate, see Ref. 8 for a further discussion on the increase of jitter from voltage differential across magnetic switches. With this change in volt-seconds, the start of the pulse is not at the same point on the B-H curve, resulting in variation from pulse to pulse. While most testing is done with the T3 secondary open, tests are done on an uncharged three electrode spark gap using a trigatron switch; Fig. 5 is one of the waveforms generated. The pressure is adjusted so the gap breaks at the peak voltage that is generated from a 1:1 transformer (19 turns on both primary and secondary of T3). In this instance, the trigatron switch breaks at 5.8 kv and quickly drops to 0 V and oscillates FIG. 5. Spark gap test with an atmospheric pressure triggered spark gap, gap breaking down, no breakdown (gap pressure intentionally increased).

5 Barnett et al. Rev. Sci. Instrum. 86, (2015) around zero for a few cycles. For charged three electrode gaps, a step up transformer can be used to increase the secondary voltage to a level that reliably triggers the spark gap. IV. CONCLUSION A solid-state HV trigger pulser has been demonstrated to operate reliably up to 2 khz with a risetime of 30 ns, with relatively low jitter. Being battery operated and optically controlled ensures portability and system isolation. The system reliably triggers spark gaps switching up to 40 kv. The overall operation can be improved using this design by stacking the high voltage section enabling a voltage increase with relatively the same risetime. Although a 60 W power supply is used to maintain a small package size, stable operation past 1 khz can be achieved by utilizing a larger power supply to maintain the 4 kv bus. This would provide stable and continuous operation up to the limit of the power supply. ACKNOWLEDGMENTS This material is based upon work supported by the Test Resource Management Center (TRMC) Test and Evaluation/Science and Technology (T&E/S&T) Program. This project is funded by the T&E/S&T Program through the U.S. Army Program Executive Office for Simulation, Training, and Instrumentation (PEO STRI) under Contract No. W900KK- 12-C Any opinions, findings and conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the TRMC T&E/S&T Program and/or PEO STRI. Distribution Statement A. Approved for Public Release Distribution is Unlimited. 1 J. Weihua, N. Oshima, T. Yokoo, K. Yatsui, K. Takayama, M. Wake, N. Shimizu, and A. Tokuchi, Development of repetitive pulsed power generators using power semiconductor devices, in IEEE Conference on Pulsed Power (IEEE, 2005), pp V. P. Gubanov, S. D. Korovin, I. V. Pegel, A. M. Roitman, V. V. Rostov, and A. S. Stepchenko, Compact 1000 pps high-voltage nanosecond pulse generator, IEEE Trans. Plasma Sci. 25, 258 (1997). 3 L. Yi, L. Fuchang, F. Xibo, Z. Qin, Z. Heqing, L. Hua, and D. Ling, Design and construction of a trigger generator based on pulse transformer for spark gap switch, IEEE Trans. Plasma Sci. 39, 3378 (2011). 4 X. Fan and J. Liu, A 70 kv solid-state high voltage pulse generator based on saturable pulse transformer, Rev. Sci. Instrum. 85, (2014). 5 J. Weihua et al., Compact solid-state switched pulsed power and its applications, IEEE Proc. 92, 1180 (2004). 6 J. Lin, J. Yang, J. Zhang, and X. Chen, An all solid-state high-voltage ns trigger generator based on magnetic pulse compression and transmission line transformer, Rev. Sci. Instrum. 84, (2013). 7 P. M. Kelly et al., Operation of a compact, modular pulse-forming network based Marx generator used to drive a high-power microwave source at 500 Hz repetition rate, Rev. Sci. Instrum. (to be published). 8 J. Lin, J. Zhang, J. Yang, H. Zhang, Y. Qiu, and X. Yang, Jitter characteristic of series magnetic pulse compressor employed in ns trigger generator, Rev. Sci. Instrum. 85, (2014).