Whodve thunk it -- that gas control, so conceptually trivial, would be so hard to get right, and so important? Not me at first, but it seems to be the case. After fixing my leak, and trying a new change on the almost destroyed main feedthough (which I will report on separately, this time might have been the charm), I could start working on that one again, and it's paying off. Now that I have reliable HV again, and a real tight system, this became the limiting factor, again. The new inlet valve is working pretty well, albeit at the low end of the input regulator pressure range -- I will add some sort of orifice and volume reducing insert to this next time I have the system open, and think that will make it "just so". At this point it dumps just a bit too much gas per pulse for the best control. I was able to tame down the exhaust part by slowing the turbo down further, it's now at 25% of the nominal rotation rate, and a single valve pulse only changes the last digit on the pressure gage. Couldn't do that small a step conveniently while I had a leak, and had to flow more gas anyway then to keep it reasonably pure. This whole thing appears to be something kind of divergent -- when it gets worse, everything gets worse and has to work harder, but when it gets better, everything is easier, in kind of a cascade.
At any rate, without the arcing around the feedthrough (inside the tank), the ballast resistor doesn't get so hot -- so the new cooling system improvements handle much higher input powers easily now, and I'm able to run 20-30 ma at full voltage with fine enough gas control. This merely moved the overheating issues to the tank itself, and now I have to do short runs or melt the neutron oven. More input power is not the way I'm intending on going longer term, but it was nice to be able to take some data up there to confirm that it's probably not really the way to go, and to make sure that everything else can hack it -- and it can, except for melting the oven.
For once, everything worked right. The new detectors and preamps were flawless, the vacuum system tight and right, the power supplies solid, no arcing, the new gas control box (some one shots to actuate the inlet and outlet valves) is handy and very stable, and I mostly got to just sit there and watch it go, and tune around gas levels and voltage/current levels at will.
The net activation results for the (timed, 5 minute) runs were:
Run 1 -- silver to 1550 or so cpm, indium to about 150 cpm (total, that's not super far above the background here which is about 70-90 cpm itself).
Run 2 -- silver to 2064 cpm, indium to 240 cpm.
These counts are from 2" square samples, on a 2" diameter pancake geiger, and are the first 10 second count intervals after getting one 10 second count for background after turning off the fusor. We do the silver first, as it decays so fast, and now have the process more or less "down" to where the delay between shut off and start of count is in general about 15 seconds -- I get one background count while I'm fishing the stuff out of the neutron oven, one count partial background/sample as I put the sample on there, then the next full sample count is what I report.
I believe that second run is a lab record, or close enough for me. Only this time it wasn't chance or luck as perhaps it was before.
I noticed on both runs a kind of serendipitous correlation between the 3He count rate and the resulting silver activation rate. The 3He was counting in the 1.5 khz range during run one, and in the 2-3 khz range for run two. This system counts much faster than the B10 tube, which is both less sensitive and farther away at the moment, but it is still not saturated. Listening to it, with pure DC on the main grid, and nothing on the secondary grid, it's pure uncorrelated white noise from the stereo amplifier. When I "jiggle" things with some sort of AC into the second grid, you can start to hear correlations and bursts that are in sync with the jiggling, and it makes for some nice scope screen shots too.
I didn't run my fancy data aq stuff, which still has some usability issues -- I'm working on that, but I did use the human observer to the max, and what came out of this set of runs was the fact that it is far more important to keep the voltage up than the current. The only reason to set to current limit up to 20 or 30ma is to keep the voltage from dropping too much when there's a bump in gas pressure. In other words, simply pouring the raw power to this is not the way to go for highest Q. The silver activation especially is very sensitive to the last minute of the run, and any serious time wasting glitching there (like letting in too much gas) really wrecks the silver results, but not so much the indium, which integrates better (longer half life).
So this is *almost* perfect. There is the possibility that hitting the exhaust valve takes out too much gas, which requires hitting the inlet button, which puts in too much gas, and a little backing and forthing is still required to get it "just so", and it will then stay there for minutes on end. This tends to mean there's a glitch near the end of a 5 minute run as described above, with a little fast button pushing to get back on the sweet spot again, which roughly speaking, is just a tiny amount more gas than the "lights out" level. This seems especially true without another ion source in there - and that's how this was done this time.
Most of these runs were at currents well below the current limit I'd set -- the best production of neutrons was in the range of 10-15 ma input. To get that while holding something above 48kv is moderately touchy, and the main function the high current limit serves is to keep the volts up during momentary situations of a little too much gas, which keeps neutron output from dropping drastically. The more we do this, the more we find that output is a very steep function of input voltage above 40kv or thereabouts. It seems an almost exponential curve from the spot where we get the counters solidly above background at about 20kv. Every added kv adds more neutrons than the last one, and the curve is very steep near my present limit of 53kv. If I were to keep going down the standard fusor path, the first new thing I'd be doing is upping the power voltage -- the Q and output are both rising fast as I hit the power supply limits on voltage, I've found no peak or other side of a peak going in that direction yet. Higher power (via more gas and more current) is kind of a "meh" result -- things get hotter, not that many more neutrons come out, it's not to scale in that parameter -- doubling current does not double neutrons, so I'm losing Q in that regime. I'm very gratified that even with a kilowatt input, the composite W/graphite grid only gets incandescent (red heat) in the middles of the tungsten rods. The better grids are like that -- they don't get as hot as a poor design that intercepts more ions, and produces a poorer visual focus. This is truly a case of subtlety over brute force it seems.
I've been noticing now that I have pretty stable gas control, that how hot the tank is (and thermal history in general) is now a major drift factor. There's always a certain amount of D in the tank walls and grid parts, which gets driven back out as the temperatures go up, with some time lag. So some sort of ongoing control is required, it's another divergent thing. As the tank heats up, pressure goes up a little, which makes it draw more current, creating more heat....this is a small effect on the pressure gage, but a real big one on the real pressure indicator -- current at voltage, which is much more sensitive than my gage is. Any automation I do for this with some microprocessor will probably want to see the HV supply numbers as part of its input data.
The radiation shielding seems to be working pretty well. There are some leaks in it, pointing away from the operator position, which then scatter around the lab, but at pretty acceptable rates. The Canberra military detector/dosimeter shows not much more than 2-3 times background where I sit, or about 100 urem/hour, more or less (both this and background are variable here). If I put it near one of the leaks, it's up in the few millirem/hour range (the detector sees gammas, betas, and neutrons according to the specs). That final layer of shielding, the lead glass, really helps with exposure at the operator position, yet allows very close inspection and pictures to be taken of the fusor innards, which I like a lot. In this case, it seems to be showing me that Q does go up with a finer visual focus, which is well past what my video camera can see -- looks like the Tech guys have me beat there with a higher quality video camera.
Anyway, finally I've got enough stability to want to start taking real data and establish a baseline with some confidence to compare future results with. At this point, it's not luck anymore, finally, gheesh. So that is the plan -- get a real solid baseline over a set of parametric conditions to compare future changes and improvements against. Then I can start talking more intelligently about what an ion source, or some scheme to get a lower neutral to ion ratio will do, and what various perturbation schemes do for things.
Here's a rather enticing scope capture. Here, we are putting in some 60hz AC on our second grid, and note how it's "herding" the fusion to a particular part of the AC cycle. What interests me about this is that the average neutron rate does not go down, even though the main grid current does, somewhat, and the time during which neutrons are produced is obviously less. So a little fooling around with things like this is obviously warrented. Can I for example increase the duty cycle during which it's making neutrons? In that case I get more output and less input, which of course is the direction to be going in. I kind of seriously doubt that a 60 hz sine is the best perturbation signal, or that this off center randomly aligned grid is the best way to do this -- but again, it does produce improvements even so. Time to whip out a huge audio output transformer I've got (LTC lin standard!) and the arb waveform generator and explore at least the audio range, and perhaps a different type and placement of the secondary electrode, eh?
- Top trace, 3he analog., purple trace after thresholding (and is the scope trigger), blue trace is the b10 tube, and the green trace is the applied AC on the second grid.
