This could go in several places, but since most of this is a digital project, I'm putting it here, with the software under embedded software.
This is actually the analog part of this project. This is the HV power and signal conditioning for the standard counter based on some NOS Russian pancake geiger tubes we got surplus. Hope we find more, they're really nice and extra sensitive to betas, which is what most activation products give for detection, a nice match. The tubes seem very stable and repeatable, making them ideal for this use.
Here's the schiz:
Edit: Due to real world testing there have been a couple of changes to the above. The 1 meg resistor from the HV supply has changed to ten megs, and the order of the 1 meg resistor and 47 pf cap after the geiger tube has reversed so there's less stuff up at HV on the board to attract dust and damp. Also, we found that by tying a 1 meg resistor (not shown) across pins 1 and 3 of this tube, the sensitivity goes up 25% or so, so I added that. I'll redraw this and re-post when I get a chance. /edit
There is more skull sweat in this than meets the eye at first. Sometimes it takes extra work to get to real simplicity. I am using an off the shelf CCFL inverter well below its normal supply voltage (which would have been 12v) and even then, with only a half wave rectifier, it gives a little too much voltage. Also, these are noise generators, and we'd rather not kick a ton of noise back into the main 5v power, nor put a big capacitor directly across the USB power bus - the spec says that's a no no. So a 47 ohm resistor handles both of those requirements. This one needs to be a half watt so it can't fry no matter what, and it gets very slightly warm in use. More could have been done to reduce power draw, but hey, that's coming from a computer that presumably has plenty, and it takes a lot more parts to do it -- a voltage doubler and down regulator for the CCFL are required to cut the power to about 1/4 of what it is, but really since a bigger drop in a regulator also uses some wattage, you only get to about 1/2 this unless that's an efficient switcher. Nope, we KISS around here, and this page draws right around 20 ma off the 5v bus.
I show the 1 meg resistors as quarter watt. In this case it has nothing to do with power, but size and voltage standoff ability. A too tiny package risks internal breakdowns with high voltage drops. I like the blue 1% resistors for digi here. The 47k resistor is a 5v pullup, and the load for the attenuator that takes the 100 or 200 volt pulses down to size (about 10v), so the diodes can clip them to the digital rails and hand the PIC counter a very nice, clean, waveform to count. In the prototype, I didn't add the 1n4148's at all - there are already some wimpy diodes in the PIC itself, but for production, these should be in there to protect the PIC better. A dust and humidity induced arc across the second 1 meg resistor would take out the PIC without them.
The CCFL as shipped has two capacitors in series with the transformer secondary, used as ballasts when used with a CCFL lamp. We don't need and can't allow them here with a simple half wave rectifier, so I pulled them out, did a little creative PCB track cutting, and put the diode and filter cap back on the CCFL board itself. You can put them separate if you like, this just saved space and made my sorry bench kludge a little more robust during development. As the geiger tube draws almost no current even when counting, and the CCFL runs about 50khz, this simple power supply is actually quite noise and ripple free -- way better than actually needed for this, and since the geiger tube is very unpicky about voltage, no further regulation is needed -- it would just add parts that could fail, and definitely cost money.
I'll get the digital part here as another post as soon as I get it drawn. It's pretty simple too.