Last cap in CW always blowing up

[JoeBusic]

I have an issue with a CW setup. I'm not sparking. I need it for science. So the clasics...PSU-ZVS-HV transformer-CW 5 stage. As I need very low current there is a 12GOhm resistor. Subject in the matter is a coil thingy that influences "something" and it's measured.
Now I'm building my 4th CW with higher rating caps (50kV), as I tried 30kV and 40kV. Transformer output is 15kV peak. I got a suggestion to put a 1MOhm resistor before CW. Parts are on their way.
But to be sure what's going on, if anyone knows, as it's the last cap that blows, should I put a resistor (say 10kOhm) before that cap? Or maybe before all of them. FALSTAD says the difference is small, it just takes longer to achieve full voltage.

[JoeBusic]

So I got a couple of suggestions. One is to put a 10k resistor before the last cap and another is to put 1k resistors before all of them and a 10k before the last one. I'll try both methods when parts arrive.
Chip in for other suggestions.

[klugesmith]

It would help if you can show a schematic for _your_ "5 stage CW" and the adjacent bits of input and load.
Even if it's just a snip or link to something published, e.g. by Voltage Multipliers Inc.

You say transformer output is 15 kV peak. Can you tell us about measuring that? And how are you measuring CW output voltage?

[JoeBusic]

I had a electronics student come over with a DIY apparatus for measuring HV output. I believe it's a rectifier and load resistors in it. Didn't ask. We set up the PSU output to get it to a smudge under 15kV.

[davekni]

Pictures of your physical setup might be helpful too.  Unusual to have problems with a capacitor at HV end of a CW multiplier.  Caps at the start experience higher voltage and current.  I'd guess your issue might be voltage breakdown through case of your failing final capacitor, from case to air or oil or whatever insulating medium you are using.

[JoeBusic]

As seen, I have potted the cascade in silicone (for electronics, very high electric resistance and good heat flow). Resistor is potted in electronics epoxy.
I have removed the silicone and will remove the last stage. Will connect all again and record what happens. First few blow outs in oil were very visible. It was a flash at the last cap. Silicone is expensive so I'll use liquid parafine again. I have checked all caps but with a multimeter and all have good capacity. Before flukes were easily spotted. So, I'm not sure what happened this time. But, as the second time I also removed a stage and got the same results, it should be visible again.

[JoeBusic]

Something is not working with attachments. Here, the second picture. And I can't modify the post...

[davekni]

Something is not working with attachments. Here, the second picture. And I can't modify the post...

Yes, attachments took me a couple tries to learn.  My first project post had ten replications of first image.  Necessary to click the "(more attachments)" line below the first attachment to add more.
Modifying is also slightly tricky.  The "modify" button below post does nothing.  No idea why it's there.  The "modify" button above your post is what to click.

I've built only one CW multiplier above 100kV, so am certainly not an expert.  Would be worth your while to look at other large CW multiplier projects on this forum.  The one thing I'd recommend is keeping all wires and anything else conductive far away from CW multiplier.  All input circuitry and wiring should be to left of your potted CW multiplier.  Or better, mount your multiplier vertically with output on top and everything else below bottom of multiplier.

[MRMILSTAR]

When you say the last cap is "blown" do you mean that its dielectric broke down and now the capacitor is useless?

I see that your multiplier is in a small box lying flat on a table. Its possible that the voltage at the final stage is high enough to break through your insulating medium to the table below especially if your components are against the edge of the box. I would try elevating the multiplier above the table and try it. I ran into this problem with my horizontal Marx generator. I have to operate it on acrylic standoffs to get it high enough off the table otherwise it will arc to the table. If we could see the inner layout of your multiplier components it would be easier to judge.

You mentioned that you used silicon and are now going to try paraffin. Although these are good insulators they can easily form voids while poring. The trapped air pockets can allow flash-over. If feasible, I would do the development with oil to ensure that there are no voids and you can possibly see what is failing. . After you get it working you can then possibly replace the oil with paraffin.

[JoeBusic]

  The one thing I'd recommend is keeping all wires and anything else conductive far away from CW multiplier.  All input circuitry and wiring should be to left of your potted CW multiplier.  Or better, mount your multiplier vertically with output on top and everything else below bottom of multiplier.
[/quote]
Cables are long enough. On the left are the PSU, ZVS and transformer. 50cm away goes the CW and a meter later is my coil thingy. Everything is layed on a piece of 8mm thick G10 got from a motor shield company, to isolate from the table (just in case).

[JoeBusic]

An epoxied CW and a picture of the silicone one before pouring.

[JoeBusic]

When you say the last cap is "blown" do you mean that its dielectric broke down and now the capacitor is useless?

I see that your multiplier is in a small box lying flat on a table. Its possible that the voltage at the final stage is high enough to break through your insulating medium to the table below especially if your components are against the edge of the box. I would try elevating the multiplier above the table and try it. I ran into this problem with my horizontal Marx generator. I have to operate it on acrylic standoffs to get it high enough off the table otherwise it will arc to the table. If we could see the inner layout of your multiplier components it would be easier to judge.

You mentioned that you used silicon and are now going to try paraffin. Although these are good insulators they can easily form voids while poring. The trapped air pockets can allow flash-over. If feasible, I would do the development with oil to ensure that there are no voids and you can possibly see what is failing. . After you get it working you can then possibly replace the oil with paraffin.

Yes, the cap got useless.
All pours are vacuumed. Silicone was 20mm thick on all sides of the CW. I have one CW in epoxy for testing. That one didn't blow up but blew my PSU and fets. Have changed the fets and got a new PSU. This one will be also tested to see what happens. For now the lack of proper discharge seems like a valid idea what's going wrong. Before my new components arrive, a few tests will be made with remainders of CW's a have. Will be smarter tomorrow.

[klugesmith]

Can we talk about your 12G resistor?   What's the datasheet voltage rating of the individual elements that you have put in series?   In case you're new to HV resistors, the voltage rating is independent of power rating, and is not affected by oil immersion or solid dielectric potting.  Overvolting can cause significant resistance changes, temporary or permanent, or internal breakdown.

It looks like your application has a 1 mH inductor conducting 12 uA DC while sitting at 140 kV with respect to "ground".
How much voltage appears across the inductor due to 40 kHz ripple on the CW output node? In the presence of parasitic capacitance on both sides of the inductor.

What's the corona glow view when you operate in the dark?

[JoeBusic]

Can we talk about your 12G resistor?   What's the datasheet voltage rating of the individual elements that you have put in series?   In case you're new to HV resistors, the voltage rating is independent of power rating, and is not affected by oil immersion or solid dielectric potting.  Overvolting can cause significant resistance changes, temporary or permanent, or internal breakdown.

It looks like your application has a 1 mH inductor conducting 12 uA DC while sitting at 140 kV with respect to "ground".
How much voltage appears across the inductor due to 40 kHz ripple on the CW output node? In the presence of parasitic capacitance on both sides of the inductor.

What's the corona glow view when you operate in the dark?

Resistors are 2GOhm 20kV a piece. I potted them as it sparked between the connections. No more sparking.
As for the 1mH, it's just for the simulation. Didn't measure it as it's probably way lower. Due to very low current I don't expect more than 100kV output from the CW. Couldn't measure it ass it always blows in the first 3s. Even tried with only 2 stages, with same results. Simulation says the Vd is 500uV and power is 6mW. I don't see it doing anything of significance.
No coronas. Solders are flooded and all connections on the coily thing and it self are all well isolated. I had only one small corona on the first resistor (inside of epoxy) as the solder was not flooded enough. Fixed that for next test.

[davekni]

Cables are long enough. On the left are the PSU, ZVS and transformer. 50cm away goes the CW and a meter later is my coil thingy. Everything is layed on a piece of 8mm thick G10 got from a motor shield company, to isolate from the table (just in case).

How are the long cables routed?  The black overall enclosure looks like it is hinged and intended to close.  That might get cables or other components near CW multiplier.  For example, transformer secondary wired in picture extend all the way along CW multiplier.

Even tried with only 2 stages, with same results.

If two stages fail, why not try one, or at least figure out the issue with two rather than five?  Cheaper to repair and easier to measure at lower voltage.
With two stages failing, kludgesmith's earlier question becomes important, measuring 15kV input peak.  ZVS oscillators often generate momentary high output on startup.  May be simple over-voltage issue.  Though that usually fries first stage capacitor rather than final stage.

[JoeBusic]

Cables are long enough. On the left are the PSU, ZVS and transformer. 50cm away goes the CW and a meter later is my coil thingy. Everything is layed on a piece of 8mm thick G10 got from a motor shield company, to isolate from the table (just in case).

How are the long cables routed?  The black overall enclosure looks like it is hinged and intended to close.  That might get cables or other components near CW multiplier.  For example, transformer secondary wired in picture extend all the way along CW multiplier.

Even tried with only 2 stages, with same results.

If two stages fail, why not try one, or at least figure out the issue with two rather than five?  Cheaper to repair and easier to measure at lower voltage.
With two stages failing, kludgesmith's earlier question becomes important, measuring 15kV input peak.  ZVS oscillators often generate momentary high output on startup.  May be simple over-voltage issue.  Though that usually fries first stage capacitor rather than final stage.

The case is not there when in use. It's for transport. All is layed on a table.
I understand that first stage is under stress, but it blows after the 3rd second. By that time ZVS startup peak is way over. I have left overs from the last CW. Decided to wait for the resistors and, indeed, start checking from one stage on. I did change my PSU from the last transformer voltage check, so it has merit to check that again. I'll start with the front resistance to slow things down and see what happens. In the meantime I'll try to figure out a circuit addition, if the last cap continues to blow, to release majority of current from the cap, while prime circuit continues to work in the same manner. This is kind of a weird issue. I got enough suggestion for a thorough checkup. Now, it's a 5 day wait for parts.

[davekni]

By that time ZVS startup peak is way over.

Yes, I was just about to modify my last post.  A quick energy calculation shows that charging CW caps requires much more energy than would be stored in ZVS input inductors even with an extreme startup current spike.

This is kind of a weird issue.

Yes, does seem that way.  Even though your two-stage experiment points against HV corona issue, I'd try standing CW multiplier vertically so not against any surface other than air.  Then you could look carefully for corona in a dark room, and it is also possible it would fix this weird issue.

[JoeBusic]

This is kind of a weird issue.

Yes, does seem that way.  Even though your two-stage experiment points against HV corona issue, I'd try standing CW multiplier vertically so not against any surface other than air.  Then you could look carefully for corona in a dark room, and it is also possible it would fix this weird issue.
[/quote]
Good one. Now with some time to tinker things, I'll make some standoffs from acrylic for all components to raise them cca 2" from the table. That should be enough for visual inspection. IF that's the issue, it's an easy fix.
Well, smarter every day.
Thanks

[davekni]

Even though not a normal ZVS startup transient issue, have you measured your +-15kV transformer output?  At 40kHz, transformer secondary interwinding capacitance and CW multiplier diode capacitance make secondary a resonant circuit.  That resonance might be causing output voltage to be much higher than would be calculated by simple transformer turns ratio.  (The more frequent issue is that output resonant frequency is well below primary, resulting in very little output.  But the opposite is possible.)

[MRMILSTAR]

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.

[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

[JoeBusic]

Didn't like how the first resistor suffers. Practice is different than what I was told. I was told that the Vdrop has to be accounted for in the resistor chain. All was good that way but that massive coronas from the first resistor got me thinking. It's not just the Vdrop but the total V that pass thru. So, instead of getting 30kV I have 80kV, thru the resistor. To fix the issue I got some massive resistors. BTW, single digit uA, double digit uA and triple digits uA were tested. Now comes the mA. And my PSU broke when going for triple digit uA, so more distance will be made between the PSU and the ZVS/transformer pack. Going for full 2m. Now everything is stretched to 5m. If this thing turns into a prototype, I recon steel cases and shielded cables with virtual ground.

[JoeBusic]

Well, all is good. Nice to hear...nothing. Good connections and solders creates peace I have a question. Now I would like to reverse the polarity from 150kV positive to to 150ishkV negative. I understand that the diodes have to be flipped. Question is, do I have to make any changes in the circuit? Simulation seems to agree with no change, so just trying to confirm before power on.

[JoeBusic]

Update: I am not satisfied with the output. Reading here on the forum about other peoples troubles and solutions (several days of reading, good stuff), I decided to check everything from the start before building a new CW from scratch, as new parts have arrived, and 5kg of wax. I made a measuring graetz for the output of the transformer and planed to check from the PSU output, ZVS in/out and HV part. Bought a couple more multimeters for the task. As I was checking up does the graetz work properly, the error has been pinpointed. I measured before the output of the transformer to be 15kV, but it wasn't loaded. Now with full load, it was just 8.5kV. Based on the spark test of my CW output, the CW is working almost as theoretical. All the ideas of bad cabling and various other ideas just went down the drain. After reading some more here, new understanding of secondary resonance was clear. Tomorrow I'll buy a fistful of caps and do some lengthy testing. Plan is to sweep the voltage from 18-30V (my PSU limits) with caps from 20nF up to 2uF on the ZVS. One full resonance should be made, possible two. Would like to achieve full resonance at higher frequencies, like above 100kHz.
Some photos:






Couldn't ignore the urge to poke into Tesla coils. This one is almost a toy but should give me a good start. For full AC HV, I'm considering transformer cascades. Will order some stuff soon and will do the build and testing in another thread.

[davekni]

One full resonance should be made, possible two. Would like to achieve full resonance at higher frequencies, like above 100kHz.

Not sure what you mean by "full resonance".  This is a dual-resonant system, primary with ZVS cap and secondary with inter-winding and diode and other stray capacitances up CW multiplier.  Combination forms two resonant frequencies, upper pole and lower pole.  In my experience, ZVS oscillators (and typical DRSSTC drivers) will run at lower pole if primary frequency is under secondary frequency and at upper pole if primary frequency is higher than secondary frequency.  (Here primary and secondary frequencies refer to their resonant frequencies if not coupled.)

Secondary resonant frequency is likely far below 100kHz.  Hopefully it's above 20kHz to minimize audible noise.  Even that can be tricky with high voltage transformers and resulting high secondary inductance.  If you have a signal generator of some sort, I'd measure secondary resonant frequency first with no primary capacitor.  More chance of success being intentional rather than just experimentation alone.

[JoeBusic]

By full resonance I meant when they are coupled. Don't have lab instruments. Anyways, I believe resonances are power related. I have to check 3 loads, so it's 3x the testing. Purpose of this marathon is to understand things and get to min 95% efficiency. After that repeat with 12 stages and possibly find a calculable difference between them. I need this for a power supply, not for sparks.

[JoeBusic]

High frequency is out. Drop is enormous. I have burned my measuring graetz under 20kHz. Make another bigger graetz and continue with pinpointing. Now I'm at 2uF and slowly going higher. I was checking out only HV transformer output up to now. I'll add a meter to check ZVS input V and CW output V. Now goes nitpicking.

[JoeBusic]

I was wrong. Chasing resonance is just dumb. I have neglected the ability of the HV transformer.
Thinking out loud:
Caps are static charge thingies. Current first pulls out, then the voltage.
Ergo,
If the necessary current is pulled and there is not enough charge left for voltage (simplifying power equals A*V is charge), voltage drop will be significant in the multiplier.
SO , my purchase of 2.2nF caps was just an ignorant doing.
It seems that the setup should be made from the goal and calculated backwards. I'm no electronics dude so I'm learning this the hard way. OK with it!
That said, my initial problem would be solved very easily with smaller caps. Way smaller caps. Talking about 50pF caps for a 10uA load. Tho the resistor solution is not bad, it is a secondary tuning thing, not a solution.
As my intent is to make 3 tests with loads of 10uA, 1mA and =>5mA, there is a need for three setups. One setup and just changing the resistor bank is hell. I understand it now.
SO, my HV transformer is at optimum @35kHz to 40kHz. Instead of chasing resonance, it has to be made with those frequency margins.
Mid setup, for 1mA can be made with components at hand.
10uA setup will need higher frequency, but I'll try to make it happen <45kHz and see if it works.
High current setup will need WAY bigger caps and it's on hold now.
For each setup goal is 230kV, so there will probably be management of stages, not just resonance. Seems like 7 stages will do, but after resonance tuning and considering voltage drop with low currents, I'll se how it goes.
Reading about bifilar coils for resonance tuning. A lot of ways to tune.
Will buy an inductance meter. Seems like a handy thing in this situation.