Science/Engineering/Technical forums

These are now available for sale over at DJ's on this board. Joe worked up a very nice schematic and stuff chart (along with doing the PCB layout) which makes it a lot easier to read and describe how it works.
Here's the stuff: As usual, click on the picture to get it blown up bigger.

Shown are the correct parts values for a 6-9v supply (actually it works to much higher voltage as is) as in a 9v battery. Gain is 150 as shown, and the LF rolloffs were pretty empirical for the signals I plan to put through this - and are easily changed. And now that Joe has made it easy for us - I can do a simpler and better description on how it works with real component callouts that aren't ambiguous!

Taking the simple first - we filter your power supply a little just in case. That's a 47 ohm series R and those two bypass capacitors. They're not big enough to take out a lot of 60hz type ripple, they are just meant to handle picked up EMI, such as you have around anything using high voltages. Believe it or not - that 22 uF cap, C1, is the most expensive part on this board, or close. We provide a filtered power output to use as a coax return if you've got a bit of coax between this and your sensor - assuming you're not making any connection to the other end...it's a little bit quieter that way, and our signal ground is actually that filtered power rail.

The input circuit provides for AC or DC coupling, and on the AC coupling you have a choice of LF rolloff frequency by changing the value of C5 - I just pulled a number out of my hat for that - your requirements might vary. Since this is an SMD board, and the gaps are small - don't expect this capacitor to stand off HV, it won't - keep it under 100v or so, or board leakage if nothing else will get you. For most things, you should just use the DC input, which is floating up near the power rail. This is good for a phototube with a floating anode and a negative cathode/dynode supply (my favorite) or anything else that puts out negative pulses where you provide the load resistor and coupling capacitor, like say a BF3 or 3He tube.

We are using a LED, LD1 to generate a bias signal that is more or less, one Vbe plus a little bandgap voltage. The Vbe tracks the tempco of the input transistor - temp compensation is feedwforward here, and the rest provides a little voltage for it to turn on and bias up the rest of the circuit. When set up right, all the transistors are just barely on, and the 220 ohm output resistor has about .25v across it; not critical, some is required, any will work but more draws more power. We keep that small so we can have nearly the full supply voltage as output swing and to keep the DC drop across that 47 ohm R1 low too.
We feed the input this bias with a 100k resistor, R2. This sets the effective input impedance around 66k as mentioned above, since the transistor itself is a little bit of a load as set up. If you want lower input impedance, you can just reduce this value to whatever you want.

The first stage is where the rubber starts to meet the road. Collector current is about 150 uA, a little on the high side for optimum noise at 100k input impedance, but required to get the drop on low enough value resistors to keep the speed up. Trade-off alert! Since we don't have exactly right temp comp - we need for this to have fairly low DC gain, and we get that by splitting up the emitter resistor into R4 and R5, with some of it bypassed for AC, so only R5 sets AC stage gain. Again, if you want a different LF rolloff, you can adjust C3 here.

The second stage has both voltage and power gain, again, to keep speed up we are trying to use low value R's to get a high rolloff frequency with any stray capacity. The last stage is just an emitter follower to reduce the output impedance further - enough to drive coax and take no prisoners. We AC couple that output (C4), and put in a series termination resistor, which serves two functions - one is to kill ringing on the coax to your other gear, and the other is to prevent the emitter follower from becoming a one port oscillator at some lengths of coax, and yes, I ran into that rather esoteric caveat in testing. We AC couple because we don't have any idea what your gear wants, but likely it's not 9v DC on the signal!

This is all optimized for negative going input pulses, that have faster fronts than backs (hard for me to say rise time for the beginning of a negative pulse...sigh). If you need the other polarity, the best way to get there is to flip all the transistor sexes (2n5088 for that first one) and diode polarites, and the power supply polarity. Rebiasing this for full drop across that 220 ohm output R will vastly increase the power drawn, and lose the match of signal type to preamp "filter".

Negative feedback is via the 15k R6 here - to change the gain, just change that value. If you've done rebiasing due to strange-different parts or a way different supply voltage, you should do that rebiasing either at high gain (as shown) or with no negative feedback at all - get it right that way, and it will be happy no matter what you change the gain to later.

This is now a done deal, and you can now buy them from us over at DJ's - or get a board and parts, or just build from this open source hardware thread - your choice. They do work better as this SMD board, I must say, but it does take the right tools and skills to build one like this. At this input impedance and gain - it can be hard to squeeze the most out of this design building it big on perfboard - more noise pickup and slower speed, with more tendency to oscillate can be expected on a big prototype, though it's pretty good either way.

But hey - if you're in the fine wine business - everything should be as right as it can possibly be, right?

Doug, I don't want to split hairs with you, or wotnot, but I'm interested in understanding how this design has evolved. When I was trying to pick up the measly 10pC worth of charge from the pulses from my SNM18-1 tube with the original 15k input impedance in your first circuit and it didn't work, you were discouraging me from using the 100k, later 220k, input resistors I ended up trying [which worked]. I did that because I could see the input impedance was dragging the voltage of that pulse down into the noise. Now I see you are - also - using 100k's worth on your input fet. You previously cautioned me that I was at risk of creating an oscillator. What's changed here?

I haven't tried this with those tubes yet, but I will soon. I don't bother unless I have a known working source of neutrons and at present, I have a little work to do there. I'm having some arcing on my main fusor that I want to fix before it breaks another feedthrough, and the new accelerator isn't online yet - still leak hunting on that one. I like to be sure I'm seeing a real neutron signal rather than a cosmic hit - they look different on the scope - or EMI, which is a pain to eliminate for each test, though at least the EMI looks so different on the scope there is no mistaking what you're seeing.

I was doing fine before, just using their suggested test circuit and time constants, and was wondering if there wasn't something different going on at your place that was unreported - wrong cap size or something (seen that one - sometimes they don't match the markings! Especially old HV ceramics.). At the time, I was having issues with the Russian 3He tubes with oscillations that never did show up on the B10 ones. Which is it you have? It appears, BTW that those oscillations are there in the 3He's at random and apparently are not sensitive to the load and supply conditions - they just come and go seemingly no matter what - but leaving the tube on for a long time (minutes to hours) seems to cure it until the next time you power up. A big coupling cap could make it worse, or actually damage the tube dumping charge in there when the tube momentarily tries to act like a short - and that's what I think we had discussion over. I believe they use quite a fine wire in there that won't take a bunch of joules dumped into it from a coupling cap with the other end effectively grounded to the protection diodes in the context of kilovolts on the other end.

If you're hanging a scope probe right on the tube to see that - then you're putting a ten meg R to ground at that point from the probe, which will throw things off - reduced tube corona current before you even put another load on it due to DC loading from the probe - which might have an effect on what you see. Of course, any load will reduce pulse size, and it doesn't matter if you make it up with gain, which you need anyway. Once you have that good noise figure out of the first stage, more gain is pretty easy to do.

I went with the highest impedance the circuit would work well with because it's easy to reduce when you want, at any rate - all you need is a smaller resistor there or one across the input. I think what's probably important here is getting all the time constants right for the signal, which is a function of both that and coupling cap size (as well as the effective impedance of the tube - I think we can neglect the 50 meg supply resistor safely here.

Note that with the 100k R, the computed input impedance is still about 66k, due to finite transistor gain. In reality, it's probably a little better than the oversimplified calculation I gave above because of the negative feedback bootstrapping the emitter of the input stage - at least at low gain settings where there's a lot of "loop gain". But actually, this circuit will have better noise figure at around 10k input impedance, due to noise current at the transistor base at the collector current we have set now. Ideal collector current for 100k noise figure would be so low, we'd have to use much bigger R's in the other leads, which would in combination with the strays, slow the thing way down. Luckily it's a smooth shallow curve that hugs "perfect" so it's not a major issue.

I should send fusor.net a schematic for a truly charge sensitive preamp - which would be a perfect integrator with a signal that would only increase with time. What ...(unprintable)...they can be over there.
From one cargo-cult myth to another...I sometimes despair of teaching people who think they already know everything. Crap, I've spent more years in front of a scope looking at stuff like this than they have hours - probably combined.

Doug;
Thanks for the help yesterday. I cut the NFB resistor to 10k and reduced LED resistor to 15k. Also added filter caps (6.8mfd electrolytic and 0.1 mfd ceramic and serial 47ohm resistor to the positive battery lead. Also boosted battery up to nine volts DC as suggested. This amp is truely fine wine now. I tested again with the Bicron 1.5 x 1.5 NaI PMT and now I get excellent negative going pulses with appropriate positive decay tail off. I also now get real response to my Am241 test source at least count wise that is inversly variable with distance from tube. I don't have any MCA spectra yet but that will follow. My current test sources are limited to K40, Am241 and Radium daughter sources (old CD geiger check sources, Anton and Lionel) so hopefully I can get some spectral analysis using this prototype. Thanks again! Larry Upjohn

P.S. Is the price on DJ's for the pre-amp for a kit of parts including the PC board or just for the board itself? The price is quite reasonable either way and will be added to my budget wish list. More later, LRU.

The price is for a built and tested complete unit - with any gain etc you specify (though I might have to take the time to get the right resistor value for that - I don't have them all in stock in 0805 smd). If you're in the US, shipping is by mail and pretty cheap. I got some fiddilng negative reactions to posting this over at fusor.net, sigh...you never know, I think this thing is pretty cool myself. It should be a good spectroscopy amp if you're using one of the common heads that have around microsecond pulses. It probably would fail miserably with something super fast, alike BaF, but NaI is nicely slow as are most of the other types.

I'm surprised you needed to change the led resistor for 9v operation....but oh - you don't have the same led I have, and they are all a little different, so yeah - makes sense. My tiny SMD led probably needs less current to get to voltage than yours does. It actually lights up a little with that 100k in series! In my original, I wound up using a yellow one to get it right with low current draw. What I was shooting for there is to have a 6-9v range with a least a little bit of output transistor turn on - basically the range for a 9v battery over a useful life. You should get a few hundred hours use out of one when this is set up right. We look like a 5 to 10k load on the battery when there are no pulses (depending on the bias you set, very roughly). At 100khz type pulse rates, the load is more (lots) - but that's not a real common use-case. Note, if you wanted to go back down to one Li cell, you could do it without too much trouble. You'd use a higher intrinsic voltage LED instead and adjust from there. The disadvantages would be reduced output dynamic range, but that's about it - you'd want to set the bias very carefully to just barely on for that output transistor so as not to waste potential output swing range. The advantages would be that an Li cell changes much less percent over its life - so you can set it closer and it'll stay that way, it's probably more power dense for its size, and probably smaller.

The disadvantage is they aren't as cheap...not sure how that'd work out in pennies per hour though - might be fine. I just had to pick something to do the design with, and I wanted TTL level outputs for counting applications. I've done a few with two Li cells that fit perfectly into a one - AA cell holder - finding a one cell holder can be a trick with those Li's. They work nicely and so far, haven't had to get into the boxes and replace them so the life isn't terrible.

Doug and Friends;
it has been a while since I last posted here but I have been slowly churning away on RAD projects between outside pressures. I have just done up my second breadboard FINEWINE preamp with better results than the previous version. I followed your schematic with out reducing the LED resistor as I noted earlier and stuck with the 100k (R3 I think) part as noted. Works great on the bread board so hopefully the hardwired version is with similar results. I am of course using through-hole parts for now but have had to substitute a 2N3906 for the 2N5087 which appears to work okay. I do have several questions though as follows. R12 and R13 are listed as optional pull-up and pull-down. What starting range do you suggest? My concern is that I have a large rapid negative pulse returning above ground to positive with an appropriate "tail off" decay. My scope is aging (1984 TEK 7804 storage scope) and I am no longer able to capture random pulse due to excessive flare in the CRT. It doesn't appear to me variable pulse height that would be necessary for Spectral work even though I realize my eyes are probably not up to the calibration they use to have. Will the pull-up/pull-down improve the pulse shaping? My second observation is that the D1/D2 1N4148 diodes became quite warm to touch on the bread board and I have pulled them. They read okay on the DVM however...This is enough for now. Thanks for your time even though this thread is probably not on the radar right now. Larry Upjohn.

P.S. Any more progress on the Home grown Hornyak buttons? LRU.

P.S.S. FINEWINE No. 1 Prototype is my go-to preamp now and has run for over a year on the original 9volt battery. I forgot to mention that I use it with both the BICRON 1.5 X 1.5 NaI detector and now with a newer XP2102/Plastic Scint block (Don Orey similar to BC410 I think.) and I get good response from both, the BICRON being the more robust. The DO plastic scint was sentive to neutrons (probably thermalized) but still responds to higher energy gamma ( CDV700 test sources and Thorium TIG rods) with little sensitivity to K40 or AM241 sources. I have no access to any neutron sources right now so can't attest to anything but possible background. LRU.

Doug;
I answered my own questions by doing some further research on back discussion over several threads. Best wishes and I will save you having to answer my previous post for now. Thanks again, Larry Upjohn.

Doug;
I saw the light. I have been using through-hole parts to assemble my FineWine prototypes including the 1N4148 diodes for input protection. I noted them heating and discovered that my PMT input signals were being totally attenuated on both the breadboard and prototype PC board. I forgot that 1N4148 diodes in the transparent glass body are light sensitive. The current from the PMTs is being shunted away but my pulser signal gets through (the negative pulse is markedly attentuated though). Removing the diodes from the circuit of course fixes the problem. My next attempt will be to cover the diodes with heat shrink and re-insert them to see if my thought solution is correct. I note that your SMT board has opaque diodes so this problem may not occur on them. I hope this finds you finds you all well out east there off the grid. Larry Upjohn.

I doubt that light based conductivity is responsible for your issue. Remember, the input of this, unless you provide your own coupling cap, floats at near the (filtered) positive supply rail.
Therefore, any DC load that tries to "pull" it away from there gets in trouble - and if it tries to go outside the rails, the diodes will get hot, but only then - the tiny fraction of a microamp these make as "photodiodes" just isn't enough to warm them.

When used with a phototube, I use a floating anode, and a negative HV supply to the cathode and divider string. The anode has no connection other than to the DC coupled input of the preamp - none at all - the divider string opposite the cathode end goes to ground, or ideally, the filtered rail we provided in the PCB. There should be zero current into or out of the DC coupled base of the input pnp. You might therefore have a little other looking to do to see what's wrong here. You should also check the output with a fast scope - the 3906 is fairly faster than the spec'd transistor and might make the loop unstable and oscillate, which could explain a lot.

When used with a positive supply 2 wire phototube connection, which seems to be industry norm, you're going to have to provide your own, high quality coupling cap, and I'd suggest still using the base (DC coupled) input for that. The coupling cap we show was not really big enough, and in SMD form, not high enough volt rating to really handle that case anyway. Here I use about a 10k series R from the positive supply to that kind of setup, and couple from the PMT/divider into the preamp for that situation, using about a .002uf cap on down to about 1000 pf, depending on the pulse shape I want.

Yup, all's well here, and I've been slacking on writing, but not doing. I'm about to put up a major update to the solar system here, having just done it in real life. I was so sore after moving things I can't lift into places I can't reach (with help from a local farmer and a crane) that I've been kinda too pooped to do much lately. But it's done now, and wow - it really improved things subjectively. I'll have better, quantitative numbers later once I measure them in real-world use.