magrid configuration brainstorming

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

Moderators: tonybarry, MSimon

olivier
Posts: 155
Joined: Thu Feb 14, 2008 5:21 pm
Location: Cherbourg, France

Post by olivier »

I found some nice Java applets on the web, designed by Paul Falstad.
I have played with the 3D Magnetostatic Fields Applet. Select the "loop paired stack, opp." and "Display: Particles (Vel.)" options and within seconds, you will have built one third of a minimalist Polywell and will observe some confinement in the middle. Many losses from the sides, are you surprised ?
You can rotate the picture, adjust the B-field or the coil size and see the movement of the particles change in real-time.
The Java code is available and, from what I have seen, it would not be too difficult to add four more coils. It is really nice for educational purpose, but might be more difficult to turn into the visualization tool drmike is longing for.

Betruger
Posts: 2321
Joined: Tue May 06, 2008 11:54 am

Post by Betruger »

Field type "monopole attempt" looks like it only needs the coils to change from square to round to make a polywell.

drmike
Posts: 825
Joined: Sat Jul 14, 2007 11:54 pm
Contact:

Post by drmike »

Yeah, Java plugins don't run on my browser because the background is Suse Linux. I've tried. I've got a friend who does 3D animations, I should ask for help on this stuff. Generating the data is hard enough, there really is no point in re-inventing the display wheel.

I think Simon pointed me to the same web site once before. It really is good work!!

seedload
Posts: 1062
Joined: Fri Feb 08, 2008 8:16 pm

Post by seedload »

drmike wrote:I can't find any direct on line references, but the baseball coil was used at the end of tandom mirror fusion devices in the 1970's. The idea has been around for a long time. The simplicity of the Bussard shape makes things cost effective.
Thanks for the info. I can't find any references either.

I can't figure out whether you were indicated that the shape was tried and failed or just letting me know that it was not an original idea.

If the shape failed to adequately contain ions in prior fusion devices it doesn't rule the shape out for containing electrons in a polywell design. I am sure six coils pointing in at each other would have a tough time containing ions too.

To your final point, I am not sure that six or more simple coils in whatever configuration is more cost effective than one shaped one especially when considering cooling systems, electrical systems, etc..

... this is just my two cents in a brainstorming thread.

drmike
Posts: 825
Joined: Sat Jul 14, 2007 11:54 pm
Contact:

Post by drmike »

Failed isn't qute the right word - they never got a chance to fire things up full blast to find out. Partially to point out it has been tried though, so there are good references in a library somewhere.

We really need to model the plasma at some level that gives us some confidence it is at least within an order of magnitude. That's not so easy for simple fields, let alone complicated ones. I think trying things "on paper" is a good idea, and the more things we can try, the higher the probability we can find something that will help make things work.

I'm definitly not trying to say "it won't work", more that it is "ahead of its time".

MSimon
Posts: 14334
Joined: Mon Jul 16, 2007 7:37 pm
Location: Rockford, Illinois
Contact:

Post by MSimon »

olivier wrote:I found some nice Java applets on the web, designed by Paul Falstad.
I have played with the 3D Magnetostatic Fields Applet. Select the "loop paired stack, opp." and "Display: Particles (Vel.)" options and within seconds, you will have built one third of a minimalist Polywell and will observe some confinement in the middle. Many losses from the sides, are you surprised ?
You can rotate the picture, adjust the B-field or the coil size and see the movement of the particles change in real-time.
The Java code is available and, from what I have seen, it would not be too difficult to add four more coils. It is really nice for educational purpose, but might be more difficult to turn into the visualization tool drmike is longing for.
I wrote the author an e-mail and asked if he would add the Bussard Configuration to the applet. He said it was a good idea.
Engineering is the art of making what you want from what you can get at a profit.

rcain
Posts: 992
Joined: Mon Apr 14, 2008 2:43 pm
Contact:

contra-circular subjects

Post by rcain »

...thats a coincidence (or maybe not) - quite independently i too just came across http://www.falstad.com/mathphysics.html:

and in particular

3d magnetostatic fields applet
http://www.falstad.com/vector3dm/
loop pair stacked opposing - particle velocity

plus

3d vector fields applet
http://www.falstad.com/vector3d/
1/r^2 sphere

very pretty visualizations - as you say 4 coils short of a Bussard, but virtually there (if you'll forgive the pun).

i actually came across these posts afterwards to cross check the forum for a reference.

to MSimon - I was about to suggest the very same thing , re asking the author if he wouldn't mind 'Bussardisising' his code. If you have no joy, I might be able to help out with the Java, since he kindly makes source code available.

In actual fact however, i originally started out trawling the net for info Halbach Sphere's.

got a reference to O. Cugat and F. Bloch (PhD researcher)
Laboratoire d'Electrotechnique de Grenoble - circa 1998 - who apparently got up to 5T out of one (or rather, around the middle of one), albeit within a pretty smallish volume.

the orig paper/website seems to have disappeared but google has a cache of the text at least (no pix unfortunately), here - http://64.233.183.104/search?q=cache:MY ... =firefox-a

... which also cross references to an earlier discussion on this forum, here - viewtopic.php?p=879&highlight=halbach#879

... as well as some some toys i was looking at buying here - http://www.shinyshack.com/product.php?p ... tation-Kit
- though nowhere near powerful enough and totally the wrong shape.

strange how circular these 'closed' subject hunts can become.

can anyone remind me what sort of field strength WB7 is working with?

olivier
Posts: 155
Joined: Thu Feb 14, 2008 5:21 pm
Location: Cherbourg, France

Re: contra-circular subjects

Post by olivier »

rcain wrote:...thats a coincidence (or maybe not) - quite independently i too just came across http://www.falstad.com/mathphysics.html:
Great minds think alike. :wink:
drmike wrote:A good visualization tool would allow moving the data so a better idea of how it really looks can be more easily grasped. I can see the null spots are near coil structure, and that would cause problems. An interesting idea to play with though.
This week, I came across Mayavi2. This scientific data viewer is free for academic use. It is bundled in the Enthought Python distro. As an example of what you can get, the picture below displays a transparent sphere (E-field iso-surface) and two perpendicular plane cuts (one with vectors and the other with field lines). The data set represents the electric field of a point charge stored in a VTK file which I built with a few lines of C.
Image
Among its many features, Mayavi2 allows you to interactively :
  • -rotate the whole system in 3D
    -rotate or slide the cut planes,
    -slide through time series,
    -display tensor fields,
    -add filters,
    -write your own scripts.
Really cool.
Last edited by olivier on Sat Jun 21, 2008 9:12 pm, edited 1 time in total.

drmike
Posts: 825
Joined: Sat Jul 14, 2007 11:54 pm
Contact:

Post by drmike »

Very nice!

I spent the last few days trying to get my computer to run any 3D program. I have learned a hell of a lot more about Linux than I wanted to - my system didn't reboot after I futzed with a few things (several times actually....)

I really need to get a new computer. For what it costs, the ability to do full scale supercomputing on a desktop is well worth it. As it is, I can't do the 3D stuff because my
grapfics card isn't fully accelerated.

But it can certainly be a useful node on a network....

rcain
Posts: 992
Joined: Mon Apr 14, 2008 2:43 pm
Contact:

Post by rcain »

To Olivier - very pretty. Do you know what the completeness, accuracy, precision of your code, is - ie. is it good for real numerical simulation/prediction or just visualization?

The thing I particularly liked about the Falstad code, was indeed its 'particlulates': Seems like the author has got some good true object oriented (non linear) simulation going on - i managed to get a wonderful high density 'cylinder' of ions though the middle of my model - could have sworn i saw colission products flying off as a result ;)

(though i realize the mag field isn't what we are looking at to cause this - need to add the the electro sphere/well for that bit - but pretty impressive anyhow - especially since its written in Java, which isn't renowned for its performance.)

collisional modeling (and other 'fine' details) shouldn't be too had to add to it either i would imagine - we could end up with quite a complete little model of the Polywell.

To DrMike also: this brings me to the next idea - suppose we got the Falstad code pretty much complete, and providing its a fairly accurate representation, then getting the precision and the performance required for real experimental work would also be quite possible - I should imagine.

Vis: strip out all the unnecessary code (unused front-end config stuff, etc), recode the core engine in C (or perhaps compiled REX to get arbitrary precision arithmetic and other nice features), then put the whole package out as a Grid-Computing App.

I have the source code for the SETI package which would do the job. This way you could get really nice front-end demo's, etc for all the grid-participants on their home PC's, (and free promotion just like SETI), PLUS all the scenarios, configuration variations, materials assumptions, data volumes, details and precision you might care for from the back-end combiner/grid server - without requiring the expense of a supercomputer to handle it all.

You could do some really detailed modeling of the WB construction and operation; of particular interest (I noted) what happens when things get too close to physical corners and edges and the like within the physical machine itself; seems to me that details of that order are going to become critical at some near point in our journey.

Perhaps I am talking rubbish here; perhaps there a better modeling tools out there for the detailed numerical science? Am I wrong, or might it be a feasible approach?

(Seems to fit with the sort of 'garage-shop' ethos we have happening here - but hey, why not; its as good as better already...or seems to be, pretty easily?)

olivier
Posts: 155
Joined: Thu Feb 14, 2008 5:21 pm
Location: Cherbourg, France

Post by olivier »

Mayavi2 has some postprocessing capabilities, but it is basically meant for interactive visualization. However, it can be embedded in a scientific application written in Python (there are scientific libraries), C, Fortran,... Parallel programming is addressed too.
I did not have time to look at Falstad's algorithms. My guess is that the B-field is calculated once for all, then the particles are moved in the field without taking into account the B or E-fields they generate. But I may be wrong.
I think drmike is using high-precision quad-float libraries and is developing a much more sophisticated model. Perhaps particle interactions would make it difficult to parallelize, but that is a good idea if it is feasible.

rcain
Posts: 992
Joined: Mon Apr 14, 2008 2:43 pm
Contact:

Post by rcain »

Hi Olivier, think you may well be right about the B/E-fields - think it might have totally ground to a halt on my machine if otherwise. Obviously a major limitation if true. Ill grab the source code and take a quick look.

If as you say Mayavi2 has parallel built-in too, then that is a plus already for that. And no point in duplicating any precision stuff if its already on the go.

Maybe just leave Falstad with doing cute animations - good for popular visualization certainly and relatively easy on most desk tops.

Hope the original author can turn something up, before I start fiddling with it tooo much - hes bound to do a much better job.

I was looking to see what sort of maths Mayavi2 uses at its core - I was wondering maybe Cifford Algebras (like the better gaming engines) - but couldn't find it on their site - just MVC, delegation and some object wrapping as architecture. Any idea?

Seems a productive environment though from what you say already.

olivier
Posts: 155
Joined: Thu Feb 14, 2008 5:21 pm
Location: Cherbourg, France

Post by olivier »

Mayavi has no parallel built-in, but it is bundled with other Python libraries SciPy & NumPy that have, to some extent. It is good to know if who choose to stay in the Python environment.
Mayavi will be happy with any other source of data feeded in either VTK or Plot3D file formats. More details in the users guide.

drmike
Posts: 825
Joined: Sat Jul 14, 2007 11:54 pm
Contact:

Post by drmike »

I'm using standard double precision, but looking at how all the fluids interact with the external fields. Simple models don't cut it, and even complex ones are not really good enough - you have to make assumptions somewhere.

But I've got to build experiments as well as compute, so I am finding the cheapest way to do both. I finally figured out that my computer won't do 3D rendering because my graphics card isn't supported. I can get a cheap graphics card and fix that problem - and I will try to generate standard format 3D objects which anyone can look at using programs like Blender.

I think being able to move the fields on your own screen will make understanding what is going on a lot easier. If I can get my 5 year old computer to do it, pretty much any computer can do it too.

Next step is to build a welder so I can build a vacuum system. Since welding is a plasma, I don't know that I need a vacuum system to play, but at least I'll have a goal :D

tombo
Posts: 334
Joined: Fri May 09, 2008 1:10 am
Location: Washington USA

Post by tombo »

Well, here is a picture I drew of a Truncated Cube like the MPG machines.
This is easier than beating my head against Blender.
This picture shows a 3 meter diameter Magrid with 8" thick conductors.
The same proportions hold for a 12” cube with 7/8” diameter tubing for a scale unit.
I think the conductors can be stabilized with conductive metal supports because the outer layer is all at one potential.
The current driven voltage drop is minimal so only a thin insulator need separate it from the outer sheath.
Or, the supports could have a very thin insulating break hidden from most particles by a recess.
Or a thicker insulator would support the full 30kv so the coolant and current layers could work at near earth ground potential.
30kv is no more than automobile ignition stuff.
We could make it thinner with higher tech materials.
Of course the 8" thickness is to support many more layers of functionality.

This is a parametric drawing.
I can change the cube size, coil thickness, and bend radii etc. by changing numbers in a table. (If anyone wants to see it in a different configuration just ask.)

Image

I think the MPG machines are an avenue that needs further exploration.
I understand that they were abandoned partly due to the fireworks from the insulating wire ties, and partly due to power supply difficulties.
I believe that both of these difficulties (and more of course) can be overcome.

DrMike, when you're done with welding you can use the welder to power the coils. High current, low voltage, robust, short tolerant. Not too bad if you can do what you want with a few hundred amps. It would take a huge parking lot full of them to jump start a pw100. You probably have though of this too, but it might make a nice dual use.
-Tom Boydston-
"If we knew what we were doing, it wouldn’t be called research, would it?" ~Albert Einstein

Post Reply