Steps towards FF-1 Feasibility

[delt0r]

As i said, magnetic storage could be something to consider *if* you have a fast opening switch with high stand off. Which we *don’t* have. Magnetics can be fast if the opening switch is fast which also implies high voltages.

[asymmetric_implosion]

If you charge an inductor with the necessary energy (~100 kJ) at the drive current (~2MA) you need 50 nH. That is much less than I thought. I was thinking something like 1 uH. This adds to my argument not to do math in my head. It would be difficult to couple a 1 uH load to the <100 nH pinch but 50 nH has some hope. This is one problem with plasma opening switches. You would generate a high voltage spike at switch opening well in excess of the necessary drive voltage of ~40 kV. You might be saved by the fact that the gas will breakdown and dissipate the voltage before the switch is overwhelmed. Still at 100 kV switch is required and if it's an opening switch you are doubly screwed because it has to stay closed before suddenly opening. Solid state is the only answer. Diamond will not work because you cannot keep it on. SiC has some hope but a 100 kV switch is a challenge. Turn off is far worse than turn on. Another question is how to charge the inductor with 2 MA to start with. You would still require a cap bank because I don't know of a slow pulse 2 MA source to charge the inductor. I guess a fly wheel could do it for low rep rate operation like they used to do on Tokameks. I guess you could charge many units in parallel but you wouldn't want to see the source inductance get too low because you couldn't match to the load.

[delt0r]

You seem to be having a completely different conversation. POS are crap…which is why i keep saying IF you have a opening switch that we currently DON’T have. Of course it can’t work now…THAT is what i said like every time. I also said the current would be a challenge without superconductors.

But a compulsator with some magic solid state switch it would really work without superconductors. With potentially higher energy density than caps. Even higher than fast water caps.

[asymmetric_implosion]

Sorry if you think I’m having a different conversation.

I was just doing the math on storing energy in an inductor to drive the pinch. My employer has a patent on inductive energy storage system for powering arcs. You basically use a DC supply to charge the inductor and then close a switch. You get the voltage needed to breakdown the arc and drive the current for like 100 us. That is my very limited background with inductive energy store (mainly from lunch room chats) so I was taking it from that perspective.

POS are crap above 1MA but some do work at the <1MA level. A machine at NRL, Hawk, uses a POS with reasonable success instead of a water line. They drive 700 kA in 200 ns or less.

I’m badly out of touch with compulsator tech, but I thought you need to brake the compulsator to extract energy from it. I take your comment to mean it is not true. I was also under the impression that 2 MA was beyond a compulsator with fast pulses like 10 us. Can you recommend any reading material so I can get caught up?

I’m not sure energy density is the biggest problem in pinch type pulse power systems. I think folks would use smaller pulse power if it was available but I don’t think anyone in the pinch community sees that as a driving area of research. Most of the problems seem to focus on faster rise times and load design.

[delt0r]

Rail guns and emag guns for next gen tanks are driving compulsator work. Self excited, something on the order or MA is achievable. Heating is an issue. IEEE is where i have found the literature. Don’t want to sound too harsh, but a patent on inductive energy storage for a arc is ridiculous.

Energy density is not really the problem, but switches are and physically large caps means its hard to keep inductance down without going to a 2 stage design or whatever.

Personally I see a commercial plant going for LVA, since they can be designed to add in both voltage and current.

[asymmetric_implosion]

Well, the patent is more complicated than just an arc. That is the simple story. I agree it is a bit overkill but it was granted. To be fair it has to operate in conditions that are not agreeable to the bench top so that is really the innovation.

The size of the caps haven’t hampered the inductance of PF pulse power. A typical cap has a low inductance of 20-50 nH per unit with several units in parallel (~10). The switches are typically the big inductance hogs at 20-300 nH per unit but they are that value regardless of the cap geometry in any practical case. Some large PF systems like DENA have a bank inductance of <5 nH by massively parallel architecture. My three PF devices operate at ~20 nH in the bank with three different types of switches and three very different architectures. In our sources, the switches and caps typically contribute only 5-7 nH (lots of little things in parallel). Most of the inductance comes from bus bars near the load that must accommodate the transition from air to vacuum. It's not universally accepted but the idea has floated around the PF community that you need some finite inductance in the bank to make the PF work at optimum. The value of inductance goes up with current. It does not agree with power transfer theorems but to accommodate certain plasma conditions to confine and sustain the pinch. The inductance relationship I see thrown around most frequently is the pinch inductance should match the axial phase inductance. The bank inductance should match the axial phase inductance. I don't know if this is the true optimum but it shows up consistently over the years since the 1970's. The pinch inductance is going up with current and in most cases the axial phase inductance is going up with current. Therefore, the bank inductance is going up to sustain the ratio. I can tell you this much, the PF-1000 machine has a low source inductance at 1.8 MA and it performs as well as FoFu-1 at 900 kA. By any pinch scaling law, PF-1000 should be up by more than 10X over FoFu-1.

The LVA is a linear voltage adder?? The high current pulse power equivalent is the linear transformer driver (might be same thing with different name). Modules that operate at 1 Hz are currently available with 100 kV output at 1 MA into a matched load. You might need 3-5 of these units in parallel. Sandia National Lab is handing them out to universities for numerous studies. I know U. Michigan and U.C. S.D. have modules. The key questions are the scale up of the PF physics at ~3 MA with source inductance. The LTD is a low inductance source so it might not be able to drive the additional inductance that might be needed to optimize the pinch without significant loss. Time will tell.

[delt0r]

LVA is a LTD. Same thing different name. Matching does not seem hard. They can infact match to more or less all sorts of different loads as well as anything else can. Pulse times from micro seconds to ns are also fairly typical. They however are not that cheap. The ferrite tends to be the bulk of the cost. More so if you go for metglas. However it does offer a way to add smaller current and voltage modules that could use solid state switching to more or less the MA levels and MV levels (add a transmission line transformer for more design flexibility too). For a commercial operation the extra capitol cost could well be justified.

[asymmetric_implosion]

The one bummer with the LVA is the switch count. You have many switches and if their lifetime is too short you have a huge replacement cost. Last I saw a Sandia LTD it had something like 40 switches in it. You need three to five modules so the switches could be a killer. Just throwing it out, but we use a voltage step down transformer to operate the primary at high voltage and low current while operating the secondary at high current and reduced voltage in a small PF. The big advantage is we can drive 60-80 kA using two thyratron switches (~10 kA each) on the primary. It minimizes the switch count (i.e. consumables) but the cost of the cores is significant. In the case when you are replacing switches frequently, it might be a path forward. The real problem is the primary voltage at high current. You need something like 60 kV just to drive the coaxial rundown section. So you need more like 100 kV on the secondary. For a reasonable reduction in switch count you want a 3:1 or 4:1 stepdown. A 400 kV primary is not impossible….

[delt0r]

I can’t see 10s or even 100ns rise times on a 400kV primary transformer. That is their issue. 50Hz transformers at the 250kV range are horrendously expensive. Can see a 400kV high frequency one being much cheaper.

In a LVA design you don’t need to worry about 1 or 2 switches per unit failing if designed right. Also they are solid state, so should have very long life times if rated correctly. Its often better to spend a lot up front to prevent downtime.

[Joeviocoe]

Would any voltages greater than 60 kV be even relevant to LPP’s DPF?

[asymmetric_implosion]

The Sandia LTD’s use gas switches and have rise times of 100 ns at 1 MA per module. They operate with +/- 100 kV on the input and give 100 kV on the output with 2:1 current enhancement [Mazarakis, PRSTAB 12 050401 (2009) for 0.5 MA LTD with 100 ns rise, new work was presented last year at the 1 MA level]. The switch lifetime is poor in the SNL LTD and solid state is better in that respect, it becomes a problem of the number of units and triggering the switches. The little I’ve run across on solid state LTD technology is modest currents (100 kA) but it takes many switches per module to get the ~1 us rise times that matter. The voltage is pretty low per module at 2-5 kV. I’m not saying it won’t work but it will require a milliion or more switches. Can you trigger a million plus switches at low jitter? I think so. Can the solid state switches survive the pinch voltage at ~ 3 MA, which is 750 kV to 1.25 MV, when it feeds back to the pulse power (inductively divided of course)? I don’t know. Is the cost of the front end more than the gas switch solution over the long run? Probably not.

I don’t have the ref for it but I am familiar with the work, people use what are called load current multipliers (LCM) on ~1 MA Z-pinches to optimize the matching between the load and the source. The LCM is nothing more than a 2:1 current step up transformer on 100 ns machines. Our PF system operates at 700 ns using a 6:1 current step up transformer. The transformer cores were like $1000 per unit. We use a total of twelve cores. The Thyratron switches we use cost $3500 per unit with another $4K for the controller box per switch. It was a huge cost savings to use the transformer solution and reduce consumables. It fires at 10 Hz as long as the anode can survive. Admittedly, the V-s product for a 3 MA system would be significant by comparison but the cost may not be as bad as you think. The problem again is switches.

[asymmetric_implosion]

Joeviocoe wrote: Would any voltages greater than 60 kV be even relevant to LPP’s DPF?

It depends on the optimum load design for the PF. As I said before, every PF regardless of size has to contend with to 10-20 mOhm of impedance in the flowing plasma. It is an artifact of the self similar physics of the PF rundown. You desire a roughly 100 km/s plasma speed during the coaxial rundown. The geometry is coaxial so the time rate of change of inductance (dL/dt) is 2E-7*ln(b/a)*v_axial. b is the cathode radius and a is the anode radius. For most PF devices, the ratio of b/a is between 1.5 and 2. Thus ln (b/a) is between 0.4 and 0.7. This leads to 8E-3 to 14E-3 Ohm. Some additional impedance always crops up so it is common to use 20 mOhm as the upper limit for machine design. If you use 20 mOhm, you need 60 kV just to drive the coaxial section during the axial rundown at 3 MA. Consider the other impedance like the bank resistance (<10 mOhm), bank inductance (~20 nH), bank capacitance (10-1000 uF) and inductance/resistance of the plasma (time varying). At a minimum you need to double the minimum voltage to achieve your desired current with a reasonable tolerance. Using this data, it is clear you might need more than 60 kV even at the 2 MA level. The last published paper from LPP has data at 1 MA and ~40 kV on the bank [PoP paper]. The scaling to 2 MA is fairly linear unless the cap bank size went up.

I was suggesting a scheme to minimize consumables (switches) by using transformers. To minimize the number of switches, you need to increase the primary voltage. Step downs of 2:1 are known at the 1 MA level but 4:1 would be better. If you require 100 kV to drive the load, then you need 400 kV on the primary. It seems a bit ridiculous at first glance, but >1 MV is very common in large pulse power. The trick is gas switches are used so you have the lifetime problem. The difference is by using a 4:1 transformer you need 1/4 the switches if you have a 400 kV switch (not easy). By replacing consumables with non-consumables (Transformer cores) you might find an advantage in cost in the long term. This approach works well at low currents but it might now work as well at high currents. More to be done.

As delt0r has said, there are other alternatives like the LVA. Each system has merits and problems. It will come down to cost and lifetime in the end. If the LPP’s cost model works for 1 month operation, then you need 5E8 shots between shutdown. Solid state can do it if it doesn’t fault. In fact it could possibly run for a few months which leaves the anode as the limiting factor. State of the art solid state pulse power for a PF is at 260 kA, 8 kV and 80 Hz. It is a big leap to 3 MA, 60 kV and 200 Hz but one that might be necessary.

[andrewmdodson]

http://www.evincetechnology.com/tech_overview.html

possibility of diamond switches evolving soon into something useful for the focus fusion project?

[asymmetric_implosion]

andrewmdodson wrote: http://www.evincetechnology.com/tech_overview.html

possibility of diamond switches evolving soon into something useful for the focus fusion project?

The website is very vague. My comment on using field emission electron sources is the limited current in the trigger. One should note that diamond is not a tickle trigger like Si. It usually takes a substantial trigger to turn diamond on and the trigger needs to stay on to keep the switch conducting due to the short carrier lifetime (~1ns). Laser triggered diamond switches have this problem as the necessary laser intensity is difficult to maintain for more than a few ns. It seems they are targeting power grids which is a very different beast than pulse power. I know diamond has come a long way due to work by groups like Diamond Detectors Limited, but the trigger is still the weakness. This is the reason diamond is a radiation detector right now. Last I checked and it has been a while, these type of switches are limited to ~1 kA. Going to a 3 MA system will take many components in parallel. The missing piece is the voltage hold off. Ideally, you do not want to have switches in series. Switches in parallel are not a much of a problem. The switch is going to have to hold ~60 kV. Again, it’s been a while but diamond has been tested successfully to 50 kV in photo conducting configurations.

[Joeviocoe]

Has the new Tungsten crown regularized the filaments, and yielded any higher gains?

[Author]

Posts