This is the big'un: an 8-stage, full-wave, negative-polarity, air-insulated 120 kV Cockroft-Walton stack. Tonight I finally got a driver cobbled together, and some results to share, puzzle over, and discuss. Basically, though, it's a healthy fire-breathing stack.
The drive circuit follows the SLR (series-load-resonant) topology, using an H-bridge of HGTG20N60AD IGBTs (D denoting the built-in antiparallel diodes) to switch a load leg that has L = 24 uH, C = 0.34 uF, and a big ol' ferrite-core transformer. I won't even bother to put a circuit diagram up, because Steve Ward has already done it here:
http://www.stevehv.4hv.org/ccps1.htm
Mine differs in some insignificant ways (I put 440V TVSs across each IGBT, for instance, and came up with different values for many components), but Steve has been my knowledge base on this project so far.
I will go through the details as I describe the photos attached:
(1) This is the stack itself. On top is a water dummy load made by taking a 16" length of 1/2" dia. PVC pipe, capping the sucker with copper repair stubs, and filling with distilled water and a few crystals of CuSO4.5H2O to taste (in my case, 2 MOhm). The upper copper stub has a little hole drilled in the top, enough to keep water from pouring out if accidentally tipped, but to allow electrolytic gas escape. Don't forget that detail! Such a water resistor is very cheap for what it does: it's non-inductive, can stand off high voltages, and because of the high heat capacity of water, you can temporarily run very high currents through them. Never flash down a stack like this without some sort of current-limiting load in series.
(2) Drive electronics, in base of cart on which stack sits. In front left in the box is the gate-drive VFO, a TL494 circuit. In back is the H-bridge and line rectifier and filter caps. At right is the transformer. In the middle you can see a coil of heavy speaker wire. This is most of the inductance for the load leg (about 20 uH of it). The other ~4 uH is leakage inductance in the transformer. The transformer primary has 7.5 turns of heavy litz wire salvaged from an HFMOT on a huge ungapped core obtained from Steve Hansen. The secondary is a 400-turn coil of #28 magnet wire, center-tapped. Thus the step-up ratio is about 53.
(3) Delivery of sauce: a 6-8" arc in progress. They get more impressive than this, up to about a foot long, limited by the little 1.4 kVA variac feeding everything and soon to be replaced with a beefier variac. The arc starts at around 5".
(4) Gate drive and primary current waveforms with stack loaded down hard (free end of water resistor grounded). Classic sinusoidal SLR waveform can be seen, in proper phase with the drive signal such that the zero-current points occur when either the IGBTs proper or their antiparallel diodes are commuting.
(5) I have a hard time explaining this primary waveform, which results when the stack is completely unloaded. Explanations welcome. I don't like the look of it, because it implies non-zero-current switching. That said, the IGBTs do not perceptibly heat up during operation, either under load or not. So I'm guessing the current itself is not that bad. I just don't want heavy dI/dt ruining things. I should mention that the source of these signals is a VERY crude current transformer (ungapped core salvaged from an HFMOT, wrapped with ~6 turns with an alligator clip lead, put around the primary conductor, loaded with a big 10-ohm resistor.)
More to come later on this big killer. I hope to enter it into service with Carl's Jr. very soon, but before that I need to figure out the unloaded waveform and make sure it's safe on my circuit.
-Carl
|
2007-24-01 01:58 120 kV / 100 mA supply sees first light 
(Carl Willis) [Latest: 2007-17-05 23:40]
(Carl Willis)