Discharging C2 with parallel B mosfets probably with open up the body diodes of the A mosfets.

So connecting to C4 makes more sense.
the dc offset energy is then again stored (partly) in C4 for reuse.

I should make a new drawing again

best to postschematic ang geber And diagram as a open source summary os people can try it DD

Quote from securesupplies on February 9th, 2023, 10:49 AM

best to postschematic ang geber And diagram as a open source summary os people can try it DD

indeed, good idea

pcb's will arrive today, parts should be here tomorrow or Wednesday.
meanwhile sitting in the sun.

death can be surprisingly beautiful. bringing transformation in deep levels of life.
grief is only a part of it.

People come and go.
until a next life time. :rolleyes:

looking good as always. I hope they will work as intended.

nice boards
 can you post gerbers or put on
OSH Park ~https://oshpark.com

like to try replicate as it will be important for other to do same
and re validate

DD

Quote from securesupplies on February 13th, 2023, 06:24 AM

nice boards
 can you post gerbers or put on
OSH Park ~https://oshpark.com

like to try replicate as it will be important for other to do same
and re validate

DD

yes if itt works
first debug

parts arrived. building testing soon

looking great

found small mistakes only, in cap placement

All voltages are correct (5.1V) no chips overheating.
Led is on. output A1 and A2 are working, from the inverting 4049 buffer.

When I connect a 4017 output to the hef4528 input, and the next pin of 4017 to reset, the led goes out, and the on the output B1 B2  no signal.
So something is wrong.

lets see if I connected it right.
this is a screenshot of the cascaded 4017 counter IC's connections

ok the led works, blinks. the reset pin works, so I can make delays.
The out from 4017 arrives at the 4528.
but there it goes wrong.
input pin 4 shows the pulse.
but the cap on pin 2 doesn't show a pulse.

it works! the PCB has no faults, I just needed to adjust the trimpot to get a clear out.
the minimal output to the B mosfets is now a 500ns pulse.
I can delay it up to 25 impulses, enough (hopefully) to charge the DC offset.

that is, with a 5k trimpot, ant a 470pF cap on the 4528

ready for testing
will probably rearrange the settup, to shorten the discharges wires. also use other wires.
but for first (low voltage) testing this is fine.

I made the setup so I have positive voltage impulses from L1
and a negative dc offset on L2.

C2 is discharged to C4 negative side.
long leads would also make it ring (bad)

As expected there is a lot of ringing, from the high currents, and long cables.
Ideally C4 C2 and the B mosfets would be very close together, with short leads allowing high currents without to much ringing.

I still can't see the discharge of the DC as I would like to see it.

I recorded my first test, an made an unlisted video here

https://youtu.be/Z1iR2Opdul0

I'm also considering discharging the DC offset of L2, on the C6 side at the time of the impulse.
the impulse on the C2 side of L2, already discharges the dc offset during the voltage impulse.
If I can discharge the C6 dc, at the same time, it would create an interesting voltage over L2.
C6 is at the outside rim of L2.
But... C6 is large, much larger so it will produce much higher currents.
I wont dare to put mosfets to conduct these.

but there is another option:
Series SCR (thyristor), simultaneous switched by a transformer

my delayed pulse pcb works but can be simplified.
only one out from the 4017 is needed which can go directly to the 4528
 and only one reset is needed which can directly go to the pin 15 (reset) of the first 4017
So instead of 3 rows, I only need 1

Because the dc offset created on L2, by the L1 impulse, is slightly less in voltage due to losses, The peak of the impulse on L2 becomes positive, which in turn opens up the body diodes of the B mosfets.
greater isolation of the high voltage DC might make it better.
I don't expect to much problems from this as it only is the top of the impulse.
The impulse is used to excite L4, which should still work. (needs further testing at higher voltages)
So the positive peak of the impulse is passing though the body diodes of the B mosfets to the negative side of C4, which in turn will be slightly deminished in voltage. But I assume this wont be a problem either.

B mosfets, should be placed between C4 (drain) and C2 (source), with short thick leads.

the DC offset is greatly deminished during these first tests, when the B mosfets are working.
Instead of seeing a DC discharge and recharge, the voltage stays low.

the impulse has less peak voltage, but still should charge the DC up. Where is the energy going?
Why is the DC so low? Maybe 4 steps is simply to little. lets double it to 8 impulses per discahrge

when the impulse becomes positive it opens up the B MOSFETs body diodes...
does that mean there is a path for the dc to discharge through?
is that whu the dc doesn't charge up as high as it does without the mosfets connected?
hmm

Im more and more in favour of discharging the c6

but then I need to place the dc offset diode at c2, which is fine, until that also means I get less dc. maybe it will be enough...

I measured at the pwm cap resistor junction of the 4528
scr275 in green shows what it should be. this is with the power off to the coils.

scr276 shows what happens when I insert enough power into the coils.
noise.... way too much noise.

I had this once and solved it with a PI filter. but wonder if I also should need a faraday cage box.

its clearly visible in the yellow trace that the b mosfets are false triggered.
What a mess....

I did use ground plane, and even used a +5 plane on the other side of the pcb
the white and green wires are from the reset and delayed out from 4017.
I already replaced the white, making it a wire bridge.
green... might be ...
oh well, lets check it all, and see where the problem comes from. its what it is

ok, the noise comes directly from the power supply.
The 12V is very unstable,
and the 5V just follows it...

So... lets start with upgrading the power supply.
its a wonder the mosfets even work. but they have filters on the pcb, so that makes sense.

green shows the noise on the 5v and 12V

pi filter made of 75uH and 2x 4,7uF on the 12V input of the switch board.
measured the 5V noise... still a lot
scr281 shows the new ripple
scr279 shows the old.
still room for improvement.

I am only having an inductor on the positive line.
makes no sense to me.
going to add one on the negative line as well (75uH)

I added another 75uH on the negative psu line.
scr282 with, scr281 without.

so that's not working,
instead I can better put more pi filters in series.

now the pulse out of the hef 4528 is around 500ns minimal.
might want to investigate if the 4047 is quicker
https://www.ti.com/lit/gpn/cd4047b
it has 2 less pins, is newer and probably a better option overall

scr284 shows pulse at 4528
scr285 shows 5V at lm317
scr286 shows 5v at pin 16 of 4017

made another pi filter at the 5V out of the lm317.
used 22uF with 2x 330nF
capacity is to low. but it also has 1uF at the IC's side.

I lifted the middle pin of lm317 out, and placed a 22uH inductor in series, with 330nF to ground on each side of the inductor

just a little bit of attenuation, maybe even nothing if I look at the scope shots...
might have connected it wrong somehow?...
dinner break  time.