Video output transformer design

High frequency, antennas

Video output transformer design

Postby Doug Coulter » Mon Aug 23, 2010 9:44 am

I am preparing to build some "video amps from hell" here, and am trying to find some good data on broadband transformer design without having much luck. I need to match say a push pull parallel quad of pentodes running about 1.5kv on the plates to an output that can make 50-60kv and have pretty wide bandwidth -- say hundreds of khz to tens of mhz. The requirement will probably ease considerably once I learn what I truly need for my fusor drive experiments, but I'll never find that out if I can't kind of sweep the space with arbitrary waveforms first.

The best advice I've gotten (seemingly) so far is just make the best air core tranny I can manage, then stuff RF ferrites in it to get it to go lower in frequency. At the turns ratio needed, self resonance issues look like a big problem to me and the solution isn't obvious (to me). Fortunately this can be fairly low duty cycle, so I can really abuse the tubes, which will make 2 amp peak currents apiece if they don't have to do it too long at a time. So I've got 1.5kv @ 4a peak per side on push pull in my back of envelope design. I could perhaps go higher in main DC input volts, but that gets me into troubles with getting the tubes really turned off well enough to not burn up just sitting there with no signal. The tubes I plan to use, 6KD6, will take 7kv (and larger) peaks on the plates for very short times, and in general have favorable grid drive requirements which is why I chose them. There was a time when these were used a lot in amateur radio linear amplifiers, which is where I got my experience with them. They will really take a licking and keep on ticking. I plan to drive them with CRT cathode driver chips used on old CRT computer monitors (80v peak, about a watt each) and supply those with a Hantek DDS-3x25 arbitrary waveform generator.

I can't use the good old Pi network most hams use to match things, because at this point I don't think narrowband sine waves are going to get the job done. I could of course be wrong about that, and it often happens in research that at first you need better gear to establish the answer to a question, and after that, something a lot simpler will do.

But even I can't afford one of those monster broad band amps from AR or ENI -- heck, they cost much more than a new car, used!
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: Video output transformer design

Postby Crispin Metzler » Thu Oct 13, 2011 1:05 pm

The tomatoes are going gangbusters so I'll briefly post my ideas on this problem. Longer posts after the snow flies :-)

1) To heck with transformers; Get a bunch of old color TV shunt regulator triodes and build up a distributed amplifier to deliver the KV directly. Hopefully P-P will be OK? If so you're not even pushing 'em. They did a better job baking gas out of the plates than on the sweep tubes which gas and develop brown getters if much current flows in plate breakdown country. We went through quite a few similar 6LF6's doing a 6 hour commemorative bootleg broadcast back in the '80's. Plate modulated Class C into a poor antenna match. You need to choose Z for the lumped L/C line as in 2) below of course....

2) If you can't go direct, use my trick: "Z" is everything. At MF-HF I'm pretty comfortable if 10 Ohms < Z < 1000 Ohms depending on BW. ZBW is as important as GBW.
Without an estimate of Z the problem is undefined, with it I calculate a simple 3 up Butterworth with the desired bandpass. It'll have 3 resonators and if there's a 1 Hy inductor at 30 MHz (for example) I'm screwed, so I chop it into bands or adjust Z or whatever and try again. Once a realizable network is found one can usually tap L or C at one end or the other to achieve the desired transformer ratio or employ other tricks. At higher frequencies you get to use transmission lines and cavities and such for the network.

3) Stacked 1:1 transformers? Probably stacks the parasitics too and won't work. (Some guy has a patent on this using cores).

4) Beware of ferrites. Manufacturers give out loss formulas that make powered iron cores look bad by ignoring the amount of material needed to achieve the desired permeability. Often the real life powdered iron core has so little magnetic material in it that its losses are lower. You get to run 'em closer to saturation for a given linearity too. Run ALL the numbers or just experiment.
Crispin Metzler
 
Posts: 5
Joined: Thu Oct 13, 2011 11:36 am

Re: Video output transformer design

Postby Crispin Metzler » Thu Oct 13, 2011 1:15 pm

5) You did say 'sweep' the frequency? Would you believe a large (Tesla coil sized) Variometer with a nice long plastic rod to the rotating part? A simple p-p oscillator using those 3-500Z's could drive the thing. Fire it up and twiddle the knob. A frequency counter with the cap off the input jack tells you what frequency you're on. Ought to make a nice display at night too.
Crispin Metzler
 
Posts: 5
Joined: Thu Oct 13, 2011 11:36 am

Re: Video output transformer design

Postby Doug Coulter » Thu Oct 13, 2011 6:29 pm

The actual problem is a bit worse than that (though the above ideas are very cool!). I need to strike a plasma in a near-vacuum. Initial Z infinite. Then goes fairly low once there are some ions there - say some 10's of milliamps at 50kv (or more). Now that I have ions, they are the load between a central fusor grid, and the grounded shell. Now, I want to be lble to move them around at will -- accelerate them toward the shell, slow and stop them there, bunch them up, then continue bunching them while driving them back away from the shell and towards a central focus. The ions themselves act as a reactive load across the stray C of the grid-shell system as they pass through the grid and out the other side.
This load will vary all over the place within a cycle - and will depend on how good a job of bunching I've been able to do, and how much that gets scattered in interactions at the grid focus in a fusor.

So, not merely swept, but broadband within a single cycle - it won't be anything like a sine wave. I got an arbitrary waveform generator of about the right specs which is essentially a re-purposed VGA card with a computer interface that lets you draw waveforms at will to generate.

I strongly suspect that once such a system is working, the actual demands won't be anywhere near as bad as described above. The problem is, nothing out there can mathematically predict or model very well what this load is going to look like -- the space charge of the ion cloud is comparable to the net charge we put in there "imposing" the driving field, and can even more than cancel it if bunching is done well. The fancy ion-motion simulation programs can't eat this input (so they say, and I believe them). So, until I find out what works, I don't know what it's going to take beyond some fairly broad guesses. That first one is hardest. Might be that once we get it working the first time ever, we find out that a simple waveform with a couple harmonics does it fine - but at this point, we can't even guess at the rep rate or frequency closer than a couple octaves. We can easily guess that with the ions trying to cancel the applied field, the net field will usually be less, and the speeds slower, than what you'd predict using Newton's laws to predict what would work with so few of them their charge doesn't interact much with the other ions.
What little data I have shows speeds as low as the few tens of khz -- but with some risetimes in the tens of ns...broadband, but those modes were "accidental" and kind of "relaxation oscillator-like" and otherwise non-ideal for what we want to achieve here.

What I'm trying to accomplish here is to control a bunch of self-replling ions, letting them spread out in space and time when it doesn't matter, and bring them back together in a focus, tightly focused in space and bunched in time, just when they need to be to interact. Many will scatter out, and defocus in space and time (and velocity), so each cycle gathers them back up again, then repeats bunching them and firing them back through the grid focus.

Think of a triode, inside out, with the plate virtual at the grid center, and playing with transit-time games, with klystron-like time bunching only when absolutely required, just before they hit focus -- ions are hard to cram a bunch of into a little space and keep them that way, so I just don't try for that. Unlike photons, a charged particle beam repels its own contents, and focus can't be maintained over any large time or space (or at least, no one ever has) -- but if you set up the ducks right, you can get it where/when you need it., recover the leftover energy not lost on each pass, and go again at some rep rate. We're just trying to bang some very uncooperative rocks together here.

There's a lot of lore in the fusor world about ion recirculation. Well, actual measurements here show that as very minimal and disorganized at best - not even a good damped oscillation - Q is max of a couple, maybe. You've got the spring(s) and the masses, and a DC applied force. That's not an oscillator (there's no identified gain mechanism, but there is friction), just a fairly oddball tuned circuit. What I want is to drive this thing and get recirculation "with malice aforethought", and I need to be able to reduce the entropy produced by near miss scattering on every pass along the way.

There is nothing currently on the market at all that claims to be able to do this. While outfits like Amplifier Research make amps that are broadband enough into say 50 ohms (or near), there is nothing that can do this at the KV levels into a ??? changing load at all. And a used AR amp at 1/4 the power I need costs more than a new fancy automobile!
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: Video output transformer design

Postby chrismb » Thu Oct 13, 2011 6:54 pm

These relate to some issues I thought over when I was constructing an HF power supply to put out 500V@~kW. It would have been nice to stick it in to a transformer to get the volts, but I wanted frequency control from kHz to MHz and, as you, to feed both low and high impedance loads. So I just went for direct power switching with a series of FETs which seem to handle open circuit down to 100 ohms or so. Not tried (not needed) it at lower impedance yet.

A little after this, I got hold of an original 50 yr old book by a guy that worked on cyclotrons in the early days. I wasn't expecting to find much I had not hear of, but given my efforts at the time I found a few interesting nuggets that I've never seen repeated in any subsequent works, one of which might be of, at least, interest if not useful:

In the early days when they were matching high powers to the range of drive frequencies they needed (at 10+ kV drive), they used two coaxial air cores to match loads in the main tube power circuits within which was a rotatable ferrite core. By rotating this ferrite in line with, or out of line with, the induced field they got a range of reactance with which they could tune the load match.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 7:32 pm

Re: Video output transformer design

Postby Doug Coulter » Thu Oct 13, 2011 7:10 pm

Right, that's what Crispin was talking about with the variometer, in essence. Still, they only needed a sine for that stuff, and I'm not yet sure what I need, but kind of doubt a sine will be it - that's the crux of this biscuit, the thing that makes it hard, else a ham transmitter swept and tuned manually would have found it already, or at least within just a few weeks of trying.

So unlike that situation where even in a narrowband thing, the tank circuit losses were the limiting factor(!) -- I have it even worse unless I'm just being too pessimistic about what's needed. I've already tried some big sines at various speeds and feeds, no good so far. What I'm trying to do along these lines is make it possible for a human to sweep a much bigger parameter space than just frequency and amplitude of a sine in less than a lifetime. While there is something magic about what happens with the NST drive, it's obviously quite chaotic and in practice, very non ideal and hard to repeat and to control. What I'm trying to pull off is like what I had here: viewtopic.php?f=9&t=58&p=91&hilit=ion#p94 but at a far higher rep rate, and much more control over what and when, and without a secondary off axis grid at some random spot in the works -- that just can't be right! But that's where our magic extra high Q mode is, which I've managed with just the one grid and some various weird series impedances, but not as repeatably. When that latter works, though, it gives roughly 1k more net fusion/watt input (for Crispin, that's what we fusion guys call Q, it's damn confusing to an RF EE , the overloading of the jargon this way).

I bet Crispin can help you with your stuff, though...I got to see his resume when he applied here - and that was that. Good stuff indeed (reminds me of me a little bit but different focus - excellent). Thinking about your cylinder/doughnut thing, I'd bet sines actually work better than square waves re bunching. See the stuff on the drift tube under accelerators for why that is.
You need a slope, rather than a fast risetime, to get the bunching to re-gather the spread out ions. If that's what you want, that is. That project is of course, "on you" and we're all rooting for it too.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: Video output transformer design

Postby chrismb » Thu Oct 13, 2011 8:08 pm

I'm fairly sure that my device requires square waves, or more towards that than sine. This is because if you were to run a sine in a circular manner, what you are actually doing is rotating a 'static' electric field whereas I need a 'switching' field so that work can be done on ions.

I've been thinking this through for several years and I am fairly sure of this. Similarly, it might well also be like that for the drift tube - it is, 'after all', a circular accelerator that has been straightened out (depends on your perspective, eh!?!). I suspect you need square switching too, else the particles will slide backwards in relative phase to the cycles much more readily. You want the particles at the front of the 'wave' because they are much more stable there in phase space - so that if they miss the acceleration field because they are getting ahead of it then its no big deal because they will miss out on some energy and the rest will catch up with them. Having it the other way, which sines encourage, is that the particles will slide backwards in phase so they experience the 'back end' of the cycle, which is a smaller field, then get left behind yet more so.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 7:32 pm

Re: Video output transformer design

Postby Doug Coulter » Thu Oct 13, 2011 9:04 pm

Please read the Haliday I scanned at the link above, it explains how bunching works (and why it won't with a square wave) as well as anything I've seen so far. And it DOES work with sine waves, or it wouldn't work in cyclotrons, linacs, CERN and so on - except, it does work in all those places - the design documents are in our library too, and their results speak for themselves. You need a time varying field that accelerates the laggards more, and the ones that are ahead less, to keep the stuff bunched. A vertical risetime does not satisfy the conditions. Ion traps use sines too...
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: Video output transformer design

Postby johnf » Fri Oct 14, 2011 2:27 pm

Doug
some helpful books
Transmission line transformers
by Jerry Sevick
ISBN 1-884932-66-5

and Ferromagnetic core design & application handbook
by Doug DeMaW W1FB
MFJ PUBLISHING

Jerry's book shows up to 1:25 turns ratio transformers that will go 40 ohms to 1000 ohms over 0.5 -30MHz
INXS make some mighty good high voltage FETs that are very quick which would make great tube drivers which would allow direct coupling for you to preserve phase / amplitude into your load
johnf
 
Posts: 376
Joined: Sun Aug 08, 2010 4:51 pm
Location: Wellington New Zealand

Re: Video output transformer design

Postby Doug Coulter » Fri Oct 14, 2011 3:25 pm

Thanks John, I'll look into those. Unlike some, I know I can always learn more, and that it will be useful at some point.

Depending on the application, I'd be going from several K ohms (xmit tube plates) way up to the kv/ma levels a fusor runs at (megs)....and the problem there is hitting self resonance too low. Cores help a lot (need fewer turns so less stray C), but don't fix everything. Could be why there is no such thing available out there, even at "government" prices. This is why I temporarily ditched the arb waveform multi grid fusor idea and went to the simpler beam stuff to get some more learning to take back. It could well be I don't need what I think I need if I can exclude some of the search space some other way, so that's what I'm doing on that topic right now.

For low power tube drivers, I've found (and built up) some old CRT cathode driver chips - DC-100 mhz, 80v output at about a watt. Good to drive the tube that drives the really big output tube.
I hadn't thought of fets directly for the output (they can have their own drive issues as well) yet, but yeah, that could work. These big tubes tend to need about 200-300v grid drive for full class C, and some tens of milliamps grid current, about 20-60 watts per. Not horrible to come up with. It's worse broadband, as there's no tuned circuit to handle the fact that to get that couple tens of mlls grid current there is none most of the time, and an amp sometimes.

All this is of course trivial if I'm just driving a drift tube accelerator at a single frequency with a sine, which is the project du jour. All I have to deal with there is tank circuit losses from the high circulating currents due to the capacity of the drift tubes, but at least I can try to tune that out with standard narrow-band techniques. And I only need a small step up from the existing plate waveform for that -- maybe 35% at most, not tens or hundreds to one.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Next

Return to RF

Who is online

Users browsing this forum: No registered users and 1 guest

cron