thread for segments files and parameters for simulation runs

Discuss how polywell fusion works; share theoretical questions and answers.

Moderators: tonybarry, MSimon

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

icarus wrote:
Yes, but we do have some idea how the experiments "should" work,
speak for yourself einstein, you do know that mother nature will make you look like a complete idiot when/if it is ever known how this thing behaves?

You are definitely not a physicist to be so categorically cock-sure of how this thing will work .... but enough of that, just stop putting unsupported conjecture into a theory section please, for once.???
Sheesh, namecalling? Really? Can you try to make some minimal attempt to be civil? This isn't gradeschool.

Wb-6 is not "conjecture." Wells are not "conjecture," they have been measured. Rick is a physicist, and so is Bussard, which is why I keep referencing them. I don't know what you're flailing at exactly, but maybe take a few deep breaths or something?
Last edited by TallDave on Wed Nov 24, 2010 10:31 pm, edited 1 time in total.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

happyjack27 wrote:
TallDave wrote: No, there really is nothing keeping them there. The coils don't keep them in the center of the WB, they just keep them in the WB.
that's what i mean. i don't mean the exact center. in fact from my sims it looks like the WB is pretty hollow. i'd imagine due to mutual electrostatic repulsion.
OK, then, that seems to make sense. Except... why would you need to aim them there? Once they're in the WB, they should stay there, at least for 100K transits or so.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

krenshala wrote: What I think would be an interested thing to check, and I believe would help with what you guys are talking about above, would be to start the sim with zero electrons "in the system" and the magrid active, then power up the e-guns and slowly start shooting electrons into the well.

This should test both whether you can get enough electrons into the system with the guns outside the magrid (firing into/through cusps), but it should also be another good way to verify whether they will spontaneously form that electron shell/sphere near the center.
no way vary the particle count right now. (though you can vary how many real particles each one represents) that would take some deep re-coding and possibly hurt performance. but you can start them all of in no man's land and they'll cycle through the eguns when they go out of bounds.

anycase, that's something i plan to do but in order to do it i need to calculate how much energy i need to fire them off at to get them into the wiffleball. you can see at the end of the video you're referring to i was trying to do that manually and it was just too sensitive.

until then there's really only two useful well-capable modes: start them off randomly throughout the magrid, or randomly inside WB.
Now if only I had a better video card to run some of these myself ... Did you ever post up the exe for it? My system may be slow, but I can always kick stuff off and check it after waking up/getting back from work. :D
having posted the exe yet. it's constantly evolving and there are some parameters that you still have to set in the code. e.g. whether it's all-electron, all-ion, or mixed, and if you need to make major adjustments to the time scale. (i could add in an exponential slider for that, but then if you hit it by accident you could totally reset the sim.)

at minimum you need a CUDA-capable card. Which means we're talking Nvidia processor GeForce series. although 6800's are technically cuda-capable, i'd recommend against using them for anything CUDA as they have very few cores. 8800 GTX minimum. (which should actually be pretty cheap by now.) also i can compile for 32-bit or 64-bit. and i'm not sure but i think you need some library files so you might need to download the nvidia sdk. but maybe i can find hot to turn them into static libraries so they're compiled into the exe. tell you what i'll compile a 32-bit version and a 64-bit version and upload both. but you might not see it till after thanksgiving.

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

TallDave wrote:
happyjack27 wrote:
TallDave wrote: No, there really is nothing keeping them there. The coils don't keep them in the center of the WB, they just keep them in the WB.
that's what i mean. i don't mean the exact center. in fact from my sims it looks like the WB is pretty hollow. i'd imagine due to mutual electrostatic repulsion.
OK, then, that seems to make sense. Except... why would you need to aim them there? Once they're in the WB, they should stay there, at least for 100K transits or so.
"Once they're in the WB..." yes. and there precisely lies the problem: getting them in the WB in the first place.

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

So what is the representation ratio right now?

I've got cross-fired 5700s, I'd be curious to see how they do. Dunno about CUDA though.

Are you familiar with Joel's simulation work? He seems pretty open to sharing via email if you're interested.
Last edited by TallDave on Wed Nov 24, 2010 10:50 pm, edited 1 time in total.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

"Once they're in the WB..." yes. and there precisely lies the problem: getting them in the WB in the first place.
Ah, I see what you mean now. I don't think that's an issue IRL, though: my understanding is they get fired in along the field lines. It's an open "recirculating" (or maybe just cusp-plugged) device.

I wonder about the hollowness... it's been mentioned before that if you look at it in LTE the electrons collapse into a Debye sheath, but a driven system won't do that...
Last edited by TallDave on Wed Nov 24, 2010 10:57 pm, edited 1 time in total.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

TallDave wrote:So what is the representation ratio right now?

I've got cross-fired 5700s, I'd be curious to see how they do.

Are you familiar with Joel's simulation work? He seems pretty open to sharing via email if you're interested.
i like to keep it at 1. that's what you see in most of the videos. though sometimes i'll max the slider out to 1000 just to see what happens. you can see the sliders in the upper left in the video. might be a little difficult to make out their labels, thou, esp. at low res.

the high ratios doesn't seem to hurt the accuracy too much (though how would i really tell). so i've changed the sliders to work on total particle charge in coloumbs (separate for ions and electrons), and use logarithmic units, (1*10^x) so i'll be able to simulate pretty much arbitrary deep well depths then. i suppose we shall see how well that works.

unfortunately 5700's are ATI cards. There's a version of the source code i started out with compatible with ATI cards, but it'd be a lot of work to port it over and the gpgpu code for that is all in OpenGL which i'm nowhere near as fluent in as CUDA so i don't even think i could write it without climbing a very steep learning curve.

Is that ephi? If so, i've looked at like every page on that website a few times. if not, then i'm not familiar with it. either way i'd be happy to talk with another simulation writer via email. swap ideas, perspectives, methods, code, what have you.

icarus
Posts: 819
Joined: Mon Jul 07, 2008 12:48 am

Post by icarus »

dave:
Wb-6 is not "conjecture." Wells are not "conjecture," they have been measured.
Maybe not, but if YOU knew what you were talking about you would be dangerous. Luckily you merely just crap the place up with your mis-understandings. As much as you pretend, you don't know better than anyone else how the 'model' output should look.

And as I have said from the beginning if the model is set-up right you shouldn't be tweaking and adjusting to get the results you think you should ... that would be climate science ... it should simply be a true representation of the physics.

You're assumptions could be as faulty as anybody's I don't see why they should go anywhere near happyjacks model! Leave the lad's model alone.

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

Yeah, I figured the 5700s were a longshot. That's probably more coding than I'd want to do.

Don't know if he's on ephi. Here's a few links from the IEC conferences.

http://fti.neep.wisc.edu/iec2009/talks/ ... elroge.pdf

http://fti.neep.wisc.edu/iec2009/agenda

http://www.plasma.ee.kansai-u.ac.jp/iec ... ogers2.pdf

I'll pm you his email, not sure if he's still publishing/simulating stuff but he sends something out when he does.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

TallDave wrote:
"Once they're in the WB..." yes. and there precisely lies the problem: getting them in the WB in the first place.
Ah, I see what you mean now. I don't think that's an issue IRL, though: my understanding is they get fired in along the field lines. It's an open "recirculating" (or maybe just cusp-plugged) device.

I wonder about the hollowness... it's been mentioned before that if you look at it in LTE the electrons collapse into a Debye sheath, but a driven system won't do that...
in sim life it appears i need very precise velocities, which i understand electron guns are capable of in real life. but i don't know what those are. i'm working on having the sim calculate them on the fly but i'm stuck at how to calculate the electric potential (not to be confused with field) from a finite charged wire.

I don't know what a Debye sheath is. presumably its a thin layer of electrons, but beyond that it's greek to me.

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

TallDave wrote:I'll pm you his email, not sure if he's still publishing/simulating stuff but he sends something out when he does.
thanks. is saw one of those from a link posted on the prometheus fusion site. looks like this could be helpful.

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

icarus wrote:dave:
Wb-6 is not "conjecture." Wells are not "conjecture," they have been measured.
Maybe not, but if YOU knew what you were talking about you would be dangerous. Luckily you merely just crap the place up with your mis-understandings. As much as you pretend, you don't know better than anyone else how the 'model' output should look.

And as I have said from the beginning if the model is set-up right you shouldn't be tweaking and adjusting to get the results you think you should ... that would be climate science ... it should simply be a true representation of the physics.

You're assumptions could be as faulty as anybody's I don't see why they should go anywhere near happyjacks model! Leave the lad's model alone.
Heh, I guess civility was rejected as beyond your capabilities.

Icarus, I'm always happy to entertain the notion I'm wrong about something, and I'm always upfront about the sources of my information and their trustworthiness can be evaluated on their respective merits. Rick and Bussard may be wrong on some things. I don't claim to know anything more than anyone else, I'm just presenting my understanding of what they've told us. You don't seem to be able to back up your complaints with anything more than invective and appeals to ignorance that don't stand up to a moment's scrutiny. Feel free to consider yourself smarter/better/whatever than me, I really don't care, but your opinions on that score add little to the forum.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

happyjack27 wrote:
TallDave wrote:I'll pm you his email, not sure if he's still publishing/simulating stuff but he sends something out when he does.
thanks. is saw one of those from a link posted on the prometheus fusion site. looks like this could be helpful.
Always glad to be useful. I'm very interested to see the different conclusions different models arrive at. (If only we could see EMC2's code. Having direct access to the real data is obviously a huge advantage.) Joel ended up finding a breakeven size around 150m iirc.

You might also be interested in the Chacon/Miley/Barnes/Knoll paper. They did a full bounce-averaged Fokker/Planck simulation. Parabolic wells apparently gave very different answers than flat ones.

http://fsl.ne.uiuc.edu/IEC/Miley_Phys.o ... 02000).pdf
Last edited by TallDave on Wed Nov 24, 2010 11:23 pm, edited 1 time in total.
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

icarus
Posts: 819
Joined: Mon Jul 07, 2008 12:48 am

Post by icarus »

Okay, all banter aside, lets get some of the basics agreed upon;

1) Can we all agree that the electric potential in the precise center is always zero? (By Gauss theorem and assuming all possible surrounding charge distributions are spherical enough)

2) Can we also all agree that the first spatial (radial) derivative of the electric potential in the precise center is also zero (i.e., either a local maxima or minima)?

TallDave
Posts: 3152
Joined: Wed Jul 25, 2007 7:12 pm
Contact:

Post by TallDave »

Oh good, we're going to be civil now.

Do we agree potential wells have been measured?
n*kBolt*Te = B**2/(2*mu0) and B^.25 loss scaling? Or not so much? Hopefully we'll know soon...

Post Reply