Fusor Computer Modeling

It may be difficult to separate "theory" from "application," but let''s see if this helps facilitate the discussion.
User avatar
Carl Willis
Posts: 2841
Joined: Thu Jul 26, 2001 11:33 pm
Real name: Carl Willis
Location: Albuquerque, New Mexico, USA
Contact:

Re: Fusor Computer Modeling

Post by Carl Willis » Sat Aug 31, 2013 12:14 am

The keys to a successful (physically meaningful) model or simulation are, first of all, having a specific problem that is within the domain of validity of the model, and second, furnishing a relevantly-accurate representation of the system being studied.

Often, our discussions about modeling don't begin with a clear statement of any problem. Instead, some powerful computer code or collection of codes are conceived of as virtual "black boxes" inside of which one can, apparently, synthesize reality to suit any and all information needs. And that's a total pipe dream:
You can't begin to talk about "how close" a model would come if a problem isn't defined. You can't meaningfully consider how to represent the system (does one ignore or not ignore gravity, the Uncertainty Principle, and all that other stuff in the list) if a problem isn't defined. And you can't propose what tool to use (SIMION, Poisson Superfish, etc.) if a problem isn't defined.

SIMION and Poisson Superfish are indeed applicable to SOME fusion-related questions. I have never heard of a "cloth simulator," though; Maybe it helps crackpot free-energy inventors pull the proverbial wool over their investors' eyes?

-Carl
Carl Willis
http://carlwillis.wordpress.com/
TEL: +1-505-412-3277

User avatar
Monroe Lee King Jr
Posts: 105
Joined: Fri Aug 23, 2013 3:57 am
Real name: Monroe Lee King Jr
Location: Lampasas, Texas
Contact:

Re: Fusor Computer Modeling

Post by Monroe Lee King Jr » Sat Aug 31, 2013 1:08 am

Me personally I like to watch the pretty simulations and play with the possibilities the software provides :) It gives me ideas. It's fun :) and educational. Better models mean more meaningful results! I'm sure if your looking for magic in the software you wont find it! Because we would have to know the magic to program the software for it. It just a tool SIMION works if your doing beam simulations for sure. The idea is to advance the software for our thing.

Monroe
Today's the Day! We go into Space!

User avatar
Carl Willis
Posts: 2841
Joined: Thu Jul 26, 2001 11:33 pm
Real name: Carl Willis
Location: Albuquerque, New Mexico, USA
Contact:

Re: Fusor Computer Modeling

Post by Carl Willis » Sat Aug 31, 2013 1:22 am

SIMION works if your doing beam simulations for sure.
Again, it depends entirely on the problem. It's useful for SOME beam transport simulations (often those without significant space charge effects or collisions or transient fields).

A hammer is useful for some repair tasks, but not for repairing fine bone china.

-Carl
Carl Willis
http://carlwillis.wordpress.com/
TEL: +1-505-412-3277

User avatar
Monroe Lee King Jr
Posts: 105
Joined: Fri Aug 23, 2013 3:57 am
Real name: Monroe Lee King Jr
Location: Lampasas, Texas
Contact:

Re: Fusor Computer Modeling

Post by Monroe Lee King Jr » Sat Aug 31, 2013 1:59 am

So there is no possible chance for useful software for a fusor? BS There's plenty that could be done interesting and educational just based on fusor experiments you could make a simulator for various sizes voltages and currents and pressures with various grids ect.. that model what you can and cant do well. Hell you could make a spreadsheet that would help most guys make their first fusor. It doesn't have to be perfect lol just a guide.

Something that could be built upon made useful. It's not quite as easy as a tesla coil but it's not that big a deal you cant model it to help others decide what they want to build in a general way. These other tools like SIMION are far advanced for what I'm talking about anyway.

Why make it a big joke around here to do something useful?

At least make it easy to design a good neutron source. There's nothing hard about that!
Today's the Day! We go into Space!

Ross Moffett
Posts: 89
Joined: Tue Aug 27, 2013 2:24 am
Real name: Ross Moffett

Re: Fusor Computer Modeling

Post by Ross Moffett » Sat Aug 31, 2013 2:51 am

Carl, is there some specific reason for the crackpot joke you'd like to elaborate on? I understand that there's a large number of them out there, and I'll probably come to understand this site is a magnet for them, but there's nothing particularly crack-potish about computer modelling of physical phenomena that I know about.

The problems of ICR are well-defined, I think, and I'd like to simulate all of the major ones, whatever the major ones may be. Finding out what those may be and how to do that is the point of this discussion, right?

I'd like to start by modeling a spherical apparatus with some disturbances to the sphere where there may be a window, gas valves, vacuum pump port, high-voltage feedline for the grid and any other disturbance others may think is necessary to model. I'd then like to introduce, with user-defined controls available, some number of deuterium atoms, some number of unwanted particles (oxygen, nitrogen, CO2, Argon, oil particles, whatever else experts here think would be relevant). I would like to have adjustable software controls on voltage and current on the grid and then set everything loose to interact. Each "step" of the program, the particles each judge the forces on them from each other item in the system, and set a velocity vector. Next step, re-evaluate based on present position and velocity. Probably Heisenberg's uncertainty will have to become involved there, with some randomizer algorithm tweaking the positions and/or velocities to simulate real uncertainty. That's what the purpose of the list is - which interactions are most important, and which don't matter? In my college days when I took Semiconductor Device Physics most all of our calculations showed that gravity was completely irrelevant. So maybe it's irrelevant at a range of 1 mm from another particle but relevant when there's some free space. Or maybe there's never more than a mm of free space. If no one knows, then that's fine, software can be designed to turn these controls on/off as their importance is shown or disproven (but it takes probably some significant computing time to show that).

The significance is that if this is feasible, 3D videos can be produced with open-source software that at least I understand (Blender 3D) and give some visibility to the interactions going on. I feel the most valuable question this would answer that so far it doesn't seem like is known, is where does the fusion occur most? In the transit zone between the shell and grid? Inside the plasma ball? On the grid itself? Answering those questions, which are of particular interest to me right now without having yet studied the problem much, would give me ideas about what changes to the apparatus might be effective to increase fusion.

Note: I am not a computer programmer, nor do I have personal experience with particle simulation, but I have read about it a lot and do understand math/physics/programming. Specialization is for insects. :)

EDIT:
I'm doing some investigating to see what the viability of this sort of model is with modern computers. This is not meant to be accurate, but a general, quick back of the envelope study.

At 10e-4 Torr, a 12" spherical chamber completely filled with deuterium following the ideal gas law has 9.6368727e16 deuterium atoms. We'll round that to 100 peta-particles, or if calculating for every particle on every step, 100 petacalculations per step. The Intel I7 4-core processor is available in a 3.9 GHz package with 187 GFLOPS of power. I'm going to make an assumption that floating point precision is useful for the model - it may require double floats. Who knows how many operations are required per particle-step without setting up an actual model, so I'm going to use a large number and assume 1000 operations. That reduces the number of particle-step calculations that can be done in a second to 187 million. That's 17 years to model one step of the entire system on a particle-by-particle basis on a crazy good home PC. The IBM roadrunner in Los Alamos would take around 28 hours per step.

That's a significant enough roadblock that I see significant modelling and simplification of a specific problem would need to take place, and free-particle simulation is not really viable. Maybe if we reduce to the scale of Iron Man's fusion power cell, it's something that could more easily be simulated. :)

I also found what may be a useful book to help model the problem.

User avatar
Carl Willis
Posts: 2841
Joined: Thu Jul 26, 2001 11:33 pm
Real name: Carl Willis
Location: Albuquerque, New Mexico, USA
Contact:

Re: Fusor Computer Modeling

Post by Carl Willis » Sat Aug 31, 2013 8:41 am

It seems to me that to be meaningful, all these modeling discussions ought to start with:

What do I want to know, and how accurate or precise is good enough?


Instead, we invariably get into talking about tools. Sometimes very specific ones.

My point remains that you can't select (or build) the right tool without knowing what the job is.

Russ, toward the end of your own tool discussion, you hinted briefly at a real question: "where does the fusion occur most?" Let me focus on that as an example. The specifics of the question are of huge importance to the tractability of answering it with existing models or having a reasonable chance at a meaningful result from a custom approach. If one happens to know as an input specification the steady-state angular (w), energy-dependent (E), reactant particle flux F(r, w, E) at all locations r within the assumed system, then computing a spatial reaction rate at all r from the well-developed nuclear cross-section databases is likely to give predictions that are accurate enough for most practical information needs. But usually one does not know F(r, w, E) and implicitly wants to model this, too. In some apparatus, e.g. high-vacuum colliders with ion sources, the particle sources are well-defined; one can predict the fluxes using various techniques or can directly compute spatial reaction rates with a Monte Carlo approach. But what if you are interested in the "simple" Hirsch-type fusor that almost all hobbyists build? Holy shit, do you have a difficult problem to deal with! Fusors are often characterized by only a few macroscopic measured attributes, like voltage, total current, and pressure of a background gas. The flux of reactant particles is being generated throughout the reaction volume by a self-sustaining glow discharge, a complex soup of atomic physics processes. Predicting F(r, w, E) and somehow benchmarking that prediction against reality are massive challenges. Under certain favorable conditions, particle-in-cell codes can provide useful information in such a problem, but the capability is not generally well-developed. In fact, the cross-section measurements needed to support the atomic process simulations often don't even exist!

The question one asks and the input information one can provide about the physical reality of interest are central to any modeling exercise. Without specifics, there is nothing to talk about, no meat on the bone.

-Carl
Carl Willis
http://carlwillis.wordpress.com/
TEL: +1-505-412-3277

User avatar
Monroe Lee King Jr
Posts: 105
Joined: Fri Aug 23, 2013 3:57 am
Real name: Monroe Lee King Jr
Location: Lampasas, Texas
Contact:

Re: Fusor Computer Modeling

Post by Monroe Lee King Jr » Sat Aug 31, 2013 4:47 pm

Why not simulate the observed and forget about the physics that are so difficult then. I'm talking about a tool to help new folks build a fusor so they can make the observations themselves? Something that will advance what IS known. Something open source something that can be built upon. Rather than waste everyone's time with vagueness about generalities something quick to get to the point. If you make it so and put in X energy you can expect operation to be so.
I would hope it would get developed as computers advance and simulators advance to the point where they can simulate the harder physics. As I am sure without a doubt that will occur at some point maybe sooner rather than later. You have to start somewhere and starting with observed operation is better than nothing.
Right now the hardest thing to accomplish to launch big high powered rockets is 3 sigma dispersion analysis the FAA requires for launches. The software that does a mediocre monte carlo 3 sigma cost a grand. If your connected to DOD you can obtain TAOS (which the FAA uses) for free.

I'd rather get something started now to avoid similar things in the future here. Some of these young guy's are pretty sharp and they might pick up the ball on the harder programming if they have a place to start. The 3 Sigma dispersion analysis you can buy has been developed from a simple program called Rocksim and the original 3 Sigma was a free program called Splash.

All I know is nothing from nothing leaves nothing. I'd rather encourage someone to try the hard thing rather than say Oh it's too hard and Oh it's just not worth the effort. Because I think it is always better to try the hard thing even if you fail. You tried at some point someone will succeed! Standing on the shoulders of those before him. Give em something to stand on or all the research is for nothing! Today you need programs to teach because that's what the young use rather than books today like it or not. It needs to be fun to learn! Doesn't make it right it's just the way it is. You have to work with what you've got.

Monroe
Today's the Day! We go into Space!

Ross Moffett
Posts: 89
Joined: Tue Aug 27, 2013 2:24 am
Real name: Ross Moffett

Re: Fusor Computer Modeling

Post by Ross Moffett » Sat Aug 31, 2013 5:13 pm

Carl, you're absolutely right, as I was able to discern from my back-of-the-envelope calculations. I need to read up on fusion modelling to decide for myself whether any sort of simulation at all is worthwhile to pursue for my purposes. Do you know of any particularly good books?

User avatar
Monroe Lee King Jr
Posts: 105
Joined: Fri Aug 23, 2013 3:57 am
Real name: Monroe Lee King Jr
Location: Lampasas, Texas
Contact:

Re: Fusor Computer Modeling

Post by Monroe Lee King Jr » Sat Aug 31, 2013 6:29 pm

http://www.efda.org/fusion/focus-on-jet ... n-plasmas/ Here's a nice basic look at fusion modeling.

Monroe

What has happened here is I was talking about some basic modeling for fusor design and it got turned into a discussion about what someone else wanted. Indeed complete modeling of plasma physics is difficult. What I meant was something much simpler. However I would also like a complete model indeed and the more the better and eventually so.
Adding some of the effects from SIMION and such was just an idea I did not intend the model to be based on such advanced work, but I did get caught up in the idea. I guess I'll just have to write my own modeling software.

Sorry I even brought it up
Today's the Day! We go into Space!

User avatar
Carl Willis
Posts: 2841
Joined: Thu Jul 26, 2001 11:33 pm
Real name: Carl Willis
Location: Albuquerque, New Mexico, USA
Contact:

Re: Fusor Computer Modeling

Post by Carl Willis » Sat Aug 31, 2013 7:06 pm

some basic modeling for fusor design
What do you mean by this? If you can communicate clearly what you want out of this "basic modeling," then it will be infinitely easier for people in the community to point you in a viable direction (or tell you what the challenges are).

-Carl
Carl Willis
http://carlwillis.wordpress.com/
TEL: +1-505-412-3277

Post Reply