High Voltage Forum
High voltage => Voltage Multipliers => Topic started by: Peregrine on February 18, 2020, 10:40:58 AM
-
Help!
Hi everyone, I need to use this circuit, but after a while the capacitor burns out!
the power supply comes from a voltage multiplier, and the capacitor is 50kv 10nF, the coil 10uH
what can I do to protect it? thank you.
-
It sounds like you are using a capacitor beyond its voltage, current or frequency specifications. Please provide all details about the application and capacitor.
-
the capacitor is a doorknob 50kv/10nf , I need a quick discharge on the coil...
the oscillation frequency is about 500khz but the discharges on the sparkgap are few hertz
-
If this is the schematic for a spark gap coil, you need to swap the position of the capacitor and spark gap. I don't think this is causing your capacitor problem but it could cause problems for your transformer. As Mads stated, it sounds like you are exceeding the voltage rating of your capacitor. You want at least a 2x and preferably a 3x voltage rating for your capacitor.
-
I stand with Steve about it being easier on the transformer to drive across spark gap instead of across capacitor. Maybe not a lot easier, because of the 10M resistor.
As Mads said, there's more to learn about how you are abusing the capacitors, or if the capacitors are over-rated fakes (?).
What is the RF current in each spark, the tank circuit energy, and the average power?
Power input is partitioned somehow between losses in L, C, and spark.
Do L or C get hot when you operate the circuit at power below the destructive level?
-
If this is the schematic for a spark gap coil, you need to swap the position of the capacitor and spark gap. I don't think this is causing your capacitor problem but it could cause problems for your transformer. As Mads stated, it sounds like you are exceeding the voltage rating of your capacitor. You want at least a 2x and preferably a 3x voltage rating for your capacitor.
no, the position of the sparkgap and coil are exact. This circuit is not to make a tesla coil but a therapy device with pulsed magnetic fields (as in the picture)... and then I don't use a transformer as a power supply but a 30-40.000v direct current generator.
then for the voltage of the capacitor maybe you are right but I used doorknob which told me that they are not really suitable for fast impulses while I should take the polypropylene ones, but apart from this problem, would you have other tips to avoid damage?
I stand with Steve about it being easier on the transformer to drive across spark gap instead of across capacitor. Maybe not a lot easier, because of the 10M resistor.
As Mads said, there's more to learn about how you are abusing the capacitors, or if the capacitors are over-rated fakes (?).
What is the RF current in each spark, the tank circuit energy, and the average power?
Power input is partitioned somehow between losses in L, C, and spark.
Do L or C get hot when you operate the circuit at power below the destructive level?
the power is less than 10J, but before the capacitor broke, I saw a flash of lightning go through it ...
in tesla coils, a so-called "terry filter" is often used but there is an alternating transformer and it is used to avoid RF or that could damage it, instead mine is a DC circuit....
-
Ceramic doornknob capacitors with a 50kv 10nF rating are uncommon. Do you have a picture you could share? I suspect that the capacitor could also be unsuitable for tesla coil use.
-
Ceramic doornknob capacitors with a 50kv 10nF rating are uncommon. Do you have a picture you could share? I suspect that the capacitor could also be unsuitable for tesla coil use.
-
Even though this isn't a spark gap tesla coil I would still swap the positions of the spark gap and the capacitor. Doing so will make it less likely for any high frequency high voltage oscillations to get back into your power supply and possibly damaging it. From your schematic, the functionality should be the same other than the capacitor charging through the additional wire resistance of the inductor. The additional inductor wire resistance is negligible compared to your 10M ohm charging resistor.
I would want at least a 2x voltage margin on that capacitor. With a rating of 50 KV you aren't close. That flash that you describe sounds like the capacitor breakdown voltage was exceeded. Four of those 50 KV capacitors in series-parallel would yield a 100 KV rating with the same capacitance which would give you a comfortable 2.5x voltage rating.
-
Hi :)
capacitor failed maybe from resonance ringing, as this phenomenon boosts voltage and current in the circuit, leading to failure.
So I recommend you to use capacitors in series to have a large voltage headroom, to avoid flashover, as for example using 2* 50kV capacitors in series to have a rating of 100kV.
I also recommend using pulse film caps as they can withstand high pulse current. ;)
-
Just confirmed your estimate of ringing frequency, 500 kHz.
Associated impedance sqrt(L/C) is about 32 ohms.
That means that if spark gap fires at 32 kV, the RF current hits 1000 amperes in first cycle.
How fast does the ringing decay? You could measure that with oscilloscope and a sense coil or pickup antenna loosely coupled to the tank circuit.
If spark gap fires at 32kV, initial energy in C is about 5 joules.
If a discharge happens 4 times each second, that would be 20 watts being dissipated in C+L+spark
(and I think another 20 watts in the charging resistor).
How hot do the R, C, and L get, and how fast do they get hot?
-
Thanks friends for your advice ... I looked better at my circuit and saw that, since I had already burned many, I had put 2 40kv capacitors in series, thus obtaining a total value of 80kv 5nf ... So the problem I don't think it is the sizing of the capacitor ... Keep in mind that during the discharge a discharge bang occurs so strong that at 1m away the ears hurt ... therefore it may be that this kind of condenser is not suitable for these discharged or is it another problem? I add that the device must be kept on for at least 20 minutes, ... For a short time it is good for over 20 minutes it is not reliable
-
Sounds like your caps are overheating internally due to the discharge currents and then something goes bang as it takes some time.
/Rickard
-
Overheating seems a likely cause. Do you have ESR specs for the cap or a tan(d) ? If not, place a small pickup coil near the inductor and take a scope shot of the voltage. From the ring down of the amplitude you can get a rough estimate of the dissipation. You'll loose a few hundred volts amplitude in the tank for each cycle due to the spark gap voltage drop. At your high voltages, this might be a few percent initially. If the ring down is much faster, than the cap ESR will take most of the energy.
If that is the case, paralleling caps instead of serialising them might be the better idea.
-
You haven't told us what you are trying to do with this circuit. If you are only charging and discharging this capacitor infrequently, once a minute for example, I would disconnect the power supply from the circuit before firing in order to protect it. If this is firing rapidly then obviously you can't do that and will have to rely on the charging resistor for protection.
-
your capacitor is not rated for pulse duty if you can find some high voltage mica caps you will be better off. The foil atached to the ceramic slug on the inside typicaly will degrade with pulse duty on door knob caps. they are better for marxes or tripplers
-
You haven't told us what you are trying to do with this circuit. If you are only charging and discharging this capacitor infrequently, once a minute for example, I would disconnect the power supply from the circuit before firing in order to protect it. If this is firing rapidly then obviously you can't do that and will have to rely on the charging resistor for protection.
the capacitor does not charge and discharge it once a minute :) but continuously, as I said, even for 20 minutes ... in fact the resistance heats up a lot but now I will try to divide it into more values to have less power dissipated ...
-
do you know if there is a circuit in LTSpice that can simulate a sparkgap well? it would be very useful ...
-
I have a probe that reduces the voltage read by 1000 times ... do you think I could use it to connect it to the oscilloscope to see the signals? or to read the currents at stake?
-
You don't need a probe that touches the circuit.
I would question the accuracy of any ordinary 1000:1 voltage probe at 500 kHz.
As a couple of us have suggested,
* Any probe close to your pulser will provide a signal with which you can accurately measure the frequency and damping.
* You can get the initial voltage accurately enough from spark gap geometry, or from known power supply voltage and measured repetition rate.
* You can get the first-cycle current accurately enough from voltage and tank impedance sqrt(L/C).
-
Simulating spark gaps is a bit tricky. For most uses, a simple switch is sufficient. Instantaneous elements such as switches are problematic for analog simulation. LTSpice does reasonably well with instantaneous events, but still sometimes gets stuck (step time becomes too small until the simulation errors out). LTSpice does have switch elements along with arbitrary behavioral voltage and current sources that can be used to control the switch. Sometimes I've used thyristors instead (NPN and PNP connected into a latching circuit) with diode breakdown defining the trigger voltage.
-
At the bottom of this page is a scope shot, that shows, what you might see in your circuit. http://www.laushaus.com/tesla/vortexgap.htm It is the voltage in a primary tank of a spark gap Tesla coil with the secondary removed. In this example, the decay of voltage is mostly due to spark gap losses and approximately linear. Your waveforms would look similarly, if only a small fraction of the energy is burned up in the cap ESR. If the decay is faster and exponentially sloped, you probably dissipate the energy in the cap.
-
I detected this signal with a probe, what does it make you understand? What are those peaks on each ridge? It is not seen well but at the beginning of the oscillations there is a signal of 17mhz while the resonance frequency is about 400khz
-
The 17MHz is probably resonance of wiring inductance and capacitance (wiring from the cap through the spark gap and coil and back to the cap).
The spikes appear at the voltage peaks, which are the current zero-crossing points. At those points, the spark gap has no current, so partially extinguishes (ions recombine). It then requires a bit more voltage drop to return to normal arc conduction. Those momentary spark-gap voltage increases are likely what you are seeing as the spikes.
Polypropylene capacitors would be best for your application, a long string of lower-voltage caps. A 33-long string of the Chinese induction-cooker caps, 0.33uF 1200V, would last longer than any ceramic. (I've abuse tested a couple of these 1200V caps with bursts of +-1700V at 80kHz. Your frequency is higher, but your bursts are shorter.) They are available multiple places, but here's one EBay link:
https://www.ebay.com/itm/10PCS-New-BM-Capacitor-MKPH-0-33uF-630VAC-1200VDC-for-Induction-cooker-P-30-5/272271654633?hash=item3f64a7bee9:g:PJ4AAOSw9eVXXQsg
If a single 33-long string doesn't last long enough, 132 parts in a 66s2p array would almost certainly survive long-term. (You may find that the ring lasts longer using PP caps, as the ceramic parts are likely a significant part of your total power loss.)
Is the sharp initial start to the ring waveform necessary for your application? If not, a DRSSTC-style circuit (series-fed resonance from an H-Bridge drive) would be much quieter, more efficient, and more stable.
-
thank you David for your answer and competence. I would also use another circuit, but in your opinion, with the electronics you are talking about, you can get large peaks of energy (30.50w) discharged on a coil in a few microseconds or better, nanoseconds? or are there other techniques? thank you
-
It looks, like you're loosing about a factor of 2 in amplitude for each cycle. That cannot be expected from losses in the spark gap alone. The damping you see corresponds to a series resistance in either the cap or the coil of about 7 ohm. This is probably also much larger, than can be expected from your coil. My guess is, that a very high percentage of the input power is dissipated in the cap, which would be about 20W assuming 5J per discharge and 4 discharges per second. Too much for a doorknob.
Handling 10's of kV at a 1000A with silicon is probably not very economical. I'd stay with a spark gap, but make sure, that you have flat or better rounded electrodes. That will allow to move the electrodes closer together for a given breakdown voltage, which will reduce losses after the gap fires. And, of course, heed the the previous advice regarding the caps.
-
It would help to know your goals more precisely. You refer to "peaks of energy", but list 30.50w, where "w" usually refers to watts, which is a unit of power and not energy.
The energy stored in the coil is proportional to the square of the magnetic field strength for a given coil geometry. The spark-gap into resonant circuit is making an oscillating field with multiple magnetic-energy peaks of alternating polarity (alternating magnetic field direction). Once you are using good PP caps, the oscillation will continue well longer. Are those multiple peaks helpful? Or, are you after a single peak? If the latter, then it will be much more efficient if the oscillation can be terminated after a half-cycle, saving most of the energy from the coil back into the capacitors for the next pulse. (Thyristors would be great for that case, but most are too slow in reverse-recovery time.)
An H-Bridge series-driving your resonant circuit will make the current ramp up much as it ramps down after the spark. So, the whole burst will be roughly twice as long, having a ramp-up followed by a ramp-down. If each magnetic-energy peak of the 400kHz resonance is useful, there's twice as many in each burst. If you are after a single half-cycle (single uni-polar magnetic field pulse), then an H-bridge isn't a good topology. BTW, if you do go with an H-Bridge, it needs to handle the ~1000A, but only a few hundred volts, as in DRSSTC drive. There's lots of information here on building DRSSTC H-Bridges using IGBT bricks.
-
Uspring wrote:
>>Handling 10's of kV at a 1000A with silicon is probably not very economical. I'd stay with a spark gap...
I was about to say: or use a coil of the same size, with more turns, so less current for the same magnetic field energy. Proportionally smaller capacitance, to keep the frequency unchanged.
But then the voltage would need to be even higher, by the same factor. This is not helping to make a semiconductor-switched solution workable.
I like Dave's suggestion: drive the tank with AC voltage source in series, at the resonant frequency. Can get tank voltage much higher than the driver voltage. Tank impedance, sqrt(L/C), can be chosen to suit the driver.
As others have said: What is this for?
Suppose that with much better capacitor and spark gap, your 5 joule discharge rings for 40 cycles (100 us).
That's a loss rate of 50 kW. To ramp it up DRSSTC-style in 40 cycles would need a 50 kW driver.
-
Is high voltage necessary? It sounds like you just want to create a pulsed magnetic field. A simpler solution might be to use 1000V instead of 40000V, and use an SCR to discharge the capacitor (along with a diode if you want it to oscillate more). You could get a 1000V capacitor that is specifically designed to deal with pulsed applications like this for very cheap.
Here is something you might find interesting:
https://www.repairfaq.org/sam/other/sgtms1/sgtms1.htm (https://www.repairfaq.org/sam/other/sgtms1/sgtms1.htm)
-
I mentioned thyristors (SCRs) too, but they won't work at the frequency of interest here, 400kHz. The "critical rate of rise of current" specification is just too low for 400kHz. (I've searched many parts, as I'm planning to use such for increasing MMC capacitance during each burst on my DRSSTC eventually, at 80kHz dropping to 55kHz.)
-
I mentioned thyristors (SCRs) too, but they won't work at the frequency of interest here, 400kHz. The "critical rate of rise of current" specification is just too low for 400kHz. (I've searched many parts, as I'm planning to use such for increasing MMC capacitance during each burst on my DRSSTC eventually, at 80kHz dropping to 55kHz.)
OP never said he needed 400kHz, he just said he wanted to make a pulsed magnetic field for a therapy device (probably something related to TMS). In that case, a lower frequency would work fine.
-
Ive used Behlke High voltage switches before to switch ion beams before an inflexion magnet. A modified half bridge constructed out of two SPDT switches shifting + 10kV to 0 volts then to minus 14kV and back to 0 volts into the cap plates in the magnet. Cap = 1200pf. I used non inductive 200 ohm resistors to limit current to around 20 amps. This switching was done at several kHz and we needed the voltage to settle sub Microsecond so that we could do our measurements
Behlke make solid state switches up to 1600 amps and up to at least a 100kV
-
As a quick fix what about a low ohm resistor or/and ferrite bead to limit the peak currents from destroying the capacitor?