Just some pix while testing a little harder - when the sun is up, I have power to burn, in this case, somewhat literally.
New meter from one of those old CDV-715 boxes. 2500v full scale now (not Roentgens).
Of course, with two of those 100-something watt resistors as load - around 100ma at this setting...there is smoke, but it's only the load smoking!
For once, it's the stuff that is supposed to smoke.
With a bonus cheap IR cam pic of the load Rs.
Yep, they get hot.
load.png (14.41 KiB) Viewed 5869 times
Only one transformer got warm enough to come out of the background here. It's the GE one that needed extra turns on the primary to match the other. I think that hot spot in the foreground is just a reflection from a bench lamp. I couldn't feel anything warm after turning it off there. The transformer was just barely warm to the touch.
Where most of the other 25 watts went.
xfrmr.png (11.75 KiB) Viewed 5869 times
So, I'm going to button this up and make provision to use some nice coax to push this HV to the amplifier, where there are mA meters for both screen and plate current. Between them I should be able to hit the kill switch if something goes wrong before a lot of bucks go bye-bye.
Oh, there are around 20v of ripple at ~ 100mA. The two transformers don't match perfectly under all loads at all drives...which is probably most of it. If things go right at the fusor, it might even be something you could hear modulating the neutron output.
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.
Well, nothing goes as fast as you wish, I suppose. I'm trying to get within around 6dB of what I think the design parameters should have been...I know I'll have to change stuff, but I see no point in starting way the heck out either. C'mon good solar weather!
So, here it is now, having its output isolation transformer being tested for impedance at the tube end vs load at the fusor end, with a dummy load that approximates what it would see tuned off to the capacitive side a little (since I didn't have quite the right one for the dummy and I don't figure I'll be out of range).
Looking pretty good, the coupling is better than I expected, resonance on the secondary shows up clearly on the primary side as it should, resistive load shows up as it should, voltage ratios closely match turns ratio...I have to hope that 3/8" thick teflon will stand off 50kv at least...and now I'm just making all the little accomdations for the connections it will need in the real world prior to power tests on the dummy load...which can't last long as that load (10 100 ohm 10w resistors) will go poof pretty quickly at the predicted power. But it should only take a few seconds at power to see if I CAN get to power with the other values inside the amp for protective bias and grid leak value - pretty sure I erred on the safe side with those.
Transformer primary - 60t #14 around 4 6" long, .5: diameter type 61 ferrite rods Secondary - 30t #14 on the outside of the pipe, centered as shown. Right now loaded with 100pf and 1k for testing. Awaiting a new input cap for the main pi network from HV stuff. The one I have in there, while correct value for the F of interest, is at best a kludge of 20kv caps in series-parallel, and at best 40kv peak. The new one will replace that 100pf cap with a 1000pf pi network input cap right there on the chassis where the current is. Then the only remote is the pi inductor and fusor RC load. I'll be tuning the C out, and the R looks like most of a megohm, depending on how many ions are present. Thus making this match a bit of a bear - kilovolts of RF and "no" current - it'll be hard to measure the actual load as losses will be high in comparison. Maybe I will get the QSO with Bob going...
Messy bench, but fun.
First try at this pic I thought someone had taken my Kodachrome away. The old Kodak camera with "real glass" had internal lens fogging after being left in the unheated space for awhile. My panasonic camcorder (which also has "real glass") has more pixels, but horrible color balance for this...and the 150+ page manual is "somewhere" - I couldn't find out how to fix that from the GUI in it. But by the time I got frustrated, the old Kodak had cleared up, just like a cheap rifle scope brought in from the cold.
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.
Well then. Testing this thing is the greatest challenge so far. First light - using 5 100ohm 10w "chinese" watt resistors - smoked the dummy load in 10 seconds or so, barely even managed to get a reading on the pk-pk output on that.
So I rummaged and found a 50w Ohmite "american watts" 500 ohm R and fired it back up, along with a higher voltage rated freshly calibrated scope probe, since I couldn't tap down on this load. I must have gotten 20 seconds or so of operation before the signal generater started beeping frantically, putting out wrong stuff, and blowing its fuse. I know I'm not backfeeding its output, must have somehow gotten into the AC line, which isn't a 3 terminal thing, has a very wimpy looking common mode xfrmr, smallest I've ever seen...and what looks like 1n4007 diodes of which two are now 0 ohms either direction. Which I found out after I put a "penny" in the fuse box - on strand from stranded wire, which survived - the line cord blew...(!) and this computer reset from across the room, as well as the printer in the next room. The closer-in raspberry pi didn't blink, of course - which is one reason I use one at the fusor for data acq and some of the control. (all things running on the same 120v circuit). The pi has a somewhat larger capacitor to power use ratio and can take short line dropouts too.
So...I guess I either fix this siggen power supply and cord, or go to the good old "eat lighting without a burp" GR tube siggen, and maybe test on the porch, only bringing out the test gear required. Whoda thunk that a mere 1.5kv pk pk or so 1.6-2.0 mhz sweep would do so much to things here - not one wire in the system, even the 120v, is even a tenth wavelength, and the dummy was on 6" leads hanging from the transformer....zowie. I never did get above around 1kv on the main plate supply.
Now I'm going to have to run around and see if anything else blew. Usually not, but...it pays to check before those things are needed and trusted to tell right answers next.
And maybe look for a lot better dummy load in case I want to test at power for long enough to set and read scopes and suchlike.
And for what it's worth - I may need a smaller grid leak R, a little, but resting with screen volts on and plate off - maybe 5 ma screen current. But with grid drive - holy cow, it goes to the moon (pegs my 100ma meter). At least with that sweep it didn't fry them with the average (the drive xfrmrs don't quite do that range flat). So I guess I need an xor of HV present and drive to work the interlock relay. Both have to be there, or not there for things to be safe.
Not sure how I could shield this a crapload better - the parts that have the big RF on them are already "tight" and not too big as it is. I could build a weird looking screen cage around it all...I guess. But then getting 50kv in and 50k out (plus RF of a few kv)...gets interesting indeed.
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.
I guess you aren't even to the point about worrying about how to cool the tubes yet. In the photo I don't see any kind of fan or provision for cooling the tubes. Eimac sold a glass chimney for putting around the bottles, and you pressurized the chassis and had some air come through the socket, but most through holes between the tube and chimney. Heathkit took another approach, just had a fan blowing across the tubes in the SB-220 and the successor HL-2200 amplifiers.
The more reactive the materials, the more spectacular the failures. The testing isn't over until the prototype is destroyed.
I guess I wasn't concentrating on showing that aspect. The person who went before on this chassis didn't do anything (but my guess is it never worked). I changed that with a big noisy fan from a microwave oven up top, and a smaller (with an ouch price) pressurizing the bottom chassis where the 6146a and filament transformers live.
Top fan, kind of blends into the scenery
I did actually get the plates to glow dimly at one point, but everything else went poof right around then. I didn't get the feeling the tubes got hotter than design values. The paint is still red. The scope didn't show any parasitics (1 ghz scope). Output waveform as expected - not real pretty since it was basically just an inductive load with a large inductance so the top half of the wave (the kickback part) was a little short in duration. Didn't get time to push the "save me a picture" button on the scope. I also push air into the bottom, at the 5v transformers, which exits past the 6146a into a hole under the topside fan. A little probably gets through the holes in the power tube sockets.
Smaller new fan - that was expensive (rare size)
I maybe could reduce the grid leak resistor on the mains. I have 5k there across the otherwise not very loaded bias supply that makes around 60v protective bias. It seemed during sweeps before the dummy load and signal generator went up that only near its happiest center frequency was there plenty of net grid drive. But I really didn't have time to check or even put a probe down there. The signal generator blew its switcher supply diodes. Dunno how.. replaced them and other things are bad in there too now, it's toast. A Feeltech FY-3200s that was popular for awhile in places like eevlog. Not the world's greatest thing - just cheap and handy. But with a foot of line cord (gee, it's like a microwave oven in that regard) and plugged into the same power strip as the amp, it's hard to see how that much got into the thing - coax to the bnc on the amp front which is grounded to the AC line ground itself. Looks like the other stuff on the bench survived, though I haven't yet tested the nearby Analog Discovery 2 (expensive little toy, almost worth the money, tells lies now and then if you're not paying attention).
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.
Well, I have it mostly tested and now moved to its new home, just needing some parts from some Chinese vendors - a real dummy load that won't fry for final testing, and a new input capacitor for the output pi network which is driven by the output transformer. Same value as the older pi net input C, just some 10's more kV rating. It will mount where the little brown one is now on the amp chassis under the output transformer.
Overview.
Controls and monitor
Behind the curtains
Upper left is the shielded pipe that feeds the ion grid. Right now hooked to a 40kv 2.5ma Spellman, but in reserve is a homebrew 20kv (roughly) 45kHz or so AC supply that creates a neutral plasma but takes a bit more fussing. The DC one creates a very net-negative plasma. Foreground is the RF amp. Up to 50kv from the big Spellman SL2KW is coming in on the PE wire you see curving in to the white encapsulated bypass cap I fabricated, the secondary is on the outside, and the little brown cap is just there so I could test at all - it will soon be replaced by a larger one of the right ratings for this, at least I hope they're right.
What I saw at 2.9 mhz was that Q got better fast as I went off the edge of the plot - the limit was how low a pressure I could run and keep the plasma going. As the pressure goes up, it detunes a little - no biggie - but also loses Q in both uses of the word - things are more smeared out when there are more collisions especially those far from the grid. The result is that it "looks like" the ions are seeing a lot less field than what we put on there - at DC and 50kv it's about 1/10th what we thought we put on.
Edit - when I read my own words in the above paragraph and thought they weren't clear, I added the below. In the first sentence above, I meant fusion Q. The big surprise here was that I'd finally gotten the math for the usual ion trap stuff right, all the correct units and whatnot, and as I approached the conditions under which that math is operational, lo and behold, things worked. This took (I'm not a wonderful mathematician and sometimes the physics guys change units mid-equation without warning) some doing, and some back checking with things like existing mass spectrometers. But now it's good - under the assumption that not enough ions are present to perturb the applied field or interact significantly with one another. This assumption is of course, wrong for the conditions we desire in a fusor. We want lots of stuff and we darn well want it interacting! But it's a beginning, though obviously missing some terms (at least, see below about chaos). But it's hard to learn things when it's just getting good when you lose the ability to get more data. So you adjust the experimental conditions as I am to try and get some sort of maxima "on the paper" - it may or may not work, but we KNOW it was off the paper before, and that even agrees with settled math and technology. Now we want to advance those things. To me, the most obvious thing to try was to compensate a bit for what looks like the counteraction of our applied field by the charges of the stuff in the tank, which I'm doing with more voltage (both DC and RF) and lower frequency - my landmark bunching video was after all, taken at closer to 100khz than here, and it was pretty definite that things were doing the sort of thing desired, just not enough to fuse. In all this, remember that our geometry has far more volume outside the grid than inside, and so the interactions are quantitatively different, especially if we have any sort of synchronous motion going on in and outward. The ions are simply much farther apart when near the tank walls than they are if they wind up at a focus inside the grid...and a hope here is that their fate can be "set" on the way in, before their interaction can disturb the desired trajectories too much. In short, that "smear" effect is variable in space and time...this is non-trivial to describe, so my sorry attempts to describe my vision may be lacking...life is what it is.
I AM expecting electrical Q to go down, for ions to be lost in non-fusion scattering and hoping to compensate for those things. But it's hard to do with such a multidimensional error surface without just getting in there to try some stuff, and following the old mark I nose. So here we are.
I'm also very carefully edging back up to the sort of power levels we used to use - that one incident where evidently arc-driven drive ringing with parasitic L and C and super high fusion Q put a bit of fear into me. (and FWIW, the magic resonant frequencies measured then are real close to what we're finding are good now when we drive things deliberately!) For now, I'm running right at the threshold of input where we have neutrons well out of our detection noise, but well below the danger level, while we look for the first and second differentials on the input parameters....In some cases these are levels where with just DC drive - and the exact same stuff in the tank - we didn't have neutrons out of the noise at all, now we do with this addition, so I'm encouraged that we'll at least see a good improvement over the old ways here.
End of edit.
What I'm hoping to learn here is how far I can push the basic ion trap equations in the presence of this smear. I ran out of oomph at 2.9 mHz, and since all the voltages look "lower" to the ions than what me measure outside the tank, the obvious thing to try is more voltage, and in this case lower frequency (which in the equations somewhat compensates for lower volts as there's more time per cycle to move things). I'm of course hoping to get some sort of peak "on the paper" at all, so I can fish around better for any sweet spots there might be where we are getting recirculation of the ions at least - this scheme won't handle the electrons which would take another RF source at another frequency and so on - but hey, baby steps and maybe you get there.
Rught now, there is NO math that describes things like the area between what we call viscous flow (hydrodynamics only handles group behavior but can't tell you anything but "it's going to get unstable and make vortices" - but no details about their internals, and molecular flow where everything is ballistic. Well, since we have charged things, we're really not in the pure ballistic state either - charge "sees" each other with the usual 1/r2, so really, there's no good math here either.
The ion trap math makes the assumption of just one charged particle, more or less...at least, so few they don't interact. Of course, we want them to interact like crazy, so we're in a different world. In normal trapping experiments, you can get a bunch of things to be trapped stably, but they "learn" to avoid one another via collisions knocking some out of the stable aperture (their jargon, not mine) and being lost. But it takes a lot of scattering to lose a lot of them. What I'm hoping is that we get plenty fusion before that state obtains, and since we can feed in more ions at will...well, hope springs eternal.
And at least we'll learn something about what is now a completely mysterious parameter region. Even with just plain air - uncharged - no directed motion - Lesker and others just draw a straight line between where the viscous flow math craps out and the ballistic math craps out and calls it a day....math isn't the queen of this part of science at the moment.
This guy has a fun math channel on youtube (it's all reccomended - Grant is the best out there, and his showing of how pi can be generated by colliding blocks is killer aha stuff) - and at the end of this shows some of what we're dealing with as far as "there's no way to just feedforward this, you have to try". https://www.youtube.com/watch?v=p_di4Zn4wz4
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.
Well, after a few fits and starts, this thing is gonna fly - and fly with authority. The caps I encapsulated start acting like avalanche diodes at around 21kv, going down to 3 kv (at a current limit). I used diggalbe potting compound, so maybe I can recover, depending on what actually failed. For all I know it could have been the silicone. At any rate, replaced it with a 4800pf 50kv one from HVstuff.com, and put in a test pi input cap of 750pf@ 40kv (waiting for the real 1k pf @ 50kv to come on the boat), just to fire it up. I found out the 40kv cap actually arcs down the body at around 32kv...and in finding that out, the lightning back-drove the output transformer, which then caused the tubes to arc backwards into the grid circuit, frying the DC protective bias supply.
And it worked anyway, at least if driven right and for not too long.
But I fixed the bias supply and added an indicator to it so I'll at least notice quicker in the future should that happen again.
I have about 4k loading around a 60v bias supply - that's effectively the grid leak for the big tubes (which are RF driven with a 1::1 transformer from the 6146 plate, the cold end of the grid winding gets the -60v), and half of that load is a divider down to 27v for the 6146a. Bias increased a few volts during hard drive. Not very much, there must not be much grid current. Yet.
But it's underloaded - I need more turns on the secondary of the output transformer and more pf on the pi input cap it's driving to correct that. I can only get to around 180ma (both tubes) with increasing drive at around 2kv plate volts before screen current heads for the moon - a sign of under loading. I can, of course, get the thing toasty hot - but you know, the pi network coil, which is 115uH, wound on a beaker, gets hotter than the tube envelopes do, even when the plates are bright orange...this thing puts out the oomph, no problems there - maybe 35 or 40kv pk pk RF at the grid (after the pi net steps it up) with ions present, and with DC or not (but that gives that too-small pi cap heartburn). The two 50pf vacuum caps in series across the grid seemingly don't mind anything.
So, a few tweaks and I can move on from engineering just electrons to engineering plasmas again....I didn't see anything special on these tests, but then I wasn't even trying for that - I did have a neutron detector on and hooked to audio just in case, but nothing dramatic there. I didn't shoot for any magic numbers, even if I knew what they are for these conditions - I wanted to make sure things go up and down when I twist the knobs up and down, and they do - and don't take no for an answer. My work here is nearly dung. Some lo-fi pix from my thermal/optical camera. The thermal and optical images don't quite align, and I can't find a non-destructive way inside to fix that. The optical camera part isn't that wonderful, but it gets the idea across.
Warming them up real nicely
IMG34.png (53.04 KiB) Viewed 5592 times
Not quite smoking...skin losses?
IMG35.png (47.69 KiB) Viewed 5592 times
Tubes not that hot on the outside. That microwave oven fan is a real blaster.
IMG36.png (40.96 KiB) Viewed 5592 times
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.
Well, the new stuff works, at least without DC HV applied - I didn't get that far. In testing, I tried full power (what I could get without running up the 4.125 screen current, it's pretty obvious on the meter when it's time to stop) both with and without the ~ 25pf or so extra across the fusor grid, no ions present.
And...my HV probe fried. And did so before I could recalibrate the actual gain/loss of the probe - it's off by a factor of "several" low here - there's no scope setting for oddball probe divide ratios which this is. So I guess I'm back into making PCBs again once I find the artwork...because it was the board itself that went up, near the input - the fields in air are fairly fierce there as it's close to the point at which the fusor grid connects to the outside world. Every gas lamp and tube in the shop lit up, this thing makes real power even though I'm not yet matching it correctly.
Not sure why this failed, or why it failed *there* - insect?
The inside of that quartz tube is going to require some serious cleaning too - this really emitted some junk and made quite a stink. Even half an hour later it's stinking up this room where I brought it to take the picture. Wow.
RF amp, with vacuum caps: Eb 3.5 meter Ib 100 ma Is 16 ma drive 8v pp HV 3 divs on 4kv scale (known wrong) 6146 plate 40 ma Fs = 1.9503 Mhz
W/O caps: Eb 3.3 meter same variac setting Ib 200 ma Is 20ma drive 3v pp HV 4divs 6146 plate 35 ma Fs 2.625 mHz
HV probe smoke second capacitor traces Still undercoupled. Need more turns on secondary of output xfrmr?
The amp HV supply meter reads ~2x high - it was built with 0-5 scale, I'm using it for 0-2.5kv, roughly (it's not really linear but it is glued shut very tightly so I can't fix the scale). These output tubes are not seeing anywhere near their rated plate voltages. At 200ma, the plates do glow a little bit though not much. This was somewhere around 1750v on the plates (is 2k unloaded from supply at this setting).
4 divs on the HV probe *was* around 32kv...DC, when it last worked. This is a lot of RF voltage!
I do need to add some kind of bleeder on that gonzo plate supply filter capacitor, it stays up for a really long time, and the old screwdriver trick is really hard on screwdrivers.
Sigh, I hate to waste good solar days fixing stuff when I've got the power to do real runs.
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.
