Friday, April 2, 2010

Fast Micro-Marx Generator revisited

////// EDIT: pretty pictures added 4-2-2010 @ 9:08 AM //////

Eventually, this blog is going to earn me a visit from some three letter agency, I'm sure. That is, in part, why I am blogging about these fun amateur science projects so publicly. So if I disappear suddenly, please do drop an inquiry to the warden of the Thompson Correctional Center, Thompson Ilinois. I understand that's where political prisoners are going to be going soon.

But I digress, as usual.

I've been a bit ... not myself lately, and struggling with various challenges. I haven't accomplished much actual work on any projects (until recently) although I have been slowly setting up certain machining operations and preparing the way for various things to be done when I'm better rested, not stressed out, wide awake, and so forth. There will soon be parts at stake which have many hours of time in them. Fucking them up is not acceptable.

Oh yeah: I swear a lot in this blog, as I am attempting to write in a more chatty tone, and since I swear a lot in real life... there you have it.

I have tattoos also.

I should buy a motorcycle.

Anyway. Today I was obsessed with thoughts of the Fast Micro-Marx Generator, and not ready to proceed on the Mad Scientist Light Switch parts... so I did a lot of thinking when I wasn't having to use my brain to do my job, which was most of the day to be honest. Over my lunch hour, I dug up some papers by various really smart people about spark gaps and clever Marx generator techniques and peaking gaps and so forth and so on. There's a lot to think about for a guy who isn't very strong in math.

When I got home this evening I was ambitious for the first time in weeks, and I set to work. One of the assemblies that makes up the FMMG (cough, get used to it) is the housing which is, incidentally, also the current return and coaxial shield of the Marx Generator. The housing shall be constructed mostly from a length of 4" diameter copper pipe. The ends were saw-cut as I received it, and after cutting a piece close to the final length needed, the ends will need to be made square and circular preparatory to having flanges fitted to each end.

A problem arises: how to turn this pipe. Although I can chuck it in the lathe, it is soft, thin-wall copper and I don't have a way to support the other end.

The answer is a mandrel which can be supported at the far end by the tailstock center, fitted with two disks to support the tubing, one at the headstock end which supports the tube where it is clamped by the chuck. The other disk is fixed to the mandrel an inch or so inside the tail end of the pipe using, oh, three set screws. The OD of the disks are chosen to be only a few thousandths less than the ID of the pipe. Et Voila! The tube is now supported firmly, gripped by the chuck firmly and may be carefully turned. Which reminds me, copper is notorious for being a S.O.B. to machine, I must look up the tips and tricks and figure out what kind of tool I'm going to need. Yay insert tooling!

Tonight, I made the mandrel. It's a piece of thick walled steel tubing (pipe) fitted with inch or so long bushings turned to fit neatly into each end. The pipe had four quarter-inch holes drilled into it. The bushings were welded into place through these holes as well as around the flange left on the end. The the welded ends were annealed with a torch prior to turning. The ends were turned down, center holes were drilled into each end, and it may now be turned between centers. Except since I haven't got the headstock accessories required to do that, I suppose I'll fix one end to the headstock disk with a bunch more set screws. Hooray. I want the disks removable so I can use this for another tool some day.

Anyway, the mandrel proper is finished, but lo and behold, the fucker has a taper - that is, one end is larger in diameter than the other. A fairly significant taper. This means that the center of rotation of my lathe's head stock is not aligned axially with the center of my tail stock. This is annoying, but correctable, by adjusting the tail stock. I'll want to mount up some other hunk of metal rather than carving the walls of my mandrel ever thinner. Harrumph. This is annoying, but not surprising. I've never turned anything long before, so I wouldn't have noticed (although I've suspected for a long time).

After I decided I was done working in the shop, I cleaned up, came inside, had dinner, and then set to working on my Marx generator spreadsheet. I've started a section to help me do the calculations for the distributed capacitance and inductance of all the current paths, some of which are tortuous. If I do this thoroughly and correctly, I should be able to estimate the characteristic impedance of the thing. That will be important to know later.

After I built models in Solidworks, I was dismayed to find that my doorknob capacitors take up quite a lot of room and that I might not be able to fit as many stages (ten) as I'd originally hoped:

Marx model rework.jpg

That's the current state of my Solidworks model for the Marx bank itself - what I've come to call the "insert" since this is the bit that goes inside the housing. It was nice and tidy and complete, with all of those spheres (the stage gaps) mounted to the panel with screws and a peaking gap design mounted at one end... until I discovered a few things about peaking gaps. Then I had to tear it apart and I haven't had time at work to finish changing the configuration, design, and dimensions. I don't have a machine at home on which I can run Solidworks.

At any rate, I now realize that the output peaking gap has to be a separate assembly with a different gas pressure (and different gas, most likely). Making it a separate unit frees up room inside the Marx housing for another stage. The end with the two holes and an empty space is where the peaking gap was. We are now looking at eight stages with a charge voltage of 10kV, so an erected voltage of 80kV max. Not the 100kV I'd originally been hoping for, but c'est la vie.

The peaking gap, IF it is connected directly to the output of the Marx bank, will steepen the Marx output rise time, as it provides a slight delay ensuring that all stages in the Marx have erected before connecting the output to the load. It is possible to construct special peaking gaps which go from non-conducing to full conducting in extraordinarily short times (pS regime) by using special geometries, gas mix, and pressures versus what's used in the Marx bank.

But there are other clever tricks in the pulsed power business that I am tempted to try. One of them is the intermediate storage transmission line and peaking gap. Funny things happen when you do this right, although the dimensions - especially length - can become inconvenient quickly. Here's a block diagram of how the real pulsed power boffins typically obtain ridiculous peak powers by pulse compression:

pulse compression concept.jpg

Now _if_ (and that is a big 'if') I decide to try such advanced shenanigans, I will have to maintain a constant impedance throughout the entire structure, Marx bank, intermediate store, peaking switch, etc. So it behooveth me to know what the characteristic impedance of my Marx bank is to begin with. Knowing that, I can figure out the necessary scale required to do the more advanced stuff (and see whether it's practical or something I really want to get into). Also, it will enable me to make that peaking gap switch work better, and ultimately, match the output of the pulser to some (still hypothetical) load.

The original purpose of this Marx generator was to generate a fast enough (and high enough voltage) trigger pulse for my distortion-triggered spark gap switch:

finished switch, ready for testing (sort of)

A very fast, very high voltage trigger is needed to achieve proper field distortion triggering and hopefully operate the switch in pseudospark commutation mode.

However since I acquired the spiral generator (which I've previously blogged about):

PT-55N front.jpg

...I'm not sure I need the Marx for that purpose any more. But I do still need to build a small prototype and understand its behavior thoroughly before attempting the larger device that I plan to use for UWB impulse radiating experiments.

Thus, the new tables in the spreadsheet, and many measurements to be taken from 3D models and parts to plug into the spreadsheet formulae.

Over the weekend, I'll probably go back to working on the Mad Scientist Light Switch project.

That's how it goes,
everybody knows.

PS: Watch this space, I'll try to backfill some pictures, after I get them taken, or at least throw in some screengrabs of SolidWorks models.

No comments: