Building a new Hornyak neutron detector

[Doug Coulter]

Since I'm setting up what amounts to a new research station to test my ideas about ion trapping and recirculation in a domain previously not investigated, well, it seemed good to outfit the setup with some good sensors and the usual data aq (which I'll cover elsewhere, but it'll be similar to the one on the big system, just different stuff in the database).
Prime among the sensors needed especially for this work are neutron detectors and the usual geiger counter for general monitoring and of course safety just in case something works well.

Since the entire point of this setup is to move ions around in a manner similar to a Paul trap and look for deliberately created collisions that generate fusion and neutrons as a byproduct, it seems best to use a neutron detector that has good time resolution - one might want to know when in the phase of the recirculation the collisions are happening at. This means a fast neutron detector, not one with a bunch of time-smear due to moderation being required to slow neutrons for a slow neutron detector. Thus, the Hornyak design.

For those who don't already know, most neutron detectors depend on a capture of a slow neutron (meaning you have to slow them down to be detected - most neutrons are fast when generated) and subsequent reaction that produces enough energy to detect that single low energy (because you threw it all away in a moderator) particle. A Hornyak only sees fast neutrons, because they knock protons out of the hydrogen in plastic, which are then detected by your normal scintillator, this one happens to be best at charged particles and not so great for X rays, which is just what we want.
The photomultiplier tube has a gain on the order of a million, making detection fairly easy at that point.

This is one of those projects that's been sitting on a back burner while collecting parts for quite some time now - the smaller one used on the big rig does that job fine, there wasn't much need for a better one up till now (and actually, I still don't need a better one, but I;m about to have one with about 4x the sensitive area). I'd bought a nice 2" button from Eljen sometime back, Bill found us a very nice 10 stage Venetian blind type photomultiplier, and various mechanical pieces like the SS tubing and lead sheet this uses for an end cap. One wants to exclude X rays from this style of detector to prevent false counts - even though the ZnS:Ag doped scintillator is less sensitive to X rays than knock-on protons, it's not a perfect world, so we help it along a bit.

So, the basic design is to stick the scintillator onto the end of the phototube with index matching goop (and in this case, some kapton tape as backup), get the thing insulated and centered in the SS pipe it's going into, make that mess light tight, and somehow get HV in for the phototube, and a signal out. I plan to use a variant of the "fine wine" preamp described elsewhere here, stuffed right into the end of the SS container, and probably do the HV supply as a bump on the cable from some wall wart - big phototubes actually do draw some power in the resistive divider chain, but the CCFL's I like to use for low power HV tend to be really noisy if in proximity to microamp preamp inputs, so I don't think I'll even try to put it all in one box here. On the bench I'm using one such CCFL supply, regulated and variable, to test this thing - and it appears to work, at least on cosmic rays. Before I get too much further along, I'll fire up the old big fusor rig and verify the levels and sensitivity while it's easy to modify things.

So, the parts look like this:

sub assemblies

Some nice thickwall SS pipe is capped at one end with a cup made of 1/8" thick lead I formed on my press in a die. The inside of the pipe is painted black, and for now the cup is just taped on - it's a tight fit, but tight to the hand and light tight in the blue-green aren't the same thing.
Since the phototube and button are a little smaller in OD than the pipe is in ID (long story...but another version didn't fit) - I cut and machined a little bit of pvc pipe to tape on near where the Hornyak button joins the phototube, and cut a little copper disk to center the socket and resistive divider in the back. As you can see, I have around 3.5" left to add more stuff - probably the preamp to get the signal loud and down to low impedance right there so it won't mind going down a cable and so EMI won't be a hassle. The other end plate is a piece of aluminum I machined - that hole will be tapped to fit a BNC for the signal, more holes probably will be drilled to let the HV in and add a switch for preamp power, which will likely be a 9v battery (the preamp is easy on power).
Here's some data on the tube and divider. The divider is actually using mostly 470k resistors and has a couple of capacitors near the higher current dynodes, as is common for spectrometers.
I required it a bit from the nominal 2 wire setup, as I like to have HV come in to the photocathode and last dynode (through a resistor to let it up a couple hundred volts from ground) and use the anode as my signal output - at this point it's pure current, and isn't polluted with the current (or noise) from the divider resistors - no photons, current is essentially zero. This is good for running right into the base of a low noise PNP transistor...as everything else has a fast attack/slow decay time, this matches that characteristic, as the transistor also will turn on fast due to the current into the base, but not be quite as quick turning off. A sort of matched filter - and here what I care about is the onset, the timing of a neutron hitting this thing.

9956B_photomultiplier.pdf

Phototube specs - I rewire this a bit.

(185.85 KiB) Downloaded 311 times

The schematic as it sits now

That 1 meg load resistor was in the socket...so I used it, but it's obviously far to high an impedance to get any decent time response. No worries, the preamp that will be half an inch away will have a lower value shunt. This works out fine for the scope showing a cosmic ray hit, though.

It works! No smoke or anything. Not even noise.

Plenty of play room for a preamp, battery, power switch...

Sure is nice to have this shop to make all those little "intellectually trivial but a huge PITA" mechanical parts....

Since phototube gain is notoriously a huge function of the HV voltage, we have a handle on how to get that in the range we want - just adjust the HV. The usual method here is to create a CCFL with a voltage doubler (so there's plenty headroom) and regulate the *input* via an LM 317 sort of affair. These little CCFL inverters are quite "stiff" as far as output impedance, and even if they weren't, the resistive divider is hugely dominant as far as load goes, so the system works well. The 59jkhz or so noise of the CCFL, as long as it's out of magnetic coupling range from your other stuff, is super trivial to filter down to "absolutely quiet" (despite remarks made on another site about fusors). .47 uf capacitors are around....and not expensive. If you want to get to microvolts of noise you can use 2 RC stages there. You'll have more issues with any noise coupled into the ground by the 50khz magnetic field. Once you know that, it's an easy problem to avoid.

For reference (duplicates of good data seem wise), here's the fine wine schematic. This is designed to be low noise, DC coupled if you like, and see fast onset negative going pulses, which is what this produces. It's biased to sit right around "it's all turned off" when there's no pulse, because, hey, that's cool and it saves battery life if you use a battery.

I'll leave a few parts out that aren't needed here.

I'm right now waiting for a weird tap - 3/8" by 32 tpi. Anyone who uses those screw in BNC connectors will know that it's a pain when the nut on the back gets loose. Here we're goin for the gold, so no nut to get loose!

[Roberto Ferrari]

Hi Doug

Glad to know you will try that way.
I had an old NE 402 (92% B 10) coupled to a R6095.
Will try to get some neutrons with a Ra paint and a beryllium window (from damaged diffraction tube).
Do you have some neutron source in order to test your setup -beside your fusor-?
Roberto

[Doug Coulter]

The fusor is by far the best, most copious, and in this case, calibrated source of neutrons. I have a couple neutron sources I made with Po trom staticmaster brushes and thin sheets of Be I cut with scissors(!) - but even with those insanely hot alpha sources touching the Be...it's like 30 neutrons a minute. The problem with that is that there are more from cosmic showers (or you-name-it that will make a detector count), so it's hard to be sure if you really have it working or not, though you can at least tell how sensitive it is to gammas (most gas tubes have this issue...).

So I'll use the fusor, that's one reason I left it undisturbed. It just works, every time (hopefully I didn't just jinx myself). In theory this new one should be around 4x as sensitive as the one on there already. Though the quantum efficiency of hornyaks is low, they're really fast at the onset, so you get timing info...

[Roberto Ferrari]

Well, going to numbers my Ra-Be source would give at best 10>3 n/s.
Will wait to have my fusor running.

[Doug Coulter]

I forgot the exact conversion rate of alphas to neutrons in that kind of source, but it's really low. So of course I looked it up again. https://en.wikipedia.org/wiki/Neutron_source
"As an example, a representative alpha-beryllium neutron source can be expected to produce approximately 30 neutrons for every one million alpha particles."
That's effectively with both things powdered so all the alphas escape the emitter and hit Be before being slowed.
What I mainly recall is that the ones I made were effectively useless novelties - really couldn't use them for much.

It IS handy to have something that makes hot gamma rays around so you can test your setup to be sure it's not just being a geiger counter...
There's also an experience thing - you get used to the false rate from cosmic ray showers and can sort of know that the basic lashup works at all if you get hits at about that rate.

Which is interesting here as I have two buildings...and it's different. One is single floor, thin roof, and has a lower rate of cosmic ray showers (less for an incoming one to hit), and the two story building with massy solar panels on top has a much higher rate of showers on the ground floor...

If you get to reasonably pure D in a tank with almost any sort of grid and around 20kv actual, you should see plenty of neutrons for detector testing. Here we get solidly out of the noise around 18kv.
It goes up super fast after that.

[Doug Coulter]

It's coming along nicely, and if things go well, and the mailman brings a couple parts I want in here (like nice connectors for 9v batteries I was out of), then it'll be done today.
That preamp design...I'd forgotten how nice it is for things like this, or in general negative going pulses with fast attack and ??? decay such as this produces (most gas tubes as well).
It's especially nice when you have a floating pure current source, the otherwise unconnected anode of a phototube run with the cathode at a high negative potential, the last dynode off ground
via another resistor in the usual chain, and the anode around ground, but just tied directly in here. As long as there are no light leaks...it's hard to be more perfect. Not that there is any issue,
but phototubes are notorious for gain vs supply voltage variation. But it's quite easy to filter the output of a CCFL inverter to millivolt ripple (or less) since the frequency is so high, and
a decent capacitor size with a little series R gets you to "DC + silence" pretty easily. Yes, they are noisy electromagnetically so you can't put one in the same enclosure with out "extreme heroic" measures to stop coupling magnetically. So, just don't do that (it was done in the other hornyak, but I DID do those measures - separate internal enclosure etc). Here it's going to be external.

A couple of pro-tips:
1. There's such a thing as a 3/8-32 tap. Get one and most issues with screw in BNC class connectors go away - a little loctite and no more loose nuts ever!
2. You really can use a regular cheap kludge board with pads on .1" centers for 0806 components and sot23 transistors (you mount those a little crooked and they hit 3 pads).

You probably can't even see most of the thing in this picture. I didn't go for small or particularly for clean, just built the design. And it works flawlessly (again, gotta love it). Interestingly
it only draws around 140 uA at anywhere from 7 to 10v input with the input open, and when there's a meg to ground on the input - which I tested as the phototube socket had one, and
I felt like leaving it in there for modularity/safety - it goes down to around half that due to a little less base bias. This entire thing is biased almost at turnoff of all the stages
and draws extremely low current at idle - I might not have needed the power switch! Before putting it together, I'll add the output coupling capacitor and a 47 ohm series resistor
in the output to prevent emitter-follower oscillations I've seen in the past and to make the waveform nicer on long coax runs (which I'll never need, but..).

If I do this again, I'll just put something like an arduino nano on there, run from USB power, and use a toggling arduino pin and voltage multiplier to make the -9v or so this preamp likes.
Maybe. If you don't need accurate timing on the hits...just counts...that'd be the way to go.

I note that in tests with cosmic rays that the waveform is exactly as predicted by the attack-decay timings given for the phosphor used. It doesn't get any better than that.

The build, messy but 'purty don't make it run good'

[Doug Coulter]

First light - and it works! Happy dance and all that. Only testing with cosmic rays at the moment, just to get "in range" on the gain, which it seems I've got plenty to burn.
Not super surprising as the previous one worked fine with a pretty numb 5 stage phototube, and this one is "the big guns". Only needing around 440v to get to the approximate gain required
for a real neutron (well, I'm guessing some, but I'm used to what cosmic rays look like on these) - to trigger, say, the cmos counter in a C-Lab Standard Counter, or the counter in some arduino.
People coming from the video - scroll and read up-thread to get what's going on here. This is a special detector that mostly only sees fast neutrons, as are produced in fusion, and it has good time resolution for situations where that matters, vs any slow neutron detector that gives a lot of time-smear from the slowing down process you need if you started with fast neutrons.
https://youtu.be/EmuPjKLqKfk
And...here we are. Lookin good on the bench, anyway.

[Doug Coulter]

Annnnd - it works in real life, and is at least close to the expected sensitivity with -440v on the phototube (I use the site to backup my own documentation just in case, for later).
This of course is how we go about making and calibrating the tools we use to get the data I publish that some s**tposter on youtube says I don't generate. Yeah, right.

(I kind of suspect that since he posted from a content-free profile that it's a burner account for someone whose pet theory I've blown out of the water - I don't get many of these nastygrams, but there is one guy - recognizable via word usage, grammar, syntax, and hobby horse - who seems to think it's his mission to put me down for not contributing enough - even resorting to ad-hom attacks, while he, anonymous, himself contributes precisely nothing)

While this itself isn't useful fusion data, it is useful engineering information on how to get such a sensor working.
And oh, here's some nice data on it working. This was plotted with the old "standard counter" software, which was also running its little geiger counter. It's gratifying that at least at floor level where it happened to be to make the "grab the first cables you see" reach - the geigers are staying right around the natural background, so I'm not taking any big hit here from x rays or gammas.

I didn't run the neutron up really high either in this particular test - just far enough to get a feel for the correct adjustments. There will be more to follow, but this standard counter/hornyak will go into the new setup as both metrology (once I have a real calibration, perhaps later today if the solar power is good), and as safety gear to warn me if things suddenly start working "too well" for my safety. Gas tubes can (and have here) saturate and stop counting at really high fluxes. These, well, they're pretty numb as things go - see the dox linked above from Eljen - and hard to saturate without it being obvious. I also now know that if I stop hearing a natural background - something isn't right.

Anyway, here's a nice screenshot from the lab machine. I plan to do more later on, perhaps some video, but you know that without a cameraman I'm well past one-armed paper-hanger turf and it may not work out. The lab is cramped, there's no way to put the camera on a tripod and catch it all, and I need my hands for valves and mouse clicks to control things. While the old girl has become really reliable as these things go, sometimes it takes a bit of a nudge for all those data acq computers to be happy in concert. Humans are still needed in the loop here. Maybe Bill's next visit he can play cameraman.

Screenshot from local control computer - mild conditions in fusor, just a nice cal source this time.

Here you can see where I first turned on main HV, and then reduced the gas pressure and added more input power. You can also see some background from cosmic rays (I assume). If we were working down in the noise, most of those could be rejected via waveform differences - some of them are very sharp fast pulses from things hitting the phototube itself, for example - so they don't look like real neutron hits on the detector scintillator. Also, most pulses with noise on the decay side indicate you're seeing a cosmic induced shower from an altitude well above you - and the different parts ot the resulting particle shower get to you at different times, as they're going different speeds. Nice to know, this must have been pretty confusing to early workers.

[Doug Coulter]

I'll probably wind up editing this post....
Due to real life interrupting with reality and whatnot, I've discovered that the sensitivity of this thing is just too doggone dependent on the phototube gain, which is itself super dependent on voltage when we're at the low end of the curve, which we are right now. The upshot is around a 10 to 1 sensitivity variation with around a 20v difference (440 to 460) on the tube. While I could regulate it well inside a percent, that's just too touchy and prone to error down the road for my taste. It looks to me like I should run on a flatter part of the tube gain curve, and cut the preamp gain down to match, and get something a lot more stable in the longer term. A real pain right now - but thinking down the road, the systems I'm working with are so complex that I simply can't tolerate much error rate in any one thing, so some extra sweat now will save a lot later on...wrong results reported are especially embarassing.
So while I took pretty pix and videos that would take a day or two to upload even after re-rendering down to lower bitrates, other than as perhaps dark humor at my own expense...the real thing to do is just grit the old tooth and fix this up correctly for long term use. It's too nice and we're too far into this effort-wise to blow it now.

I realized calibration is a little tough with two different display programs. Their times don't match because I can't start them simultaneously, and while one uses a 10 second moving average of 1 second samples, the other uses an IIR filter on 1/10th second samples and wiggles a lot more - so when going for precise ratios, well, it's a little too much up to what the human sees for my taste, and I'm fairly trusting in my abilities. So, as usual, a little more work to do.

One cause of this is in the very nature of hornyak buttons. A neutron may graze the thing and most of the energy just escape out of it - analagous to scattering out of a gamma spec sensor.
Even a direct hit could be in the plastic between the phosphor and by the time the knock on proton hits the phosphor, the energy is largely gone, or conversely, the hit could be inside one of the phosphor bands, which as you can see in the picture in the datasheet linked above, is more or less opaque - that's why there are alternating rings of white and clear - to let the light out.
So...it's kind of a judgement call where you draw the threshold line here even in the best of times, and a little bit of gain change can make a big difference in whether a pulse is declared to be a neutron or not...Thus it's incumbent on the instrument builder to pay close attention to that one indeed. So...I will, as I don't want wrong measurements, they mess up my ability to solve problems via the use of factual observations.

Well, OK, here's a little eye candy if you aren't me looking at my mistakes, anyway.

Normal data aq on top, standard counter with new hornyak on the bottom, with bonus silver activation at the end.

The lower two plots are from the old standard counter program. The seconds elapsed DO NOT MATCH the upper two, which are our current data aq/database system controlled by the raspberry pi, whose remote screen (VNC) is shown on the left. The standard counter stuff uses 1 second measurement intervals and adds a 10 second moving average to the plot (FIR). Our big data aq uses 1/10th second samples and uses a FIR filter to smooth the results on the plot, making comparison a bit difficult. An error on my part - I must have kicked the HV switch on the standard counter sitting on the floor - is the cause of that gap in the lower right plot, and the loss of around 100 seconds of the silver activation data, which would otherwise be a lot more impressive...sigh.
Doing all this while attempting to also run a camera is evidently a bit past my speed....I'll get more up when I have stuff worth looking at. Perhaps I could use some of this footage in the "it almost woked" section as an example of failure and what causes it - maybe how to avoid or overcome it in future efforts. No one like to put stuff there for some reason, but it's potentially the most informative/instructive part of the board, if we'd just use it.

[Doug Coulter]

Ok, I did as I told myself I'd do, and generated a calibration with a new dedicated regulated supply, and the magic numbers are that it's 3.96 times as sensitive as one half the diameter, and it's 3880 cpm per million neutrons/second. This tracks down to pretty low and up to pretty high numbers, though this has a detectable/countable background count of around 10 cpm and the old one has what looks like nothing in this particular run (but I've seen it count background somewhat on other days...).
In a fairly long run I got a good number of calibration points from both detectors and did the math. This also seemed to agree with what I heard on the 3he and b10 tubes on audio, which I always run with for "comfort noise" if nothing else. All is good, it can do upstairs now.

I just realized I should do one experiement with it before that, though. If I make a moderator with say, a small hole in it (a few calibers deep) then it's directional - and I've never done that little test to see if in fact the neutrons are evenly produced all around the tank. It'll be easier in the other one, but two tests is far better than one - or none. I recall back to Tyler Christensen showing what looked like localization of neutron production near his tank walls, and I think I saw something similar once, not enough to make a real scientific claim yet. Well, questions are what we do here - we attempt to answer them.

Here's one from the old days when he and I first noticed this:
viewtopic.php?f=25&t=52&p=74