[Prev][Next][Index][Thread]

Re: Propagation velocity in long helical coils.

Hi Bob,

At 09:41 PM 05/17/2000 -0400, you wrote:
>Hi Terry and all,
>
>As you can read my mathcad files thats going to be helpful.  I suggest you
>start with the zero internal C version as thats the one I have the equations
>for.   Also I have not included ground reflections or top load mag
>reflections. The ground reflections may be important for end effects at
>least thats what Malcolms measurements indicate.  When you or Malcolm has
>some comparative L measurements on the ground and isolated we will know more
>about it.  You will also need inverse FFT.  As in many of these types of
>problems you solve the transformed equation then convert back to get the
>answer you wanted. One more thing I assume you have complex FFT routines
>i.e. unless its a symmetric bipolar coil you need the sin and cos terms. Or
>I can add the ground reflection for symetry that probaly best but a bit more
>work on the equations and we need those measurments as a check on the
>reflection equation.

I have been sent two FFT algorithms.  I seem to remember writing complex
function routines was part of the fun of doing FFTs in BASIC ;-))

>
>Could Malcolm get hold of Wheeler's paper for his L formula so we can check
>its got no inapprorate assumptions ie it is valid for short coils.
>
>I will need to tidy up the work sheet and put it in an algoritmic form.  Can
>I assume the resusltant programme will be freely distibuted?   I am looking
>forward to seeing the current profile of a short coil.  Do we have any
>volinters for some anamated graphic output rotines.

All my stuff is free and public domain...  QBASIC can do graphics (crude by
today's 1600 x 1200 x 16M 30 FPS stuff but...  ;-))  It may be best to
generate files and have one of the fancy plotting programs do the pretty
frames for an animation.  I think I have all the software toys to do that...

>
>Ultimatly we need the transiant solution so we can watch the current and
>voltage profiles build.  That may require an FFT for each element of the
>current and voltage profile.  Or we could build a finite element simulation.
>Thinking about that it would have the mega advantage of showing how the I
>and V build and it may be possible to patch it into microsim.  Even a 100
>element one would be good and  the distibuted coupling to the primary would
>be a relatively simple. So which one do you want the equations for ?

If you can send me anything, I can start looking it all over and figure a
plan of attack.  The time simulation thing would be really neat!!

>
>Need to do the sums on the computation time for the simulation.  You have
>the self C and current flow in and out and voltage across each element then
>the mL and C to every other element. Just 5 equations per element So 3N +
>2N**2 calculations per run cycle. Say 50 run cycles per coil cycle (depends
>on the integration and diff algorithms).  So for one cycle and 100 elements
>its one million equations. Say 10 million floating point operations.   How
>does this compare to the self C program.  Both  would be a nice cross check
>and the maths one normally provides more insight.  I think ground
>reflections are simple for both due to symmetry.

I have a dual processor 400MHz Pentium with 256M RAM and hard drive space I
have never bothered to format.  80% of the time it is turned off...  Number
crunching is not an issue ;-))  I think it can do 10M floating point ops
faster than it can refresh the screen.  E-Tesla5 can approach a trillion FLOPs.

My lack of coherent programming skills and QBASIC may be a limitation...


Cheers,

	Terry


>
>Regards Bob
>