I made a solenoid bifilar coil, on a 8cm diameter PVC pipe.
connected it to the positive and negative DC power supply,
And used the mosfet switch, to open and close the series connection between the 2 windings.

I put a diode between the drain and the winding that connects to V+
and another diode between the source and the other winding that connects to to V-1

C1 =1980pF is parallel to the drain and source of the mosfet (as it was before)

Ieft the V+ connected to earth ground but, this should also work without earth ground.

PSU=0.15A 7.0V dc
f=73.1kc/s

scr576
yellow is measured between the diode and the positive coil
orange is measured between the other diode, and the negative coil

SCR577
yellow is measured between the source and diode, so the positive leg of C1
orange is measured between the drain and the other diode,, so the negative leg of C2.

Because the coils are coupled, they form a capacitor, when the mosfet switch is open, charged by the PSU.
the coils keep resonating after the mosfet switch has opened up the series connection.
Weird is that the C1 also shows this ringing voltage, it should be connected by the diodes, but I guess, since the mosfet is floating (isolated) it drifts along with the changing voltages of both the L1 windings. since I measure referenced to earth ground.

Hmm... I should remove earth ground (maybe only connected to the isolated gate drivers, not PSU)

edit:
WHY is the orange signal changing phase after the diode, and the yellow not?

edit2:
I can't fix the ring, by placing diodes parallel to the windings.

The DC part of the circuit, can have a diode to ground.
only the AC part (L3) needs a direct path to earth ground.
this means the drain, and the C1 can be connected to earth ground via a diode,
So the current which might be sucked in by the C1 discharge is then stuck in the system.
This would make it more clear to measure.
So to be clear, the PSU also would be grounded through a diode.
The diode should pass the negative current from earth ground into the system.
While L3 is directly earth grounded (bypassing the diode)

I tried several ways of creating feedback into L1, which all didn't work, and instead only increased the load.

L1 needs to be a pure inductiance coil, with low capacity. So I could make a air solenoid coil with large winding separation, to reduce the capacity.
This L1 coil is then to be kept separated from the coil stack.

L2-L3 keep working as the coil capacitor. L2 being parallel series resonant, and parametrical excited,
As C1 is not part of the L1C2 resonance, when it is discharged (50%) of the duty cycle).

the high currents of L3, which are not magnetic but dielectric, are then used to induce resonance in a loose coupled L4.
L2 is the primary for the L3 secondary (current impulse dielectric induction)
so L3 is then the primary, and L4 is the secondary (displacement current dielectric induction).

Let's see how L4 behaves and what happens when it is tuned properly to the L3.

so L1 not coupled,
L2 L3 close coupled
L3 L4 loose coupled

I loose coupled L4 to L3, 15mm distanced. both coils rotate in the same direction.
L4 again copies the voltage of L3, at a lower amplitude.
L4 was outside rim grounded like L3 is.
L4 had no tuning capacitor
F=71.5kc/s
psu=20.0V 0.63A 12.6W
green is L4 inside rim
orange is L3 yellow is L2

So this is just a first setup test.
C3=14nF which is small, so the current will be also low (didn't measure)
C1=C2=1980pF like before

Now I need to tune, for high currents around L3, and then tune L4 to L3

I wonder if L4 will draw in even more current when I increasd the impedance

If I remove the loose coupled L4, the L2 L3 frequency barely changes.

ok, Same setup as previous test, but, now I added C4 parallel to L4. C4=30nF C3  still is 14nF C1=C2=1980pF
f=48.5kc/s So as expected the frequency now is lowered by adding capacity. (before was 71.5kc/s)
PSU=16.1V 0.61A dc =9.8W (before was 12.6W, so less input power for this test)
L3 L4 is still 15mm separated loose coupled
SCR578
Green=L4 voltage orange =L3 L2=yellow

SCR579 now orange is the current of L4
SCR580 now orange=current L3, green= voltage L3 (!)

Note how C3 is only 14nF and C4 is 30nF, but voltages are almost equal in size. for L3 and L4
And note how, the current of L4/C4 is much higher.

This is counter intuitive, but very very interesting.
Also note, that L3 and L4 are IN phase.  (L4L3 both are wound in the same direction).
So... if this is the TEM mode
Let's see if the LMD mode is ...
 interesting...

I tuned up to the LMD mode, and found it at 102.2kc/s
I increased the PSU to 25.0V 0.58A=14.5W
scr581  green=L4 orange=L3 yellow= L2 all voltage
note how now, the L4 voltage is lower then the L3 voltage. which is as expected with a bigger cap for C4

SCR582 green=L4 volts, orange=L4 AMPS
amps look pretty decent not excessive

scr583 Green=L3 current Orange= L3 Voltage

So now all works as expected. L3 has higher voltage, lower current with smalle cap
L4 has lower voltage higher current with larger cap. (all coils are equal)

I wonder... did I make a measurement mistake, with the TEM mode, where L4 is larger instead of smaller

I should put my shield back in place.
I feel completely drained

I keep forgetting...
I can redo the video test whereby I charged a capacitor through a bottle of distilled water.

But this time use the L2 square wave.
a gold electrode.

and most importantly, No diode at the cap.
since the displacement current will flow dominant in one direction.

I could also make it more visual by having a load, but that again would demand the use of a sparkgap for more power.

I removed L4 from the stack, to see If I can get a max current point for L3, by changing C3.
Yellow= L2/C2 (inside rim) voltage
Orange= L3/C3 (inside rim) voltage
green= L3 (inside rim) current

instead of typing everything down again, I attached a photo
Note, that the L3 voltage measurement, is including the ripple from the switch distortion.
That's Why I did another measurement for the last one with the cursor.
It said, 545Vpp but really it was around 482Vpp for L3 (orange)
With 31nF there appeared to be an increase of voltage and current, So I am going to measure there again, with smaller steps (1nF C3 change)

I did more measurements, and found 35nF for C3 gave most voltage for L3.
f was then 47.9kc/s psu was 13.4V @ 0.52A (scr598 and scr594)

picture attached has all the data, not all in order!

I'll now Add L4, and see If I can find a maximum amp point for L4.
screen is back in place, and I still feel vital again today after being drained yesterday (slept during the day)

I added L4 back to the stack, 15mm distanced from L3 again as before.
C4 is now 61nF which is its largest value for the cap tuning board.
Since C3 is now much larger than before (35 instead of 14nF)
I will probably need a much larger cap for C4.
psu=13.4V @ 0,43A
f= 69.4kc/s (much higher, I want to get back to 48kc/s)

yellow= L4 voltage
green= L4 current
orange= L3 voltage

so... maybe I can add another tuning board in parallel, or... I think it might be better, to just solder 60nF to the board, and start from there.

probably need many more tuning rounds, to get a grip on the situation, and get it tuned right.
so many variables.
haven't even played with the distance of L3 L4

also, since I did the L3 max current test without L4... It probably will need to be redone.
So Ill go back and forth, to see if I can get a max power point. L4 needs to get max voltage and current.

And again. it is hard to compare, as the freqeuncies change. And the power change.
everything is dynamic...
So I am mostly looking for a non linear jump, like I saw in the voltage of L3.
But now mostly for the L4 max current.

I should make a formula, including frequency input power and L4 current, to get the best compare.
L4A/f/P kut I used to be good at these formulas, but I steered away from them so long...

i felt the L3 L4 where to close so I gave all 4 wooden screws one more turn. now the opening is around 17mm

with the same settings as before,
f=66.2kc/s
psu=0,44A 13.4V dc

freqency is lower, current and voltage is higher in L4, psu amps are a bit higher.

I increased the distance to 22mm between L3 L4, by given the wood screws 2 more turns.
psu 0.39A 13,4V dc
f=61.1kc/s

now L4 current is lower, while voltage is higher

reduced the distance again with one turn, now around 18 mm separation.
f=63.6kc/s
psu 0.41A 13.3Vdc
yellow=L4 orange is L3 in these tests. green is L4 current.

these coupling distance measurements are to vague.
I'll keep it at this 18mm (measured) for now.
c4=61nF C3=35 with these tests.

C4=45nF
C3 still 35nF
f=71.9
PSU=18.0V 0,55A dc
orange now is c2/l2
green yellow  still L4

the charge/discharge of C1 looks good, but still very low, 200V

From the theory of parametric excitation of resonance, I need the parametric change to be twice the resonant freqeuncy.
But... here the change is by C1 being resonant when charged, and not part of the resonance when discharged.
so... how do I get that right?
C1 C2 L2 now shows around 6 periods, for each period of charge/discharge.
So, I should tune much higher. hmm... or.. I should increase the C2 C1, which would also lower the c1 voltage, but I got power to spare. but then again... the mosfets can get hotter with a larger c1 cap discharge.
hmm... I dont know. I looked into thart before, and with larger C1 C2 it also got a high frequency ring, if I remember correctly

If L3 is in dielectric longitudinal mode from the dielectric induction,

And L4 is in magnetic mode,

then L3 is faster then L4
also L3 is close coupled to L2 which will make L3 faster

and L3 should have a larger cap to slow it down,
while L4 has a smaller cap to keep it fast.

then both longitudinal L3 and transverse L4 can sync up.

Pure mental fart probably, but worth looking into.

Also...
I could compare L3 L4 with both current clamps and voltages and see how they compare. that would give the best insight.
my other scope could be used to keep an eye on L2 and the square wave.

I did the test of Reply #381, and gathered more info.
This time measured voltage and current of both L3 and L4 coils. in TEM and LMD modes.
In the lower TEM mode, it is very weird, to see the L3 current rippling, and being relatively very low,
While in comparison the L4 is much higher in current.
Is the displacement current of L2/L3 directly flowing into L4? and amplifying it?

all tests, C3=14nF(orange) C4=30nF (yellow) distance between L3 L4 is 18mm

TEM:
f=49.76 kc/s 
psu=14.7V 0.56A =8.2W
scr599 green= L4 current   4.6App
scr600 green = L3 current  1.4App including hf ripple!
 LOOK at the PHASE! its 180 degrees out of  phase with voltage of L3 (orange)
This is a VERY strange reading.
it is as if the L3 current is directly used to amplify L4. Which is a good thing.

LMD:
f=93.46 kc/s
psu=14.7V 0.35A
scr601green=L3
scr602 green = L4

I will make small steps, and see if I can amplify the current in L4 more

to make it complete,
here is the data of L2.
orange= still L3 voltage
yellow= L2c2 inside rim voltage
green is L2 inside rim current
psu 0.55A 14.7V
F=49.92kc/s
C3=14nF C4=30nF
all the same as previous test

Look at that current in L2, 5.67A pp that is even bigger than in L4.
While, L2 only has 1980pf in series and 1980pF in parallel, while it is close coupled to L3 which has 14nF in parallel.

Very very strange readings!
Also look at the phase of the current again!
It is LEADING instead of LAGGING. So not 90 degrees, but 270 degrees
WTF
Current probe arrow is pointing towards coil as always

yellow is voltage
green is current of L4 inside rim.
indigo is a power measurement,
lets see if I can some of that juice out.
630W pp so 315W from cap to coil, and 315 W from coil to cap back.
This is NOT RMS power.

Still a large amount of power, for only 8.2W input....

what are you using to harness the power from the kicks?