Exact design for lossless SSTC's Discussion
Edited/Updated October 21, 2004
Date : Sun, 29 Aug 2004 20:14:07 -0600. Subject : Exact design for lossless SSTC's
Original poster: "Antonio Carlos M. de Queiroz"
Hi all:
I made some further investigations about the theory of "sstc's", and found how to
design the LC part of the system so complete energy transfer occurs without
breakout, as can be done for a capacitor-discharge system.
I considered the system:
. kab
. o----Ca---+ +----+-----o
. | | |
. La Lb Cb
. | | |
. o---------+ + ---+-----o
The objective is to apply an alternating voltage to the input, and after a
determined number of cycles have all the energy that entered the system in Cb,
without voltage in Ca or currents in La or Lb. I considered the input voltage as
a cosinusoid or as a sinusoid. A square wave input makes little difference.
I could (so far) identify three different possibilities:
a) cosinusoidal input: the two natural oscillation frequencies of the network
and the excitation frequency must be in a ratio of three integers with odd
difference (as 9:10:11 or 9:10:13), with the excitation frequency being at the
center. I found this solution first, almost by chance when looking at other
problem. The solution exists only for the 4th-order system.
b) sinusoidal input: The three frequencies must be in a ratio of three odd
integers with double odd differences (as 9:11:13 or 9:11:17). The excitation is
at the central frequency. The design is just a bit more tricky, and
produces the best networks for practical use, with practically perfect "soft
switching" if the integers differ by just 2. The design procedure allows also
the generation of structures of higher order, as an "sstc magnifier", that
operates with four simultaneous oscillations instead of the three of a
capacitor-discharge magnifier.
c) sinusoidal input: Similar to (b), but the excitation is at the upper
frequency. Less practical because the switching is not soft. There is no
solution with excitation at the lower frequency.
There are many possibilities for the choice of the three integers, as to choose
the three integers so the excitation is arbitrarily close to one of the
resonances, what results in a rise with "bumps", but the switching is not soft,
and the idealized waveforms can be quite different from what a real circuit
generates.
Interestingly, the resulting networks (b) are practically identical to the ones
generated from a Butterworth filter
(see my page at
http://www.coe.ufrj.br/~acmq/tesla/sstc.html), but the voltage gain (without
load) is a bit higher.
I will set up a page with some examples and more details, but if you want to
try, I have implemented the design procedure in the program "sstcd", that can
make the calculations and plot the waveforms. The program (for Windows) is
available at:
http://www.coe.ufrj.br/~acmq/programs
. Antonio Carlos M. de Queiroz.
Date : Mon, 30 Aug 2004 07:50:38 -0600. Subject : RE: Exact design for lossless SSTC's
Original
poster: "Steve Conner"
>The objective is to apply an alternating voltage to the input,
>and after a determined number of cycles have all the energy that
>entered the system in Cb, without voltage in Ca or currents in La or Lb.
Thanks for your post Antonio, this is very interesting. The "sinusoidal input"
case sounds almost identical to what Jimmy Hynes was doing (semi-empirically)
with his original DRSSTC.
However, if I remember right, Jimmy H. found that he got better experimental
results when the oscillator was retuned to coincide with the lower split
frequency. I suppose this would correspond to a:a:b or whatever, except that the
b mode would hardly be excited at all.
I think that DRSSTC's actually function better when one mode is excited far more
than the other, and hence Ca and Cb are both fully charged at the instant of
breakout. If the drive is kept on through breakout, then none of the energy in
Ca can return to the DC link, hence it all must feed into the discharge.
In other words I believe the energy left in Ca is not a "loss" but actually
contributes to the sparks. Hence a DRSSTC driven in this way should give twice
the bang energy (for a given capacity of MMC bank) of a DRSSTC that is optimized
for complete energy transfer according to the theory you posted. It may not give
the fastest energy transfer, but the energy transfer rate is limited by how much
current the semiconductors can stand anyway.
Following on from this, if I were trying to make a SSTC magnifier, I would use
the mode relationship
a:a:b:c
where b and c are "Don't care". So of those four simultaneous oscillations, two
(the drive frequency and the lowest frequency mode of the coil system) would be
at the same frequency and the other two would be minimized. Steve C.
Date : Mon, 30 Aug 2004 21:54:39 -0600. Subject : RE: Exact design for lossless SSTC's
Original
poster: "jimmy hynes"
Hey,
> Original poster: "Steve Conner"
>
> >The objective is to apply an alternating voltage to the input,
> >and after a determined number of cycles have all the energy that
> >entered the system in Cb, without voltage in Ca or currents in La or Lb.
>
> Thanks for your post Antonio, this is very interesting. The "sinusoidal
> input" case sounds almost identical to what Jimmy Hynes was doing
> (semi-empirically) with his original DRSSTC.
Yeah, that's what I was doing. I eventually decided to skip that approach.
Once sparks started flying, it was very difficult to keep a good notch. Getting
bigger sparks wasn't easy either. Getting bigger bangs meant more cycles
and a lower k, so tuning and
streamer loading was even more important.
At one point, my current transformer broke, so I started tuning for largest
sparks instead (too lazy to fix it :p). That method ended up giving bigger
sparks, so I decided to abandon the old idea. The streamer load does keep the
primary current down, so there isn't too much energy just going into the primary
and back to the lytics.
Jimmy.
Date : Mon, 30 Aug 2004 21:54:50 -0600. Subject : RE: Exact design for lossless SSTC's
Original
poster: "boris petkovic"
Steve wrote:
> However, if I remember right, Jimmy H. found that he got better experimental
> results when the oscillator was retuned to coincide with the lower split frequency.
---------
IMO, this is to be expected. Spark loading naturally lowers the secondary
frequency and the lower split frequency tuning option could be experimentally
better for success in energy transfer and performance efficiency. For
classical TC that effect has been known
for a long time. DRSSTC's are bit different transient vvise though.
Neverthless, the reason is likely to be the same. Take care, Boris.
Date : Mon, 30 Aug 2004 21:55:01 -0600. Subject : Re: Exact design for lossless SSTC's
Original
poster: "Bob (R.A.) Jones"
> Original poster: "Steve Conner"
> I think that DRSSTC's actually function better when one mode is excited far more than
>the other, and hence Ca and Cb are both fully charged at the instant of breakout.
I think the voltages on Ca and Cb are 90deg out of phase for each frequency of
the split mode. One leading and one lagging.
So the current is max in one L when the voltage is a max on the C on the other
side. Assuming the transient has decayed. Bob Jones.
Date : Tue, 31 Aug 2004 07:47:24 -0600. Subject : RE: Exact design for lossless SSTC's
Original
poster: "Steve Conner"
>So the current is max in one L when the voltage is a max on the C on the other side.
This could be true, but my general argument still holds- the primary and
secondary tanks are both full of energy at the same time. Steve C.
Date : Tue, 31 Aug 2004 12:06:32 -0600. Subject : Re: Exact design for lossless SSTC's
Original
poster: "Antonio Carlos M. de Queiroz"
Tesla list wrote:
>
> Original poster: "jimmy hynes"
> Yeah, that's what I was doing. I eventually decided to skip that approach.
> Once sparks started flying, it was very difficult to keep a good notch. Getting bigger sparks wasn't
>easy either. Getting bigger bangs meant more cycles and a lower k, so tuning and streamer
>loading was even more important.
This is really a problem. But even with the range of k usual for capacitor
discharge coils a large increase in voltage gain is possible. The problem is
that this may be not enough, since the input voltage is in the hundreds of
Volts, not in the thousands.
> At one point, my current transformer broke, so I started tuning for largest
> sparks instead (too lazy to fix it :p). That method ended up giving bigger sparks, so I decided to
>abandon the old idea. The streamer load does keep the primary current down, so there isn't too
>much energy just going into the primary and back to the lytics.
What kind of input current waveform did you get with this approach? I see
that there are usually three possibilities that result in "soft switching"
(input voltage in phase with the input current): One is to drive the system
exactly between the resonances (what I am
proposing). The others are at the resonances (impressive results, but somewhat
dangerous for the driver). Antonio Carlos M. de Queiroz.
Date : Thu, 02 Sep 2004 07:56:15 -0600. Subject : Re: Exact design for lossless SSTC's
Original
poster: "jimmy hynes"
Hi,
> Original poster: "Antonio Carlos M. de Queiroz"
> Tesla list wrote:
> >
> > Original poster: "jimmy hynes"
>
> > Yeah, that's what I was doing. I eventually decided to skip that
approach. Once sparks started > > flying, it was very difficult to keep a good notch. Getting bigger sparks
wasn't easy either.
> > Getting bigger bangs meant more cycles and a lower k, so tuning and streamer loading was
> >even more important.
>
> This is really a problem. But even with the range of k usual for capacitor
> discharge coils a large increase in voltage gain is possible. The problem
> is that this may be not enough, since the input voltage is in the
> hundreds of Volts, not in the thousands.
Yeah, it did work, it's just that the other way worked better :). I ended up
getting nearly twice the spark length after switching to the other driving
method.
>
> > At one point, my current transformer broke, so I started tuning for
largest sparks instead (too
> > lazy to fix it :p). That method ended up giving bigger sparks, so I
decided to abandon the old
> > idea. The streamer load does keep the primary current down, so there
isn't too much energy > > just going into the primary and back to the lytics.
>
> What kind of input current waveform did you get with this approach?
> I see that there are usually three possibilities that result in "soft
> switching" (input voltage in phase with the input current): One is
> to drive the system exactly between the resonances (what I am proposing).
> The others are at the resonances (impressive results, but somewhat dangerous for the driver).
>
I never got any good scope shots of it running at the lower pole. Slight changes
in the tuning would sorta mess up the waveform, but when it was working well, it
was sorta similar to an RC shape that never fully flattened out. Antonio
Carlos M. de Queiroz.