Mastodawn

the Z-pinch in the copper plates is one of the most bonkers things I've ever seen.

I feel like we should get IXYS or Infineon to sponsor him, and get some of their engineers to develop a parallel SCR or IGBT driver that can actually handle the load. hard to imagine a more effective way to demonstrate a product's robustness than that.

"hey man, here's an IBC container full of mineral oil and IGBTs. enjoy!"

the more I think about it then more I am certain that this thing would be far more terrifying if you could have a microsecond rise time on those 160kA, at least at the switch side (the inductance in the cables might limit it somewhat)

yeah uh at 160kA/μs the entire battery voltage would be negated by *checks notes* 375pH of inductance, which is on a par with the package inductance of an 0201 part (not including the mounting inductance)

so I don't think that rise time will be achievable at the load. but still, make the switch that fast and you'll get some wild effects.

Patrick Herd Feb 22

@gsuberland I've done engineering to get a lot of the way there before :-)

I love this, so I Feb 22

@gsuberland hell I’m wondering what kind of mechanical switchgear would be needed for that much DC current in the first place

@jpm if it exists it's probably a parallel set of high speed hydraulic rams in a pressurised SF6 environment or something.

I love this, so I Feb 22

@gsuberland now I’m wondering what kind of stuff is hiding inside HVDC substations, probably would be similar

F4GRX SÃ©bastien Feb 22

@jpm @gsuberland yes

https://library.e.abb.com/public/3b6db5ddd75590bfc1257ba50027f74d/06-13%202m309_EN_72dpi.pdf

Russ Garrett Feb 22

@jpm @gsuberland HVDC systems have their circuit breakers on the AC side because it's so much easier to break fault currents on AC.

Lack of reliable HVDC circuit breakers is what's stopped us building more complex HVDC grids so far, but there's a *lot* of work currently going into that:

✧✦Catherine✦✧Feb 22

@gsuberland @jpm I was expecting some DIY vacuum or inert gas switchgear (and a bounce off semiconductors)... not that janky thing with two hydraulic cylinders lol

robryk Feb 22

@gsuberland I was kinda confused why, for pipe squashing, he didn't put a "fuse" (a steel bolt or sth like that) in parallel with the pipe. The issue was that the rise time was too long due to, in part, inductance of the whole circuit, so shorting with a "fuse" would certainly improve on the inductive part. Do you know if I'm obviously wrong somewhere?

@robryk it may have worked, but the footage would be more difficult to discern due to the fuse exploding.

robryk Feb 25

@gsuberland I'm watching the video bit by bit, and he does mention pulse-forming networks at some point. I wonder if he discounted the one he actually has or didn't consider it.

qualia Feb 22

@gsuberland "equivalent series inductance of a car battery" doesn't seem to be a question that gets asked very often for some reason. hmmm

i think for testing pulse caps you can do a short circuit ringdown test and calculate from that, perhaps quarter cycle risetime on a single battery short circuit might surface some relevant info, but i can't imagine this is a well-controlled parameter across the bank

i am now morbidly curious and look forward to an alarmingly axisymmetric arrangement in the future. theoretically if he rearranges the damnable thing further into arcflash territory to a higher voltage it could open up the possibility of an ablative spark gap triggering type of situation

Ignas Kiela Feb 22

@gsuberland tbh I was thinking when watching "why doesn't he try moving the cables closer to reduce impedance" but then I realized that if the cables were closer together they would be trying to tear themselves apart because of the opposing currents

@gsuberland Infineon has 2.4kA rated 4.8kA repetitive, 1200V modules. Cost 30k or so in modules, probably another few grand in busbar, but would be able to do that kind of pulse all day long.

@AMS *begins drafting email*

@gsuberland The big trick will be the buswork design so it

Doesn't move.

Helps the dinner table full of IGBTs share as evenly as possible.

@AMS any thoughts on whether uneven switching would present a problem alongside the parasitics? my main thought would be dI/dt from some switching on earlier than others might cause a big voltage spike on the output, then when the later ones turn on it's much harder on them (and maybe violates SOA)

but idk how much of a concern that is versus just keeping the loads balanced in the first place.

@gsuberland I'd go for local gate drive at each one, equal length fiber optics for sync, and come up slow (1us pulse, 2us pulse, etc.). Risetime would be sub us easy for each gate. Give them a uF or ten of film cap at the terminal and you're likely able to figure out balancing as you go. I'd guess that 18V gate will saturate at 8.some kA so desat protection should help keep you safe. You'd mostly be depending on the diodes and snubber cap to keep voltages safe.

@AMS makes sense, thanks ^^

imagining what kind of TVS diodes you'd need to absorb that kind of pulse 😅

@gsuberland Steering diodes, capacitors, and pulse forming networks are probably the way to go. The average power isn't that high.

@gsuberland I'm looking at these https://www.infineon.com/assets/row/public/documents/60/49/infineon-ff2400r12ip7p-datasheet-en.pdf. The 10nH module inductance will keep you from too much badness. Probably could get another kA/module if you ask them for a IGBT top, diode bottom buck-configured thing.

@gsuberland Don't need mineral oil if the pulses are short enough. They'll just go 25C to 150C Tj in a ms or 10.

Patrick Herd Feb 22

@gsuberland i'd use the same approach we use in big offline coils, use 1x IGBT sized for each parallel string of batts, that way timing is really not critical and a random turn off or turn on doesn't blow anything up

@PatrickHerd my worry would be that the dI/dt from the earlier ones switching on would lead to a huge inductive voltage spike across the later ones, massively violating the SOA.

miiiight be fine, but idk. it's not like you can get megawatt rated TVS diodes (although it would be fun to try)

Patrick Herd Feb 22

@gsuberland works in tesla coil duty so I suspect it would be fine

Chris Petrilli Feb 22

@gsuberland there are so many moments like that, it's fabulous. Physics!

Chris Petrilli Feb 22

@gsuberland watching massive (4/0?) cables just flop around due to magnetic forces like they were wet noodles.

@petrillic he wasn't lying when he said he'd be attaining energy densities on a par with the National Ignition Facility

Greg Bell Feb 22

@gsuberland @petrillic the NIF is jealous of some of these sick experiments

Chris Petrilli Feb 22

@ferrix @gsuberland I do worry about the health impact of zinc plasma though...

Cassandrich Feb 22

@petrillic @ferrix @gsuberland Seems like he would have learned his lesson after last time...

F4GRX SÃ©bastien Feb 22

@dalias @petrillic @ferrix @gsuberland he has a mask and he's outside. Given how afraid he was of Allen Pan doing shit with lasers, I believe he actually cares about the kind of security he needs.

penguin42 Feb 22

@gsuberland Yeh glorious - longest video I've watched (ever? Certainly in ages); they should get him to do some tours of some of the Fusion places.

@penguin42 get him over to CCFE!

Tóth Gábor Baltazár Feb 21

@gsuberland
🥴

woe2you Feb 21

@gsuberland NOPE. That's a tomorrow watch, it's too late for horror.

Perrin42 Feb 21

@gsuberland

Good thing he's using that advanced Safety Tree™!

Greg Bell Feb 21

@gsuberland a feature-length video including amazing effects and unexpected twists. It was a thing of beauty

Greg Bell Feb 21

@gsuberland I never before realized how much physics abhors switching 100 kiloamps

Greg Bell Feb 21

@gsuberland "I dunno just put two plates together". Physics: "YOU THOUGHT"

@ferrix I wanna see someone build a solid state version (assuming he hasn't; I've not finished watching yet)

@ferrix a huge bank of IGBTs would probably do it, with some active circuitry to balance the gate drives, but I bet there are a ton of demons to chase down.

Greg Bell Feb 22

@gsuberland later he uses the expedient of blowing up an SCT(?) for each run. Unclear that it helps. I think there's plenty of room for improvement on his physical plate smashing setup, but I've never played at that level

@gsuberland @ferrix you'll need undividual gate drivers. What need to be tuned is possibly the timing of the driving pulses.

Bornach Feb 22

@f4grx @gsuberland @ferrix
He'd be making his own krytron switch next.
What could possibly be wrong with doing that on YouTube?

Greg Bell Feb 22

@gsuberland often watchin SP I think "with a little more money this man could really level up" and then I wonder whether it would lose the bootstrappy charm

ADHDeanASL Feb 21

@gsuberland Correct me if I’m wrong but I don’t think “now, the scientific method is great and all, but…” isn’t something one usually hears connected with genuine science.

Graham Sutherland / Polynomial Feb 21

@adhdeanasl given what he says immediately afterwards I think the point is a reasonable one.

after all, JPL and NASA exist because a bunch of nutters tried to make rocket fuel in their dorm, and rather than expel them they gave them access to an old shed and told them to keep at it. most of what they were doing was just mixing energetic compounds together to see what stuck, what exploded, and what turned into deeply offensive tar that would strip the paint off walls at twenty yards.

Tamonten Feb 22

@gsuberland well I procured 0.25% of that amount of car batteries today. I might need some time to catch up.

It was for the van though so it might be closer to 0.4%. And the guy behind the counter basically told me how to commit warranty fraud with it so maybe 0.7%? I’ll be there in like three months, four tops.