[Swlug] DIY Geiger counter success?

Mon Mar 17 08:14:18 UTC 2025

At Z+0000=2025-03-16Sun17:28:10, Rhys Sage via Swlug sent:
> I rebuilt the circuit on the breadboard again, having had no worthwhile readings from my oscilloscope. I'd been excited to see a faint wavy line on the scope. That excitement was dashed when I noticed the scope did exactly that line, on its own.

    That's why I was not at all hopeful about your 0.01v voltage variations.  This is clearly a very tricky circuit to get working.

    In your last breadboard circuit, I noticed a missing connection between the 1st transistor collector I think it was and the positive power rail.  Please can you update us with another photo so I can check the wiring for you?

    One of the potential problems that I identified was your huge loop areas on the breadboard.  Did you try my suggested mitigations?

    Did you try the idea of testing with a different photodiode and giving it test pulses of light from an LED?  Then keep shortening the pulses and see how far you get, and whether you can get it down close to tens of nanoseconds?

> I connected a speaker where it was supposed to feed to an Arduino and to the power line. I got a very quiet low frequency and consistent buzzing noise. That is welcome.

    Hmm.  Sounds like all the little loop antennae on your breadboard might be picking-up 50Hz hum EMI.  Can you try reducing loop areas as much as possible, including adding capacitors across the supply, as close as possible to where it is used, such as to truncate the effective loop areas there.  30k of gain is a very sensitive amount of amplification, so it's going to amplify all this EMI if you don't manage to get things to be a lot tighter.

    Why is a buzzing noise welcome if it is not what your circuit is designed to do?  I think somehow you need to try getting a bigger test signal in there with essentially a different photodiode rig driven by a test LED.  Then when you get an actual signal that you intended, rather than just noise, we can calculate a signal-to-noise ratio, and then get an idea of how far off is your circuit's performance from being able to detect a 20ns pulse.  This way, you would be able to monitor your progress and see whether or not each change helps or not.

> Now I'm wondering if I should be getting a continuous buzzing.

    Yeah, no.  You should be getting pulses, which might not sound like anything on the buzzer having drawn an envelope over them, but if they are audible, they should sound like clicks.  Because of this, I do not advise this as the best way to test that something is going on.

    I think what I would try next is to try testing with a voltmeter on the output, and to remove the resistor of the envelope detector subcircuit (the 10MΩ one) for now.  This will ensure that once you get something trigger the input, the voltage of your output will drop.    You should then be able to reset it back up by connecting your 1MΩ resistor across the capacitor.  Most likely, with your circuit being so sensitive to noise, the voltage will drop as soon as you remove that resistor each time, but at least it gives something to experiment with.

    Test this against a "control" photodiode that you dipped in thick black paint or something (and without the biasing resistor that I suggested -- this would be a later performance enhancement).  Try reducing loop areas and see whether you see an improvement in stability.  My understanding is that the page 78 circuit, but with the Geiger--Muller tube being replaced with a photodiode, and with the 10MΩ resistor removed, if the photodiode is painted black such that no current passes through it, the voltage on the capacitor /should/ remain constant.  The only other reason it would change is due to EMI (and a tiny bit of parasitic conductance, but I don't think it would be significant, because capacitors hold their charge for long enough to measure them, and in many cases much much longer -- months or years -- that's how flash memory storage works).

    So basically what I am suggesting is to temporarily alter the circuit such that the capacitor voltage should stay at whatever you set it to, and then see whether it does stay, or whether actually the EMI is way too much of an influence, which I strongly suspect.

> I'll play with it a bit more. I'm running it now off a 9v battery. The 10M resistor ended up as 1M as I don't have 10M.

    Of course, if you happen to have a paper-bound strip of 1MΩ resistors (or an assortment or whatever), 10 of them in series would be 10MΩ.

    But 10MΩ is quite unusually-high for low-voltage / small-signal stuff, so its mere presence in that page 78 circuit indicates to me that what you are dealing with here is probably not compatible with a breadboard, simply because of the loop area problem.  Did you try watching that video I mentioned by W2AEW about prototyping techniques that work better for radio frequencies?  Did you try any of my suggestions for reducing loop area on your breadboard to at least mitigate the problem, such as keeping the power rails adjacent, or adding capacitors across them to truncate their effective loop area?

> A noise is better than I've had so far. Better than that - a noise that wasn't accompanied by plumes of magic smoke.

    :-]  It's certainly exciting to have an occasional encounter with the magic blue/grey genie, but be sure to test that your components have not magically morphed into insulators or conductors during the genie's visit! ;-)

> I have a bit more fiddling to do with it then I had an ampule of Tritium to test it on, on the way. After that I'll code the Arduino driver.

    You be careful.  I seriously advise doing the tests that do not require dangerous substances first, like swapping the photodiode for one covered in black paint (a "control") and then for one controlled by an LED (effectively an "opto-isolator").  Then again, you can skip the black-painted one if you just were to make a optoïsolator sealed from outside light but then leave the LED disconnected for the control test, as the effect should be the same: no light to the photodiode; and therefore almost no current flowing; therefore very little amplified current; therefore very slow change in voltage of the capacitor.

    You might find that in these tests with the 10MΩ resistor removed, the small value of the capacitor still changes much too quickly.  You could try a bigger capacitor there just for the test.

    The general idea here is to keep stripping-back the function of the circuit until you find that it behaves as expected/intended, then start adding that functionality back into the circuit, step-by-step, and see where it goes wrong, or what level of sensitivity you get to before the noise starts to dominate.  When you find the edge of where it stops working, that is the ideal setting to experiment with, because each thing you try will do something different and give useful feedback as to what works and what doesn't.  If each thing you try results in an signal-to-noise ratio of zero, you are not learning much from those experiments.  It is more useful, both for motivation and for information-theoretic efficiency of learning, to have successes at least 50% of the time on average.  That's why I'm trying to break this big problem up into multiple smaller problems that are easier to actually achieve.

    One other idea -- have you set the trigger on your oscilloscope such that each pulse starts at the same place on the screen?  If you are not already doing that, then it should make it a lot easier to actually see the momentary pulses flashing up above the solid line a bit like 1940s cinema film dust noise -- very quick but you can see its presence thanks to your own persistence-of-vision.

Kind regards,
James.
-- 
Wealth doesn't bring happiness, but poverty brings sadness.
Sent from Debian with Claws Mail, using email subaddressing as an alternative to error-prone heuristical spam filtering.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.lug.org.uk/pipermail/swlug/attachments/20250317/62b17f84/attachment.sig>