"A century after Erwin Marx, the field is more interesting than it has ever been. The next decade will reshape what 'pulser' means at every voltage class."
Pulsed power is, on the surface, a mature field. The basic architectures are 50 to 100 years old. The fundamental physics has been understood since the radar era. Practitioners trained in the 1980s can read modern pulser data sheets without confusion.
Look closer and the picture is different. The switch landscape has shifted twice in the last decade and is still moving. The application set has tripled since 2010 and continues to grow. The diagnostic tools available on a 2025 pulsed-power bench would have been research equipment in 2005. The commercial pricing of devices that were research curiosities in 2015 now sits in catalog territory. Pulsed power is mature in its principles and immature in its products: the principles drive a continuing reshuffling of what the products can do.
This chapter covers what to expect over the next five years. Some of what follows is happening now. Some is in laboratories with a clear path to commercialization. Some is speculative. The line between these categories is fuzzy and gets fuzzier the further you look. Where speculation begins, this chapter says so.
Silicon carbide MOSFETs at 1.7 kV per die are commercially common as of this writing. 3.3 kV SiC MOSFETs are in production from Wolfspeed, Mitsubishi, and a few other vendors. Research has demonstrated 10 kV and 15 kV SiC MOSFETs in laboratory yields, with commercial production expected within three to five years.
The implication for pulser design is substantial. A 10 kV pulser today uses a stack of eight to ten 1.2 kV silicon MOSFETs with all the gate-drive isolation and balancing complexity that implies. A 10 kV SiC MOSFET in single-die form replaces that stack with one device, with simpler gate drive and faster switching. Cost per device will be higher than the silicon stack for some years, but reliability and switching performance will be substantially better.
The same scaling applies to GaN. Commercial 650 V GaN HEMTs are joined by 900 V parts (commercially available since 2023) and 1200 V parts (announced, low-volume production). GaN's ceiling is set by epitaxial layer thickness and substrate engineering, both of which are improving steadily. A 2 kV GaN HEMT in commercial production by 2028 is plausible.
The combined effect is that the half-bridge MOSFET pulser at 30 kV becomes practical with SiC stacking, and the half-bridge GaN pulser at 2 kV becomes a commodity. Pulser BOMs simplify, reliability climbs, and the historical reasons for using IGBTs in the kilovolt class largely disappear.
DSRDs have been research-class devices since the 1980s. The combination of Russian device-physics expertise (FID Technology, Megaimpulse) and Western packaging engineering has produced a small commercial DSRD market over the last decade.
Two trends are accelerating this. First, the underlying device physics is being implemented in new geometries that lend themselves to higher repetition rates and longer lifetime. Second, the integration with semiconductor primary switches (SiC MOSFETs feeding DSRDs) is producing pulsers with sub-nanosecond rise times at tens of kilovolts in package sizes appropriate for benchtop use, not research-only racks.
Within five years, expect commercial DSRD-based pulsers in the 5 to 30 kV class with rise times below 500 ps to be available from multiple vendors at price points appropriate for industrial laboratory use. The instrument bench gains a new performance class that simply does not exist now in commodity form.
PCSS devices have been a research and specialty market for two decades. The combination of GaAs and SiC PCSS, together with integrated femtosecond fiber laser sources, is producing increasingly practical instruments.
The commercialization barriers have been threefold: device lifetime (high-field operation degrades the bulk material), trigger laser cost and reliability, and thermal management. Each is improving at a measurable rate. SiC PCSS in particular has shown lifetimes in the millions of pulses at amplitudes that were previously unattainable.
Within a decade, PCSS-based pulsers in the picosecond rise-time class at multi-kilovolt amplitudes should be commercial products, used in ultra-wideband radar, electromagnetic-effects testing, and physics experiments needing sub-nanosecond timing. They will not replace solid-state half-bridge designs in mainstream applications, but they will own the picosecond niche cleanly.
The physical packaging of high-power pulsers is constrained by traditional manufacturing: machined metal enclosures, hand-wound transformers, point-to-point wiring of the high-voltage section. Additive manufacturing is starting to change this.
3D-printed copper now reaches conductivity within a few percent of bulk copper. 3D-printed dielectrics can be tuned in their printed-density profile to control field gradients. The combination opens a path to pulse-forming networks and pulse transformers that are designed and built as single integrated units rather than assembled from discrete components.
Practical impact is several years out. The first commercial product to use 3D-printed pulse formers will probably be a specialty research-grade pulser, not a commodity bench instrument. Within a decade, the technology should reach mainstream pulser packaging in some product lines.
Modern pulsers ship with microcontrollers that already do real-time control loop work (the PCX-7500's PID is one example). The next generation extends this to machine-learning-assisted control: closed-loop adjustment of pulse parameters based on observed load behavior, predictive failure detection from sensor patterns, and automatic compensation for component drift over device lifetime.
The hard part is not the AI: PID and adaptive control already do this work in many applications. The hard part is the diagnostic instrumentation that feeds the AI: high-bandwidth current and voltage sensors integrated into the pulser, temperature sensors at every dissipating element, and the data-logging infrastructure to retain enough history for the AI to find patterns.
Modern BNC-DEI products have begun this trajectory: USB and Ethernet interfaces, internal data logging, remote diagnostic capability. Over the next five years, expect pulsers that report their own health, predict their own end-of-life, and adjust their operating parameters in response to load characterization that happens automatically. Some of this is honest engineering progress. Some of it is marketing surface on top of what is already routine PID control. Practitioners will need to read product specs carefully to distinguish.
Several standards bodies are working on pulsed-power-relevant material that will shape the next decade of commercial deployment.
IEEE PCSC (Pulsed Component Standards Committee) develops standards for pulsed-power components: capacitor specifications, switch ratings, transformer test methods. Active work on solid-state Marx test methods and on SiC pulse-rated component characterization.
IEC 61000-4-25 addresses HPM testing and protection. The current revision is being updated to reflect the broader threat landscape (drone-mounted HPM systems, urban-scale EMP studies).
MIL-STD-188-125 (the HEMP protection standard for ground-based facilities) is being revisited for the first time in years, to update the threat model and the protection requirements.
ISO 11161 and related industrial-machinery standards are folding in pulsed-power equipment as those technologies move into industrial production lines. The pulser integrated into a food-processing plant or a 3D-printing system needs to meet the same machinery safety standards as any other equipment in those environments.
The cumulative effect of standards activity is that "the rules" for selling and deploying pulsers will be more formalized in 2030 than they are now. This is generally good for users (more predictable interoperation, clearer safety) and modestly painful for manufacturers (more certification work, more documentation overhead).
A speculative section. Quantum computers based on superconducting qubits use pulse-shaped microwave drives at millikelvin temperatures. The pulse-control hardware (arbitrary waveform generators, RF mixers, fast DACs) is closer to high-speed RF instrumentation than to high-power pulser engineering. But the underlying architectural challenges (precise timing, low jitter, multi-channel synchronization) overlap.
If superconducting quantum computing reaches commercial scale (a real if), the pulse-control hardware will be a specialty market sized in the hundreds of millions of dollars. Pulsed-power expertise transfers awkwardly: the voltages are tiny, the currents are tiny, the rise times are picoseconds. But the timing-distribution and trigger-synchronization expertise that pulsed-power instrumentation companies have transfers cleanly. BNC's precision timing line is well-positioned if this market develops.
The hype cycle in pulsed power, like any field, runs ahead of reality. Some of what gets announced as "transforming the field" is genuine progress. Some is engineering improvement at a new spec point. Some is marketing.
A short, opinionated guide to where the hype is and is not justified:
Genuine inflection. Wide-bandgap commercialization. The replacement of silicon stacks by single-die SiC at the kilovolt level changes what pulsers cost and how they fail. This is real and continues.
Real progress, oversold. AI-assisted control. PID has been doing closed-loop control of pulser outputs for decades. The improvements over the next five years are real but incremental. The "AI revolution in pulsed power" framing is mostly marketing.
Honestly uncertain. PCSS commercialization. The technology is improving and a commercial market is plausible. The timing and the price-performance trajectory are not yet predictable.
Mostly hype. Quantum-computing-related pulsed-power. The connection to traditional pulsed-power technology is weaker than the hype suggests. The pulse-control hardware for qubits is more like RF test equipment than like high-power pulsers.
Speculative but interesting. Additive manufacturing of pulse formers. The technology exists, the cost is dropping, and the integration story is compelling. Whether this becomes a meaningful market within ten years is genuinely unclear.
Through all the technology evolution, some things do not change.
The fundamentals of Chapter 2 are still the fundamentals. Voltage and current pulsers, capacitive and inductive and resistive loads, the C × V² × F power equation, the rise-time-versus-bandwidth relationship. None of this is going to change. The student in 2035 will learn the same equations the student in 2025 learns.
The pitfalls of Chapter 3 are still the pitfalls. Disconnect-under-load events will continue to damage pulsers. Wrong-cable substitutions will continue to ruin pulse fidelity. Miller turn-on will continue to bite gate-drive engineers. The fixes will continue to be the geometric and electrical remedies described in that chapter.
The safety rules of Chapter 4 are still the safety rules. Stored energy is dangerous. Floating instruments improperly is dangerous. Capacitor banks need bleeder resistors and confirmed-discharge protocols. None of this depends on which switch technology is in use.
The trade-offs in topology choice (Chapter 6) and switch selection (Chapter 5) shift as the technology evolves, but the questions stay the same: what is the voltage class, what is the rise time, what is the duty cycle, what does the application need, what does the budget allow. The answers in 2030 will differ in detail from the answers in 2025. The questions will be the same.
Erwin Marx demonstrated his cascade in 1924. In 2024 the field celebrated a centennial that almost no one outside of pulsed power noticed. The Marx generator is still the canonical answer to "how do I get to a million volts in my lab," and the topology with its 1924 architecture (charged in parallel, switched into series) is still in commercial pulsers today, with semiconductor switches replacing the spark gaps.
This is the underlying truth of pulsed power. The field accumulates technologies rather than replacing them. Spark gaps are still in service. Thyratrons are still in production. Silicon MOSFETs are not going away. SiC is the new normal at kilovolt scale, and GaN owns sub-nanosecond rise times. The pulser engineer of 2030 will have all of these tools in the toolbox, plus DSRDs and PCSS and whatever else has come along by then. Each technology has a niche. Each technology has a lineage. None of the old ones disappear when the new ones arrive.
What this means for someone entering the field today is that the investment in fundamentals pays back over decades. The half-bridge MOSFET pulser you understand today will still be selling in 2040. The energy-storage equations you learned this week will still be the same equations in any future textbook. The diagnostic tools improve, the switch technologies extend, the applications widen, but the engineering core is stable.
That is the right way to read this book. Not as a snapshot of the field in 2025 (though it is also that), but as a framework that will let you understand the next thirty years of pulsed power as they unfold.
A century after Erwin Marx, it is still good work. It is still useful work. There is still a lot of it to do.
The most likely commercial trajectory for SiC MOSFETs over the next five years is: a. Replacement by GaN at all voltage classes. b. Voltage scaling toward 10 kV and higher single-die ratings, simplifying high-voltage pulser stacks. c. Cost increases that price them out of commercial use. d. Replacement by silicon as silicon technology catches up.
The PCSS technology trajectory is most likely to produce, within ten years: a. Replacement of all benchtop pulsers. b. Commercial picosecond-rise pulsers in the multi-kilovolt class for ultra-wideband radar and similar specialty applications. c. No commercial products, only research devices. d. Pulsers with no measurable rise time at any amplitude.
The "AI-assisted pulse shaping" trend is best described as: a. A genuine revolution that obsoletes traditional control techniques. b. Real but incremental improvement over PID and adaptive control that has been used in pulsers for decades, with marketing oversell. c. Pure hype with no real engineering content. d. A privacy concern.
According to this chapter, the technology trajectory that has the strongest near-term commercial impact is: a. Additive manufacturing of pulse formers. b. Quantum-computing-related pulsed-power. c. Wide-bandgap (SiC and GaN) voltage and switching-speed scaling. d. AI-assisted predictive maintenance.
The standards body developing test methods for pulsed-power components is: a. ISO TC 22. b. IEEE PCSC (Pulsed Component Standards Committee). c. SAE Aerospace. d. ANSI Z39.
The Marx generator's continued relevance after a century reflects: a. The conservatism of the pulsed-power field. b. That the topology accumulates rather than replaces: each new switch technology can be plugged into the Marx architecture, extending its life. c. Lack of better alternatives. d. Industry standards prohibiting newer topologies.
A reader who learns the fundamentals of Chapter 2 thoroughly should expect those fundamentals to: a. Become obsolete within five years. b. Apply substantively unchanged for at least the next several decades, regardless of which switch technologies dominate. c. Apply only to legacy products. d. Apply to academic problems but not industrial ones.
The same seven questions, graded instantly with your score saved on this device.
Answer key at end of book.
IEEE Pulsed Power Conference Proceedings, biennial. The single best window into the state of pulsed-power research.
Wolfspeed and EPC roadmap publications on SiC and GaN scaling. Vendor roadmaps are not always reliable but they are the most current public information on what the next two to three years will look like.
Grekhov, I.V. and Ivanov, A.M. Picosecond Power Electronics. Specialty area, but a useful reference on the DSRD and SOS device-physics trajectory.
MIL-STD-188-125 and related standards documents from the US military's HEMP protection program. The most authoritative public information on what the threat-protection landscape requires.
IEEE Std 1366 on power-system reliability (somewhat tangential but useful for the broader picture).
Application notes from FID Technology, Eagle Harbor Technologies, and Megaimpulse on their current product roadmaps. These are the vendors driving the specialty-pulser frontier.
Bluhm, H. Pulsed Power Systems. Springer, 2006. For perspective, a book published nearly twenty years ago that already anticipated most of what this chapter describes. The fundamentals do not change.
End of Chapter 11.
The appendices follow.