600kV / 20Joule Mini-Marx-Cascade
 
The construction idea of  my little Marx  is "borrowed" from Steve Ward's specimen. By good luck I  got hold of  a bunch of 2nF/40kV Cera-Mite doorknob capacitors from a  big scrapped excimer laser. For seventy 100kOhm/30kV  HV resistors (100.4 RS 100kOhm 5% TK 100) I am indebted to Metallux  AG  , as well as for the HV Powersupply to  Eltex GmbH , which supplies  -30kV (negative voltage!). Considering, the Marx serving just for demonstration purpose, only a most basic topology was choosen, and no measures (i.e. output resistive and capacitive ballasts) for pulse shaping were taken. Therefore the output is probably in form of some damped oscillation pulse, depending mainly on the capacitive environment of the output (i.e. the sphere-gap, shown below).

Nicely controllable power supply: DC-HV cascade KNH34 -30kV/5mA from Eltex GmbH 128'013 bytes	2nF/40kV Doorknob cap manuf. by Vishay/Sprague Cera Mite corporation 20'182 bytes	Metallux  HV Resistors 31'335 bytes	Schematic diagram of the 20 stage Marx 45'134 bytes
A 5-stage 150kV Pilot Mini-Marx was buildt first 77'197 bytes	Pilot Marx, firing 18cm 71'591 bytes	20-stage Marx. Polystyrene-glass plates: 140cm x 13.5cm x 0.5cm 77'882 bytes	20-stage Marx, 67cm discharge between points 105'891 bytes
Zoom of discharge 60'870 bytes	stainless steel d=30cm "gazing balls" (slightly ferromagnetic) 72'358 bytes	22cm discharge pulse in 30cm-sphere-gap (right sphere grounded) 86'034 bytes	22cm pulse + small additional pulse to air from grounded(+)sphere 77'638 bytes
22cm pulse, with multiple starters from (-) sphere.  Note the ionized air cloud 84'117 bytes	Misfiring 22cm pulse, with multiple starters from (-). Note the ionized air cloud 65'857 bytes	Sphere-gap spark-table diagram 68'728 bytes	Simple Marx calculator following F.Fruengel Impulstechnik, 1960 ( Excel spreadsheet )
The 30cm sphere-gap allows a (raw) direct measurement of the HV pulse voltage. Following Antonio Carlos M. de Queiroz 's inca program the breakdown voltage is about 425kV for 22cm spark. An interpolation in a pubilshed sphere-gap table reveals about 400kV (see diagram), meaning the output of this Marx is probably only 2/3 of the 600kV design voltage. Probably the 19 gaps should be opened a little more, in order to charge the capacitors to a higher potential. Extrapolation of the CRC handbook sparkgap-table for point electrodes reveals about 367kV for 67cm spark, which is rawly consistent with the above.