Generic preamp for proportional type tubes

[Doug Coulter]

In line with my previous post on generic detector power supplies, I'm also working on a generic preamp for various types of proportional tubes, in my case the tubes would be 3He, BF3, and B10 types. They all need a preamp, preferably right at the tube to cut noise pickup down and reduce parasitic load capacity in cables (also to reduce microphonics). After a fair amount of futzing around, I've determined that you really don't need the worlds fastest amplifier for these -- they have finite response time, and if an amplifier slew rate limits, it really doesn't hurt things, as the eventual outputs are going to an audio amp and a TTL/CMOS counter -- in fact, what I've found is a medium speed but quiet and low current opamp acts like a nice pulse stretcher for these signals.

What has been working well in the things I've tried so far is a simple transimpedance setup. That's a fancy name for a current-voltage converter using an opamp. You simply put a resistor between the output and inverting input, the size of which determines the sensitivity. I've found 10 megs to be a good value for most things, if the opamp has low enough input current, which is fairly easy to find these days. This will give signals out from millivolts to volts in one stage from the tube, and since you almost can't get as few as 2 opamps in a package these days, you can use the other amp as a slow comparator to make full logic level outputs over some threshold.

I universally battery power these, using one or two Li primary cells; 1/2 aa is a nice size and has an amp hour of capacity at low draw. So for a design that draws a couple ma, you get decently long battery life even if you forget and leave it on overnight, and it's not even worth it to use a battery holder -- I just solder to them and stick them in the box with hot glue. These batteries also have a very nicely flat output voltage vs life, so I might even be able to skip using a voltage reference to set thresholds -- for neutron detectors, you just need something above what the hottest test source of gammas and betas will trip, and when a real neutron (or a darn cosmic ray) comes in, it will make a full scale pulse out of the preamp stage anyway, so it's not a touchy threshold at all if you control the tube voltages well (see generic power supply for these).

Like I mentioned in the other thread, this is a good candidate for a homebrew PCB, it's pretty simple and the only thing that really change from situation to situation is some parts values.
But of course you have to build those first couple of them to find that out, so I built one on a kludge board, now one closer to real to apply it and test if a different opamp would do.

Opamps known to work:
TLO-84, your generic fet input opamp, works great, but needs at least 9v to run. And split supplies as the common mode input range doesn't go to V-.
MCP-6024 -- lower power version of the same thing, a little slower but not much (more advanced tech). I'll try and use these in the final design, but didn't have a 2 amp version in stock and they draw about 1 ma per opamp -- we're trying to be efficient on batteries here. I've used this one a lot for other things, it has nice manners.
TLC-2272 is what I had in stock as a two holer in soic-8, so that's what I'm showing here. I'll probably order in some more of the microchip parts in 2 per box flavor to make more, though.

Here's what the prototype looks like. I changed the connector on the B10 tube to a type N to match what was already in the little box.

Preamp guts, very simple...not really, but it's elegant.

This was a tiny little box that basically came with two connectors, a 50 ohm small dummy load, and a diode in it. I adapted it to this use. The battery is in there, hot glued to the underside of the kludge board.
That made the panel busy, as of course, I had to have jacks for HV, analog and digital output, and a power switch. I skipped the led for this one -- would use too much power.

Busy panel

Here's what the output of the first stage preamp looks like with about the hottest source I have sitting on the B10 tube. This is a 1p22 spark gap tube (ra-226 and daughters since WW II), along with a piece of Th and a CS137 source -- I'm blasting the living daylights out of it here, and am a little nervous sitting close to this for too long. Each source is over 10k cpm on a geiger counter, some a lot more than that.

Response with a very hot gamma/beta source sitting on the tube.

Here is what a single pulse of the gamma background looks like at a quicker sweep speed:

Single pulse shape, not so shabby

B10gammaFast.gif (10.23 KiB) Viewed 6689 times

Not too bad. With no rad source present, I just see a few millivolts of noise (when the metal cover is on the preamp!)


Cosmic rays are the bane of any low background/high sensitivity situation. Here's what that looks like. I have the scope peak detecting so I can have it slow-scroll and still catch the tiny pulses, and in this case, caught it in the act of moving, but you can see the one huge spike there from a cosmic, which is about what you get from a real neutron -- 3 volts, or clipped, in this arrangement with just one Li cell driving it.

Cosmic hit with other rad source present

I am running this on a single Li cell, 1/2 aa in size (two fit neatly into an AA battery holder for 7v) that has volts over life specs that are real decent. Digikey part number is 439-1002-ND on this unit. And here's the volts vs life, the lowest curve of which is just the current this design actually draws. Pretty darn flat for a primary cell.

Li cell volts vs lifetime

As it turns out, looks like about a 350 mv threshold will be needed to keep the gammas from making this count (except cosmic ray energies) so that's a pretty simple 10::1 resistive divider off the battery, neat. That's what I'll try, at any rate, and I can use 90k/10k and not add much current drain. Heck, with this opamp I could use 9 megs and 1 meg for that matter, but sheesh, that's gilding fine gold and painting Lillys. In other words, not much point.

I'll add the thresholding stuff and post again when I've got pics of it counting real neutrons from the fusor, but having already built one of these, I'm pretty sure it's going to work just fine, and finally make solid CMOS level outputs I can plug into my multigeiger counter aux inputs to get that logged along with everything else during a run -- with a non thresholded audio output for my favorite monitoring device -- a stereo audio amp I built into a rack panel with speakers for my fusor rig.

[Doug Coulter]

As promised, I did a fusor run with this all nearby. The middle of the B10 tube was roughly 4 feet from grid center, sitting on the floor, with a couple power supplies in between it and the fusor.
Tube was moderated inside a 6" diameter, 18" long piece of HDPE. Now I know where to set the threshold -- the very pleasing answer is "anywhere" -- I don't really need the second opamp half, strictly speaking, but I'll lash that up too so as to keep a separate straight output and get a buffered output to drive logic levels.
Fusor really isn't up to snuff right now, there's still that leak (we ran out of He trying to find it yesterday) and "one more mod" on that HV feedthrough needed, but it runs at all, which is well enough for this. The second issue kept me down to the low 30kv region, actually but that's still good enough for this, and at "full snot" I may actually have to move the tube farther away!

Here's some pictures. The first is the thing just waiting for me to turn on the HV for the fusor.

Fusor not running, background. Blue trace is a faraday probe in the fusor to check for EMI/neutron counter co-incidence (and vert amp labeled wrong for this probe).

Now we turn the fusor on and try to keep it running long enough to take some of the money shots:

Setup as above but neutrons present

See above

See above

Note that I had to run the fusor in the less productive "stable" mode at high gas pressure to get these. That's really what the faraday probe is showing. It goes nuts during the higher Q pulsing modes, and ensuring a pulse type discrete counter such as this isn't just counting on EMI is one large pain -- which is why we do the thing with the silver as backup -- no way for that to ever be dead wrong. This is more useful for tuning and seeing quick things happen as you adjust something, the sliver is, well, the gold standard and final arbiter of real results. Room for just about all types of detectors here, and in fact, I forgot to turn on and look at the gamma ray spec head while I did that run, so down to the lab I go again -- that would be nice to know if it's useful as a fusion detector too (and probably easier for most than getting some sort of neutron tube going, unless you have too much money or luck on ebay).

[Doug Coulter]

Here's the wrapup on this project. The schematic as finally finished is here:

Schematic for proportional tube preamp

Click on this for a bigger version.

I used the TLC 2272 here, because I had it in stock. It's a little slow, but then, so are these tubes, so it all works out. A too-fast opamp would just allow for more troubles with oscillation and noise pickup. For a little bit faster response, I'd use the MCP 6022 in this circuit. It draws about the same power, but is a newer tech, and I've had good luck with those too in similar uses.

The schematic shows another way to generate a threshold. For this one, I just used fixed resistors because the threshold was far from critical, and a pot is bigger, less reliable, and costs more.
But if you wanted something a bit more generic, that would be the way to go. I show resistors above and below the pot so by adjusting all three R values, you can make the adjustment less touchy by having the pot cover a smaller voltage range.

Old-school opamp jocks will perhaps have kittens on some of this. No bypass! It's just not needed with an Li cell and short wires. No hysteresis on the comparator! Again, simply not needed here. You would need these things with older tech, a sloppier larger build, or much faster opamp -- more isn't necessarily better in this game. Heck, I didn't even use a reference diode to create a stable threshold. It's not needed here due to the very flat curve on the battery voltage during discharge, and the wide margin that basic tube provides.
I could have added a small series R between the input coupling cap and the diodes to limit diode peak current in "hot plugging" -- but really, you shouldn't do that anyway with HV already on, and with the tiny cap value, even 1n4148's will probably take the impulse that generates. I'd rather fry those than an opamp, at any rate. I was simply running out of room on the tiny surfboard, and didn't want to add extra complexity anyway. Really sharp-eyed old schoolers will notice that the input stage is slew-rate limiting on the front of large pulses -- so? We're not doing low inter-modulation distortion audio here, it simply does not matter. Now, if we were doing a fast gamma ray spectrometer off a phototube, with 5-10 ns pulses, it would matter, but we're not -- that design will be up here later on, and it is trickier when linearity matters and speeds are very quick.

After building this up, putting on the lid, of course I had to test it. Here's some waveforms. I had a chunk of U sitting against the tube, and a neutron generator (very weak one) also sitting on the moderator. I'd guess most of the actual counts shown here are really just cosmic rays, but even then the count rate was quite low -- several seconds between counts and clicks on the audio amplifier.

Waveforms

b10pre_cosmic1.gif (4.45 KiB) Viewed 6661 times

The upper trace is the linear output, the lower is the digital one. You can see the noise from the U gammas and daughter betas in this if you look closely.

b10pre_cosmic2.gif (4.5 KiB) Viewed 6661 times

b10pre_cosmic3.gif (4.53 KiB) Viewed 6661 times

As you can see, I got away with no hysteresis, and no bypassing here just fine. This is about as sweet as it gets in this kind of work. The digital pulses go to about 3.3v, which is a high in either TTL or CMOS even if those are 5v powered, and right on down to ground, no glitching, and the rise and fall times are nice -- not so fast as to create ringing on a decent cable, and not so slow as to create noise-induced false counts either.

Now I'll build one for our faster responding 3He tube (which also tends to count faster as it's more sensitive) and see how that plays out. Currently I'm using a two transistor circuit with a too-high threshold to catch double-hits only, and the pulse widths are on the short side for audio from that circuit, so I'm going to try this one (with the faster opamp) and see what I get there.
We also have two BF3 tubes to try with this soon. So we should be able to work up a family of these, with minimal changes needed between tube types.

Note, that 170pf input coupling cap is a "special" thing, not so much the value (47 pf to 250 pf work fine), but because it's rated at up to 8kv and has very low leakage to generate odd noises. Further, this needs to be mounted solidly, as it's in a high impedance circuit and one face is at high voltage. This creates an electrostatic microphone if it can vibrate. In all this kind of thing, one should try tapping on the preamp box while listening to the audio output to see if you've got this problem, or you'll have issues down the road. This is one of a bunch of reasons I really like to have an audio output, linear, on just about everything. I built a cheapo radio shack stereo amp into my fusor rig, with a lot of switchable inputs to accommodate this idea, and it's a very nice things to have.

I'll detail that in another post, but just point out here that a digital signal going into a digital counter can't give you all the available information. The threshold "throws away" some, and it doesn't matter to a counter whether all the counts come in very fast, with nothing in between, or in some timing pattern. But your ears can tell, and often as not, it matters. The stereo imaging can also tell you things about the coherence of two sensors -- are they counting together in some burst fashion, or both just random. It's quite striking, it doesn't take a lot of practice to make your brain into a pretty good way to notice "interesting things" going on, and is very useful in the discovery phase of things -- which is just about all the time in this game.

[chrismb]

Doug, I got the impression you were doing this using opamps so as to preserve pulse height for spectrometry purposes, rather than pulse counting alone?

I have been testing out a recent Geiger tube (seems it has next to no 'proportional' range) with an LM311 comparator, which I am lead to believe by texts [I'm only parroting, here, as I am no electronics wizz like you] is a better solution for fast, FSD pulse outputs than opamps. It's extremely easy and very sensitive - I pull the one channel up a 0.1mV above ground, then the other channel is connected to the tube by a capacitor, and to ground by a big-ish resistor to hold it low. Pulse width can be controlled by that resistor value (wrt the cap) and if you wanted to lengthen the pulse then I suspect some diode in there somewhere will make that channel act as a charge-pump.

I had a stack of LM311s for the other part of my project, but I've already got some opamps on order because I wanted to see if they could help differentiate pulse height from this tube.

[Doug Coulter]

Chris, I've been editing this while you were online -- refresh to get the whole thing,

The real reason to have a reasonably linear preamp right at the tube connection is to make the signal loud enough so that the noise on a threshold generator isn't much bigger than the signal itself. This satisfies that. The signal straight out of the tube is for example, far smaller than the noise in your typical comparator input. As an old engineer said to me once, it's a millionth of a gnat fart.

There's no point in looking at amplitude too closely in a B10 tube, actually. It's a proportional tube, true. But gammas (or betas) give relatively low outputs, as shown above in some detail -- that's why you use a hot rad source free of neutrons to know where to put threshold, wheras neutrons give anything from zero to "max" depending on where the reaction products of B10+N are produced and where they go in the tube. So the only reason to have any amplitude linearity at all for this is to reject the gammas/betas, and only see the few reactions off the B10 that dump tons more energy into the gas than they do. The amplitude doesn't mean anything much here, as some are partially trapped in the boron itself, or don't go right through the center of the tube and dump the full megavolts into the gas. So all you can say is if you get a really big pulse, it had to be a neutron (or something with megavolts of energy, like a cosmic ray muon). What that means is that you're going to miss some neutron-boron reactions that don't happen to put more ions into the tube gas than a gamma ray does -- that's just the breaks of the game here. The same is true in all gas type tubes. They work because the neutron<>whatever reaction dumps so much more energy into the gas that it makes a much larger signal than anything "normal" like gammas or betas. Some of the neutrons are also lost below threshold, it's one of those tradeoffs we see in just about all engineering, nothing is perfect.

So it's like the old problem of setting a threshold on a burgler alarm again (an awful lot of things map to this problem). If you set the threshold too low, you get false alarms, in this case, from gammas and betas. If you set the threshold high enough to reject all of those, you get some missed detects. There's nothing you can do outside the tube to make that better. We don't need perfect linearity or even perfect monotonicity to do this as well as can be done, given the signal here, just some, and it's a lot easier to set a threshold at the millivolt level than at the pico-coulomb level right at the input -- hence the need of a preamp at all, to change a charge to a voltage big enough to be out of the internal noise of the reference and comparator.

Yes, geigers have what amounts to a built-in threshold, it's an avalanche-discharge in the gas in there that makes them put out so much signal power for so little input power (power gain in the many millions). Once above the tube threshold voltage, all the pulses are the same size. They all get bigger if you raise the tube voltage, and smaller if you lower it, but still all the same size as one another if over the avalanche threshold. You can run one in proportional mode, at voltages below the geiger threshold, but at that point you need a real serious preamp like this to see any signal at all -- its micro and millivolts (into a high impedance, it's actually a charge signal -- x number of electrons) in that part of the operating range. Doing that, you can then use the tube to tell between alphas and other things, as alphas make much bigger pulses in a geiger tube, but once above threshold, you may as well just jam the signal into some digital counter direct, there's no other information other than "it's there" or "nothing there" at that point.

This preamp design will let you see signals from a geiger tube run well below the pedestal voltage (Halliday has a nice chapter on that, as does John Strong). But it's not all that useful other than distinguishing alphas from everything else as the gas isn't that great an energy absorber as you've already mentioned. They are pretty unpredictable on gammas vs energy, it's a matter of chance for those how much energy they deposit into the tube gas. Ditto betas which will only dump some tiny fraction of the original beta energy into ionizing the gas. Alphas dump quite a lot more over the same path length, being higher charged and much slower moving.

So for the purpose of using this on geiger tubes, it's completely irrelevant and off topic -- they have a built in preamp if run as designed. If you want to use a geiger at lower voltage and as a proportional ion chamber, I hear it works better with reverse polarity, but that that point -- it's just as easy to build an ion chamber, like Charles Wenzel did, and you can easily make one that works better than an undervolted geiger tube for that job. If you need real energy linearity, you use a scintillator and a much fancier preamp, as also mentioned above.

See, we know the Q of the boron10+slow-neutron reaction, and it's very high, and makes an alpha which efficiently transfers energy into the gas if it traverses any. But depending on things we can't control, that alpha may not go through all the gas -- it could be produced at the tube edge and be going sideways right back into the tube wall, so some neutron reactions won't show full energy. Therefore, trying to use this as a spectral indicator is more or less completely a waste of time, and in fact, detecting neutron energy with any confidence at all is quite a trick according to all the literature. Here, we've lost that information a long while before they get to the tube, because without a moderator (which slows an also randomized neutron energies and directions), the cross section of neutrons on boron is so low you really don't get anything at all.

Pretty much all these arguments apply for BF3 and 3He tubes as well, hence the title "generic preamp for proportional tubes". We don't run them up to the geiger/avalanche region at all, that's what the word proportional is all about. But we do run them up the the point of having some "gas gain" so they are "proportional" rather than "direct". Else a mere half a MeV of energy would be very hard to detect with the best tech there is -- at CERN, it's about e^4 electrons best case in a silicon diode at far higher energies, which ain't a lot of signal. A MeV isn't a lot of joules to say the least, heck, it's a tiny number in ergs!

Those old lm311's are going to give you some heartbreak -- think "I need hysterisis" or you're going to get some serious garbage - there are some good examples in the app notes. They also have a lot of input bias current (300 na), so you can't use them in a high impedance circuit easily at all -- it's really bad with those compared to a 1 picoamp (or even 10 na) opamp input. Back when they were designed (well before my hair turned gray), that was a required tradeoff to get the speed. The newer tech is a lot better in those regards FWIW.
That's why you see almost no proportional range. By the time you're out of the comparator noise (many times more than the opamp) you can't see anything at all at tube voltages where the tube is proportional. When it's in proportional mode, the pulses are microvolts - more or less. Well below the comparator noise for the biggest pulse in that mode of tube operation. My spec sheet on these doesn't mention noise at all -- which is the huge red flag there. It's microvolts, plural, and that's bigger than the proportional signal levels.

You need to use at least 1 meg in series with the power to the tube, or risk frying it with glow discharge under some conditions. They really don't like that. Pretty much kills them in a few seconds, due to using up all the quench hydrocarbon gas in this particular tube design.

In this case, with a slow signal (in the 25 us range) we just don't need comparator speeds, so I just used an opamp for that part too. As seen above, it obviously works.

[Doug Coulter]

So, now that the first one works so perfectly, time to go on and do one for the 3He tube. I already had one that was beautiful in some regards, made from discrete transistors. It drew precisely zero current when not pulsing -- bonus. But the pulses were much too skinny for audio, and too small for counting with TTL kinds of things, so now we make one to spread out those skinny pulses a little, and drive logic levels too, just like the one above. This time, what the heck, I went ahead and used a full quad opamp, this time the MCP 6024 which is lots faster than the one I used before, and a little easier to switch out should it get fried (stuff happens ). For a bit more speed, well, this is a much faster opamp (about 10x the slew rate), and I split up the gain into two amplifiers in the linear section, which takes better advantage of the GBW product of the amplifiers. I still used a section as a comparator, as in this case the free "low pass" is desired. Though this tube is faster than most, really, it's not an RF situation, and we'd just as soon not have it able to count at RF kinds of rates -- that would have to be an error. That left one section drawing power, and tying it off in the reccomended way on the datasheet is worst-case for that, so I decided to use that last one as gain=1 buffer for the threshold source, which is a potentiometer across the battery -- this one will have full range available. It probably has too much gain for the way I was running the tube earlier, which did produce volt level pulses into a meg impedance. I will simply run the tube HV down to get less gas gain if I can pull it off that way, else I'll be in there changing resistors. I really like the linear part (for audio and other analysis) to stay linear and not clip.

I'll have more pix and a schematic soon for this, and some report on how it works. I'll now have to pay attention and turn off the power when not in use.

Front side of kludge board

I used a socket, not ideal from a noise point of view (or parasitics) but nice if I have to change the amp out for whatever reason. Sorry about the hairs in the pic, this is not a clean room!

And, here's the other side. I instinctively do this as though it were a PCB layout even though you can really "cheat" on these. But the next time you want to change parts, and realize you used component leads for "tracks" -- you'll learn.

Back side, flipped L-R, top is still top

mcp6024.pdf

Opamp data sheet

(637.81 KiB) Downloaded 493 times

Here's the schematic:

Schematic as built

Notes:
All caps are generic .1 uf ceramics except the special input coupling cap, a special low noise 8kv 170 pf ceramic.
The resistors are mostly non critical. The one from the HV to the tube has to be large so it won't arc over in case of a short (and maybe fry the tube in the process).
I used RCA for the linear output, SHV for the HV input, BNC for the digital output, and the crazy UG-560/U for the tube (since that's what's on the tube). This will screw right on the tube in use, like the other one, this box is a little bigger so I didn't have to cram things in as tight (pix to follow, after I test and do any mods required).

The input protection diodes are a little strange here. They have to turn on quicker than the ones in the chip to make any difference (or you need a resistor between them and the chip input pin). I happened to use some very bizarre germanium power diodes (ampere ratings) I had laying around (inherited from an older engineer), but schottky should do fine here too. You want them beefy enough to handle the full peak current that 170 pf cap can put out if the input is suddenly shorted...In other words, they're not needed if that can't happen and you're careful.

10 points for anyone who notices the wire that's on the schematic but missing from the pictures (going to fix that now).Or for that matter, an error I drew into the schematic. I'll give you that one -- I actually configured the second gain stage non inverting, with the 100k to gnd and the feed from the previous stage to the + input -- and it matters, oops, as this is all based on ground levels and the opamp can't make negative of ground outputs. The way I actually wired it, I'll get positive going output pulses. If you do it like this schematic...negative going (if the + input had been biased off ground). Counters would not care one bit. I'll fix that when I redraw the final schiz after testing and any mods.

[Doug Coulter]

OK, here's the corrected schematic. Since no one bit on the 10 points, the error in the build is that the threshold input does go to the .1uF bypass cap, but there's no wire to the correct amplifier input...Could have been tons of fun to troubleshoot.

Corrected schematic for the fancy version

[Doug Coulter]

Here's the result, almost ready for use -- just have to add the missing side to the box and crank in the screws.

Build up

A couple of build notes. I added a HV bypass cap so I can do the same trick with the power supply for this I did for the other one. It's the red cap. The big green guy is the 170 pf coupling cap.
Note that in either case, I paid attention to which side of the cap was at ground (more or less) to reduce in-box corona and noise. I used an old Allen Bradely pot for the threshold. Once I find out where it would be set (in conjunction with probably lowering the ~3k HV supply to get in gain range) I might add some resistors to reduce the range of the threshold adjustment.

The battery is the thing in black tape. At the 4.7 ma measured this draws, it won't have a super long life, and I was tempted to put a led on there to remind me to turn it off when not in use, but then again, that wouldn't help the life either...can't win. They do make a full AA size of this guy (twice the capacity) and I'll order some soon, but -- not cheap, about $7 each for 2 amp hours.
Again, the reason to use this rather than say, 2-3 regular AA's is the flat voltage over life curve that saves other parts and current drain of a voltage reference.

It's dark now, no spare wattage to fire up the fusor, but I'll try to get this tested soonest and report. Hopefully, I'll now have "stereo neutron detection" off my audio amp, along with being able to jack the digital outputs of both into the multigeiger data acquisition box. I may have to add some prescaling on those counters to handle the fast count rates better. We'll see, it's a software setting in the PIC hardware there. There's only one hardware counter in that, used already for the geiger tube, these other inputs just interrupt the PIC, which could eat a lot of cycles...