Coaxial Marx Generator for EMP / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

Coaxial Marx Generator for EMP

previous 1 2 3 next

Move Thread

LAN_403

Dr. Shark

Sun Apr 06 2008, 02:13PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Yet another entertaining post by Dan about his Marx:

I killed a new FLUKE 87 handheld multimeter with my Marx Generator. I had the multimeter attached to the NST i was using to power a CW multiplier which then charged the Marx Generator to 40kV per stage. The multimeter was attached to the Primary side of the NST and was completely isolated from Marx Generator and charge circuitry. The Marx Generator even had its own isolated grounding.

Anyways, a discharge took out the multimeter.

Actually, I've killed quite a bit with that Marx Generator. My house alarm system, my house digital thermostat, a TV, and most recently, a nice multimeter. You'd think i learn, right?

It's a shame that he won't join the discussion, but I am glad I found two other people who are interested in this topic.

Now I am not really prepared to throw over the idea of a Marx generator in favor of a BlÃ¼mlein circuit, since that would mean getting all new parts, doing all the reading again etc., and I'd rather see some sparks soon

Harry, that paper you cited from what exactly what I had linked in my post above, did you notice that? I was just referring to the wave-erection as being BlÃ¼mlein-like since it involves a wavefront traveling at the speed of light. I don't know a whole lot about the particularities of that circuit apart from it's alleged use in TEA nitrogen lasers.

Anyway I said I wanted to get this going on the experimental side, so here we are. This is a bit of an "artist impression" of what I want individual stages to look like.

The spark gap (only one electrode shown, the other of course is connected to the next stage), is exactly in the middle, and there is a hole drilled down the center so all the gaps can "see" each other. The arrangement with the three capacitors in parallel is taken from the work of Mayes et al., although they don't say where the sparkgap is, and I presume it is not down the middle. I thought it was a nice idea though, because is makes the Marx even more coaxial

The whole assembly is going to slip in an acrylic tube, which will be sealed to the PCB disks with some kind of goo. I want to keep the distance between stages as short as possible to have a short discharge path.
The right photo also shows that I started putting the capacitors in blocks of 6, so I will have only 5 stages of 60kV each. My hope is that this way I will have less timing jitter and a faster pulse. With some clever design I might be able to make the whole 300kV generator as short as 30cm (12").

Machining the disks is a total PITA for me, look at my ghetto setup:

just a little bit too much feed in my "lathe" - which is a drill press and a chisel in a vice, really - and I lost an inch of plexiglass to shards. Whooops!
The PCB segments are even worse in their own way, the fiberglass dulls my cutting bit in about a second. I guess I need tungsten carbide bits here, maybe I'll grind down a drill bit for one.

Proud Mary

Sun Apr 06 2008, 03:07PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

If the gaps are accurately aligned, front surface mirrors at top and bottom of the stack should increase the amount of UV available for pre-ionization of the subsequent gaps after the first gap has fired.

So far as I can see, the fast Marx designs of Mayes and others are often arranged so as to be fitted into a pressurized (very portable) cylinder, which also provides the stray capacitance said to be needed for the sequential gap firing in the 'wave erection' mode.

There is an interesting paper Development of a 300-kV Marx generator and its application to drive a relativistic electron beam which contains lots of useful practical detail of home made parts, which you may find helpful:

Dr. Shark

Mon Apr 07 2008, 06:33PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Harry,
that last paper you cited is part of my printed paper collection too. Seems that there is only a limited amount of publications available to the "general public" (with IEEE access, that is), so the papers I have read are likely to greatly overlap with what you and others have read.

I'd like to get back to the issue of pulse rise time once more. I am happy with the fact that, regardless of inductance, you can theoretically get a voltage rise time as short as you want, e.g. in the picoseconds. But since the power delivered to the load is P=I*V, the current has to rise fast, too. This is, in my understanding, where the inductance of the generator comes to play, and I would like to go back to my example at this point:

We have the frequency omega^-2 =LC and the impedance Z^2=L/C so for a given pulse width and target impedance we can exactly calculate the required capacitance and inductance. E.g. if we want energy at 1GHz, and Z=100Ohm, we get L=20nH and C=2pF. This is completely and utterly ridiculous, so I wonder where I am going wrong?

This example may be rather over-the-top, but remember that with my proposed 10-stage generator, we would be looking at inductances in the uH(!!!), which would basically mean deliberately placing inductors in the discharge path, thereby considerably slowing down the discharge, just to get the impedance right. There must be somethings wrong here, but right now I am completely blind to what it is.

Proud Mary

Mon Apr 07 2008, 07:13PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

I am not clear what you're getting at here, old fellow.

Rise time is always limited by bandwidth, since an ideal square wave would require an infinite bandwidth.

The output impedance of the conventional Marx is capacitive and this can make matching to a specific load difficult. Generally speaking, the erected Marx appears as a single capacitor to the load.

Ultra-fast Marxes are powerful UWB sources, but are not suitable for the generation of a single frequency such as 1000MHz as you have suggested.

Dr. Shark

Mon Apr 07 2008, 07:32PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Sorry if I have been unclear, I'll try and rephrase the question. I certainly agree with everything you have said above; the point I was trying to make is the following.
Say we have given erect capacitance, like 200pF for the particular configuration I was proposing for my capacitors. Now to match this to a 100Ohm load, we wish to make the characteristic impedance equal to 100Ohm. In this case that would mean introducing a total inductance of 2uH, since
Z0 = sqrt (L / C) = sqrt (2uH / 200pF) = 100 Ohm
now 2uH is a rather big inductance, and even with very sloppy design I would probably not get up to that value, to I would have to add a series inductor to the Marx generator to make it look like a 100Ohm line. This does not make sense to me since a 2uH inductance would severely limit the bandwidth, and probably confine the output to frequencies below 10MHz.

Proud Mary

Mon Apr 07 2008, 09:06PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

Now I get you, Dr Shark! You need a Pulse Forming Line to match the Marx to the load.

According to Mayes:

Traditional Marx designs utilize a pulse-forming-line
(PFL) to deliver energy to the load at a matched
impedance. Because of the physical length needed to store
a useable amount of energy and deliver a pulse from the
ten to hundreds of nanoseconds, PFLs can range from
several feet to several dozen feet in length depending on
the dielectric used. These dielectrics also serve the dual
purpose of providing voltage hold-off between
conductors, and consequently are also used inside of the
Marx itself. As a result of the insulating media (typically
transformer oil) and the large capacitance necessary to
rapidly charge the transmission line, the system becomes
extraordinarily bulky and difficult to calibrate and control.

Dr. Shark

Tue Apr 08 2008, 04:11PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Harry, I really appreciate your efforts here, but it seems that I still didn't make myself quite clear. What you cited above is about Traditional Marx designs, i.e. a huge multi-kJ beast, whereas I am interested in a compact low energy system (I already spend $110 on the capacitors, for only 8J energy), that can be directly coupled to the antenna. I belief that this is also what both Dan and Mayes have in mind.

I'll just bite the bullet and build something, I won't be able to do any precise tuning anyway. My new scope (yay, Tek 222) only does 1MS/s single shot, I would need at least 1GS/s for some proper diagnostics. If it does not work for generating UWB radiation, I'll just use it for flash x-ray or some other easy-to-drive load.

Regardless of the application, and if this whole thing is ever going to work, I am still curious about the surge impedance though. Would it at all be valid to link the condition of Z_load=Z_0, where the highest possible amount of power is delivered to the load, to a critically damped discharge? I am trying to link the reflected wave traveling back and forth down the line to the oscillations in an underdamped RLC circuit. This would fit in with the definition of Q = Z_0/R_load, but it wouldn't it also mean that the speed of light is decreased by something like 10 orders of magnitude in a typical circuit?
Looks like I have some reading to do...

Proud Mary

Tue Apr 08 2008, 06:54PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

Dr. Shark wrote ...

Harry, I really appreciate your efforts here, but it seems that I still didn't make myself quite clear. What you cited above is about Traditional Marx designs, i.e. a huge multi-kJ beast,

Mayes doesn't mean that at all. Mayes is a business-minded scientist, and Farr a science-minded American businessman. By 'conventional,' Mayes means in comparison with his 'wave erection Marx,' which is the thin basis for his academic claim. Farr also has to promote his technology as new to sell it to his military sponsors.

Did you notice that Farr was not able to stop the transmission line to his 'impulse radiating antenna' from flashing over? That's because he couldn't match the impedance, isn't it?

Now, it seems to me that you could use a Marx pulse to excite a tuned circuit, but not with any great efficiency - as with Farr's transmission line flash-over. If you decide to do this, I would suggest a lumped constant tuned circuit such as a cavity resonator filled with oil, and immersed in oil, as a good place to start.

Notice that Farr has dumped most of his antenna energy into a network of terminating resistors, but even then the system would still flash over.

I know of no other technology but Pulse Forming Lines for impedance matching of the Marx output, and I cannot think of another matching technology that would not flash over and/or be destroyed by the output pulse.

Anyway, I shall be most interested to see how you get on with your experiment.

Dr. Shark

Thu Apr 24 2008, 02:20PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

After I finally obtained (hopefully) all of the needed materials, construction of the generator is underway, so I am starting to put some consideration into the antenna structure. I have been browsing through the Radar Handbook, a true gold mine for all things RF, and I got interested in Phased Array Antennas, more specifically planar arrays. From my (naive) point of view this would solve two problems at the same time:

Virtually arbitrary impedance by serial / parallel connections
Very high gain if the array is made large enough

The second point is really a question of size constraints, at the very low frequencies the generator is radiating, anything remotely directional would be prohibitively large. But the impedance matching point may be more interesting: My generator is largely capacitive, so it looks like a 10 Ohm source rather than a 100 Ohm source. So if I have an array with 10 antennas in parallel fed by individual 100 Ohm coax cables, presto, I have perfect impedance matching.

Now the question remains, how do I split the 10 Ohm output into ten 100 Ohm outputs? This is probably not trivial, right?

Proud Mary

Thu Apr 24 2008, 09:20PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

Why don't you connect a TEM UWB horn made out of sheet metal or PCB board directly to the Marx ouput terminal, Sharky?

previous 1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob