New Capacitors!

destructive testing of the new capacitors

Some of my first eBay capacitors wins arrived. Among the shipment was 200 units of 6.3 kV 470p Russian military capacitors. Although not ideal for Marx service, I thought I'd give them a try. Physically they are just a little smaller than my current 15kV 220p caps, but the leads are *very* thick, 1mm according to my calipers. First step was destructive testing.

The cap was connected via two 1M and a 10M resistor across the full 17.4kV coronatron supply. The two 1M resistors were to protect the supply from inductive back EMF transients, the 10M resistor is purely for charging current limiting. The legs of the cap were bent to provide an adjustable spark gap. The very ends or the wire bent back in a reasonable approximation of a constant field surface to reduce corona effects.

The gap was initially set at about 5kV, and slowing increased until the capacitor failed. That was the theory anyway, the damn thing wouldn't fail! At least not straight away, it took several minutes of operation at the full 17.4kV before a dull orange-red flash was seen through the encapsulation and the cap stopped holding charge. The failure was curious, the capacitor must have gone very leaky rather than short or open, it read 2n on my meter, but the reading was not stable, it was probably confused. I don't have a megger (must build one) and I didn't try measuring its leakage using the supply, but it was more than 20M between the leads with a multimeter. The coronatron supply drew a lot less current with the cap across it, which is normal for it when in short-circuit protection mode. However 20M is far from a short circuit for it, the fault must be voltage dependent.

reliability testing of the new capacitors

So, these capacitors appeared to be quite robust beyond their ratings. I decided to do some reliability testing, continuous operation at 7.6kV (the full voltage of the -ve supply rail). A new cap was soldered in place and the gap was set so it just fired by carefully teasing together the ends with a piece of plastic drinking straw while the cap was fully charged. I monitored the voltage across the cap with my high voltage probe, the gap setting was so fine the 200M load of the probe dropped enough voltage from the 12M charging resistance to stop it breaking down, I measured 7.6kV in this state. Unfortunately an attempt to remove the ground-side 1M resistor from the circuit (as a safety feature to keep my HV probe and multimeter at ground) resulted in the damage of my bench supply AGAIN!, when a gap discharge induced a large ground surge.

After repairing the supply (Replacing the -ve rail error amplifier IC, a LM358, I socketed all the ICs in the supply the last time it failed, as dismantling it to take out the board is quite hard. Now I can replace blown parts by just lifting the cover and swapping them.), I let the capacitor sit there sparking the gap continuously for about 30 minutes. At the end of which the capacitor was warm to the touch, but not stinking hot like the charging resistors were, they had been burning up as much power as the sparks discharged.

See how bright the flash is! There is a reflection inside the camera optics, a specular glow, plus diffraction artefacts are visible in the picture. The image was taken at full speed, it is not a timed exposure. Either the (fill) flash triggered the gap, or I was very lucky to capture it just as it self-triggered. The ESR and inductance of these caps must be very low.

Note how I bent back the ends of the leads into an eye and filled the loop with solder. This construction technique proved very effective. The smooth surface of the solder (due to its surface tension pulling it into a minimal surface, like a soap bubble) made the gap very stable and reliable. The corona losses have vanished, I can no longer feel or hear any ion wind, or pull current to a neon bulb until I am so close to the gap the distortion of the electric field makes it fire.

This gap field distortion effect is very interesting, even dielectric materials do it. It may prove very handy in the future for fine-tuning a gap. I used it several times to fire the Marx generator manually with a soda straw. Combined with my improved all-smooth-surface construction technique, the new Marx generator gives fantastic performance. With only 5 stages and under half the voltage per stage, it gives very healthy 20mm sparks. Compare that to the performance of the 16 stage unit!

new marx generator using the new caps

Note the excess solder usage and the loop-surface-blob gaps. The corona rings on the input silenced the leakage I could hear before I added them. Although not important for correct operation, they are a nice professional touch, making everything more reliable and repeatable when measured. Note also the staggered placement of the resistors, ensuring the part of the resistor closest to the gap was at the same potential. Once charged this isn't an issue, but during charging there is significant voltage across the charging resistors. In this unit 1/2W resistors proved reliable, the previous 16 stage unit used 1 W resistors and still was limited to about 7kV. The think the smoother construction and slightly better quality coating on the 1/2W resistors allows the reliable use of these parts.

new marx generator in operation

With this far more stable device, I was able to conduct more scientific tests than on the previous prototype. The coronatron supply's supply current varies with its load, I used this fact to observe the recharge time and ensure the entire stack was getting a good charge before breakdown of the first gap. I also tried probing around the top capacitor with my HV probe, which proved to be a painful experience more than once! The top capacitor is getting to at least 90% of the full voltage before the cascade starts.

If a grounded target for the discharge is not offered a very feeble crack is heard and tiny violet sparks jump all the gaps with quick repetition. I am not sure where this discharge is going, hopefully not punching through capacitor dielectrics, and rather being absorbed in the resistors or spraying some ions off the tip. There is a strong electric field generated, with some hiss and neon bulbs glow near the top of the stack, so I assume the caps are getting series connected and a big voltage appears at the top, but there is little energy actually delivered. Good, it behaved itself even in extreme conditions.

shortwave light and electromagnetic output

Here are some eye-candy shots. Although they do serve a purpose, displaying the range of EM radiation produced by they Marx generator output. The micro centrifuge tube contains a little Fluorescein solution, which as you can see in the transmitted light is normally a orange-brown colour. The short-wave light from the discharge makes it fluoresce brightly.

fluorescein structure

Note also the neon bulbs. The one nearest the spark is legless, the others further out have their legs spread wide apart to act like little antennas to the strong electric field. As you can see, the effect drops off quickly with distance. A Marx generator is truly a DC to Daylight device! In fact they produce a small amount of soft X-rays as well, nothing to worry about at room pressure however.

high density polyethylene glows a bit too

The high density polyethylene is somewhat fluorescent too. I've see this kind of bluish glow from CD cases (and these tubes) under a UV-A black-light tube. Note the legless neons are glowing inside the thick dielectric tube, dielectrics don't stop electric fields.

discharge current has a kick

There is a lot of current in that discharge pulse. The tube sitting in those turns has 14 turns around it, and is shunted by the neon inside the tube. It is a damn handy little gadget around the bench.

I tried measuring the discharge waveform. I placed a centrifuge tube over my CRO probe and cautiously approached the output terminal while sparking. The capacitive division ratio was unknown, but I saw more than 50V spikes with rise and fall times limited by the CRO and probe response. My CRO is only a 20MHz unit, and the probe 50MHz. The rise time could not be measured, the trace is too unstable I need a few Giga-sample/sec digital storage CRO with -ve triggering to observe such fast transients. I estimate that the rise time is less than 10 nS.

With an erected capacitance of 94p, or a total stored energy of about 64 mJ and assuming the discharge is all over 100ns, that works out to around 6MW of peak power output. The true figure may be higher. I'll have to work out a way of measuring the output waveform. Perhaps I can build a travelling-wave diode sampling device, but that is a significant undertaking, lots of fiddly transmission line construction. Time to troll eBay again for a gigasample DSO...

With such short rise times I am curious about the atmospheric nitrogen in the gaps. The rise time is fast enough to get a population inversion and lasing action, we can already see the shortwave light which would be much weaker if the rise time was slower (decayed N is a good absorber of UV light). There well could be super-fluorescent lasing going on but probably not enough to endanger the eyes. Another health issue that came to mind is the lead in the solder surface of the gaps. Some may be vaporised and inhaled. Using lead free solder might be a good idea, but stannous vapours are almost as bad as plumbium ones anyway. I've read that Lead wire in cheap imported candle wicks is a health danger, but they are consumed much faster than the electrodes in this Marx generator. At higher powers it may become more of an issue.

A Marx generator is not a kids toy anyway! They produce dangerous voltages, loud sounds, ozone, nitrous oxides, radio waves, UV light, X-rays and many lost hours of drawing sparks through and across various things... Guess they are irresistible kids toys then! At least I've got mine to play with until the twenty 2n2 25kV caps arrive!

Leave a comment on this article.

Parent article: Marx Generator.