Energy Balance Formula for Polywell

[MSimon]

- 20 MW of fusion power is dumped as heat into the ring magnets when reaction products including neutrons hit them.

Whatever else he's saying, this is an interesting point. Am I to assume that the magnets are getting a high pressure stream of LN2 to keep them cool while they're getting roasted with neutrons (in D-D)? I also assume that with p-B11 the helium nuclei will tend to fly off around the magnetic field lines and not at the magnets directly?

John,

First off the problem is not neutrons when running pBj. It is alphas. In fact neutrons would be better since if they are high energy they will not react much with the materials inside the reactor. This may be balanced by the greater number of neutrons produced.

Second taking out the major portion of the heat with LN2 is not going to work.

I expect to use 3 coolants.

H2O at around 120 to 350 deg C (I haven't run the numbers so this is not very exact).

LN2 at 77K

LHe at 10 to 30 deg K

It will be interesting trying to balance all the delta Ts and flows and required pressures to get the heat flows to balance at useful temperatures.

[MSimon]

I guess I'm not that interested. Fact is it's unproven technology, the physics is rather obscure, and Bussard isn't talking.

Bussard has talked plenty. It requires study. This is, as you point out, difficult to understand in detail. Took me about six months of intensive study to get a solid feel for what is going on.

I have shortened that considerably with the material collected at IEC Fusion Technology. Which is why I provided the links.

Which just goes to prove the old adage. "You can lead a horse to water but you can't make him think."

[JohnP]

First off the problem is not neutrons when running pBj. It is alphas.

Right. So now we have added added to our already complex soup of electrons and ions. I would imagine that the plasma net charge would have to be measured closely, injecting more electrons to make sure the overall balance is negative to maintain the well (???)

[MSimon]

First off the problem is not neutrons when running pBj. It is alphas.

Right. So now we have added added to our already complex soup of electrons and ions. I would imagine that the plasma net charge would have to be measured closely, injecting more electrons to make sure the overall balance is negative to maintain the well (???)

Yes. Dr. B so states. He also states that the difference from charge neutrality is about 1E-6.

I'm not sure how you would measure this.

One of the reasons this is so hard to figure out on a descriptive level (let alone the complex math) is that there is so much going on.

You have double layer phenomenon - the well
You have the Wiffle Ball Effect - from circulating currents
You have edge annealing of particle velocities

These to a certain extent are second order results of first order phenomenon (electron injection and acceleration for intstance). However, these second order effects are critical to proper device operation.

[drmike]

If the difference is 1e-6, then the electric field will be pretty huge. I suspect that's a good thing, it helps keep the positive ions at high speed so long as you have more electrons in the middle.

I don't think the physics is all that obscure or complicated, but I'm trained in the art. The hard part isn't the physics, it's the approximations to the physics. The basic physics is just

m_k^u * a_k^u = G_k^u + F_ret,k^u,n * v_k^n

Where the ^u are super scripts for 4 vector relativistic notation and _k is for particle k. G is the acceleration and coulomb term, F is the retarded field from all the other partcles.

If you've got 1e20 particles, that's hard to solve. But you can write down the physics. Since that can't be solved (yet, it's kind of amazing how fast and big computers are getting!) you need to make some approximations.

A lot of people have spent their lives figuring this stuff out. It's not that hard to get up to speed, but you really have to want to!

With luck I'll have pictures of mag fields in a week, electrons in those fields in two weeks. We'll see - sometimes life throws curve balls. While crude, I'm hoping a lot of things will be easier to understand if we can get some pictures. The problem is that we are dealing with some complexity - 3D of B field, 3D of E field, 6D for each particle species (position and velocity) so for just electrons and a MaGrid we're looking at a 12D object!

Crude models can help get crude estimates of power consumption and generation. With out even that level of knowledge, it's hard to answer any questions. Getting that knowledge isn't that hard - it's just time consuming.

[MSimon]

For a few mil I can design you a quantum computer that can solve the equations in real time. :-)

Speaking of time that gives you 13 dimensions to deal with.

[drmike]

For a few mil I can design you a quantum computer that can solve the equations in real time.

Speaking of time that gives you 13 dimensions to deal with.

Yeah, a quantum computer would be nice!

A prime number of dimensions must be a good sign

[TDPerk]

For a few mil I can design you a quantum computer that can solve the equations in real time.

Try designing that for a few thousand and building it for under $10,000.

Check out the Propeller at http://www.parallax.com/propeller/index.asp.

I'm just a fan, I have no financial interest in it.

And Ver2 comes out next next year.

2150MIPS for about $20.00 if I recall.

[MSimon]

The Propeller is discussed in Controls section of design.

For the same $20 I can get a SEAforth chip that does 24,000 MIPS.

The Propeller design itself is not so hot and they have totally hidden the processor internals re: documentation.

In addition the Propeller is video oriented. The SEAforth is Real Time/DSP oriented. A much better fit for controls.

[TDPerk]

...and but for this I will not reply further.

Forth users tend to generate write once, read never code--you may not have this as an issue.

The ASM and HLL lang. for the Prop are quite determinate and exactly intended for the control of real world hardware, at which it does quite well. Anyone who knows it can follow anyone else's code easily. The idea Parallax has anything to hide about the Propeller is frankly silly, both block schematics and microcode based instruction construction tables are in their forums and/or documentation, and many languages--including FORTH--have been implemented on it.

[MSimon]

...and but for this I will not reply further.

Forth users tend to generate write once, read never code--you may not have this as an issue.

The ASM and HLL lang. for the Prop are quite determinate and exactly intended for the control of real world hardware, at which it does quite well. Anyone who knows it can follow anyone else's code easily. The idea Parallax has anything to hide about the Propeller is frankly silly, both block schematics and microcode based instruction construction tables are in their forums and/or documentation, and many languages--including FORTH--have been implemented on it.

That is a cannard.

I had a government inspector read my mostly uncommented FORTH code for a mil project. He inspected 5 to 10 projects a week. Fortran, C, assembler, Pascal, ADA, etc. He said my code was the easiest to understand he had come across in a couple of years.

The answer of course is to avoid mediocre or worse programmers.

i.e., if you want something done well hire the best people.

BTW I also had programmers under me. I enforced discipline. And what was the most critical discipline I enforced?

CLEAR THINKING

If you can write clear concise sentences you can write good FORTH. If you can't you should find another job.

[MSimon]

TD,

Please give a link to the register and ALU layout of the Propeller. It may be on their site. It is hard to find.

BTW do you know if the FORTH is a Machine FORTH variant or is it interpreted? Interpreted FORTHs are nice from a quick development standpoint. They do cut your available speed from 1/3 to 1/10th what you can do in assembler or Machine FORTH.

In any case, for bringing up new hardware there is no better language than FORTH. Which is why Sun uses it in all their terminals.

[TDPerk]

That is a cannard.

In your case it may not be true, but it is still commonly seen that much computer code is unreadable, and that in Forth it is particularly easy to quickly develop efficient but unreadable code. The dictionaries are developed idiosyncratically for each application (or they are more general than admits of being as spectacularly efficient as FORTH can be).

You may find most (likely all) of what you are looking for in these places:

Block Diagram: http://www.parallax.com/dl/docs/prod/pr ... ockRev.pdf

Data Sheet: http://www.parallax.com/dl/docs/prod/pr ... DSv0.3.pdf

Counter App Note: http://www.parallax.com/dl/appnt/prop/A ... rsv1.1.zip

The Propeller FORTH is interpreted, and if I recall ran at 1/1 to 1/4 speed of the native assembly (depending on how closely a FORTH word matched an existing assembely function). BTW, the clocks per instruction you quoted were for the interpreted SPIN language, not the native assembly code.

Additionally, if I recall correctly, the new Prop 2 has all one clock instructions, including a multiply (not yet a hardware division, though).

[MSimon]

I have seen the first two already. The third is new.

There is still no block diagram of the register/bus/alu internal layout.

If you find one let me know.

And yes. FORTH is a programmer multiplier.

Bad programmers will be 10X worse. Good programmers will be 10X better.

The secret to good FORTH is good programmers.

BTW on the mil project we had a team of 2 1/2 programmers competing against a team of 30 C programmers at another company. We always beat them by a wide margin. What we could do in a month took them 6 to 12 months. It was very funny. To us. The competition was in constant pain. LOL.

You really ought to go over to the Control thread.

It talks about a guy who was selling FORTH. The buyer wanted to know what he was going to do with 500 C programmers when if he used FORTH he might only need 50 or 100.

One of these days industry will wise up and rise up to reduce the waste.

The reason you need languages to impose discipline is that industry has to hire lots of programmers because they use tools that are not very good for the job. What we like to call in engineering "a cascading failure".

Fire 80% of the programmers. Keep the best. Give them good tools and watch the productivity rise.

Brooks talks about this in the "Mythical Man Month". He says there is a difference of at least 10 to 1 between good and bad coders. Given the right tools I can make the difference 100 to 1.

Any way if I have a say in the matter (doubtful) it will be FORTH all the way. Better, faster, cheaper.

[MSimon]

And yes.

Every FORTH project amounts to writing an application specific language.

However, that is exactly what makes good FORTH readable and reduces waste.

You don't have to use a rivet where a nut and bolt would suit better.