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

Re: Lumped vs. T-line - You be the judge...



Hi Bob,

At 08:56 PM 04/23/2000 -0400, you wrote:
>Hi Terry and all,
>
>I can not quote your post because I deleted it by mistake.
>
>> original Poster: Terry Fritz <twftesla-at-uswest-dot-net>
>
>
>You  suggest that a secondary may look like a short transmission line with a
>series L.
>
>Do you mean that the distributed L C and R can be approximated by that  or
>.its an equivalent circuit.?

The first section of the secondary seems like it is mostly just an inductor
according to field plotting.  Near the top where the capacitance and high
energy is, I think is much more like a transmission line.  Putting and
inductor in series with a short transmission line as a model may be useful
but I have not had time to play with such a model to see if it works.

>
>The lumped model I have sent you is an attempt to represent the true
>distributed L C and R as ten lumped stages it is not an equivalent circuit
>in
>the sense that it has not be reduced to a minimum required to reproduce the
>behaviour of a coil.  A major difficulty, well at least for me, with the
>lumped view is that there is apparently no equivalent model that
>correctly represents the distributed L C and R and resonates at more than
>one
>frequency.  Your suggestion appears to have a similar  problem.

I think a model like your's has great promise.

Bob's model is at:
http://users.better-dot-org/tfritz/site/misc/BOB.ZIP

I tried a simple test with it at:
http://users.better-dot-org/tfritz/site/misc/BobMod.gif

I think getting the coupling and inductors set up is fairly easy.  I
guessed at the capacitor values from the chart at:
http://users.better-dot-org/tfritz/site/misc/Efield.gif

I tried making the things all proportional and then correcting for Fo
frequency.  My actual coil had a few harmonics this did not predict but
this is obviously all still very young...

The distribution of C along the coil may fit some closed form equation but
trying to figure out a true T-line equation with a function of C rather
than a constant C may not be easy at all.

>
>Assuming your L is at the driving end then your equivalent circuit will
>resonate when ever the transmission line looks like C which will be at
>frequencies less than its 1/4 wave length.  This  is directly equivalent to
>an end loaded line with C.  I have not done the maths but I assume the phase
>shift caused by the L/input impedance will be frequency dependent.  So I
>suspect this may be a good equivalent circuit that represents both the multi
>resonances and dispersion of a secondary as opposed to a tuned circuit.

The series caps would help account for the fast rise while still keeping
the T-line harmonics.  It has the mutual coupling and variable capacitance
that I think is very similar to the actual coil situation.  It could use a
little resistance here and their but I am running out of nodes on the
crippled version of MicroSim.

>
>Because its an equivalent circuit for the input to output and does not
>necessarily model the true distributed L C and R it will not necessarily
>model the distributed voltage of the secondary  ( just like the turned
>circuit
>equivalent circuit).   However it would have the advantage of more
>accurately
>modelling the effects of the initial transient when the gap fires.  It may
>need a fiddle factor on the output voltage as it may not be possible to
>simultaneously have the correct input impedance for the primary, the correct
>output impedance and resonate at the right frequencies with the right phase
>delay.
>
>I previously suggested one or more resonance periods for a good FF..  That
>is
>incorrect due to the dispersion and other reasons. My best advice now, is as
>you suggested, use a long time sample so the end effect is less preferable
>decayed to zero if you have sufficient samples.
>
>What is the maximum number of samples you can take?

It always takes 2500 samples of the screen length up to 1GHz

I could increase the square wave frequency and capture many cycles which
should help the FFT to have more points of interest.

>How long In samples to decay to say 10%?

I am not sure what the decay time is.  I know what you are looking for so
I'll play with it and see if I can get useful step response info for a
detailed FFT

>How does your simulation program model a transmission line? (this is a hard
>one)

Just a cut and past from the instructions....

Description: The simulator uses a distributed model to represent the
properties of a lossy transmission line.
That is, the line resistance, inductance, conductance, and capacitance are
all continuously
apportioned along the line’s length. A common approach to simulating lossy
lines is to model
these characteristics using discrete passive elements to represent small
sections of the line.
This is the lumped model approach, and it involves connecting a set of many
small subcircuits
in series as shown below:

>A section of an RCLG lumped transmision line segment repeated a few times.<

This method requires that there is enough lumps to adequately represent the
distributed
character of the line, and this often results in the need for a large
netlist and correspondingly
long simulation times. The method also produces spurious oscillations near
the natural
frequencies of the lumped elements.
An additional extension allows systems of coupled transmission lines to be
simulated.
Transmission line coupling is specified using the K device. This is done in
much the same way
that coupling is specified for inductors. See the description of
Transmission Line Coupling
for further details.
The distributed model allows freedom from having to determine how many
lumps are
sufficient, and eliminates the spurious oscillations. It also allows lossy
lines to be simulated
in a fraction of the time necessary when using the lumped approach, for the
same accuracy.


I have no way of altering the program's internal T-line algorithms... Darn!
;-))

Cheers,

	Terry

>
>Regards Bob
>
>