"Quick & Dirty" Marx generator
Update May 2003 - Marx Three is now up & running
Do you like the idea of tesla coils and other high-voltage sparking stuff, but don't have the time, money or patience to build something that elaborate? Here's a fun little project that can make big, fat, noisy sparks at least 2 inches long, and can be built very quickly and cheaply.
The Marx generator consists of an array of resistors, capacitors and spark gaps arranged as follows:
The capacitors, C, are charged up in parallel via the 1M (one megohm) resistors, so they each become charged to the input voltage. When first (leftmost) spark gaps breaks down (sparks), the voltage across the next increases, causing it to break down, and so on for all the other gaps. When all gaps have broken down, the low impedance of the ionised air in the sparks effectively connects all the charged capacitors in series, multiplying the input voltage by the number of capacitors. The ionised air path has sufficiently low resistance that the charge resistors don't have any significant effect.
Rb has a ballasting effect. This is needed to avois a continuous arc forming across the first gap after it fires - this prevents further firing. The value will depend on the type of power supply used - values around 1M usually about right. High values will reduce the maximum spark repetition rate, so you may be able to get more 'bangs per second' by reducing Rb.
Although it is possible to make a Marx generator with just an array of resistors,
capacitors and spark gaps, it can be hard to make it fire reliably, as it will
depend on the breakdown voltage of the spark gaps, and there can be a fine line between
not firing and firing before all the capacitors are fully charged. One crude solution is
to initate the breakdown of the first gap mechanically, e.g. by waving a (suitably
insulated!) screwdriver between the electrodes.
I based the component values on what I had available, and there is plenty of scope for
experimentation here. If in doubt, try building just a couple of stages first to see how
well the parts you use will work.
Capacitors should ideally be ceramic types, as these are best suited to the fast pulses in a Marx generator - polypropylene pulse-rated caps would probably also work well, but are more expensive, and harder to find with high voltage ratings. Use the highest voltage types you can find - 6.3KV types are readily available from most component suppliers, and values up to 15KV are not too hard to find. I used 4.7nF 4KV parts as I'd recently picked up a big bag of them very cheaply at a radio rally (hamfest). Larger values will give fatter sparks, but take longer to recharge! You can often run these in excess of their rated voltage, but obviously this runs an increased failure risk.
An input voltage of about 4-8KV is recommended. Above this, many problems will detract from the 'quick and dirty' approach, requiring much greater care in construction to reduce corona losses from sharp edges, multiple series resistors would be needed, and the capacitors will be harder to find. The high impedance of the charging network mean that corona losses can significantly degrade performance above about 5-6KV unless care is taken to avoid any sharp edges - e.g. by filing the wire ends and joints smooth..
This design requires a low-current source of high voltage DC in the range 4-8KV (polarity is unimportant). Small, low-current HV supplies from things like ionisers, photocopiers and laser printers should be suitable. I used a surplus 5KV 2mA photomultiplier supply.
The number of stages you use will determine the output voltage, and hence spark length. There will be upper limits on the practical number of stages, due to charging time and the speed at which the discharge propagates along the gaps. I ran out of resistors before I found the limit!
Close of spark gaps and general construction.
The spark gaps are simply formed from thick tinned-copper wire, as shown above. It is important that the gaps are curved, and the sharp ends of the wire are bent away as shown - this avoids corona loss and premature breakdown. The wire used should be at least 1.5mm dia. The output gap however should use fairly sharp points for maximum spark length. Take care to avoid spark discharge from the ground electrode to the resistors or capacitors near the 'hot' end, as these discharges could break down the components' insulation.
Gaps should be initially set to about 1.5mm. Some adjustment may be required - if the generator doesn't fire, one or more gaps are probably too far apart. If gaps are too close, especially near the supply end, you may see a continuous arc forming, preventing further operation. Increasing the gap on the first (and possibly second) stage may reduce the tendancy to arc. Increasing the ballast resistor Rb also helps, at the expense of slower recharge time. Continuous arcing can also be significantly reduced by passing some airflow, e.g. from a small fan, between the gaps.
As can be seen from the photo, a very crude 'Christmas-tree' type construction can be used, the components being simply soldered together - of course a neater assembly technique could be used, but bear in mind that whatever you build it on must withstand the full discharge voltage - Perspex (Acrylic/Plexi) or polycarbonate (Lexan) is ideal. If doing a simple 'lash-up', at least place it on a plastic sheet to avoid discharging to the table/banch - wood is not a good insulator at these voltages! Avoid sharp edges on any of the connections (e.g. don't use metal screws), as this can cause corona losses, reducing output significantly.
The size of sparks at the output gap will depend on the input voltage and number of stages - start with it set to about half an inch, and gradually increase it (careful!) until you no longer get sparks.
More on Marx generators here
This is a development from the 'quick & dirty' device above. The construction is basically similar, but it has better (i.e. some!) mechanical support, being mounted on a strip of Perspex (=Acrylic/Plexi) which was heat-folded to a 'V' section for rigidity. It produces sparks up to 9" long from an input of about 16KVDC.
To allow a higher input voltage, I used 1nF 15KV Panasonic caps. The choice of this value was dictated by what was readily available from normal component suppliers. Resistors were series pairs of 1M2 Philips (now BC components) VR37 metal glaze parts. These have a voltage rating of 3.5KV each, so the pair has a 7KV rating - a bit low for the 15KV input voltage, but this voltage is only seen for short periods, and they appear to have held up OK. Again, the value of 1M2 was simply the lowest readily available value. Spark gaps are similar to those used on the Q&D version, but with the wire ends bent over & soldered to reduce corona losses. Gap distance is about 12mm for 16KV.
I found that above about 15KV, some sparking occurred across the capacitor leads, and so had to add some rather messy insulation in the form of silicone sealant on the capacitor leads - it would have been so much easier to sleeve the leads beforehand!
No trigger circuit was necessary to get this generator firing fairly reliably - I think
the higher input voltage helped a lot with this, probably due to corona. For any given
spark gap size, adjustment of the input voltage was fairly important - too low and it
never fired, but as voltage was increased past the point where it started firing, firing
frequency increased, but output started decreasing, as the lower gaps were firing before
the caps at the end of the chain had fully charged.
>Discharging through the wall of a plastic drink bottle. If you get the spacing right, the first few shots build up a huge charge on the surfaces of the bottle (you can hear it "sizzling") , and the next one triggers a spectacular flashover.
A quick & easy way to improve spark length is to build a second 'Marx two' design, and run them as a pair, in a bipolar configuration. The overall arrangement is effectively the same, except that the input supply is fed to the centre of the bank :
In this configuration, one side generates a high positive voltage, and the other a high negative voltage, giving twice the spark length than a single tower. One advantage of this arrangement is that the voltage to ground at any point is never more than half the total output voltage, reducing corona losses. I built a second bank, consisting of 13 stages (This was simply limited by the size of plastic sheet I had to hand!), and the combined setup will produce sparks about 14" long between the two towers.
It is very important that the impulse from the gaps firing propagates rapidly to all stages, and this can cause problems when coupling two towers which are a distance apart. The main problem is the capacitance to ground of the connection marked 'transmission line' in the above diagram. To reduce this, I used a polythene insulated wire (from stripped RG58 coax cable), supported by a couple of plastic aerosol can caps (pictured left) , to ensure that the cable is held at least 40mm from the surface. Coiling the wire through two caps allows the length to be easily adjusted without the wire curling down towards the table.
This image is a long
exposure of several shots, and has been colour-enhanced to show the haze of corona
streamers that can be seen in darkness.
Another enhanced image showing the corona streamers when no output spark
You want to get the biggest LOPT you can find - the size and power of the LOPT is
usually proportional to screen size, and monitors tend to have bigger ones than TV's of
the same screen size. If you get a LOPT on its own, so can't look at how it was connected
in the monitor, it can be hard to figure out the connections, as the high-voltage
rectifier has a forward voltage too high to test easily. If you have problems figuring out
the pinouts, try this method :
If your LOPT has additional high voltage output leads (sometimes marked 'focus' and/or 'Screen), cut them off, and seal the ends of the cut wires with silicone sealant (bathroom sealant).
Lessons learnt from this Marx generator to be incorporated into "Marx Three"...:
* Extra insulation is required on the capacitor leads above about 15KV to avoid sparks
jumping across the capacitor leads, and it's a lot easier to put this on BEFORE soldering
everything together! Soft silicone sleeving, or maybe heatshrink would probably be the
best, to avoid small air gaps. Alternatively, the 'V' construction would allow the whole
thing (apart from the spark gaps) to be encapsulated, which would also reduce corona loss,
but require quite a lot of potting compound!
Marx Three is now running
Cheap HV Capacitors.
Thanks to Robert Trautman for the following info on home-made capacitors
As you know it's difficult to find inexpensive capacitors of any real
value that'll handle the higher input voltages. I'm using a 25KV flyback transformer
(LOPT) from an old computer monitor as my input, so I needed capacitors that'd handle at
least that voltage.
So, I built my own capacitors using Mylar sheets, normally used as page
protectors in notebooks, and embossing aluminum strips from a craft store. To give them
structure, I rolled these materials (three pieces of Mylar and two aluminum strips,
alternated) around 7/8" (2.2 cm) diameter wooden dowels. In order to achieve the
required voltage rating (I estimate they'll take up to about 50KV), the edges of each
chamfered aluminum strip (capacitor plate) extend no closer to the edges of the Mylar than
about 1.6" (4 cm) all around, thereby allowing 3.2" (8 cm) of distance between
adjacent plates. These materials, once rolled onto a 7/8" (2.2 cm) x 3" (7.6 cm)
dowel produce a 0.005uF (5nF) capacitor at 50KV in a 5.5" (14 cm) x 1" (2.54 cm)
space. The dowel, being made of wood, should be about the same length as the width of the
aluminum strips so as to avoid arcing to this. Obviously, much greater capacitances may be
achieved with longer aluminum strips than those I used, but this was a start.
I just wanted to pass on to you and those to whom your site might interest that 0.005 microFarad, 50KV capacitors - even if not the most efficient - can be built easily for about $2.00 in materials, each. Again, much larger capacitances may be achieved with longer aluminum strips for the plates. The only problem with longer plates is that another source of the Mylar will have to be found. These notebook sheet protectors provide only around 15.5" (39.4 cm) of useable material per sheet - just enough to accommodate 12" (30.5 cm) long plates with adequate space around them.
Update Oct 2005 - it appears that most of the problem was the power supply, not the capacitors...