A company called Science Research Labs (SRL) built a solid state DPF back in 1992 and upgraded it for years to reach 230 kA. The pulse power alone cost $1M. The operating voltage was 8 kV at most. It used saturatable magnetics. The problem is using solid state in series. SRL mastered the technology at the 230 kA level. A 10X scale up would be daunting at best. A 1 MA PF typically operates at 40 kV or more so that’s many switches in series and parallel. For a 6 kV switch unit, you would want to have 8 switches in series. With a 100 A switch you need 20,000 switches groups. A 2 MA PF with no room for upgrades in current requires 160,000 switches. Solid state if built correctly can last for 10^10 shots.

Spark gaps and railgaps can achieve both the rise time, operating voltage and current carrying capability to drive most pulse power devices. Rail gaps can operate up to 20 Hz at ~100 kV and 0.75 MA per switch if done properly. Switch life time is a problem for spark gaps and railgaps with ~10^7 shots at most.

Solid state technology has come a long way. I remain hopeful that it will work at the currents and voltages desired someday. The problem I see is managing several thousand switches at once in this regime. I’ve looked into it before and the potential down time for the pulse power and cost is a real problem. Even if the switches are managed well on the forward going pulse the pinch voltage can be many times the charge voltage and how is the reversal current going to be dealt with. SRL has worked on the technology since 1992 but they have updated when possible and they are using circa 2005 solid state technology in the last paper. The system operates at up 80 Hz at 230 kA. The switches they use don’t require low jitter because of the saturatable magnetics.

A fast switch with 1 ns jitter is not necessary for a PF. The typical current rise time is ~ 1us. A few percent jitter is tolerable on a plasma focus. The PF devices I work with have ~5% switch jitter and they work well. In many ways, the shape of the current pulse does not matter for a PF. We use the sinusoidal shape because the pulse power is cheap. A linear current ramp or a fast rise to a flat top current should work as well. Other pulse power architectures can give these shapes but they cost more than folks are willing to pay. Solid state is typically in the regime of more than folks wish to pay. Another issue with solid state for this application is the radiation produced by the PF. One the research scale the switches won’t be affected by x-ray or neutrons. However, on the power plant scale, the solid state would be bothered by radiation. You can make the components so called, rad hard, but that costs even more.

I’m sure this all sounds like I’m dumping on solid state, but these are the concerns of people building the pulse power systems. The promise of solid state pulse power has not escaped anyone. The Army, Air Force, DOE, and many other are pouring money into solid state pulse power technology for applications like radar and microwave generation at high power. The problems appear to be manufacturing limitations at the moment but there are serious physics limitations to solid state technology that are not easily overcome. The problem for LPP is many of these applications require high voltage and only modest currents 100 A to 1 kA. The dream of a 10 kA switch at high voltage isn’t interesting to those funding the technology. DTRA is probably the only US agency that was interested but it faded when physics slapped them in the face. Last I talked to someone who builds both solid state and gas switch technology professionally, he was confident that gas switch technology was the only approach to >100 kA systems with modest repetition rate (~10 Hz) in the 1 us rise time regime for years if not decades.

The rail gap is a gas switch like a spark gap. Rather than use two hemispherical or two cylindrical shaped electrodes, two rails are parallel to each other. A third electrode (trigger) lines between them but out of plane. Drawings are on-line if you look up rail gap. The switch requires a fast rising trigger pulse (4 kV/ns) to operate correctly. Basically, the switch is a multichannel spark gap. With the fast trigger you can set up many channels to carry the current. The arcs jump around the electrode on different shots so you can get reasonable lifetime and low inductance (~20 nH). A spark gap can tolerate a slower trigger as only one arc is produced. The inductance of a spark gap can be quite large ranging from 50 nH to 200 nH per switch. The rail gaps tend to carry higher current than most spark gaps but operate at similar voltages. The rail gap can operate in a flowing gas mode that keeps the switch relatively clean. I use four rail gaps in a PF at the 0.25 MA scale and they are great switches. The lifetime of the switches is ~10^7 shots but you can extend it using refractory electrodes instead of brass which is the standard. The gaps are for sale on line for ~$20K per switch. In a 2 MA machine I would guess you need twelve switches at a minimum. The trigger does cost and it needs some thought but the technology exists to make a 4 kV/ns pulse without too much trouble. I have operated my rail gaps for thousands of shots without cleaning, replacement or problems.

The other option is the thyratron. Thyratons are wonderful switches for ~ 10 kA, 40 kV applications up to 1 kHz. I use a pair in an experiment. They are the easiest gas switch to trigger that I know of. Particle accelerators wouldn’t exist without them. They are still the main stay of SLAC and other large systems so the infrastructure to produce, maintain and operate is in place. For the lab scale PF, the thyratron is costly (~$4000 for the switch as described) and they require auxiliary equipment like heaters which make the switch more like $7500. They have low jitter (~10ns) with almost no effort and can operate with ~2ns jitter for the dedicated pulse power master. The problem is these switches aren’t good with reversal and they carry such low currents relative to the desired peak current. I priced a thyratron based PF at the 200 kA level and it cost nearly $250K in parts. One might argue that $2.5M is not so bad in the context of a 2 MA power plant and buying in bulk helps reduce the unit part cost. I see this an an option moving forward. Switch lifetime is ~10^8 shots so you can probably make a power plant work with ~200 switches. I believe SLAC operates thousands of switches. Again, know how is there.

In the end cost, lifetime, reliability and the operating characteristics will determine the switch choices. If the FoFu power plant operates at ~10 Hz I find it hard to believe that solid state technology would be used. If the plant must operate at 100 Hz or more, get out the check books and prepare for a costly technology development cycle.

andrewmdodson wrote: My understanding is that the switch only has to close extremely rapidly… the capacitor bank discharges completely into the reactor, expending functionally all the current and then the system resets for the next pulse…

Why does there need to be reverse conducting? Are you using the same electrical pathway to recharge the cap bank?! I thought that no power had yet been extracted from the DPF?

The switch needs to close rapidly but it doesn’t have to be 1 ns or so. The bank discharges completely but the PF is an underdamped ringing circuit. As a result some current flows back into the cap bank. Some solid state switches can’t handle reverse conduction from the ring. I know SCR type switches turn off when the voltage and current reverse. So what happens to the current that is trying to flow back into the bank? You have created an opening switch device. Opening switch devices are well known for pulse compression applications that need to generate high voltage in a fast pulse. The PF circuit is a charged inductor without an outlet when the switch opens after a forward going pulse. Inductors like to keep current moving so they generate voltage at the expense of stored magnetic energy. A fast closing switch can more than double the voltage left in the PF. The SRL group had the same problem so they implemented a massive collection of fast, high voltage diodes to channel the reverse current safely away from the switches to ground. Most gas switches don’t care about the direction the current flows so they happily ring until the resistance in the circuit converts the energy to heat or the system voltage drops below the sustaining voltage (<1 kV for most gas switches).

In an ideal PF, the reverse current would be small but it doesn’t always work out that way. In a good pinch att 0.25 MA I still see a negative ring of ~70 kA. That is a small fraction of the forward pulse but it could easily take out solid state components if the voltage rings up.

This assumes the voltage at pinch time doesn’t take out the solid state system. When a plasma focus ‘pinches’ the plasma to achieve fusion conditions, the impedance in the circuit increases dramatically. I can’t speak for FoFu-1 specifically, but a 1 MA machine typically generates ~250 kV at pinch time. The voltage is divided back to the bank by series impedance relationships. If the switches are a moderate to large impedance in the circuit which isn’t a bad assumption, they will see a substantial voltage drop. Suddenly you have a 40 kV circuit with a positive voltage discharging to near zero and a 250 kV spike appears at the the pinch. About 50% of the voltage will appear across the switches. Can a 40 kV hold off switch in forward bias deal with a 125 kV voltage spike in reverse bias? I don’t know for sure but I have my doubts.

Don’t hate the solid state guys, hate the universe. I’m pretty sure the solid state guys would love for the rules to be more favorable. 🙂