Working Kapanadze Generator Circuit!

[Dog-One]

I have a few components around here I could
experiment with and see for myself.  That's what we're here for right?

Well I tried it.  Used a 7KV step-up transformer for the input and an
ignition coil for the output transformer, configured step-down.  I used
an OPA549 amplifier to drive the input transformer and took my
sine wave signal from my signal generator.  Used about a 1.5mm
spark gap for the arrestor.  For the capacitor, I happen to have
a few 20KV 1000pF pieces.

Once I had the circuit hooked up, I was able to find the resonant
frequency at 26.4kHz.  I slowly up'd the input until the spark started
jumping.  What I saw was the output lamp get dimmer and the
current draw went up a lot.  There were some high voltage spikes
on the output, but they were so short it didn't boost the output
lamp at all.

So at the moment, I'm not really convinced.  I tried probing around
with a ground wire and could get a small arc to jump to it, but no
power increase.  Maybe this thing would work if I was messing with
500kV, but like most of us, that's way beyond what I can work with
on my bench.  It could also be the driver circuit has to be modulated
instead of being a steady sine wave.  I didn't see any reference to
that in the patent, but to me it sounds like something that would be
necessary and not disclosed--basically to hide from anyone all the
necessary ingredients needed to make this thing work.

Not sure what else to try, if anyone has some good ideas, I'll leave
this thing setup for a few days.

[Dog-One]

There's more going on here than is explicitly stated in the patent.
There are three capacitors in this circuit for a reason and that "signal"
line between the two capacitors in series isn't just for sampling voltage.

I think we are looking at a parametric resonator solution.  The driver
box is effectively shunting and releasing C₂ in-sync with the AC
frequency it is driving the input transformer.  When C₂ is shunted, the
overall capacitance of C₁ & C_X goes up and conversely, when C₂ is not
shunted, the overall capacitance goes down.  This is where you would
get a voltage gain at each cycle.  At the zero cross of the cycle, you
shunt C₂ and allow the voltage to increase and just before the peak,
you open the shunt, the capacitance drops and the voltage has to
increase.

It's a similar method to coil shorting, only instead of locking the
current in an inductor, here we just remove one of the series capacitors
from the circuit and add it back in right when we are near peak voltage
in the tank circuit.  I would suppose then, the voltage seen by the
secondary of the input transformer also increases and being a transformer,
the primary side voltage would also increase, which explains the
arrestor on the secondary side, which is there to prevent the control
box from exploding.

So what I can't really figure out is how the control box is driving the
input transformer.  It states its input is AC220V and if rectified, you
only have a limited voltage there and I would suspect the current
is fixed to some reasonably low current.  Otherwise you're just pouring
in watts which will eventually lead to self destruct.

In the past, I played around with parametric oscillators, but I did my
switching on the inductor side, kind of like Jim Murray's SERPS unit.
It didn't work well because you end up dealing with extreme back EMF
that squirts out and smokes transistors unless the transistors are super
fast and even then nature finds a way to bite you in the ass.  This
Chinese patent is rather novel in the way it uses capacitors to get the
same effect, but it surely isn't clear about the principal of operation.

I'll do some more experimenting with this.  I think I'll need to do
it with lower voltages though.  I don't have anything that will switch
4000 volts for sure.

[Dog-One]

What we need is a "Smart Switch".

I'm thinking a power JFET would work nicely since it can switch
either polarity.  We place this switch across C₂ and control it
as a self-contained module.  What it would need to do is when
it senses zero voltage across the cap, the switch shunts and starts
a delay timer.  When the timer expires, the switch opens and the
cycle repeats.

In operation, we adjust the delay timer proportionally to the
frequency the whole system is running at.  With a parametric
resonator like this, there are two frequencies:  One is with C₂
shunted and the other is with C₂ in-circuit.  The overall run
frequency will be something in between, tweaked by the delay
timer.  Because of this, we wouldn't want the Q factor to be too
high, so some resistance is a good thing.  Too much resistance
and we'll burn up all the power gain as heat.

Another possibility that might work is to use lossless clamps
on the input transformer and just recycle the power that way
without any output transformer.  I would imagine we still need
some kind of inductor in series with the capacitors.  Kind of
doubt it would work using the secondary of the input transformer
as the only inductor.  Using lossless clamps, the whole system
is by-design looped, if it works, you'd know when the power
supply amperage drops to zero.

[Dog-One]

Here's a starting point for experimenting.

Q Factor close to 10, with overlapping frequencies.  I would expect
with this configuration, one might actually see some per-cycle voltage
increase if you get the timing right.

If we set the base frequency to 30kHz (33.3us/cycle) and the delay
timer to 1/8th of a cycle, it should be roughly 4.2us--obviously adjust
to best performance.

Anyone good at putting together a simulation to see what this
might actually do?  SolarLab, this is right up your alley.   

[Dog-One]

Not so good.  The simulator doesn't calculate capacitors being
switched in and out of the circuit dynamically.  It tries, but the
timing pretty much gives it away that it can't handle it.  Those
pulses on the green (V1) trace should line up at the start, exactly
on the zero cross, which they do if I run the simulator with the
switch disabled.  And since that doesn't work right, I'm pretty
certain messing with the component values isn't going to show
me the parametric voltage rise I'm looking for.  I actually think
the simulator is trying to modify the frequency without taking
into account the stored energy in the system, which is probably
not realistic.

Anyway, here's the idea and the sim file to explore.

To really test this, I think it will take real hardware.