Last cap in CW always blowing up

[JoeBusic]

Did you measure the transformer peak output as connected to the multiplier stack or was the transformer output disconnected? You can get very different measurements with and without a load on the transformer.

It was measured with 16GOhm resistor load. I saw your apparatus. Nice idea to make a permanent voltmeter on the transformer output. Will make that, too. Now, while thinking about it... I have a 24V PSU and turned it down to lowest voltage (21V). I'll get another one at 20V to manage adjustments between 16V and 22V. It should be easier to manage the transformer output.

[MRMILSTAR]

The meter I think you are referring to is the one mounted in the wall of the acrylic chassis. That meter only measures the input voltage to the ZVS circuit to insure that it will not be damaged by insufficient voltage level. That meter is indeed permanent. To measure the flyback output voltage, I built a stand-alone full wave bridge rectifier that I only connect temporarily to make the measurement. That is also explained in my posting.

[JoeBusic]

The meter I think you are referring to is the one mounted in the wall of the acrylic chassis. That meter only measures the input voltage to the ZVS circuit to insure that it will not be damaged by insufficient voltage level. That meter is indeed permanent. To measure the flyback output voltage, I built a stand-alone full wave bridge rectifier that I only connect temporarily to make the measurement. That is also explained in my posting.

Got it. Anyways, I measured the voltage before and got 15kV with 19.5V from the PSU. Now I have 21V from the PSU. Output from the transformer is probably a little higher than 15kV. As I ordered 50kV caps, that voltage is of no issue any more. I had 40kV caps blow in the same manner, which is why I think that voltage is not the source of trouble. I will make a permanent voltmeter to check the CW continuously. In that way I'll always know is all OK. I'm thinking about making a box out of polycarbonate and fill it with oil to isolate it fully.

[JoeBusic]

I thought of making this to balast the current. Just in case the idea of too little discharge with full charging can destroy a cap. But the resistors are way to expensive for me now.
Plan is to check now the corona possibility. Two tables joined and the circuit is stretched lengthwise. Every stage is higher by 2" with stretched out cables mid air. I have parts for a 3 stage CW so tomorrow I'll be smarter. Solders will be flooded even more, ergo, ball solders all over. If coronas are not seen nor heard and the cap blows again, only one thing remains.

[klugesmith]

I don't understand how that last simulation got its numbers.   
There's over 1000 watts of power in the 20 megohm resistor. 
Does the multiplier work at that power level, even if the diodes and capacitors are ideal?

[MRMILSTAR]

Making an adequate load resistor can be expensive. My load resistor, rated for 400 KV and 300 watts, cost almost as much to build as my CW multiplier. My load is air-insulated so I had to use special construction techniques to avoid flashover. I had to use a lot of spacing, smooth curves, and acrylic separators to prevent flashover. My load is about 1 meter high.

[JoeBusic]

I don't understand how that last simulation got its numbers.   
There's over 1000 watts of power in the 20 megohm resistor. 
Does the multiplier work at that power level, even if the diodes and capacitors are ideal?

Yes. I got it down to 140W but still expensive.
But it's not usable power, it's calculated power at the effect. It's necessary for the choice of components. My PSU is 400W, ZVS is 300W, so the CW can output max 250W. Calculated effect power can be quite high and some people think they got an overunity device. Essentially, CW power is unusable for work. It's made for special physical effects per need.

[NyaaX_X]

Maybe some photo of broken capacitors part? Are they cracked? or burned?

You said "blow", so maybe they were burnting in the epoxy cover?

Also, I am watching those circuit boards with copper pads..

[JoeBusic]

Maybe some photo of broken capacitors part? Are they cracked? or burned?

You said "blow", so maybe they were burnting in the epoxy cover?

Also, I am watching those circuit boards with copper pads..

No signs of burning. They just flash and stop working. I have several now not working. Some have lost majority of capacity but some have (by multimeter) full capacity but don't work. I don't have HV cap testers.
I'm also tinkering about that board. My next one will be from polycarbonate with drilled holes, just in case. But if it were the copper to blame, I believe it would be visible. I was literally staring at close distance when it flashes. I only saw oil moving, but nothing else besides the flash.
Btw, the flash occurs at the in leg of the cap. It suggests overcharging, I assume.

[NyaaX_X]

I made one without board before.
https://youtu.be/Up69LFPi8-M?feature=shared

^^^^^ Loud Noise ^^^^^

And the ZVS I used 8V input. I think the arcs were quite short. But I only have a small size AC transformer. That is the reason...

[JoeBusic]

Parts are here. Building time. In the meantime another CW, made of remainders of the last one, stopped working. I have learned a few things in the process. Like if multistrand cables are used, when removing the insulation, none of the strands should be cut. If only one is cut, there will be coronas. I tested one CW potted in epoxy and got a lamp. Solders were not good and there were many sharp edges from the caps and diodes legs. I have repeated soldering with care and no coronas were there. In a dark setup, there were no coronas anywhere. But after 2s, the CW stopped working.
Got some acrylics to get things in better shape. 50kV caps will remove any doubt about the output from the transformer. Resistors are here for testing and to slow things up. I need it for 3min to work. So I was thinking if I get to 1min, make a 2s pause and punch another minute. 2s should be enough to discharge caps. 3 times 1 min and I got my 3minutes.

[JoeBusic]

It works. Had some situations but all were resolved except the meter. Will do that on Monday. So, what I did...
Install the resistor.


Make the Vmeter.



It didn't work, so I removed it. ChatGPT says the resistance was too low for my setup and should go 10x higher.
Some solder balls to kill coronas.


I still have some connection with slight coronas, so resoldering will be done. Interesting to see that HV connection without insulation has no corona when the soldering is good. I'm still a novice...

[JoeBusic]

Working for now. I did have coronas with resistors and decided to bath them in oil. It helped. Interesting to see the oil movement in the CW and the resistor bank bath. Sounds like an interesting idea to invent a HV mixer